id stringlengths 3 9 | source stringclasses 1 value | version stringclasses 1 value | text stringlengths 1.54k 298k | added stringdate 1993-11-25 05:05:38 2024-09-20 15:30:25 | created stringdate 1-01-01 00:00:00 2024-07-31 00:00:00 | metadata dict |
|---|---|---|---|---|---|---|
269148660 | pes2o/s2orc | v3-fos-license | Linear Cross-document Event Coreference Resolution with X-AMR
Event Coreference Resolution (ECR) as a pairwise mention classification task is expensive both for automated systems and manual annotations. The task’s quadratic difficulty is exacerbated when using Large Language Models (LLMs), making prompt engineering for ECR prohibitively costly. In this work, we propose a graphical representation of events, X-AMR, anchored around individual mentions using a cross-document version of Abstract Meaning Representation. We then linearize the ECR with a novel multi-hop coreference algorithm over the event graphs. The event graphs simplify ECR, making it a) LLM cost-effective, b) compositional and interpretable, and c) easily annotated. For a fair assessment, we first enrich an existing ECR benchmark dataset with these event graphs using an annotator-friendly tool we introduce. Then, we employ GPT-4, the newest LLM by OpenAI, for these annotations. Finally, using the ECR algorithm, we assess GPT-4 against humans and analyze its limitations. Through this research, we aim to advance the state-of-the-art for efficient ECR and shed light on the potential shortcomings of current LLMs at this task. Code and annotations: https://github.com/ahmeshaf/gpt_coref
Introduction
Event Coreference Resolution (ECR) involves identifying events that refer to the same real-world occurrence both within and across documents.Traditionally, ECR is performed on pairs of event mentions in a corpus through the use of rules, features, or neural methods to generate similarity scores (Kenyon-Dean et al., 2018), with neural methods such as Transformer-based encoders (Devlin et al., 2019;Liu et al., 2019;Beltagy et al., 2020) achieving state-of-the-art performance on various ECR benchmarks (Caciularu et al., 2021;Held et al., 2021).However, the quadratic nature of pairwise approaches makes it challenging to scale up to large corpora of thousands of documents.
Figure 1 presents three event mentions (m 1 , m 2 , and m 3 ) with their respective event triggers highlighted.m 1 and m 2 are examples of coreferent events, while m 3 is a related yet non-coreferent event.While m 1 and m 3 contain sufficient information required to make a negative coreferencing decision between them, additional extrasentential context is needed to determine the coreferential relationship between m 2 and the other two mentions.
The challenge of ECR stems from the inherent issue of establishing singular terms for event mentions (a, b) that can be compared for identity (is a = b?; Davidson (1969)).Consequently, pairwise methods resort to approximations of the coreference relationship between each mention pair by leveraging either the sentence or the entire document for contextual information.The methods that need to rely on the entire document for each pairwise decision (as in the case of m 2 ) are intractable on large corpora.We propose that by extracting Target Mention (m 1 ) HP today announced that it has signed a definitive agreement to acquire EVT EYP Mission Critical Facilities Inc.
Target Mention (m 2 )
Financial details of the acquisition EVT were not disclosed.
Target Mention (m 3 )
Earlier this month Hewlett-Packard unveiled a bid of nearly $14 billion bid to purchase EVT Electronic Data Systems.the key semantics of mentions and by introducing a graphical structure between each mention, we can compress the information.This way we not only are able to create identifiers for the mentions that can be compared for sameness, but also make ECR completely linear in complexity.Figure 2 illustrates the graph structure based on our new Cross-document Abstract Meaning Representations (X-AMR), inspired by Abstract Meaning Representation (AMR; Banarescu et al. (2013)).X-AMR captures event triggers and arguments linked using VerbNet Lexicon (Schuler, 2005) and a Knowledge Base (KB), resulting in corpus-level event graphs that we use to symbolically perform ECR without the need for pairwise scoring.Our specific contributions in this paper include: • X-AMR annotations using the annotation tool provided by Ahmed et al. (2024) on the ECB+ corpus (Cybulska and Vossen, 2014).
• A novel ECR algorithm over the X-AMR graphs that avoids the computational cost of traditional pairwise approaches.
• An evaluation of the algorithm using goldstandard annotations for the ECB+ corpus.
• And finally, an evaluation of the approach with automatically generated X-AMR graphs using GPT-4 in zero-shot and few-shot settings with prompts based on a condensed version of the annotation guidelines of X-AMR.
Together, our annotations and findings suggest a promising path toward extending robust ECR to real-world applications where exhaustive pairwise approaches are not feasible.
Annotation Guidelines for X-AMR
We aim to annotate key event semantics with four arguments, ARG-0, ARG-1, ARG-Loc, and ARG-Time, capturing agent, patient (and theme), location, and temporal information.The selection of these arguments is to circumscribe an event by its minimal participants (Lombard, 2019;Guarino et al., 2022).We use the guidelines presented in the next section to hand annotate the roleset and argument information for the ECB+ train, development, and test sets using the standardized split of Cybulska and Vossen (2014).Following the annotation guidelines, we provide the enriched annotations of the ECB+ corpus by two Linguistic students.We use the prodi.gy-basedX-AMR annotation tool provided by Ahmed et al. ( 2024) 1 .
PropBank & AMR
Semantic role labeling (SRL) centers on the task of assigning the same semantic role to an argument across various syntactic constructions.For example, the window can be the (prototypical) Patient, or thing broken, whether expressed as syntactic object (The storm broke the window) or syntactic subject (The window broke in the storm).
1 Readers are encouraged to check the original paper for details about the annotation tool 2022)) has over 11,000 Frame Files providing valency information (arguments and their descriptions) for fine-grained senses of English verbs, eventive nouns, and adjectives.Figure 3 gives an example Frame File for agree as well as an instantiated frame for HP has an agreement to acquire EYP.
The resulting nested predicate-argument structures from PropBank style-SRL also form the backbones of AMRs, which in addition includes Named Entity (NE) tags and Wikipedia links (for 'HP' and 'EYP' in our example).AMRs also include explicit variables for each entity and event, consistent with Neo-Davidsonian event semantics, as well as interand intra-sentential coreference links to form directed, (largely) acyclic graphs that represent the meaning of an utterance or set of utterances.
Our enhanced X-AMR representation follows AMR closely with respect to NE and coreference, but stops short of AMR's additional structuring of noun phrase modifiers (especially with respect to dates, quantities and organizational relations), the discourse connectives and the partial treatment of negation and modality.However, we go further than AMR by allowing for cross-document coreference as well as multi-sentence coreference.X-AMR thus provides us with a flexible and expressive event representation with much broader coverage than standard event annotation datasets such as ACE2 or Maven (Wang et al., 2020).
Roleset Sense Annotation
The first step in the annotation process involves identifying the roleset sense for the target event trigger in the given text.Annotators, using an embedded PropBank website and the assistance of the tool's model, select the most appropriate sense by comparing senses across frame files.Handling Triggers with No Suitable Roleset: If there is no appropriate roleset that specifies the event trigger, particularly in cases when the trigger is a pronoun (it) or proper noun (e.g., Academy Awards), the annotator must then search for a roleset that defines the appropriate predicate.
Document-level Arguments Identification
Next, we identify the document and corpus-level ARG-0 and ARG-1 of the selected roleset.Annotators use the embedded PropBank website as a reference for the roleset's definition, ensuring that the ARG-0 (usually the agent) and ARG-1 (typically the patient) are consistent with the roleset's constraints.For arguments that cannot be inferred, the annotators leave those fields empty.
Within-and Cross-Document Entity Coreference Annotation: Annotators perform within-and cross-document entity coreference using a dropdown box of argument suggestions (suggested by the model-in-the-loop), simplifying coreference link establishment.In difficult cases like m 2 (Fig 1 ), where ARG-0 and ARG-1 are missing, the dropdown box helps by suggesting "HP" and "EYP" from the m 1 sentence.Similarly, in m 4 (Figure 4), the drop-down box assists in resolving ARG-0 (it) as "HP", using the information earlier within the sentence.Annotators are also allowed to input multiple values separated by "/" as needed, (e.g., if two people performed some action together, "Person 1/Person 2").
Nested ARG-1: In many cases, the ARG-1 may itself be an event.In such cases, the annotator is tasked with identifying the head predicate of the ARG-1 role and providing its corresponding roleset ID.We then search for the annotations for such an ARG-1 and connect it to the target event.Fig 4 has an example of a mention with an eventive ARG-1.For this, the annotator needs to provide the roleset for the predicate of the ARG-1 clause (agree.01)as the ARG-1 in this annotation process.
ARG-Loc & ARG-Time Identification
Annotators may also utilize external resources, such as Wikipedia3 , or Google-News, for the accurate identification of temporal and spatial arguments.This is required when the document does not explicitly mention the location and time of the event.
Human Annotations
To perform the X-AMR annotations, we employ two annotators, and we execute this process in a systematic two-step approach.In the initial phase, these annotators are responsible for identifying the roleset ID associated with each event trigger.We aggregate all event mentions for which both annotators have concurred on the same roleset ID.For those instances where there is a lack of consensus between the annotators, we enlist the assistance of an adjudicator to resolve the discrepancies.The annotations that have been finalized, either through agreement or adjudication, are then collectively advanced to the subsequent task of identifying the arguments.
Annotation Analysis
We have currently annotated all the mentions in the corpus with their Roleset IDs and 5,287 out of the 6,833 with X-AMR.In the three splits, only the Dev set has been fully annotated.We calculate the interannotator agreement (IAA) on the common Roleset predictions.The IAA is highest for the Dev set at 0.91, as depicted in Table 1.Consequently, we utilize the Dev set as our benchmark for experiments in the following sections.Arguments: Our analysis reveals a significant presence of mentions with nested ARG-1 annotations, as highlighted in Table 1 (w/ Nested ARG-1).This underscores the importance of capturing nested event relationships effectively.Additionally, our annotations for location and time modifiers successfully capture this information for nearly all mentions (w/ X-AMR), thanks to the assistance provided by drop-down options and the model-in-the-loop approach.These tools are particularly valuable in cases where date references are not explicitly mentioned in the document.
Graph-based ECR Algorithm
Our proposed approach for ECR builds upon previous research efforts that use minimum participants.Cybulska and Vossen (2013) utilize heuristics to ascertain event relationships based on various factors, such as location, time, and participant compatibility.Choubey and Huang (2017) employ iterative techniques to identify event relations, both within and across sentences.It's important to note that both of these approaches are pairwise methods and do not incorporate cross-document entity coreference into their methodologies.In contrast, our approach with X-AMR not only leverages cross-document entity coreference but also capitalizes on AMR's nested event structure for ECR.
EID: Event Identifiers
We generate EID using the roleset, ARG-0, and ARG-1.To evaluate the influence of location and time, we produce EID lt by incorporating ARG-Loc and ARG-Time.These identifiers facilitate comparison between two events, allowing coreference resolution by matching the identifiers.Specifically, two events (m i , m j ) are deemed coreferent (where coref(m i , m j ) is true) if any of their identifiers match in EID(m i ) and EID(m j ) match, as illustrated in Equation 1.
Even though Eq 1 is represented pairwise, we design the clustering algorithm by first creating buckets of mentions with the same identifiers.This way we generate a sparse binary similarity matrix of only the pairs of mentions in the same buckets representing the EIDs.
EID Generation
We generate the identifiers differently for Standard events (m, like m 1 , m 2 , m 3 ) and Nested Events (m e , such as m 4 and m 5 ).For standard events, the identifier (EID 0 ) is constructed by merging the ARG-0, roleset ID (PB), and ARG-1 as shown in Equation 2. For instance, EID(m 1 ) is denoted as 〈HP, acquire.01,EYP〉.
In the case of Nested Events (ARG-1 is also an event), we employ a recursive strategy to generate identifiers.Specifically, we produce multiple EIDs by traversing the arguments of nested events up to a maximum depth, N, as delineated in Equations 3 and 4.This method aims to connect the root event to a standard ARG-1 within the event chain.This procedure is denoted as EID n , where n indicates where × is the Cartesian product for generating all the concatenations of the tuples.
Using Eq 3 and 4, we generate the identifiers for m 4 and m 5 (Figure 5) as shown below: The 2-hop identifier of m 4 is exactly same as the 1hop one of m 5 , except for the roleset IDs.To detect synonymy between the rolesets, we use VerbNet (PB HVN ), and we maintain a KB to link the arguments.
With the combination of all these components, we can infer that m 4 is coreferent with m 5 .We also use the ARG-Loc and ARG-Time to separately generate an identifier, EID lt , for both kinds of events as shown in Equation 5. (5)
Clustering Methods
We generate the adjacency matrix of the mentions by using certain baselines and the event identifiers.The adjacency matrix is then used for hardclustering the events by finding the connected components.
Baseline (LEM):
Clustering mentions with the same lemma for their triggers serves as our baseline method.EID N ∨ EID lt : We cluster the mentions when they have either the same EID N or EID lt .
Rolesets IDs (PB
We also include the VerbNet class versions of the final two methods. In addition to the individual methods listed above, we also employ combinations of methods on the annotations of the two annotators (A 1 , A 2 ).∧clustering employs the rule that two mentions should have the same annotations from A 1 and A 2 (A 1 ∧ A 2 ).∨-clustering employs the rule that any of the two annotators's annotations could be the same for two mentions (A 1 ∨ A 2 ).2019)).To evaluate recall, we compute the mean recall values from MUC and B 3 (R avg ).Similarly, our precision metric, P avg is derived from the average precision values of MUC and B 3 .Our primary measure of overall performance is CoNLL F1.We applied various algorithmic methods to the ECB+ development set, which has been annotated using the X-AMR framework by A 1 and A 2 .Each annotator's performance is independently assessed, along with the ∨-clustering We report the baseline results using only the lexical information, and, the ECR performance of the proposed graph-based algorithm on the X-AMR annotations of A 1 and A 2 , and, a union of A 1 and A 2 (A 1 ∨A 2 ).Boldened are the interesting results.
ECR Results of A
approach, A 1 ∨ A 2 .We collate the results in Table 2 4 .
From the table, it is evident that utilizing the roleset IDs (PB H ) achieves a better result than lemmas (LEM).Even though PB HVN has the highest recall of 91%, the overall performance is quite low.The 9% recall error indicates the gap in the VerbNet class annotations for all the PropBank rolesets.This suggests there may be room for refining the VerbNet-Pro annotations for better compatibility (Spaulding et al., 2024).
Comparing annotators, A 1 provided more accurate annotations than A 2 , particularly in identifying location and time elements.Both annotators performed best with the EID N ∨ EID lt setting, with A 1 recording the best CoNLL F1 score of 79.8%.A 2 's annotations would need further refinement in order to match A 1 's recall.When considering precision, the EID N ∧ EID lt method stood out, with A 1 registering the highest precision at 92.8%.
For A 1 ∨ A 2 , the results are mixed.Although the recall is consistently higher than any individual annotator, it does not beat A 1 's best CoNLL.This method achieves the best recall of 86.9 when used in conjunction with the VerbNet classes while also having a CoNLL F1 greater than 70%.The mixed results for the combined method underscore the complexities involved in integrating and harmonizing annotations from different sources.
A CoNLL F1 of 80% seems to be an upper bound for a purely symbolic approach for ECR.However, we want to stress that after annotating X-AMR, we are in effect collecting free ECR annotations (eg.75% coreference links with 93% precision).ECR annotations are traditionally done in a pairwise manner, an approach that is tedious and error-prone (Song et al., 2018;Wright-Bettner et al., 2019).In contrast, X-AMR has an annotator-friendly methodology where an annotator would need to read a particular event mention typically only once.It also avoids annotation errors cascading into subsequent mentions as demonstrated by the high precision of our method.
GPT-4 as Annotator
Recent work in prompt engineering converts a textbased natural language task to a corresponding structured prediction task.In this spirit, we create prompts for extracting the X-AMR graph for a specific event, by providing the instructions for the task, exemplars for the structure of the response, and the right context for in-context zero-shot learning.Due to budget and time constraints, we make a smaller subset (120 mentions) of the development set (dev small ) to run our experiments on.We then assess the performance of GPT-4 (September 27, 2023 version) against the human annotations for this subset.We try two prompt engineering techniques (G 1 , G 2 ) to extract the X-AMR graphs of the events, and use the EID generation and clustering methods from §4.2.
G 1 : Prompt Engineering
For G 1 , we use a straightforward approach to generate the prompts.We start by generating a list of instructions.As shown in Figure 7, we arrive at five instructions by condensing the relevant sections of the annotation guidelines.We adopt a semiautomated way of generating the instructions, in which we first pass the relevant sections to Chat-GPT and then hand-correct its output.
Structured Prediction: Label Definitions
We then prompt GPT to produce a JSON output as the response.We offer detailed definitions for the keys in the JSON string, as illustrated in Figure 6.Additionally, we incorporate the coreference key and prompt GPT to generate Wikipedia links in the format "/wiki/Title_Name".Labels for Chain of Thought reasoning (Wei et al., 2022) are also
Label Definitions
Here are the definitions of the keys in the JSON output:
Annotation Instructions
You are a concise annotator that follows these instructions: 1. Identify the target event trigger lemma and its correct roleset sense in the given text.2. Annotate the document-level ARG-0 and ARG-1 roles using the PropBank website for the roleset definitions.3.If the ARG-1 role is an event, identify the head predicate and provide its roleset ID. 4. Perform within-document and cross-document anaphora resolution of the ARG-0 and ARG-1 using Wikipedia.5. Use external resources, such as Wikipedia, to annotate ARG-Loc and ARG-Time.
Figure 7: The condensed annotation instructions serve as a guide for GPT-4 in its generation of X-AMR event extraction.
included, addressing questions like "Is it a Nested Event?", "What is the event trigger?", "Who are the participants?",and "When and where did the event take place?".The final key in the list is "Event Description" that is a way to prompt GPT to produce a concise sentence encapsulating the event arguments including Time and Location.Finally, we add the entire document of the event and the sentence with the marked trigger (phrase in the sentence sorrounded by <m> and </m>) as context, and then prompt GPT to generate the corresponding JSON response.
G 2 : Prompt Engineering
A challenge observed in G 1 is its inability to determine specific pieces of information, particularly, 'Location' and 'Time' when they are absent within the source document.To address this shortfall, we introduce a complementary method, G 2 .
Event Descriptions: In G 2 , instead of relying solely on the document's raw content, we incorporate additional context derived from the event descriptions of what we term as complete events.These complete events are identified across all documents related to a specific topic at the prediction stage.A complete event is characterized by having all its requisite arguments, including Time and Location, predicted by G 1 .
De-duplication:
To enhance the quality and relevancy of this list, any coreferent events (duplicates) are eliminated.
Event List in Context:
With the refined list, we pivot from using the entire document as context (as practiced in G 1 ) to utilizing this labeled list of Event Descriptions.Furthermore, the description of the current target event is also included.
Best Matching Event Description:
We introduce a label called "Best Matching Event Description" at the beginning of prediction.This label pinpoints the most comprehensive and relevant description in correlation to the target mention.The intention behind this is to direct GPT's attention to a singular event description, enabling it to supplement the arguments not identified by G 1 .
In essence, G 2 furnishes a richer context, combining aggregated information from various documents, to rectify the limitations observed in G 1 .
ECR Results of G 1 and G 2
We compare the methods G 1 and G 2 (The cost for running G 1 was $4, and G 2 was $6.) separately with A 1 (the annotations with better quality among the two annotators) on dev small .As shown in Table 3, the roleset identification by GPT-4 is impressive, thereby we only see a 3 point difference between PB H and PB G .In dev small , we observe the results are bounded by recall, therefore we use the VerbNet class approaches.
When using EID methods, A 1 achieves the best CoNLL F1 of 83.6.When it comes to GPT-4, both G 1 and G 2 , fell terribly short of A 1 and do not even surpass the Roleset ID baseline (PB G ), with G 1 's best performance is short by 25 points and G 2 by 15.The shortcoming of G 2 can mainly be attributed to the failure of capturing nested events (only 5 of the 26 nested event arguments were identified).Interestingly, these methods consistently improve performance over the VerbNet baseline (PB GVN ).Between G 1 and G 2 , we see a large performance increase (G 2 over G 1 by 10 points), emphasizing the benefits of using corpus-level Event Descriptions in the prompts.
The results reveal the limitations of GPT-4 on this task.However, efficient usage of corpus-level information in generating X-AMR graphs lays out an exciting path forward for future work.Table 3: ECR results comparing A 1 with the two prompting methods G 1 and G 2 , on dev small .We report the baseline results using only the lexical information, and, the ECR performance after leveraging X-AMR.Boldened are the interesting results.
Algorithm Complexity
We conduct an artificial experiment to empirically demonstrate the ECR algorithm's linear complexity when coupled with X-AMR.In this experiment, we expand the annotated event mentions of the development set (by duplicating) to create a sizable collection comprising 200,000 mentions.Next, we systematically execute the algorithm across varying ranges, from 60,000 to 200,000 mentions.For each iteration, we measure the time the ECR algorithm 60 00 0 80 00 0 10 00 00 12 00 00 14 00 00 16 00 00 18 00 00 Mention Count Explanation: In ECB+ these events are falsely labelled as coreferent.
Incorrect VerbNet Class 9.1 [...] who was gunned down at their office Christmas party.
Table 4: Qualitative Error Analysis of the ECR Algorithm for A 1 on dev small .In Snippet, the event triggers are in bold font, and the key texts that help recognize the errors are underlined.
takes to run and depict the plot in Figure 8.The figure shows that the algorithm's running time is roughly linear.Efficiency-wise, the algorithm would take under 30 seconds even when the number of mentions surpasses 100,000, thus presenting a tractable solution for ECR at scale.
Error Analysis
We analyze the errors made by our system by examining the clustering decisions based on the A 1 , G 1 , and G 2 annotations of dev small (121 mentions).
In the human-annotated A 1 , we observe that ∼1.6% (2/121) of the misclassifications were due to annotator misinterpretations of the passages.In another ∼5.8% (7/121) of cases, an incorrect cluster is assigned due to errors in the original ECB+ dataset's labels, which are made evident by mismatched X-AMR arguments.For example, the ECB+ labels erroneously consider football coach Jim O'Brien's separate terminations from Ohio State University and the Philadelphia 76ers as the same event.Finally, ∼9.1% (11/121) are misclustered due to PropBank labels which do not yet exist in VerbNet and so do not belong to a class; e.g. the newer "opening fire" roleset wasn't identified as being of the same class as "shoot".Examples of each type of error are provided in Table 4.
In addition to the problems faced by human annotators, the machine-created annotations G 1 and G 2 suffered heavily from their inability to access external resources to resolve relative times, locations, and references to entities, in addition to inconsistent annotations (which the human annotators did not suffer from due to the saved argument drop-down), e.g ['South_Richmond_Hill,_Queens', 'Queens', 'Richmond_Hill,_Queens'] all refer to the same place 5 . 5For a more comprehensive list of examples, please refer to the provided Excel file in the repository
Limitations & Future Work
One limitation of our approach is that we require the PropBank resource for a particular language.In addition, the annotation tool is for-pay software.However, PropBank now has annotations for Chinese, Arabic, Urdu, Hindi, French, German, Spanish, and Catalan, and the annotation tool also works on a variety of languages.Our future work involves annotating X-AMR on a Spanish corpus.
The annotation tool released by Ahmed et al. ( 2024) omits a lot of AMR information (e.g.modality and negation), sticking strictly to the concept of minimal information for ECR.We also do not empirically demonstrate the efficiency of the modelin-the-loop annotations in this work.We leave the tool enhancements, including the incorporation of GPT-in-the-loop and a thorough analysis of the annotation efficiency (like Cai et al. (2023) andAhmed et al. (2023b)) for future work.
The results for both human and GPT-annotated approaches fall short of state-of-the-art techniques for ECR that involve heuristics (for filtering) and fine-tuning BERT in a pairwise manner (Held et al., 2021;Ahmed et al., 2023a).We hypothesize that the X-AMR annotations might be beneficial to the heuristic-based filtering step in these methods.The Event Description generated by G 1 could also be employed while fine-tuning BERT which we believe is an interesting direction for neuro-symbolic methods for ECR.
Two main issues of using GPT-4 in our work are Data contamination (Magar and Schwartz, 2022;Wu et al., 2023) and reproducibility.Since both PropBank and ECB+ are publicly available resources, it is most likely that the test sets might be part of its pre-training data.We argue that since our task is vastly different from the pretraining task, the effect of contamination is minute as demonstrated by the results.Reproducibility, however, is a much greater limitation.By providing the GPT-4 output on the train set (will release this upon acceptance), we set a mechanism to distill the knowledge into smaller in-house reproducible models like LlAma (Touvron et al., 2023) (Or even much smaller traditional auto-regressive models like FLAN-T5 (Chung et al., 2022)) for future work.
Finally, we limit the scope of our work to gold mentions instead of predicted mentions (Cattan et al., 2021).As a result, we could not compare X-AMR directly with the output of standard AMR parsers (Flanigan et al., 2014).Future work can approach this in a two-step way, with the first step being trigger identification, and then we can employ X-AMR on the predicted mentions.
Related Work
Document-level event extraction and event extraction with prompts (Li et al., 2021;Yang et al., 2022a;Xu et al., 2022) has been a major source of inspiration for our work.We extend this methodology for a more comprehensive cross-document level extraction by taking into account the named and unnamed arguments from previously seen documents into the annotation framework.
The Generative Pre-trained Transformer (GPT; Radford et al. (2018Radford et al. ( , 2019))) is an auto-regressive Transformer (Vaswani et al., 2017) language model developed by OpenAI, demonstrating exceptional performance across various natural language processing tasks.It uses a unidirectional, selfattention mechanism for effective context representation and is pre-trained on extensive unsupervised text corpora.The model follows a two-stage process of pre-training and fine-tuning, allowing it to adapt to specific tasks with minimal labeled data.GPT has undergone several iterations, with GPT-4 (OpenAI, 2023) being the most recent.
In recent years, research has increasingly focused on evaluating GPT's performance in multitask and zero/few-shot learning scenarios (Brown et al., 2020;Kojima et al., 2023).For instance, the study conducted by Radford et al. (2019) assesses the effectiveness of various LLMs in a zero-shot learning setting.Their findings imply that these models have the potential to equal, if not exceed, the performance of existing baselines on a range of NLP benchmarks.
Our objective is to underscore the importance of X-AMR with a focus on event coreference resolution, which integrates PropBank (Palmer et al., 2005;Pradhan et al., 2022) SRL as an intermediate phase.This approach is motivated by GPT-4's capability to produce free-text SRL for individual events (Zhang et al., 2022) instead of directly generating interconnected event graphs, as necessitated by AMR.By leveraging GPT-4's strengths, our suggested method can offer a more thorough and effective representation of events in a given text while preserving their structure and relationships, and therefore facilitate ECR.
Besides the hybrid approach and prompt engineering, we also stress the need for a linear algorithm over a quadratic ECR method, utilizing the generated graphs.Quadratic ECR with GPT (i.e., binary coreference decision between mention pairs) has produced negative outcomes, as evidenced by Yang et al. (2022b).Furthermore, this method would be expensive, potentially costing hundreds of dollars to execute using GPT-4.By adopting a linear algorithm, we aim to address these limitations, offering a more cost-effective and efficient solution for ECR.We propose a linear graph-based method for ECR using the generated key semantic information for the event mentions.
Over time, efforts have been made to enrich event datasets, such as the Richer Event Descriptions (RED; O'Gorman et al. ( 2016)) corpus and the Event Coref Bank plus corpus (ECB+; Cybulska and Vossen ( 2014)).The RED corpus enhanced ERE (Song et al., 2015) annotations by marking coreference for entities, events, and times, as well as temporal, causal, and subevent relationships in partial coreference through a multi-stage pipeline.In ECB+, Cybulska and Vossen (2014) expanded event descriptions by adding event classes with specific entity types and times, as well as inter-/intra-document coreference, to better represent the events within the ECB corpus (Bejan and Harabagiu, 2010).In a similar light, we enrich the ECB+ corpus with the X-AMR annotations with the goal of making ECR efficient and as a way to assess the performance of GPT-4.
Conclusion
In this paper, we introduced X-AMR, a corpus-level version of AMR.We provided a new model-in-theloop tool with which we enriched the ECB+ corpus with X-AMR annotations.We then introduced a novel linear graph-based ECR algorithm that leverages the nested event structure and the crossdocument entity coreference of X-AMR.The annotations coupled with the algorithm serve as a way for linearly generating cross-document event coreference annotations, cutting through a very challenging task.Finally, we developed two prompt engineering approaches for GPT-4 to automatically produce X-AMR graphs.We then compared the results against human annotations and showed limitations of GPT-4 on this task.We also provide comprehensive and concise GPT-generated event descriptors in this process that we believe have a lot of utility in other event tasks.Collectively, our contributions pave a path toward efficient ECR methods and their corresponding annotations.
Figure 1 :
Figure 1: m 1 and m 2 are examples of coreferent mentions.m 3 although related to m 1 and m 2 , is a different acquisition event.
Figure 2 :
Figure 2: Compressed event semantics graphs of m 1 , m 2 , and m 3 .The graph for m 2 is generated using the entire document in which it appeared.Verb-Net classes are used for synonymous predicates.KB is used for argument coreference resolution.
Figure 3 :
Figure 3: The PropBank roleset definitions of agree.01 and the expected annotations in X-AMR.
Figure 4 :
Figure 4: Eventive ARG-1 in m 4 for the roleset sign.02.The ARG-1 clause is annotated as the connecting event with roleset ID agree.01
Figure 5 :
Figure 5: Event Identifier and Coreference for events with eventive ARG-1.The EID 2 (m 4 ) is equivalent EID 1 (m 5 ) with the help of VerbNet to detect synonymy and KB to link arguments.
Roleset ID : Figure 6 :
Figure6: Label definitions for the event's Roleset ID and the Arguments that include the Wikipedia links.Event Description is a single sentence encapsulating the key components of the event.
Figure 8
Figure 8: X-AMR ECR Algorithm running time on synthetic mentions.The dotted line is used as a reference to check the linearity of the algorithm (red line).The complexity is roughly linear.
Table 1 :
Corpus statistics for event mentions in ECB+ and the mentions annotated with X-AMR ( * Annotation in Progress).Inter-annotator agreement for the Roleset ID is highest for the Dev set.
We cluster mentions based on the strict similarity of these PropBank Roleset IDs.We use the human (PB H ) and GPTgenerated (PB G , see §6) roleset IDs separately.EID N : We cluster with EID N using PB H or PB G EID lt : We cluster with EID lt using PB H or PB G EID VN N : EID N only, but with PB HVN or PB GVN to identify roleset classes for PB H or PB G EID VN lt : EID lt only, but with PB HVN or PB GVN to identify roleset classes for PB H or PB G EID N ∧ EID lt : We cluster the mentions when they have the same EID N and EID lt .
Table 2 :
ECR results comparing the annotators on the Development Set of the ECB+ Corpus.
The man is thought to have fallen much earlier in the day.m2:[...] Duncan Rait died after slipping and falling[...] | 2024-04-16T06:44:19.366Z | 2024-03-25T00:00:00.000 | {
"year": 2024,
"sha1": "d35228bdefa5af70b1b3db93e665b78bb17028db",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "ArXiv",
"pdf_hash": "d35228bdefa5af70b1b3db93e665b78bb17028db",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
90593337 | pes2o/s2orc | v3-fos-license | Early proof of T cell involvement in ovarian cancer
Volume 3 | Issue
Immune Checkpoint and Ovarian Cancer
Bose CK * Consultant Gynaecological Oncologist, NSC Bose Cancer Research Institute, Kolkata ISSN: 2454-3284 Although the idea, called "cancer immunotherapy", is very appealing and has previously been shown to work in several mouse models of cancer, it has in general been very difficult to translate cancer immunotherapy approaches in human.Because of this frustration, by the 1990s, many scientists and biotechnology companies had given up on the idea of cancer immunotherapy.Eight years after first detection T cell suppression effect of CTLA4 the identity of the antibody molecule was established.Antibody to CTLA4 could increase T cell which in turn destroys cancer cell.This started a completely new age of tumor immunology.Besides CTLA4 many immune checkpoints are now shown which pave new ways in manipulation of immunological control over malignant tumor cells.It has lent an important measure to manage especially recurrent and refractory cancers and those cancers where there is an unmet need like recurrent melanoma, renal cell carcinoma and recurrent ovarian cancer.As a new development this subject is experiencing rapid progress and multiple avenues are opening up.Though there are many hurdle to overcome this needs constant updating specially for students of ovarian cancer who are looking at it with much hope.
More than three decades back T cell infiltration was noticed in ovarian cancers [1].In 2003 Zhang et al appreciated their role in improved survival.They observed at least 60% benefit in 5 year survival in a cohort of 74 patients.They were treated to obtain complete clinical response after debulking and platinum-based therapy.CD3 + T cells within their tumor and IFN-γ plus macrophagederived chemokines made the difference [2].Reason of such heterogeneity is still unclear though subject has grown hugely and gained much strength.Regulatory T-cell subsets (CD4+) had confusing immune suppressing role.However, improved survival is noticed in patients who had higher numbers of intraepithelial CD8+ T cells compared with patients without intraepithelial CD8+ T cells (median survival 55 vs 26 months) [3,4].Strong positive correlation was observed between levels of CD8+ T cells and granzyme B within tumors [5].MHC related interferon regulatory factor (IRF)-1 and metastasis related chemokine receptor (CXCR) are the two genes differentially expressed in tumors with high versus low CD8+ T-cell infiltration.Heterogeneity in the tumor microenvironment among patients with EOC, and various immune cell populations those have been associated positively or negatively with clinical prognosis, included tumor-infiltrating lymphocytes (TILs), [6] myeloid-derived suppressor cells [7], and tumor-associated macrophages [8].
Early proof of T cell involvement in ovarian cancer
Tumour infiltrating lymphocytes (TILs) express the negative regulatory immune receptor, programmed cell death 1 (PD-1), [9] which were upregulated on T-cell activation and suppresses T-effector functions.Several cellular populations, including cancer Journal of Gynecology Research cells and tumor-associated myeloid cells, express its ligand PD-L1 [10][11][12][13].Expression of PD-L1 by tumors has been associated with decreased intraepithelial TILs and poor overall survival in EOC [13].
Immune checkpoints in ovarian cancer
In an immune-competent murine model of EOC, PD-1 and PD-L1 blockade has led to eradication of tumors through the expected reprogramming of the tumor microenvironment, [14] which suggested potential benefit from PD-1/PD-L1 inhibition for patients with EOC.The attenuation of T cell function by cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) was also noticed in EOC [15].Although the binding of peripheral membrane proteins B7-1 or B7-2 to CD28 provides an important costimulatory signal, the engagement of CTLA-4 by these ligands induced cell cycle arrest and diminished cytokine production [16][17][18].Documented effort to test ovarian cancer started with Hodi et al [19] who gave two patients a single infusion and then nine cases [20] upto 11 infusion of 3 mg/kg ipilimumab after an autologous ovarian tumor cell vaccine transduced with granulocytemacrophage colony-stimulating factor (GVAX).Only three patients had SD of >2 months.
Checkpoint Inhibitors in ovarian cancer
A phase 2 clinical trial of ipilimumab in relapsed platinum-sensitive ovarian cancer with measurable disease is ongoing though not recruiting any more (NCT01611558).Some trials either in post transplantation or along with other checkpoint molecules like nivolumab and pembrolizumab are recruiting.There are other two trials which are either suspended or terminated.The PD-1 antagonist nivolumab is tested at 1 or 3 mg/kg every 2 weeks in 18 patients with relapsed platinum-resistant disease regardless of PD-L1 expression; there was a 17% overall response rate (ORR) and a 44% disease control rate (DCR=CR+PR+SD), with 2 CR, 1 PR, and 5 patients with SD [21] A phase 1b study tested the PD-1 antagonist pembrolizumab at 10 mg/kg every 2 weeks in 26 patients with heavily treated PD-L1+ ovarian cancer chemotherapy [22].There was a durable ORR of 11.5%, and a DCR of 34.6%, with 1 CR, 2 PR, and 6 patients with SD.Now pembrolizumab has highest number of clinical trials in ovarian cancer.The PD-L1 antagonist avelumab was given at 10 mg/kg every 2 weeks in a phase 1b study of 75 patients with platinum-resistant or chemotherapy-refractory ovarian cancer regardless of PD-L1 expression [23] with an ORR of 10.7%, and a DCR of 54.7%.A phase 1 study of another PD-L1 antagonist, BMS-936559, revealed one objective response in 17 ovarian cancer patients [24].A summary of trials in ovarian cancer is given in Table 1.As a whole benefit in OS with ICIs was significant in both younger and older patients with a cut-off age of 65-70 years in a total of 5265 patients from nine RCTs of ICI [25].When patients are dichotomized into younger and older groups with an age cut-off of 65-70 years, ICIs improved OS in both younger (HR, 0.75; 95% CI, 0.68-0.82)and older (HR, 0.73; 95% CI, 0.62-0.87)groups.An improvement in PFS was observed in younger (HR, 0.58; 95% CI, 0.40-0.84)and older (HR, 0.77; 95% CI, 0.58-1.01)patients.Subgroup analyses according to ICI and tumor type showed a consistent survival benefit in both younger and older groups except for the subgroup of older patients treated in 4 trials of anti-programmed cell death protein-1 (PD-1) monoclonal antibody (HR, 0.86; 95% CI, 0.41-1.83).
These drugs generally contain warnings with regard to increased risks of severe immune-mediated inflammation of the lungs, the colon, the liver, the kidneys (with accompanying kidney dysfunction), as well as immune-mediated hypothyroidism and hyperthyroidism.
Journal of Gynecology Research
In clinical trials for melanoma, the following side effects occurred in more than 10% of subjects and more frequently than with chemotherapy alone: rash and itchy skin, cough, upper respiratory tract infections, and peripheral edema.Other clinically important side effects with less than 10% frequency were ventricular arrhythmia, inflammation of parts of the eye (iridocyclitis), infusion-related reactions, dizziness, peripheral and sensory neuropathy, peeling skin, erythema multiforme, vitiligo, and psoriasis.In clinical trials for lung cancer, the following side effects occurred in more than 10% of subjects and more frequently than with chemotherapy alone: fatigue, weakness, edema, fever, chest pain, generalized pain, shortness of breath, cough, muscle and joint pain, decreased appetite, abdominal pain, nausea and vomiting, constipation, weight loss, rash, and itchy skin.Levels of electrolytes and blood cells counts were also disrupted.
Conclusion
It is evident from discussion that much obstacles are remaining in establishing immuno therapy in ovarian cancer as response rate is low and no FDA approval is in the offing.Expectedly Professor Maurie Markman has commented at the 33rd Annual Chemotherapy Foundation Symposium "The checkpoint inhibitors are not ready for prime time yet in ovarian cancer.It's not because there's evidence that they don't work -it's just that there's no evidence at all, " But that will in no way undermine this important effort of treating first the relapse and refractory cases.It will certainly and find out the hindrance and to make best use of developed knowledge to find ways to subvert shortcoming of this important measure of treatment.
Volume 3 |
Issue 1 Journal of Gynecology Research Submit your next manuscript to Annex Publishers and benefit from: Submit your manuscript at http://www.annexpublishers.com/paper-submission.php→ Easy online submission process → Rapid peer review process → Open access: articles available free online → Online article availability soon after acceptance for Publication → Better discount on subsequent article submission → More accessibility of the articles to the readers/researchers within the field 24.Brahmer JR, Tykodi SS, Chow LQ, Hwu WJ, Topalian SL, et al. (2012) Safety and activity of anti-PD-L1 antibody in patients with advanced cancer.N Engl J Med 366: 2455-65.25.Nishijima TF, Muss HB, Shachar SS, Moschos SJ (2016) Comparison of efficacy of immune checkpoint inhibitors (ICIs) between younger and older patients: A systematic review and meta-analysis.Cancer Treat Rev 45: 30-7.
1. Ipilimumab Monotherapy in Recurrent Platinum-sensitive Ovarian Cancer.Active, not recruiting.Has Results Phase II Study 2. Nivolumab With or Without Ipilimumab in Treating Patients With Persistent or Recurrent Epithelial Ovarian, Primary Peritoneal, or Fallopian Tube Cancer Suspended 3. Safety Study of MGA271 in Combination With Ipilimumab in Refractory Cancer Recruiting 4. A Study of Nivolumab by Itself or Nivolumab Combined With Ipilimumab in Patients With Advanced or Metastatic Solid Tumors Recruiting 5. Monoclonal Antibody Therapy in Treating Patients With Ovarian Epithelial Cancer, Melanoma, Acute Myeloid Leukemia, Myelodysplastic Syndrome, or Non-Small Cell Lung Cancer Terminated 6. Ipilimumab After Allogeneic Stem Cell Transplant in Treating Patients With Persistent or Progressive Cancer Completed 7. Safety Study of MGA271 in Combination With Pembrolizumab in Refractory Cancer Recruiting Blockade in BRCA-deficient Ovarian Cancer Recruiting 2. Durvalumab and Tremelimumab in Combination With First-Line Chemotherapy in Advanced Solid Tumors Recruiting 3. A Phase 1 Study to Evaluate MEDI4736 in Combination With Tremelimumab Recruiting 4. Study of Tremelimumab Alone or Combined With Olaparib for Patients With Persistent EOC (Epithelial Ovarian, Fallopian Tube or Primary Peritoneal Carcinoma) Recruiting Study of the Safety, Tolerability, and Efficacy of Epacadostat Administered in Combination With Nivolumab in Select Advanced Cancers.Recruiting 4. Nivolumab With or Without Ipilimumab in Treating Patients With Persistent or Recurrent Epithelial Ovarian, Primary Peritoneal, or Fallopian Tube Cancer.Suspended 5. Study of FPA008 in Combination With Nivolumab in Patients With Selected Advanced Cancers Recruiting 6.A Phase 1 Study of AM0010 in Patients With Advanced Solid Tumors Recruiting 7. A Study of Nivolumab by Itself or Nivolumab Combined With Ipilimumab in Patients With Advanced or Metastatic Solid
Table 1 :
Pembrolizumab in Combination With Chemotherapy in Frontline Ovarian Cancer Not yet recruiting 2. Pembrolizumab/Carboplatin/Taxol in Epithelial Ovary Cancer Not yet recruiting Phase II: 3. Study of Niraparib in Combination With Pembrolizumab (MK-3475) in Patients With Triple-negative Breast Cancer or Ovarian Cancer (KEYNOTE-162)Recruiting 4. A Study of Pembrolizumab With Standard Treatment in Patients With Recurrent Platinum-resistant Ovarian Cancer Recruiting 5. Efficacy and Safety Study of Pembrolizumab (MK-3475) in Women With Advanced Recurrent Ovarian Cancer (MK-3475-100/ KEYNOTE-100) Recruiting 6. Dose Dense Paclitaxel With Pembrolizumab (MK-3475) in Platinum Resistant Ovarian Cancer Recruiting 7. ACP-196 Alone and in Combination With Pembrolizumab in Subjects With Recurrent Ovarian Cancer (KEYNOTE191) Recruiting 8.A Combination Clinical Study of PLX3397 and Pembrolizumab To Treat Advanced Melanoma and Other Solid Tumors Recruiting 9. Pembrolizumab and Ziv-aflibercept in Treating Patients With Advanced Solid Tumors Recruiting 10.Study of the Effects of Pembrolizumab in Patients With Advanced Solid Tumors Recruiting 11.A Study of Pembrolizumab on the Tumoral Immunoprofile of Gynecologic Cancers Not yet recruiting 12. Study of CM-24 (MK-6018) Alone and In Combination With Pembrolizumab (MK-3475) in Participants With Selected Advanced or Recurrent Malignancies Recruiting 13.Safety Study of MGA271 in Combination With Pembrolizumab in Refractory Cancer Recruiting 14. Pembrolizumab in Subjects With Incurable Platinum-Refractory Germ Cell Tumors Recruiting 15.A Phase 1/2 Study Exploring the Safety, Tolerability, and Efficacy of Pembrolizumab (MK-3475) in Combination With Epacadostat (INCB024360) in Subjects With Selected Cancers (INCB 24360-202 / MK-3475-037 / KEYNOTE-037/ ECHO-202) Recruiting 16.A Phase 1 Study of AM0010 in Patients With Advanced Solid Tumors Recruiting 17. Study of Pembrolizumab Plus Chemotherapy in Patients With Advanced Cancer (PembroPlus) Recruiting 18. MK-3475 Immunotherapy in Endometrial Carcinoma Recruiting 19.Phase IB of Selinexor in Combination With Standard Chemotherapy Recruiting 20.Phase 1 Study of Intradermal LV305 in Patients With Locally Advanced, Relapsed or Metastatic Cancer Expressing NY-ESO-1 Recruiting Ligand 1 (aPDL-1) Antibody Atezolizumab, Bevacizumab and Acetylsalicylic Acid in Recurrent Platinum Resistant Ovarian Cancer Not yet recruiting 2. A Study of Atezolizumab Administered in Combination With Bevacizumab and/or With Chemotherapy in Participants With Locally Advanced or Metastatic Solid Tumors Recruiting Matched Paired Pharmacodynamics and Feasibility Study of Durvalumab in Combination With Chemotherapy in Frontline Ovarian Cancer Not yet recruiting 2. A Phase 1/2 Study of Motolimod (VTX-2337) and MEDI4736 in Subjects With Recurrent, Platinum-Resistant Ovarian Cancer for Whom Pegylated Liposomal Doxorubicin (PLD) is Indicated Recruiting 3. Phase 1 and 2 Study of MEDI4736 in Combination With Olaparib or Cediranib for Advanced Solid Tumors and Recurrent Ovarian Cancer Recruiting 4. TPIV200/huFR-1 (A Multi-Epitope Anti-Folate Receptor Vaccine) Plus Anti-PD-L1 MEDI4736 (Durvalumab) in Patients With Platinum Resistant Ovarian Cancer Recruiting 5. Durvalumab and Tremelimumab in Combination With First-Line Chemotherapy in Advanced Solid Tumors Recruiting 6. MEDI4736 (Durvalumab) in Patients With Brain Metastasis From Epithelial-derived Tumors Not yet recruiting 7. Study of Azacitidine and Durvalumab in Advanced Solid Tumors Not yet recruiting 8.A Phase 1 Study to Evaluate MEDI4736 in Combination With Tremelimumab Recruiting 9. A Phase I/II Study of MEDI4736 in Combination With Olaparib in Patients With Advanced Solid Tumors .Recruiting Study Of Avelumab Alone Or In Combination With Pegylated Liposomal Doxorubicin Versus Pegylated Liposomal Doxorubicin Alone In Patients With Platinum Resistant/Refractory Ovarian Cancer (JAVELIN Ovarian 200) Recruiting 2. Avelumab in Previously Untreated Patients With Epithelial Ovarian Cancer (JAVELIN OVARIAN 100) Recruiting 3. Avelumab in Metastatic or Locally Advanced Solid Tumors (JAVELIN Solid Tumor) Recruiting Safety Study of MGA271 in Combination With Pembrolizumab in Refractory Cancer Recruiting 2. Safety Study of MGA271 in Combination With Ipilimumab in Refractory Cancer Recruiting 3. Study of Immune Response Modifier in the Treatment of Breast, Ovarian, Endometrial and Cervical Cancers Completed Has Results 4. BMS_PD-L1_onco : Assessment of the PD-L1 Protein as a Biomarker in Oncology and Hematology Recruiting Study of DPX-Survivac Vaccine Therapy and Epacadostat in Patients With Recurrent Ovarian Cancer Recruiting 2. Safety and Efficacy of CRS-207 With Epacadostat in Platinum Resistant Ovarian, Fallopian, or Peritoneal Cancer Recruiting 3. Epacadostat Before Surgery in Treating Patients With Newly Diagnosed Stage III-IV Epithelial Ovarian, Fallopian Tube, or Primary Peritoneal Cancer Active, not recruiting 4.A Study of the Safety, Tolerability, and Efficacy of Epacadostat Administered in Combination With Nivolumab in Select Advanced Cancers m,Intervention: Drug: Nivolumab + Epacadostat Recruiting 5. A Phase 1/2 Study Exploring the Safety, Tolerability, and Efficacy of Pembrolizumab (MK-3475) in Combination With Epacadostat (INCB024360) in Subjects With Selected Cancers (INCB 24360-202 / MK-3475-037 / KEYNOTE-037/ ECHO-202) Recruiting 6. DEC-205/NY-ESO-1 Fusion Protein CDX-1401, Poly ICLC, and IDO1 Inhibitor INCB024360 in Treating Patients With Ovarian, Fallopian Tube, or Primary Peritoneal Cancer in Remission Recruiting Immune checkpoints in ovarian cancer trials Volume 3 | Issue 1 | 2017-08-15T00:05:00.111Z | 2017-05-01T00:00:00.000 | {
"year": 2017,
"sha1": "065ffd7f18f7124e1766d6cb7614ed91f1431b11",
"oa_license": "CCBY",
"oa_url": "http://www.annexpublishers.com/articles/JGR/3104-Immune-Checkpoint-and-Ovarian-Cancer.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "065ffd7f18f7124e1766d6cb7614ed91f1431b11",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
221345245 | pes2o/s2orc | v3-fos-license | Sickness absence trajectories following labour market participation patterns: a cohort study in Catalonia (Spain), 2012–2014
Background Previous studies have focused on the relationship between employment pathways and health-related outcomes based on cross-sectional or longitudinal approaches. However, little is known about the cumulative effects of employment status mobility on sickness absence (SA) over time. The aim of the present study was to examine the association between prior labour market participation (LMP) patterns and SA trajectories from a life-course perspective. Methods This cohort study was based on a sample of 11,968 salaried workers living in Catalonia and affiliated with the Spanish Social Security system, who accumulated more than 15 days on SA in at least one quarter during 2012–2014. Individuals were grouped into three different working life stages: early (18–25 years), middle (26–35 years), and late (36–45 years). To identify LMP patterns, we applied sequence analysis and cluster analysis (2002–2011), and we used latent class growth modelling to identify SA trajectories (2012–2014). Finally, we applied multinomial logistic regression models to assess the relationship between LMP patterns and SA trajectories. Results The analyses yielded six LMP patterns: stable employment (value range: 63–81%), increasing employment (5–22%), without long-term coverage (7–8%), decreasing employment (4–10%), fluctuant employment (13–14%), and steeply decreasing employment (7–9%). We also identified four SA trajectories: low stable (83–88%), decreasing (5–9%), increasing (5–11%), and high stable (7–16%). However, the only significant association we identified for LMP patterns and SA trajectories was among young men, for whom an increasing employment pattern was significantly associated with a lower risk for increased days on SA (adjusted odds ratio: 0.21; 95% confidence interval: 0.05–0.96). Conclusions SA trajectories are generally not related to prior 10-year LMP patterns at any stage of working life. To disentangle this relationship, future research might benefit from considering working life transitions with a quality-of-work approach framed with contextual factors closer to the SA course.
Background
The ongoing phenomenon of increasing life expectancy and job insecurity affects the stability of the labour market structure [1]. Unstable employment pathways might entail the loss of economic resources, a lower likelihood of future employability when unemployed, and most important, loss of health. Experiences during working life are usually characterised as changes in employment and working conditions, the transitions in employment status that may influence an individual's future health course. Such events and transitions may occur independently or be part of a cluster or causal chain, as each working experience increases the risk for other events or transitions. "Working life trajectory" references the cooccurrence of events through the life course, covering critical points in working life such as transitions in and out of the labour market and time spent in each state as a whole unit [2][3][4].
Previous research has mainly focused on risk and prognostic factors (individual, socio-demographic, occupational) for sickness absence (SA) [5][6][7], but a less considered aspect is the extent to which transitions in and out of the labour market may affect an individual's work-related health outcomes. In the labour market context, a Norwegian study showed an inverse correlation between unemployment rate and the probability of having SA lasting longer than 14 days [8].
Together with socioeconomic, occupational, gender, and other individual factors, labour market transitions might differ with the stage of working life. For instance, young workers entering their working lives have greater mobility in terms of more gaps and a larger number of transitions between different employment statuses (i.e., employment, unemployment, or inactivity spells) compared with those in a late career stage [9]. In keeping with this pattern, a study of employment patterns among women in Germany showed that younger cohorts followed a trend towards discontinuous and part-time careers, whereas careers involving continuous full-time employment or being a housewife were becoming unusual [10].
Traditionally, for health and health-related outcomes, occupational epidemiology has assessed exposure-outcome association measures, whether simple or accumulated, at one point in time. Few studies have involved longitudinal analyses of the effect of prior working life on health [11][12][13][14][15][16]. From the standpoint of occupational epidemiology, because a person lives in a multidimensional and multilevel context, an approach from a lifecourse perspective could elucidate relevant events in a person's working life that manifest and shape health status over time [17]. The extent to which prior labour market participation (LMP) patterns affect risk for future SA remains unexplored [2,4].
Previous studies have examined sex and age differences in health and health-related outcomes and in LMP [18][19][20][21]. The authors of one review found that women have higher short-term SA rates than men, with variations by country and age cohort. Likewise, higher proportions of SA affected the beginning and middle parts of working life among women ages 20-54 years compared to men of the same ages, which could be related to pregnancy-related health problems or psychological distress when employed in occupations where women are underrepresented [19]. Because cohorts reflect exposures to past socioeconomic circumstances and institutional contexts, considering several cohorts may uncover different influences across a life course [22]. For this reason, a separate analysis by sex and by cohort, defined by working life stage, could contribute to disentangling the relationship between LMP and SA over a lifetime.
We hypothesised that prior highly fragmented LMP patterns would be associated with more unfavourable later SA trajectories compared to LMP patterns suggesting more stable employment pathways. We also hypothesised that this relationship might differ across working life stages and by sex. In this analysis, we viewed fragmented LMP patterns as being characterised by multiple transitions from employment to unemployment, either with benefits or without Social Security coverage.
The aim of this study was to investigate the relationship between prior LMP patterns at the early, middle, and late stages of working life and subsequent SA trajectories among individuals who had accumulated more than 15 days of SA in at least one of the quarters during 2012-2014, considering occupational and socioeconomic characteristics and SA medical diagnosis categories as potential confounders.
Study population
The study population was part of the Spanish WORKing life Social Security cohort (WORKss cohort), an annual random representative sample of 4% of affiliates with the Spanish Social Security system. The sampling has taken place at least one day a year starting in 2004, and the information includes employment history register data from 1981 [23]. The sample is updated annually following an algorithm, which selects the same individuals if they continue affiliated with the Spanish Social Security system. Those lost because of administrative inactivity or death are replaced with members from the target population until the sample again reaches 4%, which ensures sample representativeness. We obtained data related to working life from the WORKss cohort information and SA records for 2012-2014 from the Catalan Institute of Medical Evaluations, which contains data only for people residing in Catalonia. SA records include information on the starting/closure date of SA episodes and medically certified diagnoses coded using the International Classification of Diseases 10th version (ICD-10).
This study was based on a cohort of 11,968 salaried workers living in Catalonia and affiliated with the Spanish Social Security system, who accumulated more than 15 days on SA in at least one quarter during 2012-2014. Their prior working lives from 2002 to 2011 were reconstructed, and they were grouped into three working life stages according to their age in 2002: early working life cohort (WLC; 18-25 years, 33.3%); middle WLC (26-35 years, 37.5%), and late WLC (36-45 years, 29.2%). Approximately 75% of the SA episodes included in these data were shorter than 15 days, mainly represented by acute diagnoses (e.g., infectious and respiratory diseases). The rationale for selecting only workers who accumulated more than 15 days on SA is that SA episodes of this length are more likely to represent severe rather than mild or moderate diagnoses (half of SA episodes lasting longer than 15 days are the result mainly of mental and musculoskeletal disorders) [24]. For this reason, severe SA episodes may better reflect the potentially long-lasting effects of prior LMP patterns. This duration of episode also allows for exclusion of an influence from SA monetary benefits on a person's SA behaviour because long-term SA represents serious illness [25].
Variables
LMP patterns were defined based on weekly transitions among six labour market states: employment, unemployment insurance benefits, means-tested unemployment benefits, transition, without coverage, and without longterm coverage. The Spanish unemployment benefit scheme distinguishes two categories of unemployment, unemployment receiving benefits and means-tested unemployment benefits. Entitlement to unemployment insurance benefits requires previously paid contributions into Social Security of at least one year over the last 6 years. The duration depends on the paid contributions, and the amount is based on the average wage before becoming unemployed and the replacement rate (70% in the first 6 months, 50% afterwards). Means-tested unemployment benefits can be claimed after unemployment insurance benefits are exhausted or when individuals do not fulfil the conditions for receiving entitled benefits. The duration depends on the time contributed, and the amount is lower than the minimum wage [26,27]. We maintained the separation of the two unemployment categories, with means-tested benefits as the least generous, which may allow for capture in the LMP patterns of situations involving more vulnerable workers. For situations of employment and unemployment receiving unemployment benefits/means-tested unemployment benefits, individuals keep contact with the Social Security system. Periods without records in the Social Security registry were categorised ad hoc into three states. The state transition was defined as a period between employment states up 30 days (i.e., administrative transition) [16]. The state without coverage refers to periods between employment states longer than 30 days. The state without long-term coverage accounts for the first labour market entry or return to work (i.e., left censorship) and/or a labour market withdrawal or a temporary leaving (i.e., right censorship). SA trajectories were based on the sum of days on SA each quarter between 2012 and 2014 of workers who had accumulated more than 15 days on SA any quarter.
Included covariates for the period 2012-2014 were type of contract (permanent and temporary), working time (full-time, part-time 51-99%, and part-time up to 50%), occupational category (skilled non-manual, skilled manual, unskilled non-manual, and unskilled manual), income (average monthly income categorised into quartiles), and SA medical diagnosis (grouped according to the ICD-10). The diagnosis groups included were mental disorders (F00-F99), digestive diseases (K00-K93), musculoskeletal diseases (M00-M99), pregnancy (O00-O99), injuries and poisoning (S00-T98), and others (the remaining codes). Diagnosis groups that included few people or none in the SA trajectories, such as acute (i.e., infections, circulatory diseases, respiratory diseases) or unusual according to their nature (i.e., pregnancies in women from the late WLC and neoplasms) were excluded.
Because workers could transit among different employment conditions and diagnosis groups during the SA trajectory during 2012 to 2014, we assigned them to the category where they spent most of the time for the whole follow-up period. Similarly, because workers had daily labour market states during the prior LMP, 2002-2011, we assigned them to the state where they spent most of the time for each week of the period.
Income was based on the monthly salary and unemployment benefits, subject to legal limits. The average monthly income was the total remuneration a worker received divided by time employed monthly. Additionally, time employed measured in the years 2002-2011 was used as an adjustment variable.
Statistical analyses
All the analyses were performed for each sex and WLC group separately. In a first step, we reconstructed the prior 10 years of LMP patterns by applying sequence analysis based on the six previously defined states. This methodology allowed us to describe individual sets of state sequences [28]. An optimal matching approach supported development of a matrix of distances between each sequence given indels (i.e., insertion and deletion) together with substitution costs. A custom substitution costs matrix was defined according to the characteristics of our data and study population, with a higher weight given to the transitions that were considered less frequent (i.e., from employment to means-tested unemployment benefits, or from employment to without coverage, and vice versa). Then, we used hierarchical cluster analysis to group similar sequences (i.e., LMP patterns) [29]. We based our selection of the optimal number of clusters on the average silhouette width (ASW), which allows for the assessment of clustering validity (ASW > 0.5 indicates reasonably well-separated clusters) (Supplementary Table 1) [30,31].
In a second step, we identified SA trajectories using latent class growth analysis (LCGA) based on accumulating more than 15 days on SA quarterly (if any) during the 3-year follow-up, specifying a linear functional form. An assumption of LCGA is that individuals within a trajectory are homogeneous, and individuals are assigned into subgroups with similar characteristics, according to a membership probability [32]. The optimal number of trajectories was assessed considering the Bayesian information criterion (lower best fit) and the Lo-Mendell-Rubin adjusted likelihood ratio test and bootstrap likelihood ratio test. A minimum of 1% of the total sample is recommended for each class [33], although this rule depends on the sample size and how meaningful a small group is for the study aim [34]. Additionally, high posterior probabilities (> 0.7) and researcher criteria were applied to determine the final number of trajectories. In cases of similar values for compared goodness-of-fit indices, we selected the one with the highest entropy (near 1.0) (Supplementary Table 2) [34]. We also assessed model adequacy indices using odds of correct classification (> 5 for all groups indicates high assignment accuracy) and mismatch scores (values close to 0 indicate individuals were assigned to groups with high certainty) (Supplementary Table 3) [35].
In a final step, we examined the association between LMP patterns and SA trajectories, using multinomial logistic regression models adjusted for potential confounders. We applied TraMineR package in R for the Sequence Analysis, MPlus v.7 for the LCGA, and Stata v.13© for the multinomial logistic regression analysis.
Results
The analysis identified six LMP patterns ( Fig. 1): stable employment (value range: 63.3-81.3% of workers), increasing employment (5.6-22%), without long-term coverage (7.5-8.2%), decreasing employment (4.3-10.5%), fluctuant employment (13.6-14.7%), and steeply decreasing employment (7.4-8.8%), with 3-4 patterns in each sex and WLC group. The stable employment LMP pattern showed a higher proportion of women in the early WLC compared to proportions in the late WLC (76.9% vs 70.7%), whereas the opposite pattern was observed for men (63.3% vs 81.3%). We also identified other different LMP patterns by sex, including a pattern without long-term Social Security coverage in women only, and a steeply decreasing employment pattern only among men.
Regarding the SA trajectories, we identified four ( Fig. 2): low stable (82.9-88.1% of individuals), decreasing (5-9.4%), increasing (0.8-11.3%), and high stable (6.7-16.3%), with three trajectories in each sex and WLC group. The mean of accumulated days on SA ranged from 30 to 40 days in the low stable trajectory to 60 to 70 days in the high stable trajectory.
Across SA trajectories, we also found significant differences by type of contract and diagnosis group. Among women, compared to other groups, a higher proportion had temporary contracts in the high stable SA trajectory in the early WLC (28.5%) and the increasing SA trajectory in the middle WLC (26%). The increasing trajectory had the highest proportion of women who accumulated days on SA because of mental disorders in the early and middle WLCs (24.5 and 28%, respectively). The decreasing and increasing SA trajectories showed a similar proportion of workers on SA because of mental disorders among women in the late WLC (20 and 21.1%, respectively) and men in the middle WLC (15.9 and 16.5%, respectively).
Otherwise, the distribution of LMP patterns was mainly homogeneous across SA trajectories. Among women in the middle WLC and men in the early WLC, the increasing SA trajectories showed a higher proportion of workers who had a prior stable employment LMP pattern (78.1 and 71.4%, respectively) compared to other trajectories. For men in the middle WLC, the decreasing SA trajectory showed a lower proportion of individuals with the LMP pattern of increasing employment (8.4%) compared to the low stable (14.7%) and increasing trajectories (14.1%) (Tables 1 and 2).
Adjusted regression models did not show significant associations between the prior 10 years of LMP patterns and subsequent SA trajectories (Table 3). Only men in the early WLC who had an increasing employment LMP pattern showed a significantly lower risk for increased accumulated days on future SA compared to those who had continuous stable employment LMP patterns (adjusted odds ratio: 0.21; 95% confidence interval: 0.04-0.96).
Discussion
Our main finding was that the four SA trajectories (low stable, decreasing, increasing, and high stable) were not related to the six prior 10-year LMP patterns across the three stages of working life considered. More than 80% of workers showed a low stable accumulation of 30 to 40 days on SA on average during the 3-year follow-up.
The result was unexpected, according to our hypothesis, because it is reasonable to consider that the accumulation of adverse employment status (frequent entries into and exits from the labour market) during a previous working life could be related to an increase in SA days because of more severe or longlasting pathologies (i.e., chronic diseases such as musculoskeletal and mental disorders). However, the result persists after adjustment for socioeconomic and employment-related factors and medical diagnosis during the SA course, with this last emerging as a leading determinant of the SA trajectory [36].
As far as we know, no previous studies have examined the relationship between prior 10-year LMP patterns and future SA trajectories from a life-course perspective. A Finnish study focused on early exits from employment, identifying and describing 10-year working life participation patterns and determining the cumulative incidence of SA within these patterns. Those authors found that individuals with long-term labour market exit had the highest cumulative incidence of SA because of mental disorders [20]. Similarly, one investigation of the effect of precarious employment on SA in four Nordic countries showed that precarious employment was associated with SA of 7 days or more, and another report on health-related outcomes (including long-term SA) in 28 countries in Europe cited an association with SA of more than 15 days [37,38]. In both studies, however, the authors measured precarious employment as a multidimensional construct based on indicators obtained from several dimensions (e.g., employment instability, lack of power and rights, reduced salary), which captured the accumulation of unfavourable aspects of employment quality [39]. Therefore, the results derived from these studies may not be directly comparable to the current findings from our analyses of LMP patterns based on employment status mobility across different stages of the working life course. Previous groups also have analysed the possible association between certain labour market situations and several health-related outcomes. One group found in a follow-up of 15 years that unemployment at an early age had a dose-response relationship with increased risk of SA, disability pension, and death [40]. Another study showed that a high number of periods without Social Security coverage was the main predictor of early retirement for permanent disability [41]. In that case, the outcome was the disability pension, which is closely related to SA [42].
In contrast to these previous studies, we found no relationship between the SA course and 10 years of prior transitions in the labour market. A first alternative hypothesis from these findings could be that SA trajectories may be more related to a course that is nearer to labour market transitions (for instance, the previous 2 or 5 years). Furthermore, because a given health-related problem may interfere with an individual's ability to meet current job demands, an imbalanced situation between these two conditions might be more likely to lead to accumulation of more days on SA compared to a prior adverse labour market transition pattern.
A second alternative hypothesis that could explain the results is that we did not measure employment quality during the prior working life. It is possible that studying a more comprehensive set of employment arrangements when defining the LMP patterns, including transitions among types of contract, working time, and occupations, could lead to a more specific LMP pattern that might relate to future SA behaviour. In this vein, one study has shown that working life patterns characterised by longterm exposure to blue-collar occupations, which are physically demanding and more subject to work accidents, have an effect on health outcomes that is similar to that of intermittent joblessness [14]. However, it should be considered that including a higher number of employment situations could increase the heterogeneity of patterns, which might attenuate the effect of an association between LMP patterns and health outcomes. Furthermore, as mentioned above, in some studies, the quality of employment was measured in terms of precarious employment using a multidimensional approach, yielding stronger associations with health outcomes than obtained by measuring one dimension [15]. However, most of those studies did not have a longitudinal perspective that could allow for consideration of the cumulative effects of employment quality over time [39]. Another hypothesis could be related to the impact of contextual variables, such as the economic crisis that began in 2008, Social Security coverage, and its relationship with the health system throughout the SA benefit system in the country. The impact of the economic crisis, which implied a general worsening of the labour market context, thus might have influenced the course of future SA, regardless of the individual's employment pathway. Previous studies have focused on how contextual factors such as the unemployment rate or poor local economy affect SA. One such study showed a negative correlation between the unemployment rate and the probability of having long-term SA [8]. Another study showed that a poor local economy in terms of low municipal revenue and high unemployment rate was related to decreased self-certified SA [43]. In both cases, the threat of becoming unemployed because of a high level of unemployment discouraged workers from taking SA. However, the 3-year follow-up in our work could be too short to uncover lasting effects of the Great Recession on SA levels, especially when the most acute phase was in 2013, with an unemployment rate reaching 27% [44].
Finally, as several European studies have shown, differences in the probability of having SA exist according to characteristics of the sickness benefit system, such as the eligibility conditions and the level of generosity of SA compensation (i.e., wage replacement amount and benefit entitlement duration). For example, the probability of being on SA is higher in countries where employees are entitled to receive a full wage replacement in case of illness [45]. In contrast, another study showed a lag effect between levels of SA benefits on SA incidence, so that countries with relatively generous SA benefits show lower levels of SA in the long term. According to those authors, the generosity of SA benefits provided sufficient income support that helped beneficiaries overcome economic hardship and recover health [46]. Spain has specific conditions for access to SA (i.e., 180 days of paid contributions during the prior 5 years), relatively generous benefits (e.g., 60-75% wage replacement from the Social Security budget), and access to medical care provided by the National Health System. The cost of SA in Spain is shared by the employer (4th-15th day) and the Social Security system (16th onwards), unlike other countries where the employer pays the entire cost of SA [47]. However, because our study population was based exclusively on workers who have had SA, they were all entitled to full SA benefits, so our results may not be attributable to the eligibility conditions for access. In the present study, employees who did not accumulate days on SA were not included. Last but not least, a general practitioner grants medical certification from the National Health System, regardless of the Social Security system and companies. In this sense, a possible adverse effect of unstable employment on health may be buffered by the provision of universal health coverage by the National Health System. Nevertheless, we found that men in their early working life stage who followed a pattern of increasing time in employment were less likely to increase their accumulation of future SA days. This result suggests that a stable transition into employment at the beginning of working life may have a protective effect on future health. However, according to several studies, young workers tend to enter into more insecure and precarious employment compared to older workers [48,49], and temporary workers report higher job insecurity than those with permanent jobs and show lower SA rates [50]. In line with previous studies, we found that one-third of men at the early working life stage were employed in temporary jobs (Supplementary Table 4). An alternative explanation would be that low levels of SA are a reflection of an earlier return to work that can be stimulated by the loss of financial resources through receiving SA benefits instead of a full salary. This situation may particularly affect temporary workers because their jobs are often more precarious and employment conditions often worse, which could induce them to reduce SA length out of fear of job dismissal and being unable to fulfil their financial needs [50,51].
One of the main limitations of this study is that workers without SA were excluded, which could have meant exclusion of those with better health status, those who had poor health status but were engaged in presenteeism, or those who did not qualify for SA benefits during the previous period, leading to selection bias. Also, we have classified workers according to their employment condition, medical diagnosis group, and the labour market state where they spent most of the time as a representative of the entire period, which could have led to a potential classification bias. A weekly labour state measure could have led to underestimation of short employment spells because approximately one-fifth of these were less than four days (results not shown). Another limitation comes from the use of two modelling methods based on simplifying assumptions, which may have implications for the robustness and reliability of the results; therefore, any conclusion derived should be considered carefully [17,52]. In this regard, LMP patterns in the early working life cohorts may not represent homogeneous groups because their global ASW showed lower values than recommended. Likewise, the low stable SA trajectories showed the lowest values of assignment accuracy across WLCs. Nevertheless, the assessment of cluster analysis and latent classes showed adequate values for most of the groups. Also, the results are representative of salaried workers living in Catalonia and require caution with comparisons to other countries because differences in labour market regulations, social security systems (e.g., SA benefit generosity), and sociocultural context could shape SA in a distinct way [47]. Moreover, this study could not account for working conditions (only occupational category as a proxy) or health status before the SA course because the information was not available; therefore, a potential confounding bias might have affected our estimates. Notwithstanding, not controlling for initial health status is less an issue for those workers at early working life stages because bad health increases with age. The decreasing SA trajectories could represent not only workers who reduced accumulated SA days over time but also those who exited the labour market because of permanent disability or retirement, or who died during the 3year follow-up. Nevertheless, in our study, workers in such situations represented only 4-8% of the population, depending on the trajectory group and working life stage. Furthermore, the administrative registers do not record a worker's status when they are without contact with Social Security. Individuals could be unemployed without benefits and actively seeking a job, jobless and not seeking a job (i.e., outside the labour force or inactive), or in informal employment (i.e., working off the record, without a contract or social protection), and we do not know how these situations affect their health [53].
The study also has some strengths. First, we incorporated a life-course approach into the design. To our knowledge, few studies on occupational epidemiology have examined the relationship between 10 years of prior working life and the course of future health. Prior evidence on this issue has been based mostly on crosssectional or longitudinal studies without a life-course approach [41,54,55]. In practice, the construction of patterns of LMP allows for consideration of the timing, order, and duration of given labour market states. Second, the study population comes from a Spanish workforce cohort (the WORKss cohort) [23], which has a large sample size and comprises high-quality registers. Third, employment history and SA data are registerbased and medically certified, which allows for calculation of the exact starting and closing dates for each employment status period and SA episode. Fourth, the SA medical diagnosis certification is issued by a general practitioner from the National Health System, which implies higher validity of the diagnosis than self-reported SA measures [54].
Conclusions
In conclusion, patterns of LMP during a 10-year prior working life are not related to future SA trajectories, regardless of the stage of the working life. Future studies should consider working life transitions closer to the SA course, with an employment quality approach framed with contextual factors, which may help in understanding of this relationship. | 2020-08-28T14:24:16.901Z | 2020-03-19T00:00:00.000 | {
"year": 2020,
"sha1": "1c6fa79b6042adba88816b6956e10bce21e8ff77",
"oa_license": "CCBY",
"oa_url": "https://bmcpublichealth.biomedcentral.com/track/pdf/10.1186/s12889-020-09396-9",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "afa3a35ddbc9801574f18f66cf8ff7144eb3d6c1",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
119657862 | pes2o/s2orc | v3-fos-license | Algebraic entropy of amenable group actions
Let $R$ be a ring, let $G$ be an amenable group and let $R\ast G$ be a crossed product. The goal of this paper is to construct, starting with a suitable additive function $L$ on the category of left modules over $R$, an additive function on a subcategory of the category of left modules over $R\ast G$, which coincides with the whole category if $L({}_RR)<\infty$. This construction can be performed using a dynamical invariant associated with the original function $L$, called algebraic $L$-entropy. We apply our results to two classical problems on group rings: the Stable Finiteness and the Zero-Divisors Conjectures.
Introduction
Consider a Grothendieck category C, for example, let C be the category R-Mod of left R-modules over a ring R. In this paper we are interested in studying some particularly well-behaved numerical invariants of the objects of C called length functions. This notion was introduced by Northcott and Reufel [26] (in case C " R-Mod) in 1965 and it was also studied by Vámos [35,36] and Zanardo [39]. Let us be more precise: a length function is an invariant L : C Ñ Rě0 Y t8u such that (1) Lp0q " 0 and LpBq " LpAq`LpCq if 0 Ñ A Ñ B Ñ C Ñ 0 is a short exact sequence, that is, L is additive; (2) LpM q " sup I LpMiq, if M is the direct limit of a directed family tMi : i P Iu of subobjects, that is, L is upper continuous.
Classical examples of length functions of the category R-Mod are the composition length ℓpM q of a left R-module M , for R an arbitrary ring, the (torsion-free) rank rkpM q of an arbitrary left R-module M over a left Ore domain R, and log |M | when R " Z and M is an Abelian group.
In Section 2, we give some generalities about torsion theories and localization in Grothendieck categories; using this formalism, we recall the definition of Gabriel dimension for Grothendieck categories proving some useful properties of this invariant. Gabriel dimension was introduced by Gabriel [14] under the name of Krull dimension; its name is due to Gordon and Robson [18]. One of the main results of Vámos' Ph.D. thesis [36] was to give a description of all the length functions of R-Mod whenever this category has Gabriel dimension. Our Theorem 2.37 is an analogous statement for general Gabriel categories (i.e., Grothendieck categories with Gabriel dimension); in particular, we show that any length function in such a category is a linear combination of some "atomic length functions". The philosophical point is that any length function on a Gabriel category is a linear combination of functions which are induced by the composition length in a localization of the original category at some point of the Gabriel spectrum. We have made the choice to give complete proofs for various reasons. First of all, Vámos' thesis is not easily accessible as it was not published (excluding some particular cases [35]), furthermore it seemed natural to use the formalism of localization and torsion theories to give more direct arguments; I believe this approach clarifies some points of the theory. Nevertheless, many ideas and techniques come directly from [36].
Given a left Noetherian ring A with a distinguished central element X P A, there is an important length function of the category of A-modules called multiplicity of X (see for example Chapter 7 of [25]). This invariant is based on the notion of composition length of modules. The classical notion of multiplicity was used by Vámos as a model to construct an "L-multplicity of X" based on a given length function L of A-Mod (see Chapter 5 of Vámos' thesis). Let now R be a left Noetherian ring, let A " RrXs be the ring of polynomials in one variable with coefficients in R, and let M be a left A-module. Given a length function L : R-Mod Ñ Rě0 Y t8u, one can extend trivially L to A-Mod (just forgetting the action of X) and then take the L-multiplicity of the element X P A, which we denote by multL. This defines a map mult : tlength functions on R-Modu ÝÑ tlength functions on RrXs-Modu L Þ ÝÑ multL .
The values of L can be recovered via the formula LpM q " multLpRrXs bR M q @M P R-Mod .
This procedure of extending length functions from a given ring to its ring of polynomials is useful in many situations but it has the disadvantage that is works just in the Noetherian case. In the recent paper [32], Salce, Vámos and the author studied the problem of the extension of a given length function on a category of modules R-Mod to a length function of (a suitable subcategories of) RrXs-Mod, without any hypothesis on the base ring. The key idea is to see a left RrXs-module as a left R-module with a distinguished endomorphism, roughly speaking we see RrXs-modules as discrete-time dynamical systems, and then we apply an invariant, called algebraic L-entropy, of the endomorphisms of left R-modules (see Subsection 5.3 for some more details and more references on these matters). It turns out that the algebraic L-entropy and the L-multiplicity coincide whenever they are both defined.
In this paper we study the following question, which is asking for a deep generalization of what we briefly explained above: Question 1.1. Let R be a ring, let G be a group and fix a crossed product R˚G (see Subsection 4.3). Is it possible to find a map tlength functions on R-Modu ÝÑ tlength functions on R˚G-Modu L Þ ÝÑ LR˚G satisfying the formula LpM q " LR˚GpR˚G bR M q for any left R-module M ?
We will give a partial answer to the above question using a notion of entropy similar to that used in [32]. More precisely, we will show that there are "many" length functions of R-Mod which can be extended to a "large" subcategory of R˚G-Mod if we assume that the group G is amenable (see Subsection 4.1). Let us also remark that, even for R " Z and L " log 2 |´|, Elek [10] proved that it is not always possible to find a length function of the category RrGs-Mod with the above properties (1)-(4) if G is not amenable. (The result of Elek is for compact Abelian groups but his statement implies what we said by Pontryagin-Van Kampen duality.) This shows that the answer to Question 1.1 is negative in general.
In the rest of the introduction we try to describe the main ideas of the paper and we expose our main results, including a couple of applications to classical problems. Indeed, let us start with a ring R, a left R-module RM and a group G. Analogously to the case of polynomial rings mentioned above, to specify a left RrGs-module structure on RM we have just to choose a group homomorphism λ : G Ñ AutRpM q , g Þ Ñ λg .
Looking things from a different angle, one can see the above map as a dynamical system where the ambient space is M , the time is the group G and the evolution law T : GˆM Ñ M is such that T pg, mq " λgpmq. If one chooses to work with a general crossed product R˚G and a left R˚G-module M , then the analog of the above map λ induced by the R˚G-module structure of M is not a group homomorphism, so one cannot speak properly about an action of G. Anyway, G still acts on the monoid LpM q of R-submodules of M in the sense that there is a group homomorphism λ 1 : G Ñ AutpLpM qq , g Þ Ñ λ 1 g , where AutpLpM qq is the group of monoid automorphisms of LpM q. Consider now a length function L : R-Mod Ñ Rě0 Y t8u. In order to attach to λ 1 a well-behaved dynamical invariant based on L we need a compatibility condition, that is, we need that LpN q " Lpλ 1 g pN qq for all g P G and N P LpM q. In Section 3 we study this kind of compatibility conditions in a categorical setting.
In the first part of Section 4 we introduce the class of amenable groups. Then, inspired by the notion of amenable topological entropy introduced by Ornstein and Weiss [27], we associate to a length function L : R-Mod Ñ Rě0 Y t8u satisfying suitable conditions a notion of algebraic L-entropy entL for left R˚G-modules. For doing so we need some results about quasi-tilings and non-negative real functions on the finite parts of an amenable group, that are recalled in Subsection 4.2.
More precisely, given a ring R (no hypothesis is needed on the ring) and a finitely generated amenable group G, fix a crossed product R˚G and a discrete length function L : R-Mod Ñ Rě0 Y t8u (where discrete just means that the finite values of L form a subset of Rě0 which is order-isomorphic to N) which is compatible with R˚G (see Definition 4.6). We define lFinLpR˚Gq to be the subclass of R˚G-Mod consisting of all left R˚G-modules R˚GM such that LpKq ă 8, for any finitely generated R-submodule K P LpM q. For example, lFinLpR˚Gq contains all the left R˚G-modules R˚GM such that LpM q ă 8. Furthermore, a consequence of the continuity of L on direct limits of submodules is that, given a left R-module RK such that LpKq ă 8, then the left R˚G-module R˚G bR K is in lFinLpR˚Gq. The algebraic L-entropy is a function entL : lFinLpR˚Gq Ñ Rě0 Y t8u. The main result of Section 6 is the following partial answer to Question 1.1: Theorem 6.1 In the above hypotheses, the invariant entL : lFinLpR˚Gq Ñ Rě0 Y t8u satisfies the following properties: (1) entL is upper continuous; (2) entLpR˚G bR Kq " LpKq for any L-finite left R-module K; (3) entLpR˚GN q " 0 if and only if LpRN q " 0, for any R˚G-submodule N ď R˚G bR K; (4) entL is additive.
In particular, entL is a length function of lFinLpR˚Gq.
In Section 7 we present two possible applications of the machinery introduced in the present paper. In particular, we show that, making use of the algebraic L-entropy, one can prove a very strong form of the Kaplansky Stable Finiteness Conjecture, stating that group rings with coefficients in a field are stably finite (see Conjecture 7.2), for amenable groups. Recall that, given a ring R, a left R-module M is said to be hopfian if any endomorphism φ : M Ñ M is surjective if and only if it is bijective, while M is hereditarily hopfian if any of its submodules is hopfian; it is known that a ring R is stably finite if and only if the left R-modules of the form RR n (with n P N`) are hopfian. We prove the following Theorem 7.8. Let R be a left Noetherian ring, let G be a finitely generated amenable group and let R˚G be a fixed crossed product. Then, for any finitely generated left R-module RK P R-Mod, the left R˚G-module R˚G bR K is hereditarily hopfian.
In particular, EndR˚GpM q is stably finite for any submodule R˚GM ď R˚G bR K.
The hopficity of all the submodules of RrGs bR N , with N a finitely generated left R-module, is a very strong property. In fact, we show in Example 7.11 that it does not hold when G is non-commutative free.
We conclude the paper showing, in Subsection 7.2, that a deeper understanding of the algebraic entropy would have implications also on the Zero-Divisor Conjecture, stating that the group algebras of torsion-free groups with coefficients in a field are domains. In fact, given a division ring K and a finitely generated amenable group G, we show that a fixed crossed product K˚G is a domain if and only if the algebraic entropy associated to the dimension of left K-vector spaces takes values in N Y t8u (see Theorem 7.14). In other words, the algebraic entropy detects the zero-divisors in K˚G. As an immediate consequence, we obtain that, in the above hypotheses, K˚G is a domain if and only if it admits a flat embedding in a division ring.
Torsion theories, length functions and Gabriel dimension
All along this section, when not differently specified, C denotes a Grothendieck category, that is, a cocomplete (i.e., Ab.3) Abelian category, with exact colimits (i.e., Ab.5) and with a generator. For unexplained terms and notations we refer to [34].
Remember that any Grothendieck category is well-powered (see [34,Proposition 6,p.94]) and that the set of all subobjects of a given object M is a bounded and bicomplete lattice, which we denote by LpM q; the minimum of LpM q is the 0-object and the maximum is M itself. For every family tNi : i P Iu of subobjects of M we denote by ř I Ni (resp., Ş I Ni) the join (resp., meet) of the Ni in LpM q. If tNi : i P Iu is a directed system of subobjects, ř I Ni is called the direct union of the Ni and it is sometimes denoted also by Ť I Ni. With this notation we can state the equivalent form of Grothendieck's axiom Ab.5 stating that, given an object M in C, a subobject K of M and a directed system tNi : i P Iu of subobjects of M , Let us recall the following fundamental properties of Grothendieck categories: (1) C is complete; (2) C has enough injective objects.
Given a family tMi : i P Iu of objects of C, we denote by ś I Mi and À I Mi the product and the coproduct respectively. Given an object M of C, we denote by EpM q the injective envelope of M . Given an endomorphism φ : M Ñ M , a subobject N of M is φ-invariant if φpN q ď N . Any subobject of a quotient (or, equivalently, a quotient of a subobject) of an object M is generically called a segment of M . A series of M is a finite chain of subobjects The factors of σ are the segments of M of the form Ni`1{Ni for some i ă n. If all the factors of σ are simple objects, then we say that σ is a composition series for M . If The following fact is well-known.
Length functions
In any category C it is possible to define (real-valued) invariants to measure various finiteness properties of the objects. In general, we call invariant of C, any map i : C Ñ Rě0 Y t8u such that ipXq " ipX 1 q whenever X and X 1 are isomorphic objects in C.
If we make some stronger assumption on the structure of the category C, we can refine our definition of invariant in order to obtain a more treatable notion. Indeed, suppose that C is an Abelian category (or, more generally, an exact category). The information that one usually wants to encode is about homological properties. In this setting it seems natural to ask that, given a short exact sequence in C, we have ipX2q " ipX1q`ipX3q. In this case, we say that i is additive on the sequence (2.2). If i is additive on all the short exact sequences of C and ip0q " 0, then we say that i is an additive invariant (or additive function).
In the following lemma we collect some useful properties of additive functions. (1) ipXq ě ipY q for every segment Y of X P C; (2) ipX1`X2q`ipX1 X X2q " ipX1q`ipX2q for every pair of subobjects X1, X2 of X P C; (3) ř jodd ipXj q " ř jeven ipXj q for every exact sequence 0 Ñ X1 Ñ X2 Ѩ¨¨Ñ Xn Ñ 0 in C. A natural assumption in the context of Grothendieck categories is the upper continuity of invariants, given an object X P C and a directed set S " tXα : α P Λu of subobjects of X such that In what follows we generally denote length functions by the symbol L. We remark that the usual definition of length function is given in module categories, which are in particular locally finitely generated Grothendieck categories. In this special setting, the usual definition of upper continuity is different (see part (3) of the following proposition). We now show that we are not defining a new notion of upper continuity but just generalizing this concept to arbitrary Grothendieck categories (similar observations, with analogous proofs, were already present in [36] for module categories).
Recall that, given an object M of a Grothendieck category C and an ordinal κ, a set tMα : α ă κu of subobjectsis of M is a continuous chain if (1) Mα ď M β , provided α, β ă κ and α ď β; (2) Ť αăλ Mα " M λ for any limit ordinal λ ă κ. Proposition 2.5. Let C be a Grothendieck category and L : C Ñ Rě0 Y t8u be an additive function. Consider the following statements: (1) L is a length function; (2) given an object M P C, an ordinal κ and a continuous chain tMα : α ă κu of subobjects of M such that M " Ť αăκ Mα, we have that LpM q " suptLpMαq : α ă κu; (3) for every object M P C we have that LpM q " suptLpF q : F finitely generated subobject of M u.
Proof. (1)ñ(2) is trivial since continuous chains are directed sets. On the other hand, consider a directed set pI, ďq and a directed system tMi : i P Iu of subobjects of M . If I is finite then I has a maximum, so there is nothing to prove. On the other hand, if I is an infinite set, one shows as in the proof of [17,Lemma 1.2.10], that pI, ďq is the union of a continuous well-ordered chain of directed subsets, each of which has strictly smaller cardinality than I. One concludes by transfinite induction that (2)ñ(1). Assume now (3) and consider a continuous chain as in part (2). For every finitely generated subobject F of M , there exists α ă κ such that F ď Mα and so we obtain that LpM q " suptLpF q : F f.g. subobject of M u ď suptLpMαq : α ă κu ď LpM q .
To conclude, notice that if C is locally finitely generated, then any object is the direct unions of the directed system of its finitely generated subobjects. In this situation, (1) clearly implies (3).
The notion of length function on a Grothendieck category is quite formal, this is why it seems useful to stop for a while and describe some concrete examples of length functions. We start with two easy ones.
Example 2.6. Let D be a left Ore domain, let Σ " Dzt0u and let Q " Σ´1D be the ring of left fractions of D. For any M P D-Mod, the torsion free rank rkpM q is defined to be the dimension of the left module Σ´1M " Q bD M over the skew field Q. The torsion free rank rk : D-Mod Ñ Rě0 Y t8u, where rkpM q " 8 whenever the rank of M is not finite, is a length function. In particular, the dimension of vector spaces over a field is a length function.
Example 2.7. The logarithm of the cardinality log |´| : Z-Mod Ñ Rě0 Y t8u, where log |G| " 8 whenever G is not finite, is a length function.
In what follows we define the length of a lattice. This can be applied to the lattice of subobjects of a given object in a Grothendieck category obtaining the so-called composition length, which is the invariant inspiring the abstract notion of length function. Definition 2.8. Let pL, ď, 0, 1q be a bounded lattice. A sequence of elements σ : 0 " x0 ň x1 ň¨¨¨ň xn " 1 is said to be a series of L. Furthermore, the number of strict inequalities in a series σ, is called the length of σ, in particular the above series has length n; we write it ℓpσq " n. The length of L is ℓpLq " suptℓpσq : σ a series of Lu . Example 2.9. Let C be a Grothendieck category. Notice that a series of LpM q is just a series of M . We can define a function ℓ : C Ñ Rě0 Y t8u, ℓpM q " ℓpLpM qq. Clearly, ℓpM q " 0 if and only if M " 0, ℓpM q " 1 if and only if M is a simple object and ℓpM q ă 8 if and only if M is both Noetherian and Artinian. It is a standard result that ℓ is a length function (see for example [12] for a proof in module categories).
Other examples can be obtained lifting a known length function along a localization functor as shown in Section 2.4. Another strategy to produce new examples is that of "linearly combining" some known length functions. We now describe this technique. Definition 2.10. (1) Given a length function L of C and λ P Rě0 Y t8u, we consider the function λL : C Ñ Rě0 Y t8u such that λLpM q " λ¨pLpM qq , @M P C , with the convention that 8¨0 " 0¨8 " 0; (2) given a set I and additive functions Li of C for all i P I, we consider the function It is an easy exercise to prove that the sum of length functions and the multiplication of a length function by a constant are again length functions.
Localization and torsion theories in Grothendieck categories
In this subsection we summarize some of the classical results about localization in Grothendieck categories. Indeed, given a Grothendieck category C, we study the pairs pL, ηq, where L : C Ñ C is a (additive) functor and η : id C Ñ L is a natural transformation such that Lη : L Ñ L 2 is a natural isomorphism and Lη " ηL. It is noticed in [21,Lemma 2.2] that the existence of such a pair is equivalent to the the existence of an Abelian category D and an adjoint pair of functors Q : C Õ D : S, where S is fully faithfull. In this situation L " S˝Q and D is equivalent to LC; furthermore, η is determined by L (see [21,Remark 2.3]).
In what follows we focus on the case when L is right exact, equivalently, Q is exact. In such situation, one can prove that the category D above is again a Grothendieck category and that KerpLq " tX P C : LpXq " 0u " tX P C : QpXq " 0u " KerpQq is what is usually called a hereditary torsion class. Recall that a subclass T of a Grothendieck category C is a hereditary torsion class (or localizing class) if (TC1) for every short exact sequence 0 Ñ X Ñ Y Ñ Z Ñ 0 in C, we have that Y P T if and only if X, Z P T ; (TC2) T is closed under coproducts.
If T satisfies (TC1) (but not necessarily (TC2)), then T is called a Serre class. Dually, a class F is a hereditary torsion free class if (TF1) F is closed under taking subobjects, extensions and injective envelopes; (TF2) F is closed under products.
To better understand localizations in Grothendieck categories, it is useful to introduce one more functor, that is, the right adjoint to the inclusion KerpLq Ñ C. This functor is usually called the torsion functor and it is associated with what is usually called a hereditary torsion theory. A hereditary torsion theory τ in C is a pair of classes pT , Fq such that (TT1) the class T of τ -torsion objects is a hereditary torsion class; (TT2) the class F of τ -torsion free objects is a hereditary torsion free class; (TT3) T " tM P C : Hom C pM, F q " 0 for all F P F u and F " tN P C : Hom C pT, N q " 0 for all T P T u.
Since all the torsion theories in the sequel are hereditary, we just say "torsion theory" (resp., "torsion class" and "torsion free class") to mean "hereditary torsion theory" (resp., "hereditary torsion class" and "hereditary torsion free class").
Given a torsion theory τ " pT , Fq in C, one can construct the following localization sequence: where both pairs of functors are adjunctions. In particular, inc is the inclusion of T in C (and it is therefore fully faithful); -Tτ is an idempotent left-exact radical called the τ -torsion functor, which is constructed as follows: for every object X P C, Tτ pXq is the direct union of all the the subobjects of X belonging to T , while it is defined on maps by restriction. It is the right adjoint of inc; -the Grothendieck category C{T , which is called the localization of C at τ , can be identified with the full subcategory of C given by the τ -local objects, that is, the objects X P F such that EpXq{X P F. The functor Sτ is called τ -section functor and it is the (fully faithful) inclusion of C{T in C; -the exact τ -quotient functor Qτ can be constructed as follows: Qτ pXq -π´1pTτ pEpX 1 q{X 1 qq, where X 1 " X{Tτ pXq and π : EpX 1 q Ñ EpX 1 q{X 1 is the canonical projection, for any object X P C. The τ -quotient functor is left adjoint to the τ -section functor.
Notice that, KerpQτ q " T and ImpSτ q Ď F. Notice that there is a canonical way to associate a length function to any torsion theory: Example 2.11. Let C be a Grothendieck category and let τ " pT , Fq be a torsion theory in C. Then the function L : C Ñ Rě0 Y t8u such that is a length function.
Gabriel dimension and Gabriel categories
The Gabriel filtration of C is a transfinite chain t0u " C´1 Ď C0 Ϩ¨¨Ď Cα Ď . . . of torsion classes defined as follows: -C´1 " t0u; -suppose that α is an ordinal for which Cα has already been defined. An object C P C is said to be α-cocritical if C is τα-torsion free and every proper quotient of C is τα-torsion, where τα is the unique torsion theory whose torsion class is Cα. In the sequel, we will just say α-torsion (resp., torsion free) instead of τα-torsion (resp., torsion free). We let Cα`1 be the smallest hereditary torsion class containing Cα and all the α-cocritical objects; -if λ is a limit ordinal, we let C λ be the smallest hereditary torsion class containing Ť αăλ Cα. An object M in C is said to be cocritical if it is α-cocritical for some α. For any ordinal α, we let Tα : C Ñ Cα and Qα : C Ñ C{Cα be respectively the torsion and the localization functors. Abusing notation, we use the same symbols for the functors Tα : Cα`1 Ñ Cα and Qα : Cα`1 Ñ Cα`1{Cα, induced by restriction.
Remark 2.12. By definition, a Grothendieck category C has a generator G. Furthermore, G has just a set (as opposed to a proper class) of subobject, equivalently, it has just a set of quotients. One can show that Cα`1 is the smallest torsion class containing Cα and the α-cocritical quotients of G. As a consequences we obtain that there is an ordinal κ such that Cα " Cκ for all α ě κ (just take κ " suptα : there are α-cocritical quotients of Gu).
Consider the unionC "
Ť α Cα of the Gabriel filtration (this makes sense by the above remark). An object belonging toC is said to be an object with Gabriel dimension. The Gabriel dimension of M is the minimal ordinal δ such that M P C δ , in symbols G.dimpM q " δ. In general, it may happen that C ‰C; if C "C, we say that C is a Gabriel category with Gabriel dimension G.dimpCq " κ, where κ is the smallest ordinal such that Cκ " C.
In the following example we specialize some of the above notions to categories of modules. Even if the example is stated for modules, part (1) holds for any Grothendieck category, taking a generator in place of R R. Part (2) can be generalized to any Grothendieck category with a set of Noetherian generators.
Example 2.13. Let R be a ring and let C " R-Mod be the category of left R-modules.
(1) If C has Gabriel dimension G.dimpCq " κ, then G.dimpRRq " κ. In fact, the inequality G.dimpCq ě G.dimpRRq is trivial. For the converse inequality just notice that if α is an ordinal for which RR P Cα, then Cα " C. This is because Cα is closed under arbitrary direct sums and quotients, so it contains any quotient of a free module, that is, any module.
(2) If R is left Noetherian then C is a Gabriel category. We have to prove that, given an ordinal α, Cα " Cα`1 if and only if RR P Cα. But in fact, if α is an ordinal for which RR R Cα, then the set I " tRI ď R : RpR{I q R Cαu of all the left ideals I of R such that the left R-module R{I is not in Cα is non-empty (as it contains at least the trivial ideal). By the Noetherianity of R, I has a maximal elementĪ. Then, RpR{Ī q R Cα but every proper quotient of such module is in Cα. So, RpR{Ī q is α-cocritic, in particular it belongs to Cα`1, which therefore properly contains Cα. In what follows we prove some (probably well-known) observations about Gabriel dimension. Lemma 2.14. Given a Gabriel category C, the following statements hold true: (1) G.dimpCq " suptG. dimpM q : M P Cu; (2) let α be an ordinal ă G.dimpCq and M P C, then M P Cα`1 if and only if there exists an ordinal σ and a continuous chain 0 " N0 ď N1 﨨¨ď Nσ " M , such that Ni`1{Ni is either α-cocritical or α-torsion for every i ă σ; (3) let λ be a limit ordinal ď G.dimpCq and M P C, then M P C λ if and only if M " Ť αăλ TαpM q; Proof. (1) is clear.
(2) Let A be the class of all objects which are union of a chain as in the statement. Since every hereditary torsion class is closed under taking direct limits and extension, we obtain the inclusion A Ď Cα`1. On the other hand, Cα`1 is minimal between the hereditary torsion classes which contain Cα and the αcocritical objects, thus the converse inclusion follows by the fact that A is a hereditary torsion class.
(3) The proof is analogous to part (2) and follows by the fact that the class of all objects M which can be written as M " Ť αăλ TαpM q is a hereditary torsion class, which is contained in C λ . (4) Let G.dimpM q " α, G.dimpN q " β and G.dimpM {N q " γ. As M P Cα and by the closure properties of Cα it is clear that also N and M {N belong to Cα. Hence α ě maxtβ, γu. On the other hand, C β Y Cγ Ď C maxtβ,γu and so M is an extension of two objects of C maxtβ,γu . As this class is closed under extensions, M P C maxtβ,γu , that is, α ď maxtβ, γu.
Recall that a Grothendieck category C is said to be semi-Artinian if G.dimpCq " 0. By part (2) of the above lemma, this is equivalent to say that every object in C is the union of a continuous chain of subobjects whose successive factors are simple objects. It is not difficult to show that this is equivalent to say that every object in C has a simple subobject.
Proof. As we noticed, it is enough to show that, given M P Cα`1, QαpM q has a simple subobject. By Lemma 2.14 (2), we can find a sequence N1 ď N ď M , where N1 P Cα and N {N1 is α-cocritical. By the exactness of Qα and the fact that QαpN1q " 0, we have that QαpN {N1q -QαpN q ď QαpM q. It is an exercise to show that QαpN q is simple. Corollary 2.16. Let C be a Gabriel category and N P C be a Noetherian object. Then, G.dimpN q is a successor ordinal. Furthermore, there exists a finite series 0 " N0 ď N1 﨨¨ď N k " N such that Ni{Ni´1 is cocritical for all i " 1, . . . , k.
Proof. Suppose, looking for a contradiction, that G.dimpN q " λ for some limit ordinal λ. By part (3) of the above lemma, N " Ť βăλ T β pN q and so, by the Noetherian condition, N " T β pN q P C β for some β ă λ, a contradiction. So we have that G.dimpN q " α`1 for some ordinal α ă G.dimpCq. Let us proceed by transfinite induction. If α "´1, then by part (2) of the above lemma, there exists a continuous chain 0 " N0 ď N1 﨨¨ď Nσ " N , such that Ni`1{Ni is either α-cocritical or α-torsion (i.e., trivial) for every i ă σ. Such chain has to be finite by the Noetherian condition so we can conclude. Now, if α ą´1 (successor or limit ordinal), again by part (2) of the above lemma, there exists a continuous chain 0 " N0 ď N1 﨨¨ď Nσ " N , such that Ni`1{Ni is either α-cocritical or α-torsion for every i ă σ. Such chain has to be finite by the Noetherian condition and it can be refined to a finite chain whose factors are all cocritical using the inductive hypothesis applied to each of the Noetherian α-torsion factors.
Operations on length functions
Let C be a Grothendieck category and let τ " pT , Fq be a torsion theory on C. In this section we are going to show how the length functions of C are related with the ones of T and C{T . We start with the following Proposition 2.17. In the above notation, there is a bijective correspondence Proof. The maps f and g are defined in Lemmas 2.18 and 2.19 respectively. It follows by the definitions that they are inverse one to the other. Lemma 2.18. If Lτ : C{T Ñ Rě0 Y t8u is a length function, then there exists a unique length function L : C Ñ Rě0 Y t8u such that LpM q " Lτ pQτ pM qq for all M P C. Furthermore, T Ď KerpLq. We set gpLτ q " L.
Proof. Existence follows by the fact that Qτ is an exact functor that preserves colimits. Uniqueness is clear and the last statement comes from the fact that T " KerpQτ q.
Proof. For every M P C, there is an exact sequence of the form 0 Ñ T1 Ñ M Ñ Sτ Qτ pM q Ñ T2 Ñ 0, with T1, T2 P T . As by hypothesis L is trivial on τ -torsion objects, we obtain by additivity that LpM q " LpSτ Qτ pM qq. Using this simple observation, we can define Lτ pN q " LpSτ pN qq, for all N P C{T , and verify that LpM q " LpSτ Qτ pM qq " Lτ pQτ pM qq as desired. The uniqueness statement follows by the fact that the functor Qτ is essentially surjective. It remains to verify that Lτ is a length function.
The proof that L is upper continuous follows by a similar argument and transfinite induction.
In the first part of this subsection we described how to transfer length functions along the adjoint pair Qτ : C Õ C{T : Sτ ; now we turn our attention to the adjunction inc : T Õ C : Tτ . In particular, whenever L is a length function on C, one can define its restriction to T as LaeT : T Ñ Rě0 Y t8u, such that LaeT pM q " LpM q for all M P C, and prove that it is a length function. Notice that this can be applied to any full Abelian subcategory T of C, not only to hereditary torsion classes.
On the other hand, if we start with a length function L on T , we want to find a canonical way to extend it to the bigger category C. In [36] (see also [35]) Peter Vámos introduced a technique to extend length functions which works in a more general setting. Indeed, let T be a Serre subclass of C and consider a length function L on T . We start defining an invariant (which is not supposed to have any good property but that of being useful for our constructions) L˚: C Ñ Rě0 Y t8u as follows: Given an object M P C and a series σ : Definition 2.20. In the above notation, the Vámos extension p L : C Ñ Rě0 Y t8u of L to C is the function defined by p LpM q " supt p Lpσq : σ ranging over all the series of M u, for all M P C. In the next proposition we verify that p L satisfies the axioms of a length function.
Proof. The proof of additivity is standard and it essentially follows by the definitions and the Artin-Schreier's Refinement Theorem. It remains to prove upper continuity. Indeed, let M P C and consider a directed set tMα : α P Λu of subobjects such that ř Λ Mα " M . By additivity, we readily get p LpM q ě sup Λ p LpMαq. On the other hand, given a series σ : 0 " N0 Ĺ N1 Ũ¨¨Ĺ Nn " M , we prove by induction on n P N`that p Lpσq ď sup Λ p LpMαq. We distinguish two cases: (1) if p Lpσq " 8, then there exists a non-negative integer m ă n such that Nm`1{Nm P T and LpNm`1{Nmq " 8. Notice also that Nm`1{Nm " ř Λ ppMα X Nm`1q`Nmq{Nm and so, where the first inequality comes by additivity of p L and the following equalities come by the fact that p L coincides with L on T .
(2) Suppose now that p Lpσq ă 8. If n " 1, then either p Lpσq " 0 and there is nothing to prove, or 0 ă p Lpσq " L˚pM q, but in this case M P T and the thesis follows by the fact that L is a length function on T . Suppose now n ą 1, let where the last equality comes from the additivity of p L and the fact that the sum of suprema of two monotone nets is the supremum of the sum of the two nets.
We conclude this section proving that Vámos extension and lifting of length functions via a localization functor preserve linear combinations.
Lemma 2.22. Let T be a Serre subclass of the Grothendieck category C, let Λ be a set and choose λpαq P Rě0 Y t8u for all α P Λ.
(1) By the minimality of Vámos extension proved in Proposition 2.21, we have that On the other hand, if M P C and p LpM q " 8, then 8 " p LpM q ď ř Λ λpαq x LαpM q and there is nothing to prove. It remains to show that, if p LpM q ă 8, then p LpM q ě ř Λ λpαq x LαpM q. The case |Λ| ă 8 is essentially an application of the Artin-Schreier's Refinement Theorem. Suppose now that Λ is not finite and let LF " ř αPF λpαqLα for every non-empty finite subset F Ď Λ; by definition LpM q " suptLF pM q : F Ď Λ finiteu. By the first part of the proof, x LF " ř αPF λpαq x Lα, so we have only to prove that p LpM q ě supt x LF pM q : F Ď Λ finiteu for all M P C. This follows noticing that x LF pσq ď p Lpσq for any finite F Ď Λ and any series σ of M .
(2) follows by definition of the map g.
The classification in the semi-Artinian case
All along this subsection we denote by C a semi-Artinian Grothendieck category, that is, a Gabriel category whose Gabriel dimension is 0. The main result of this subsection is to give a structure theorem for all the length functions in C. Proof. One implication is trivial, so suppose that L and L 1 coincide on simple objects. Consider an object M P C and write it as the union of a continuous chain such that Ni`1{Ni is a simple object for all i (see Lemma 2.14(2)). By hypothesis LpNi`1{Niq " L 1 pNi`1{Niq for all i. The conclusion follows by transfinite induction using additivity and upper continuity.
Definition 2.24. Given a semi-Artinian category C, we let SppCq be the set of (isomorphism classes of ) all the simple objects in C.
In a general Abelian category, the Gabriel spectrum is the set of (isomorphism classes of) all the indecomposable injective objects. One can prove that, for a semi-Artinian Grothendieck category C, SppCq is in bijection with the Gabriel spectrum. In fact, the indecomposable injective objects in C are exactly the injective envelopes of simple objects.
Let S P SppCq be a simple object. There is a torsion theory π " pTπ, Fπq such that Tπ " tM P C : Hom C pM, EpSqq " 0u (and Fπ is the maximal right Hom-orthogonal class to Tπ). One can show that π determines S up to isomorphism, in fact, S is the unique simple object (up to isomorphism) in Fπ; for this reason, we often abuse notation and write π P SppCq and denote by Spπq the simple object associated to the torsion theory π.
Definition 2.25. Let C be a semi-Artinian Grothendieck category and let π P SppCq. The atomic length function ℓπ associated to π is the lifting (in the sense of Lemma 2.18) of the composition length ℓ : C{Tπ Ñ Rě0 Y t8u via the localization functor Qπ : C Ñ C{Tπ, that is, ℓπ : C Ñ Rě0 Y t8u is the unique length function such that ℓπpM q " ℓpQπpM qq , @M P C .
By Lemma 2.18, Kerpℓπq " Tπ and, given a simple object S P C, one easily shows that ℓπpSq " In the following theorem we prove that any length function in a semi-Artinian Grothendieck category is a linear combination of atomic length functions.
Theorem 2.26. Let C be a semi-Artinian category and let
where λpπq " LpSpπqq. Furthermore, the constants λpπq are uniquely determined by L.
Proof. Let L 1 " ř πPSppCq λpπq¨ℓπ; we already mentioned that a linear combination of length functions is a length function, so L 1 is a length function. Also, we already noticed that ℓπpSpπ 1 qq " 0 for all π 1 ‰ π, so L 1 pSpπqq " λpπqℓπpSpπqq " LpSpπqq for all π P SppCq, which shows that L " L 1 , by Lemma 2.23. The proof of the uniqueness statement is analogous.
Using the uniqueness of the above decomposition, we can unambiguously define the support of L as SupppLq " tπ P SppCq : λpπq ‰ 0u .
As an immediate consequences of the above theorem we obtain the following corollaries.
Finally, if LpDq ă 8 then the above equivalent conditions are verified with α " LpDq.
Proof. The implication (2)ñ(1) is trivial, in fact Kerprkq " T . Let us prove that (1)ñ(2). By Theorem 2.17, L is the lifting of a length function on D-Mod{T -Q-Mod, where Q is the skew field of left fractions of D. Clearly, Q-Mod satisfies the hypotheses of the above corollary, thus there exists α P Rě0 Y t8u such that L " gpα¨dimQq " α¨gpdimQq " α¨rk. For the last statement, we suppose LpDq ă 8 and we verify (1). Let M be a torsion left D-module. Using upper continuity and additivity we can prove that LpM q " 0 if and only if LpDxq " 0 for all x P M . Thus, let x P M be a non-trivial element and consider the following exact sequences where the second sequence exists as AnnDpxq is a non-trivial left ideal of D (as M is torsion) and D is a domain. This shows that LpDq " LpDxq`LpAnnDpxqq ě LpDxq`LpDq; thus LpDxq ď LpDq´LpDq " 0. Finally, LpDq " α¨rkpDq " α.
The main structure theorem
All along this section we let C be a Gabriel category such that G.dimpCq " κ. The main result of this subsection is to show that, analogously to the semi-Artinian case, any length function on C can be written as a linear combination of atomic length functions.
Let us start defining atomic length functions in this broader context. Indeed, given an ordinal α ď κ and letting Cα be the α-th member of the Gabriel filtration of C, we already noticed that all the localized categories of the form Cα`1{Cα, are semi-Artinian.
Definition 2.29. Let C be a Gabriel category with G.dimpCq " κ. We let Sp α pCq " SppCα`1{Cαq and SppCq " Ť αăκ Sp α pCq. For κ " 0, the above definition of SppCq coincides with our previous definition. As for semi-Artinian categories, one can find a natural bijection between SppCq and the Gabriel spectrum of C. In fact, any indecomposable injective object in C is isomorphic to EpSαpCqq for some α ą κ and some simple object C P Sp α pCq.
For any α ă κ and π P Sp α pCq, we already defined a length function (see Definition 2.25) Definition 2.30. Let C be a Gabriel category and let α ă G.dimpCq. For any given π P Sp α pCq, the atomic length function ℓπ : C Ñ Rě0 Y t8u is defined as the Vámos extension of the lifting of the atomic length function ℓπ : Cα`1{Cα Ñ Rě0 Y t8u to a length function on Cα`1 (in the sense of Lemma 2.18).
It follows directly by the definitions that Cα Ď Kerpℓπq.
For the rest of this subsection we fix a length function L on C. We also let FinpLq be the Serre class of all the L-finite objects, FinpLq " tM P C : LpM q ă 8u .
Furthermore, we denote by FinpLq the minimal torsion class containing FinpLq.
Notice that the length function L 8 belongs to the family of functions described in Example 2.11. The fact that L f in is a length function follows by Proposition 2.21. We remark that the finite component L f in can very well assume infinite values (if L f in is non-trivial, just take M such that L f in pM q ‰ 0, so L f in p À N M q " 8); anyway its name is justified by the fact that L f in is, by definition, determined by the finite values of L.
Notice that L " L f in`L8 . This decomposition of L allows us to reduce the problem of finding a presentation of L as linear combination of atomic length functions to the same problem for L 8 and L f in . On the other hand, the decomposition of L 8 is quite easy. Indeed, for all α ă κ we let τ L α " pT L α , F L α q be the torsion theory in Cα`1{Cα such that T L α " tQαpM q : M P FinpLqu. Definition 2.32. In the above notation, the infinite support of L is the following subset of SppCq: Proposition 2.33. In the above notation, L 8 " ř πPSupp 8 pLq 8¨ℓπ.
Proof. Let L 1 " ř πPSupp 8 pLq 8¨ℓπ. Both L 8 and L 1 take values in t0, 8u, thus they coincide if and only if KerpL 8 q " KerpL 1 q. By the above observation, We can now turn our attention to the decomposition of L f in .
Lemma 2.34. In the above notation, there exists a unique family tL pαq : Cα`1 Ñ Rě0 Y t8u : α ă κu of length functions such that Proof. For all α ď G.dimpCq we consider the Serre classes Fin pαq pLq " FinpLq X Cα and FinpLq. We start defining inductively length functions L pαq : Fin pα`1q pLq Ñ Rě0 Y t8u, and their Vámos extensions It is not difficult to verify by transfinite induction that all the L pαq and L α are length functions. Let us verify the following claims by induction on G.dimpCq: If G.dimpCq " 0 (i.e., C is semi-Artinian), then LpM q " L p´1q pM q " L´1pM q for all M P Fin p0q pLq " FinpLq, proving (1'), while (2') is trivial since Fin p´1q pLq " t0u. Suppose now that G.dimpCq is a limit ordinal. If M P Fin pαq pLq for some α ă G.dimpCq, then by inductive hypothesis, LpM q " ř βăα L β pM q and so L pαq pM q " LpM q´ř βăα L β pM q " 0, proving (2'). Furthermore, given M P FinpLq, we can write M " Ť βăG.dimpCq pT β pM qq (see Lemma 2.14 (3)). Then, where the first equality follows by the upper continuity of L, the second one follows by part (1') of the inductive hypothesis (T β pM q P C β and G.dimpC β q " β ă G.dimpCq), the third one follows by the, already established claim (2'), and the last equality uses the upper continuity of ř αăG.dimpCq L α . Finally, if G.dimpCq " κ`1 is a successor ordinal, and M P Fin pκq pLq, then LpM q " ř αăκ L α pM q by inductive hypothesis and so L pκq pM q " LpM q´ř αăκ L α pM q " 0, proving (2'). Furthermore, for any M P FinpLq, where p˚q comes by the fact that Cκ`1 " C so Fin pκ`1q pLq " FinpLq and L κ " L pκq . Thus, also (1') is established.
For any α ă G.dimpCq we define the length functions L pαq : Cα`1 Ñ Rě0 Yt8u and Lα : C Ñ Rě0 Yt8u, as the Vámos extensions of L pαq and L α respectively. We can extend the above claims (1') and (2') to these new functions as follows. First of all, L f in pM q " ř αăG.dimpCq LαpM q by (1') and the fact that Vámos extension preserves linear combinations. Furthermore, for all α ă G.dimpCq and M P Cα, L pαq pM q " 0 by the construction of Vámos extension and since, by (2'), L pαq vanishes on all the factors belonging Fin pαq pCq of any series of M . This shows that Cα Ď KerpL pαq q.
The proof of the uniqueness can be obtained by transfinite induction on κ.
By the above lemma, we have uniquely determined length functions L pαq : Cα`1 Ñ Rě0 Y t8u for all α ă G.dimpCq, such that L pαq is trivial on Cα. Thus, there exist unique length functions such that L pαq pM q " LαpQαpM qq for all M P Cα`1 (see Lemma 2.19). Since all the categories of the form Cα`1{Cα are semi-Artinian and so we have a well defined notion of support for the functions Lα.
Definition 2.35. In the above notation, the finite support of L is the following subset of SppCq: With the above notion of support we can finally decompose L f in as linear combination of atomic length functions.
Proposition 2.36. In the above notation, there is a unique choice of constants λpπq P Rą0 so that Proof. For all α ă κ, we have a uniquely determined decomposition Lα " ř πPSupppLαq λpπqℓπ in Cα`1{Cα, by Theorem 2.26. Furthermore, by Lemma 2.22, this decomposition can be lifted to a decomposition in Cα`1 and then extended to a decomposition in C, obtaining that The desired decomposition now follows by Lemma 2.34.
We conclude this section summarizing the main results on decomposition of length functions in Gabriel categories in the following theorem. We remark that this statement is analogous to the "Main Decomposition Theorem" in [36] Theorem 2.37. Let C be a Gabriel category and let L : C Ñ Rě0 Y t8u be a length function. Then, there is a unique way to choose constants λpπq P Rą0, for all π P Supp f in pLq such that
Length functions compatible with self-equivalences
All along this section, let C be a Grothendieck category, let L : C Ñ Rě0 Y t8u be a length function and let F : C Ñ C be a self-equivalence, that is, (Eq. 1) F is essentially surjective, i.e., for all X P C, there exists Y P C such that F pXq -Y ; (Eq. 2) F is fully faithful, i.e., for all X, Y P C, the natural morphism Hom C pX, Y q Ñ Hom C pF pXq, F pY qq is an isomorphism.
A consequence of the definition is that F preserves any structure defined by universal properties, in particular, F commutes with direct and inverse limits and it preserves exactness of sequences. Furthermore, F commutes with injective envelopes and it preserves lattices of subobjects.
It is easily seen that for all M P C is a length function. In what follows we are going to study to what extent LF can differ from L.
The following example shows that L and LF may be very different.
Example 3.1. Consider a field K and consider the category KˆK-Mod -K-ModˆK-Mod. This category is semi-Artinian and it has a self-equivalence F : KˆK-Mod Ñ KˆK-Mod such that pM, N q Þ Ñ pN, M q and pφ, ψq Þ Ñ pψ, φq. If we take L to be the length function such that LppM, N qq " dimK pM q, then clearly LF ppM, 0qq " 0 ‰ dimK pM q " LppM, 0qq, provided M ‰ 0.
Definition 3.2. Given a Grothendieck category C and a self-equivalence F : C Ñ C, we say that a length function L : C Ñ Rě0 Y t8u is compatible with F provided LF pM q " LpF pM qq " LpM q for all M P C.
In this section we exploit the classification of length functions in Gabriel categories to find a necessary and sufficient condition on L to be compatible with F . Our motivation for studying compatibility of length functions with self-equivalences is the following: given a ring R and a ring automorphism φ : R Ñ R, we define a self-equivalence which is the scalar restriction along φ. That is, take a left R-module R M such that R acts on M via a ring homomorphism λ : R Ñ End Z pM q. Then F φ pM q is isomorphic to M as an Abelian group, while R acts on F φ pM q via the ring homomorphism λ˝φ : R Ñ End Z pF φ pM qq.
We are interested in finding length functions L such that LpF φ pM qq " LpM q.
Orbit-decomposition of the Gabriel spectrum
We start with a technical result. Let A be a subclass of C, we denote by r A the class of all the objects of C which are isomorphic to some object in A. Lemma 3.3. Let C be a Grothendieck category, let F : C Ñ C a self-equivalence and let τ " pT , Fq be a torsion theory. The following are equivalent: (2) Č F pFq " F; (3) given X P C, X is τ -local if and only if F X is τ -local; Proof. The equivalence between (1) and (2) follows since Hom C pA, Bq -Hom C pF pAq, F pBqq, for all A, B P C and by the fact that F " T K and T " K F.
(1)&(2)ñ(3). Given a τ -local X P C, one can consider the following exact sequence where T P T . It follows from (2) that F pXq P F and so Tτ pF Xq " 0. Furthermore, it is a standard fact that T -Tτ pEpF Xq{F Xq. Since F is an equivalence, it is exact and it commutes with injective envelopes, so T -Tτ pF pEpXq{Xqq which is trivial by the fact that X is τ -local (implying that EpXq{X P F) and (2). On the other hand, if F X is local, then Tτ pF Xq " 0 and Tτ pF pEpXq{Xqq, which is equivalent to say that X is τ -local, by the exactness of F and (1).
(3)ñ(1). It follows by the fact that the τ -torsion objects are exactly the objects not admitting non-trivial morphisms to a τ -local object.
Using (4), F X -Sτ Qτ F pXq, which is equivalent to say that F X is τ -local.
Recall that, for a Gabriel category C, the spectrum SppCq of C is defined as the union of all the spectra Sp α pCq " SppCα`1{Cαq, with α ă G.dimpCq. In particular, Sp´1pCq " SppC0q is a set of representatives of the isomorphism classes of the simple objects in C.
In the following lemma we show that the equivalence F : C Ñ C induces a bijection of Sp´1pCq onto itself. This fact is then applied in Proposition 3.5 to show that F induces bijections of Sp α pCq onto itself, for all α ă G.dimpCq. Proof. The fact that F sends simples to simples follows by the well-known fact that an equivalence preserves the lattice of subobjects of a given object. For the second part of the statement, just notice that, given two simple objects S1 and S2, S1 -S2 is equivalent to F pS1q -F pS2q. Furthermore, f´1 is surjective by the essential surjectivity of F . (2α) the functor Fα : C{Cα Ñ C{Cα, defined by the composition Fα " QαF Sα is an equivalence; for all´1 ď α ă G.dimpCq. In particular, via the identification Sp α pCq " SppCα`1{Cαq " Sp´1pC{Cαq, each Fα induces a bijection fα : Sp α pCq Ñ Sp α pCq , defined as in Lemma 3.4. Proof. We prove simultaneously (1α) and (2α) by transfinite induction on α.
In case α "´1, then (1´1) just says that F pt0uq " t0u, so it is trivially verified, while (2´1) is true as F´1 is just F .
Suppose now that (1α) and (2α) are verified for some´1 ď α ă G.dimpCq. Notice that Cα`1{Cα can be identified with pC{Cαq0, that is, the smallest hereditary torsion subclass of C{Cα containing all the simple objects of this category. Since Fα is an equivalence, it sends hereditary torsion classes to hereditary torsion classes so the isomorphism closure of FαpCα`1{Cαq is exactly Cα`1{Cα, as by Lemma 3.4 the image under an equivalence of a class containing all the simples, contains all the simples. Thus, given X P C, we have the following equivalences: pX P Cα`1q ô pX :" QαpXq P Cα`1{Cαq ô pDȲ P Cα`1{Cα : FαpȲ q -Xq ô pDY P Cα`1 : FαQαpY q -Xq , where the second equivalence follows by (1α). Now, to prove (1α`1) we have to show that the last of the above conditions is equivalent to pDY P Cα`1 : F pY q -Xq. The implication pDY P Cα`1 : F pY q -Xq ñ pDY P Cα`1 : FαQαpY q -Xq is trivial, as QαF pY q -QαF SαQαpY q " FαQαpY q. For the converse implication, assume that there exists Y P Cα`1 such that FαQαpY q -QαpXq. We have the following diagram with exact rows: where Y 1 " SαQαpY q and T1, T2, T3, T4 P Cα. Using (2α), one obtains Cα " Č F pCαq Ď Č F pCα`1q, so the first line of the diagram says that SαQαF pY 1 q P Č F pCα`1q, the isomorphism then shows that SαQαX P Č F pCα`1q. One concludes by the second line that X P Č F pCα`1q, proving (1α`1). In order to show (2α`1) one has to show that Fα`1 is essentially surjective and fully faithful. The former is verified as follows: takeX P C{Cα`1, let X " Sα`1X, choose Y P C such that F pY q -X and letȲ " Qα`1Y , one concludes that where p˚q is given by Lemma 3.3 (4) (using that we already verified (1α`1), which corresponds to part (1) of that lemma). To verify full faithfulness take X, Y P C{Cα`1, then where the last isomorphism follows as, by Lemma 3.3 (3), F Sα`1X and F Sα`1Y are both pα`1q-local.
Finally, let λ ď G.dimpCq be a limit ordinal and assume (1α), (2α) for all α ă λ. Then, (1 λ ) trivially follows recalling that C λ is the smallest hereditary torsion class containing Cα " Č F pCαq for all α ă λ and the same description can be given for Č F pC λ q. Furthermore, (2 λ ) follows by (1 λ ) and Lemma 3.3 exactly as in the successor case.
Motivated by the above proposition, we can give the following definition.
Definition 3.6. Given a Gabriel category C, a self-equivalence F : C Ñ C and a point π P Sp α pCq Ď SppCq, let OF pπq " tf n α pπq : n P Zu be the F -orbit of π, where fα : Sp α pCq Ñ Sp α pCq is the bijective map described in Proposition 3.5.
It is clear that each point of the spectrum belongs to a unique F -orbit. In particular, the F -orbits induce a partition of the Gabriel spectrum.
Characterization of compatible length functions in Gabriel categories
In this subsection we use the orbit decomposition of the Gabriel spectrum and the classification of length functions in Gabriel categories to give a sufficient and necessary condition for a length function to be compatible with a given self-equivalence. (2) if π P Supp f in pLq, then a. OF pπq Ď Supp f in pLq; b. λpπ 1 q " λpπq, for all π 1 P OF pπq.
Let us recall our decomposition of L : C Ñ Rě0 Y t8u obtained in Section 2.6. The first thing we did was to define a torsion theory τ " pT , Fq, where T " FinpLq is the smallest hereditary torsion class containing FinpLq. This allowed us to write L a sum of its finite and infinite components: where L 8 assumes the value 0 on T and 8 elsewhere, while L f in is the Vámos extension of the restriction of L to T . Similarly we can define a torsion theory τF " pT F , F F q with T F " FinpLF q (for the definition of LF see (3.1)). This induces a decomposition Theorem 3.7 will follow showing that τ " τF (or, equivalently, L 8 " L 8 F ) and L f in " L f in F . In the setting of Theorem 3.7, we denote by fα the self-bijection of Sp α pCq induced by F . Lemma 3.8. In the above notation, the following are equivalent: (1) τ " τF , that is, L 8 " L 8 F ; (2) π P Supp 8 pLq implies OF pπq Ď Supp 8 pLq, for all π P SppCq.
Proof. Let α ă G.dimpCq and choose π P Sp α pCq. We can now concentrate on showing that L f in " L f in F is equivalent to condition (2) in Theorem 3.7: Lemma 3.9. In the above notation, the following are equivalent: Proof. For all α ă G.dimpCq, let Lα, pLF qα : Cα`1{Cα Ñ Rě0 Y t8u be the functions described in (2.6) relative to L and LF respectively. Using Lemma 3.3, one can follow the steps of the construction of Lα in the proof of Lemma 2.34 and show that pLF qαpM q " LαpF pM qq for all M P Cα`1. Thus pLF qαpM q " LαpFαpM qq for all M P Cα`1{Cα (where as usual Fα " QαF Sα). By the structure of length functions in Gabriel categories, L f in " L f in F if and only if Lα " pLF qα, for all α ă G.dimpCq. Choose α ă G.dimpCq and let π P Sp α pCq. Then pLF qαpSpπqq " LαpFαpSpπqqq " LαpSpfαpπqqq .
Thus, Supp f in α pLq " fαpSupp f in α pLF qq and so Supp f in α pLq " Supp f in α pLF q if and only if Supp f in α pLq is fα and f´1 α -invariant, which is condition (2)a. in the statement. Furthermore, given π P Supp f in α pLq, the constant associated to π in the decomposition of L is LαpSpπqq, while the constant associated to π in the decomposition of LF is LαpSpfαpπqqq. Thus the two functions coincide if and only if Lα is constant on the simple objects belonging to the same orbit under fα, that is, condition (2)b in the statement.
Examples
Let K be a division ring and consider the category K-Mod of left K-modules. SppK-Modq consist of a single point, thus Theorem 3.7 says that any length function is compatible with any self-equivalence of K-Mod. On the other hand, we already proved in Corollary 2.27 that the length functions on K-Mod are just multiples of the composition length (which in this case is just the dimension over K) so this is not a very deep result.
The previous example can be generalized as follows. Let C be a Gabriel category and consider the composition length ℓ : C Ñ Rě0 Y t8u. Then, Finpℓq " C0 and so Supp 8 pℓq " SppC{C0q " Ť αě0 Sp α pCq, while Supp f in pℓq " SppC0q " Sp´1pCq. Clearly both the finite and the infinite spectrum are invariant under any family of self-bijections tfα : Sp α pCq Ñ Sp α pCq : α ă G.dimpCqu. Furthermore, the constants associated to each π P Supp f in pℓq in the decomposition of ℓ as linear combination of atomic functions are all 1. Thus, Theorem 3.7 can be applied to show that ℓ is compatible with any self-equivalence of C.
A further generalization of the above example can be achieved as follows. Let C be a Gabriel category, for any α ă G.dimpCq we define a length function where ℓ is the composition length in C{Cα`1. One can show that Supp f in pℓαq " Sp α pCq and Supp 8 pℓαq " Theorem 3.7 implies that ℓα (and any of its multiples) is compatible with any self-equivalence of C.
Amenable groups and crossed products 4.1 Amenable groups
Amenable groups were defined by John von Neumann in 1929 as groups admitting a left-invariant mean.
We adopt here an equivalent definition of amenability given by Følner [13]. Indeed, consider two subsets A, C Ď G, then -the C-interior of A is InCpAq " tx P G : xC Ď Au; -the C-exterior of A is OutC pAq " tx P G : xC X A ‰ Hu; -the C-boundary of A is BCpAq " OutC pAqzInCpAq.
If e P C, one can imagine the above notions as in the following picture Any such net is called a Følner net.
If G is countable, then one can take I " N and just speak about Følner sequences. Furthermore, given a countable group G and a Følner sequence s " tFn : n P Nu, we say that s is a Følner exhaustion if (1) e P F0 and Fn Ď Fn`1 for all n P N; (2) Ť nPN Fn " G. It can be proved that a finitely generated group is amenable if and only if it admits a Følner exhaustion (see for example [30,Lemma 5.3]). (1) Any finite group and any finitely generated Abelian group is amenable; (2) consider a finitely generated group G and let S " S´1 be a finite set of generators containing the unit of G. For all n P N`, let BnpSq " ts1¨. . .¨sn : si P Su and define a function: fS : N`Ñ N`such that fSpnq " |BnpSq| .
The group G is said to be of sub-exponential growth if the growth of fS is sub-exponential (it can be shown that this notion does not depend on the choice of the generating set S). If G is of subexponential growth, then tBnpSq : n P N`u is a Følner exhaustion for G (see for example [4, Section 6.11]); (3) Non-commutative free groups are not amenable. So, any group which contains a free subgroup of rank 2 is not amenable. There exist amenable groups with exponential growth.
(4) It is known that the class of amenable groups is closed under the operations of taking subgroups, taking factors over normal subgroups, taking extensions and taking increasing unions. We obtain that a group G is amenable if and only if all its finitely generated subgroups are amenable, in particular, arbitrary Abelian groups and locally finite groups are amenable.
Non-negative real functions on finite subsets of an amenable group
In their seminal paper [27], Ornstein and Weiss introduced a notion of entropy for actions of amenable groups on metric spaces. Using the theory of quasi-tiles, they were able to prove that the lim sup defining their entropy is a true limit. In what follows, we recall some of these deep results as they can be applied with just minor changes to our algebraic setting. The following terminology and results are due to Ornstein and Weiss [27] (see also [19] and [37]).
The subsets C1, . . . , C k are called tiling centers. It is a deep result, due to Ornstein and Weiss, that whenever G is an amenable group and tFnu nPN is a Følner exhaustion, for any (small enough) ε ą 0, one can find a nice family of subsets of G that ε-quasi-tiles Fn for all (big enough) n P N. More precisely:
In the rest of the subsection we recall some results and terminology about non-negative invariants for the finite subsets of G, that is, functions f : FpGq Ñ Rě0. In particular, we say that f is (1) monotone if f pAq ď f pA 1 q, for all A Ď A 1 P FpGq; (2) sub-additive if f pA Y A 1 q ď f pAq`f pA 1 q, for all A, A 1 P FpGq; (3) (left) G-equivariant if f pgAq " f pAq, for all A P FpGq and g P G.
A consequence of (3) above is that f ptguq " f pteuq, for all g P G. Thus by (2), f pAq ď ř gPA f ptguq " |A|f peq, for all A P FpGq.
The following result, generally known as "Ornstein-Weiss Lemma", is proved in [27] for a suitable class of locally compact amenable groups (a direct proof, along the same lines, in the discrete case can be found in [37], while a nice alternative argument, based on ideas of Gromov, is given in [22]). We conclude this paragraph with the following consequence of Theorem 4.3. The proof is essentially given in [19], we give it here for completeness sake, as our statement is slightly different.
Corollary 4.5. Let G be a finitely generated amenable group and tFnu nPN be a Følner exhaustion of G.
Then, for any ε P p0, 1{4q andn P N there exist integers n1, . . . , n k such thatn ď n1 﨨¨ď n k and, for any sub-additive and G-equivariant f : where M " f pteuq.
It is important to underline that the choice of the n1, . . . , n k does not depend on the function f , but we can really find a family tn1, . . . , n k u, which works for all f at the same time.
Crossed products
Given a group G and a ring R, a crossed product R˚G of R with G is a ring constructed as follows: as a set, R˚G is the collection of all the formal sums of the form ÿ gPG rgg , with rg P R and rg " 0 for all but finite g P G, and where the g are symbols uniquely assigned to each g P G. Sum in R˚G is as expected, that is, it is defined component-wise exploiting the addition in R: In order to define a product in R˚G, we need to take two maps σ : G Ñ AutringpRq and ρ : GˆG Ñ U pRq , where AutringpRq denotes the group of automorphisms of R in the category of (unitary) rings, while U pRq is the group of units of R. Given g P G and r P R we denote the image of r via the automorphism σpgq by r σpgq . We suppose also that σ and ρ satisfy the following conditions for all r P R and g1, g2 and g3 P G: (Cross.1) ρpg1, g2qρpg1g2, g3q " ρpg2, g3q σpg 1 q ρpg1, g2g3q; (Cross.2) r σpg 2 qσpg 1 q " ρpg1, g2qr σpg 1 g 2 q ρpg1, g2q´1; (Cross.3) ρpg, eq " ρpe, gq " 1 (for all g P G) and σpeq " 1.
The product in R˚G is defined by the rule prgqpshq " rs σpgq ρpg, hqgh, together with bilinearity, that is˜ÿ By (Cross.1) and (Cross.2) above, R˚G is an associative and unitary ring, while by (Cross.3) 1R˚G " e.
Of course the definition of R˚G does not depend only on R and G, as the choice of σ and ρ is fundamental for defining the product. Anyway one avoids a notation like RrG, ρ, σs and uses the more compact (though imprecise) R˚G. Of course, the easiest example of crossed group ring is the group ring RrGs, which corresponds to trivial maps σ and ρ. For more details on this kind of construction we refer to [29].
Notice that there is a canonical injective ring homomorphism R Ñ R˚G r Þ Ñ re .
In view of this embedding we identify R with a subring of R˚G. This allows one to consider a forgetful functor from R˚G-Mod Ñ R-Mod.
On the other hand, in general there is no natural map G Ñ R˚G which is compatible with the operations, anyway the obvious assignment g Þ Ñ g respects the operations modulo some units of R. Similarly, given a left R˚G-module M , there is a canonical map λ : G Ñ Aut Z pM q , g Þ Ñ λg , λgpmq " gm .
(4.2) which is not in general a homomorphism of groups. Given an R-submodule R K ď M and an element g P G, λgpKq is again an R-submodule of M but it is not in general isomorphic to RK . As described in (3.2), there is a self-equivalence of the category R-Mod, induced by the ring automorphism σpgq It follows by the definitions that λgpKq -gK. In particular, if L is a length function compatible with the equivalence F σpgq , then LpλgpKqq " LpKq. This useful fact motivates the following Definition 4.6. Let R˚G be a given crossed product and let L : R-Mod Ñ Rě0 Y t8u be a length function. L is said to be compatible with R˚G provided L is compatible with F σpgq , for all g P G.
The action of G on monoids of submodules
Let R be a ring, let G a group, fix a crossed product R˚G and let M be a left R˚G-module. The set LpM q of R-submodules of M is a bounded modular and complete lattice (with order induced by the inclusion of submodules). In fact, this lattice structure is more than what we really need: we just need to know that the least upper bound of two elements N, N 1 P LpM q is given by N`N 1 . This operation makes LpM q " pLpM q,`, 0q into a commutative monoid.
In general, given a category C, an object M P C and a group G, a left action of G on M is a group homomorphism from G to Aut C pM q. In particular, the fact that the map λ : G Ñ Aut Z pM q given by the R˚G-module structure on M (see (4.2)) is not a homomorphism of groups is telling us that the group G is not really acting on M . On the other hand, the structure of R˚G-module induces an action of G on the monoid LpM q: where AutpLpM qq is the group of monoid automorphisms of LpM q and λgpKq " tgk : k P Ku, for all g P G and K P LpM q.
Lemma 4.7. In the above notation, λg : LpM q Ñ LpM q is a monoid automorphism for all g P G and λ : G Ñ AutpLpM qq is a group homomorphism.
Proof. Given g P G, it is easily seen that λgpK1`K2q " λgpK1q`λgpK1q, for all K1, K2 P LpM q, while λ g´1 is the inverse of λg, so λg is a monoid automorphism. Now, let g, h P G and let K P LpM q, then λgλ h pKq " ghK " ρpg, hqghK " ghρpg, hq σpghq K " ghK " λ gh pKq , since ρpg, hq σpghq is a unit of R, showing that λ is a group homomorphism.
For any subset F Ď G and any K P LpM q, the F -th λ-trajectory of K is Notice that TF pλ, Kq P LpM q. The (full) λ-trajectory of K is TGpλ, Kq.
Lemma 4.8. In the above notation, R˚GM is finitely generated as a left R˚G-module if and only if there exists a finitely generate R-submodule K P LpM q such that M " TGpλ, Kq.
Proof. If R˚GM is finitely generated as a left R˚G-module then choose a finite set of generators x1, . . . , xn, so that, M " R˚Gx1`¨¨¨`R˚Gxn " TGpλ, x1R`¨¨¨`xnRq. On the other hand, if M " TGpλ, Kq with K finitely generated, than any finite set of generators of K generates M as R˚Gmodule.
Consider now an R˚G-submodule R˚GN ď M . Any R-submodule of N is also an R-submodule of M , this allows us to identify LpN q with a sub-monoid of LpM q. One has the following left actions given by the R˚G-module structures The following easy lemma will be useful in many situations. Lemma 4.9. In the above notation, let F ď G be a subset, then there is a short exact sequence of left R-modules 0 Ñ TF pλ, Kq X N Ñ TF pλ, Kq Ñ TF pλ, pK`N q{N q Ñ 0 . Proof. The non-trivial maps in (4.5) are induced by the embedding N Ñ M and by the projection M Ñ M {N . One can verify that the resulting sequence is exact noticing that TF pλ, pK`N q{N q " pTF pλ, Kq`N q{N , in fact, g 1 pk1`N q`¨¨¨`g n pkn`N q " pg 1 k1`¨¨¨`g n knq`N for all k P N`, k1, . . . , kn P K and g1, . . . , gn P F .
Let now L : R-Mod Ñ Rě0 Y t8u be a length function compatible with R˚G. We let FinLpM q " tRK ď M : LpKq ă 8u Ď LpM q .
By the additivity of L, FinLpM q is a sub-monoid of LpM q. Furthermore, by the compatibility of L with R˚G, LpλgpKqq " LpKq, thus λgpKq P FinLpM q. So, there is an induced action λ : G Ñ AutpFinLpM qq , λpgq " λg : FinLpM q Ñ FinLpM q .
The length function L is said to be discrete provided the set of finite values of L is order-isomorphic to N. In case L is discrete, the monoid FinLpM q carries redundant information for our needs, for this reason it is usually useful to reduce to a smaller monoid consisting of the finitely generated modules in FinLpM q. The following lemma, which is an immediate consequence of the upper continuity and of the discreteness of L, allows for such reduction.
Lemma 4.10. In the above notation, suppose that L is discrete and that K P FinLpM q. Then, there exists a finitely generated K 1 P FinLpM q such that LpK 1 q " LpKq.
We need to introduce a last tool on the monoid of submodules LpM q, that is, a closure operator. Indeed, we consider the torsion class KerpLq of all left R-modules K such that LpKq " 0. The torsion functor relative to this class was denoted in [32] by zL : R-Mod Ñ KerpLq, where, given K P R-Mod, zLpKq " tx P K | LpRxq " 0u ; zLpKq is called the L-singular submodule of K. If zLpKq " K (or, equivalently, LpKq " 0) we say that M is L-singular. There is a standard technique to associate a closure operator to any given torsion class (see [34]). In particular, given K P LpM q, we let π : M Ñ M {K be the natural projection and we define With this terminology we can reformulate Lemma 4.10 as follows: Corollary 4.11. In the above notation, suppose that L is discrete and that K P FinLpM q. Then, K has an L-essential finitely generated submodule.
We collect in the following lemma some useful properties of L-purifications, which follow by the fact that p´qL˚is a closure operator associated to a hereditary torsion theory (see for example [34]). (1) ppK1qL˚qL˚" pK1qL˚;
The algebraic L-entropy
All along this section, we fix a ring R, an infinite finitely generated amenable group G and a crossed product R˚G. Furthermore, we fix a length function L : R-Mod Ñ Rě0 Y t8u with the following two properties: (1) it is discrete, that is, the image of L is a subset of Rě0 Y t8u order-isomorphic to N Y t8u; (2) it is compatible with R˚G (see Definition 4.6).
We remark that, if R is a (skew) field, then L " dimR satisfies the above conditions for any choice of the functions defining R˚G; more generally, this happens for all the functions ℓα described in Subsection 3.3. On the other hand, if R˚G " RrGs, then condition (2) is trivially satisfied by any length function L.
Definition and existence
In this subsection we define the algebraic entropy as an invariant for left R˚G-modules. This invariant is modeled to encode the dynamics of the L-finite R-submodules of an R˚G-module M , under the action λ : G Ñ AutpFinLpM qq. It turns out that this notion is not well-behaved on all the R˚G-modules but just on a subclass of R˚G-Mod of all the R˚G-modules M for which FinLpM q is big enough. More precisely, Notice that lFinpLq is closed under taking direct limits, quotients and submodules but not in general under taking extensions (see [32]).
Let M P lFinLpR˚Gq and, as usual, consider the action of G on the L-finite submodules λ : G Ñ AutpFinLpM qq .
Given K P FinLpM q and F P FpGq, one can use the additivity of L to show that TF pλ, Kq P FinLpM q (see (4.4)). In particular, for any L-finite submodule K of M we have a function fK : FpGq Ñ Rě0 fK pF q " LpTF pλ, Kqq .
We now verify that the above function satisfies the hypotheses of the Ornstein-Weiss Lemma.
Lemma 5.2. In the above notation, fK is monotone, sub-additive and G-equivariant.
Proof. Let F, F 1 P FpGq. The proof follows by the properties of L, in fact, TF pλ, Kq`T F 1 pλ, Kq " T F YF 1 pλ, Kq and so fK pF q`fK pF 1 q ě fK pF Y F 1 q, proving sub-additivity. Furthermore, if F Ď F 1 then TF pλ, Kq ď T F 1 pλ, Kq and so fK pF q ď fK pF 1 q, proving that fN is monotone. Finally, since L is compatible with R˚G, for any given g P G we have LpTgF pλ, Kqq " LpλgpTF pλ, Kqqq " LpTF pλ, Kqq, proving that fK is G-equivariant.
Thus, by the Ornstein-Weiss Lemma, the limit in the following definition exists and it does not depend on the choice of the Følner sequence. The (algebraic) L-entropy of the R˚G-module R˚GM is entLpR˚GM q " suptentLpλ, Kq : K P FinLpM qu.
Remark 5.4. We defined the algebraic L-entropy for left R˚G-modules in case G is finitely generated. Anyway, the exact same procedure allows one to define this invariant when G is just countable (but not necessarily finitely generated). Furthermore, standard variations of the above arguments using Følner nets allow one to define a similar invariant in case G is not countable.
Basic properties
In this subsection we study the basic properties of the algebraic L-entropy. For simplicity we fix all along this subsection a locally L-finite left R˚G-module R˚GM . We also denote by λ : G Ñ AutpFinLpM qq the action of G induced by the R˚G-module structure on M .
Example 5.5. If LpRM q ă 8, then entLpR˚GM q " 0. In fact, if RK ď M is any L-finite R-submodule of M , then by definition entLpλ, Kq ď limnÑ8 LpM q{|Fn| ď limnÑ8 LpM q{n " 0 (for the second inequality use the fact that, as G is infinite we can take a Følner sequence such that Fn ň Fn`1 for all n P N, thus |Fn| ě n).
The following result allows us to redefine the algebraic entropy in terms of finitely generated submodules.
Proof. By definition of L-essential submodule we have that K{H P KerpLq. Furthermore, for all g P G, λgK{λgH -F σpgq pK{Hq so, as by hypothesis L is compatible with F σpgq , also λgK{λgH P KerpLq. In particular, λgH is L-essential in λgK for all g P G. By Lemma 4.12(4) and the additivity of L, LpTF n pλ, Kqq " LpTF n pλ, Hqq , for all n P N, where tFnu nPN is a Følner sequence. Therefore, entLpλ, Kq " entLpλ, Hq. We can now verify part (2). Indeed, the "ď" inequality comes directly from the definition of entropy. On the other hand, by Lemma 4.10 any L-finite submodule K of M has an L-essential finitely generated submodule H and by part (1) entLpλ, Kq " entLpλ, Hq, which easily yields the claim.
The definition of entropy in terms of finitely generated submodules given in Proposition 5.6(2) allows us to prove many important properties. In the following lemma we show that the entropy is monotone under taking submodules and quotients. Lemma 5.7. Let N ď M be an R˚G-submodule. Then (1) entLpR˚GM q ě entLpR˚GN q; (2) entLpR˚GM q ě entLpR˚GpM {N qq.
Proof. We denote by λ 1 : G Ñ AutpFinLpN qq andλ : G Ñ AutpFinLpM {N qq the actions induced by λ. To prove (1) it is enough to notice that, whenever K ď N is an L-finite submodule of N , then it is also an L-finite submodule of M and entLpλ 1 , Kq " entLpλ, Kq ď entLpR˚GM q. Let us verify part (2), using the definition of entropy in terms of finitely generated submodules. Indeed, given a finitely generated submoduleK ď M {N , there exists a finitely generated (thus L-finite) submodule K ď M such that pK`N q{N -K. Given a Følner sequence tFnu nPN , by Lemma 4.9, TF n pλ,Kq " pTF n pλ, Kq`N q{N and so LpTF n pλ,Kqq ď LpTF n pλ, Kqq for all n P N .
The above lemma has a converse in some particular situation: (2) entLpR˚GM q " entLpR˚GpM {N qq, provided LpN q " 0.
Proof. One inequality of both statements follows by Lemma 5.7, thus we have just to verify the other one. Part (1) follows by Proposition 5.6, noticing that, whenever K ď M is an L-finite submodule, then K X N is L-essential in K, thus, entLpλ, Kq " entLpλ, K X N q " entLpλ 1 , K X N q ď entLpR˚GN q (where λ 1 : G Ñ AutpFinLpN qq is the obvious action). Denote byλ : G Ñ AutpFinLpM {N qq the obvious action. Let K P FinLpM q and let tFnu nPN be a Følner sequence. The short exact sequence in Lemma 4.9 shows that LpTF n pλ, Kqq " LpTF n pλ,Kqqp TF n pλ, Kq X N q " LpTF n pλ,Kqq. The conclusion follows.
Historical notes on algebraic entropy
In this subsection we present some of the crucial steps of the development of the notion of algebraic entropy to make it clear where some of the ideas of the present paper come from.
In 1965 Adler, Konheim and McAndrew [1] introduced the topological entropy of a continuous selfmap of a compact space. In the final part of their paper, Adler et al. suggested a notion of entropy for an endomorphism of a discrete torsion Abelian group, which is usually denoted by ent. In 1974, Weiss [38] studied some of the basic properties of ent, while in 2009 Dikranjan, Goldsmith, Salce and Zanardo [9] deeply studied this invariant proving also an Addition Theorem and an axiomatic characterization for ent.
The main disadvantage of ent is that of being trivial on endomorphisms of torsion-free discrete Abelian groups. This fact comes directly from the definition, which is based on the cardinality of finite subgroups, that are all trivial in a torsion-free group. In the literature there are two main strategies to remedy this fact and find a suitable additive invariant for the endomorphisms of torsion-free Abelian groups. The first one, proposed by Peters [31] in 1979, consists in substituting the finite subgroups in the definition of ent by finite subsets. The invariant arising this way is very useful in many situations but it seems difficult to generalize this kind of approach to modules over general rings (we refer to [8] and [15] for direct proofs of all the main properties of such invariant). The second strategy was proposed in 2009 by Salce and Zanardo [33]; they suggested to substitute the invariant log |´| : Z-Mod Ñ Rě0 Y t8u by the torsion-free rank of Abelian groups, obtaining the additive invariant ent rk of endomorphisms of Abelian groups. Furthermore, Salce and Zanardo defined a notion of entropy based on some invariants of categories of modules with hypotheses weaker than that of being a length function. In 2013, Salce, Vámos and the author [32], had the intuition that the notion of length function was the missing piece of the puzzle. Using the notion of length function we were able to prove the first general Addition Theorem for the L-entropy of endomorphisms of left R-modules, with R any ring and L a discrete length function of R-Mod.
All the notions of entropy described above are invariants of endomorphisms of modules over some ring R, that is, invariants of RrXs-Mod. Entropies for multiple morphisms acting on a module also have a long history. On the topological side, the foundations of this theory can be found in the seminal paper [27] of Ornstein and Weiss. On the other hand, to the best of the author's knowledge, the present paper is the first attempt to develop a theory of algebraic entropy for group actions on modules.
The algebraic entropy is a length function
The aim of this section is to prove the following Theorem 6.1. Let R be a ring, let G be a finitely generated infinite amenable group, fix a crossed product R˚G and let L : R-Mod Ñ Rě0 Y t8u be a discrete length function compatible with R˚G. The invariant entL : lFinLpR˚Gq Ñ Rě0 Y t8u satisfies the following properties: (1) entL is upper continuous; (2) entLpR˚G bR Kq " LpKq for any L-finite left R-module K; (3) entLpN q " 0 if and only if LpN q " 0, for any R˚G-submodule N ď R˚G bR K; (4) entL is additive.
In particular, entL is a length function on lFinLpR˚Gq.
Part (1) will be verified in Subsection 6.1, parts (2) and (3) will be proved in Subsection 6.2 and part (4) will be the main result of Subsection 6.3.
We fix all along this section a ring R, an infinite finitely generated amenable group G, a crossed product R˚G and a discrete length function L : R-Mod Ñ Rě0 Y t8u compatible with R˚G.
The algebraic entropy is upper continuous
In this subsection we show that the algebraic L-entropy is an upper continuous invariant. We start with the following lemma that deals with the case when M is generated (as R˚G-module) by an L-finite R-submodule K, that is, M " TGpλ, Kq. In such situation one does not need to take a supremum to compute entropy. Proof. Given a finitely generated R-submodule H of M , we can find a finite subset e P F Ď G such that H Ď TF pλ, Kq. This shows that entLpλ, Hq ď entLpλ, TF pλ, Kqq. Now notice that, using the Følner condition, we obtain On the other hand, |FnF |{|Fn| ě 1 so limnÑ8 |FnF |{|Fn| " 1. We obtain that entLpλ, TF pλ, Kqq " lim nÑ8 TF n pλ, TF pλ, Kqq |Fn| " lim nÑ8 TF nF pλ, Kq |Fn|¨| where the last equality comes from the fact that tFnF u nPN is a Følner sequence (use the fact that, for any C Ď FpGq and n P N, one has the inclusion BCpFnF q Ď B CF´1 pFnq and apply the Følner condition for tFnu nPN ) and the definition of entL does not depend on the choice of a particular Følner sequence. Thus, entLpλ, Hq ď entLpλ, Kq for any finitely generated H P LpM q; one concludes using Proposition 5.6.
The upper continuity of entL can now be verified easily using the above lemma and Proposition 5.6:
Values on (sub)shifts
This subsection is devoted to compute the values of the algebraic entropy on the R˚G-modules of the form M " R˚G bR K, for some left R-module K, and their R˚G-submodules. Indeed, fix a left Rmodule K and let M " R˚GbR K. Notice that, there is a direct sum decomposition, as a left R-module, RM -À gPG gK. We denote the action of G on FinLpM q by The choice of the greek letter β to represent this action comes from the Bernulli shifts which are defined in ergodic theory and can be viewed as dual to the actions described here. Notice that the action β has the following properties, where F is a subset of G: TF pβ, gKq " à hPF hgK and TGpβ, gKq " M . In the following example we compute the algebraic entropy of Bernoulli shifts.
The computation in the above example shows that the entropy of M " R˚G bR K is 0 if and only if LpKq " 0, if and only if LpM q " 0. Our next goal is to show that, if R˚GN is a subshift of M , then entLpR˚GN q " 0 if and only if LpRN q " 0. This will be proved in Proposition 6.8 but first we need to recall some useful terminology and results from [3]. Definition 6.6. Let E and F be subsets of G. A subset N Ď G is an pE, F q-net if it satisfies the following conditions: (1) the subsets pgEqgPN are pairwise disjoint, that is, gE X g 1 E " H for all g ‰ g 1 P N ; (2) there exist α P p0, 1s and n0 P N such that |Fn X N | ě α¨|Fn|, for all n ą n0.
Proof. Part (1) is proved in [3,Lemma 2.2], so let us concentrate on part (2). For each n P N, let F`F n " OutF pFnq X N and notice that F`F n zpFn X N q Ď BF pFnq. Furthermore, since Fn is covered by the sets gF , g P F`F n , we have |Fn| ď |F ||F`F n |. Let now α1 " 1{|F |, thus α1|Fn|´|Fn X N | ď |F`F n |´|Fn X N | ď |F`F n zpFn X N q| ď BF pFnq .
The Addition Theorem
In this subsection we complete the proof of Theorem 6.1 verifying a very strong property of the algebraic entropy, that is, its additivity. In particular, we have to verify that, given a locally L-finite left R˚Gmodule M , and an R˚G-submodule N ď M , We fix all along this subsection the following notations for the actions induced by the R˚G-module structures: We start proving the inequality "ě" of (6.2).
The result follows by the arbitrariness of the choice of K1 and K2.
The first step in proving the converse inequality is to show that we can reduce the problem to the case when both M and N are finitely generated R˚G-modules. This goal is obtained in the following corollary (which is just a reformulation of Corollary 4.5), and the subsequent two lemmas. Proof. This is a straightforward application of Corollary 4.5. In fact, the function fK : FpGq Ñ Rě0 such that fK pF q " LpTF pλ, Kqq satisfies the hypotheses of such corollary for any L-finite R-submodule K of M , by Lemma 5.2. Furthermore, limnÑ8 fK pFnq{|Fn| " entLpλ, Kq by the definition of entropy.
Lemma 6.11. Consider a sequence N0 ď N1 﨨¨ď Nt 﨨¨ď M of R˚G-submodules of M such that N " Ť tPN Nt and letλt : G Ñ AutpFinLpM {Ntqq be the actions induced on the quotients. Then, given an L-finite submodule K ď M and lettingK " pK`N q{N andKt " pK`Ntq{Nt for all t P N, Proof. The inequality "ď" follows by Lemma 5.7. On the other hand, for all ε ą 0 there exists nε P N such that LpTF n pλ,Kqq |Fn| ď entLpλ,Kq`ε , for all n ě nε . (6.3) By Corollary 6.10, for any given ε 1 P p0, 1{4q, there exist n1, . . . , n k P N such that nε ď n1 﨨¨ď n k and entLpλt,Ktq ď ε 1¨L pKtq`1 1´ε 1¨m ax 1ďiďk LpTF n i pλt,Ktqq |Fn i | (6.4) holds for all t P N. Now, for all i P t1, . . . , ku, the set tLpTF n i pλt,Ktqq : t P Nu is a set of values of L, all smaller than or equal to the finite value LpTF i pλ, Kqq. Since we supposed L to be discrete, this set has a minimum, say LpTF n i pλt i ,Kt i qq. Let s " max 1ďiďk ti and notice that LpTF n i pλs,Ksqq " LpTF n i pλ,Kqq , for all i " 1, . . . , k.
(This follows by the additivity of L and the fact that LppN X pTF n i pλ, Kq`Nsqq{Nsq " 0, as this module is the union of modules of the form pNt X pTF n i pλ, Kq`Nsqq{Ns with t ě s, which are L-singular by the choice of s).
As we announced, we can now prove the following reduction to the case when R˚GN and R˚GM are finitely generated. Lemma 6.12. If (6.2) holds whenever R˚GN and R˚GM are finitely generated, then it holds in general.
Proof. We already proved that the inequality "ě" in (6.2) always holds, thus we concentrate just on the converse inequality. Indeed, given a finitely generated submodule R˚GK of R˚GM we claim that entLpR˚GKq " entLpR˚GpK X N qq`entLpR˚GppK`N q{N qq . (6.5) Notice that, if we prove the above claim, we can easily conclude using upper continuity as follows: It remains to verify claim (6.5). Indeed, choose a finitely generated R-submodule H of K, such that K " TGpλ, Hq. Notice also that K X N " Ť nPN TGpλ, T NnpBq pλ, Hq X N q. For all n P N, we let Hn be an L-essential, finitely generated R-submodule of T NnpBq pλ, Hq X N . By Proposition 5.6, we obtain that entLpR˚GpN X Kqq " sup nPN entLpλ, T NnpBq pλ, Hq X N q " sup nPN entLpλ, Hnq .
We let K 1 " Ť nPN TGpλ, Hnq. Notice that K 1 is L-essential in K X N (in fact, pK X N q{K 1 is the union of modules of the form ppT NnpBq pλ, Hq X N q`K 1 q{K 1 and each of these modules is a quotient of an L-singular module of the form pT NnpBq pλ, Hq X N q{TGpλ, Hnq). By Lemma 5.8 and Proposition 5.6 we obtain that entLpR˚GpK X N qq " entLpR˚GK 1 q " lim " entLpR˚GpK X N qq`entLpR˚GpK{pK X N qq , which verifies (6.5), concluding the proof.
Finally, we have all the instruments to conclude the proof of the additivity of entL. The computations in the proof of the following lemma are freely inspired to the proof of the Abramov-Rokhlin Formula given in [37]. The context (and even the statements) in that paper is quite different but the ideas contained there can be perfectly adapted to our needs. Lemma 6.13. entL pR˚GM q ď entL pR˚GN q`entL pR˚GpM {N qq .
Proof. By Lemma 6.12, we can suppose that both M and N are finitely generated R˚G-modules. In particular, there exists a finitely generated R-submodule K 1 ď N and a finitely generated R-submodulē K2 ď M {N such that N " TGpλ 1 , K 1 q and M {N " TGpλ,K2q. SinceK2 is finitely generated, there exists a finitely generated R-submodule K2 of M , such that pK2`N q{N "K2. We let K " K 1`K 2 and we notice that M " TGpλ, Kq. Finally, we let K1 be an L-essential finitely generated submodule of K X N containing K 1 . Notice that, by Lemma 6.2 we obtain that entLpR˚GM q " entLpλ, Kq, entLpR˚GN q " entLpλ 1 , K X N q " entLpλ 1 , K1q and entLpR˚GpM {N qq " entLpλ,K2q.
In the remaining part of the proof we are going to show that there exists a positive integer k such that for all big enough m P N. Applying (6.8) to (6.7), one gets entLpM q ă entLpN q`1 1´ε entLpM {N q`εˆLpKqpk`1q`1 1´ε˙`2 ε which, as it holds for all ε P p0, 1{4q, gives the desired inequality. Thus, to conclude we have to verify (6.8).
Since tFnu is a Følner exhaustion, we have that N " Ť nPN TF n pλ 1 , K1q and so, for any L-finite submodule H ď M , we can use the upper continuity of L to obtain that LpH X N q " limnÑ8 LpH X TF n pλ 1 , K1qq. Now, by additivity, this implies that LppH`N q{N q " limnÑ8 LppH`TF n pλ 1 , K1qq{TF n pλ 1 , K1qq and, by the discreteness of L, this limit is the minimum of the values. By Theorem 4.3, there exist n ă n1 㨨¨ă n k P N such that tFn 1 , . . . , Fn k u ε-quasi-tiles Fm for all m ěn. Applying the above argument with H " TF n i pλ, Kq (for all i " 1, . . . , k), we can find n P N such that
L˜T
Fn i pλ, Kq`TF n pλ 1 , K1q TF n pλ 1 , K1q¸" LˆT Fn i pλ, Kq`N N˙( 6.9) for all n ě n and all i " 1, . . . , k. From now on we suppose that m is a positive integer such that m ě maxtn, nu and |BF n pFmq|{|Fm| ď ε . (6.10) (Where the second condition can be assumed since tFnu nPN is Følner.) As tFn 1 , . . . , Fn k u ε-quasi-tiles Fm, we can choose tiling centers C1, . . . , C k . We obtain the following inequalities which imply that Applying this computation and using again (6.11), one gets: Now, let ti " p|Ci||Fn i |q{ ř k j"1 |Cj ||Fn j | and notice that ti P p0, 1q and ř k i"1 ti " 1. Then TF m pλ 1 , K1q¸. (6.13) Notice that, since tFnu nPN is a Følner exhaustion, then 1 P Fn i and so Ci Ď Fm Ď InF n pFmq Y BF n pFmq for all i " 1, . . . , k. Thus, where the first inequality in the last line comes from the fact that (by definition of InF n pFmq), F n Ď c´1Fm for all c P Ci X InF n pFmq; the last equality is an application of (6.9)). Let us assemble together the above computations: LpTF m pλ, Kq{TF m pλ 1 , K1qq |Fm|
Stable finiteness
Recall that a concrete category is a pair pC, F : C Ñ Setq, where C is a category and F is a fatihful functor to the category of sets. In a concrete category one says that a morphism φ is injective (resp., surjective, bijective) if so is the map F pφq. An object X of C is said to be hopfian if any surjective endomorphism of X is bijective. Such notion is usually introduced in categories of (Abelian) groups, rings, modules, or topological spaces.
Our interest in hopficity, comes from the following observation. Recall that a ring R is directly finite if xy " 1 implies yx " 1 for all x, y P R. Furthermore, R is stably finite if the ring Mat k pRq of kˆk square matrices with coefficients in R, is directly finite for all k P N`. Lemma 7.1. Given a ring R and a positive integer k, the following are equivalent: (1) R k R is hopfian (as a right R-module); (2) RR k is hopfian (as a left R-module); (3) Mat k pRq is directly finite.
A long-standing open question about directly finite rings is the following conjecture due to Kaplansky Conjecture 7.2 (Kaplansky). For any field K and any group G, the group ring KrGs is stably finite.
The above conjecture was verified by Kaplansky [20] in case K " C. Furthermore, it was proved by Ara, O'Meara and Perera [2] that KrGs is stably finite whenever G is residually amenable and K is an arbitrary division ring. This last result was generalized by Elek and Szabó [11], who proved the following Theorem 7.3 (Elek and Szabó). For any division ring K and any sofic group G, the group ring KrGs is stably finite.
Notice also that Ara et al. proved that any crossed product K˚G is stably finite, provided K is a division ring and G is an amenable group.
A straightforward consequence of the above theorem is that MatnpKrGsq is stably finite for any division ring K and any sofic group G. Now, by the Artin-Wedderburn Theorem, given a semisimple Artinian ring R, there exist positive integers k, n1, . . . , n k P N`and division rings K1, . . . , K k such that R -Matn 1 pK1qˆ¨¨¨ˆMatn k pK k q. This implies that, RrGs -Matn 1 pK1rGsqˆ¨¨¨ˆMatn k pK k rGsq, thus a consequence of the above theorem is that RrGs is stably finite whenever R is semisimple Artinian and G sofic. This result can be further generalized as follows: Both the proof of the residually amenable case due to Ara, O'Meara and Perera, and the proof of the sofic case due to Elek and Szabó, consist in finding a suitable embedding of KrGs in a ring which is known to be stably finite. Such methods are really effective but, as far as we know, cannot be used to obtain information on the modules over KrGs. It seems natural to ask the following Question 7.5. Let G be a group, R a ring and let R˚G be a fixed crossed product. Given a finitely generated left R˚G-module R˚GM , when is R˚GM an hopfian module? By Lemma 7.1, we know that Theorem 7.3 means exactly that, when R is a division ring and G is sofic, then any free left RrGs-module of finite rank is hopfian. This is true more generally for finitely generated projective modules: Lemma 7.6. Let R be a ring, let RM and R N be left R-modules; suppose that RM is hopfian and R N is a direct summand of RM . Then RN is hopfian as well. In particular, R is stably finite (if and) only if any finitely generated projective left R-module is hopfian.
Proof. Let RN 1 ď M be a complement for N , that is M -N ' N 1 and let φ : N Ñ N be a surjective endomorphism. Let Φ : M Ñ M be such that Φpn, n 1 q " pφpnq, n 1 q, for all n P N and n 1 P N 1 . Clearly Φ is surjective and KerpΦq " Kerpφq ' t0u. Now, KerpΦq is trivial by the hopficity of M and so also Kerpφq " 0 concluding the proof.
Notice that in general the class of hopfian modules is not closed under taking finite direct sums, for a classical (counter)example see [7,Example 3].
In [16], Anna Giordano Bruno and the author studied the concept of hereditarily hopfian Abelian group. Generalizing that situation, we say that a left R-module M over a ring R is hereditarily hopfian if and only if all its submodules are hopfian.
Example 7.7. The Abelian group of p-adic integers Jp is hopfian but not hereditarily hopfian. In fact, any Z-linear endomorphism of Jp is also Jp-linear, so End Z pJpq is canonically isomorphic, as a ring, to the commutative ring End Jp pJpq -Jp. Now, if φ : Jp Ñ Jp is a surjective endomorphism, as Jp is projective as Jp-module, there is ψ : Jp Ñ Jp such that φψ " 1 and so 1 " φψ " ψφ, proving that φ is an isomorphism and so Jp is hopfian. On the other hand, Z Jp has infinite torsion-free rank, that is, there is a subgroup G ď Jp of the form G -Z pNq which is clearly not hopfian. Hence, Z Jp is not hereditarily hopfian.
Using the theory of algebraic entropy we can now prove that a large class of left R˚G-modules is hereditarily hopfian, in case R is left Noetherian and G amenable. We remark that this is a very strong version of Kaplasky's Stable Finiteness Conjecture in the amenable case which can be re-obtained as a corollary.
Theorem 7.8. Let R be a left Noetherian ring, G a finitely generated amenable group and let R˚G be a fixed crossed product. Then, for any finitely generated left R-module RK P R-Mod, the left R˚G-module R˚G bR K is hereditarily hopfian. In particular, EndR˚GpM q is stably finite for any submodule R˚GM ď R˚G bR K.
The proof of the above theorem makes use of the full force of the localization techniques introduced in Section 2. Such heavy machinery hides in some sense the idea behind the proof; this is the reason for which we prefer to give first the proof of the following more elementary statement, whose proof is far more transparent. Lemma 7.9. Let K be a division ring, let G be a finitely generated amenable group and fix a crossed product K˚G. For all n P N`, pK˚Gq n " K˚G b K n is a hereditarily hopfian left K˚G-module.
Proof. Let n P N`and choose K˚G-submodules N ď M ď pK˚Gq n such that there exists a short exact sequence 0 Ñ N Ñ M Ñ M Ñ 0 , we have to show that N " 0. The length function dim : K-Mod Ñ Rě0 Y t8u is compatible with any crossed product, so we can consider the dim-entropy of left K˚G-modules. In particular, we have that ent dim pM q " ent dim pM q`ent dim pN q and 0 ď ent dim pN q ď ent dim pM q ď ent dim pK˚G b K n q " n .
The same argument of the above proof can be used to prove Theorem 7.8, modulo the fundamental tool of Gabriel dimension: In order to go further with the proof we need to show that, as a left R-module, the Gabriel dimension of Kerpφq is a successor ordinal whenever it is not´1 (i.e., whenever Kerpφq ‰ 0). This follows by the following Lemma 7.10. In the hypotheses of Theorem 7.8, G.dimpRN q is a successor ordinal for any nontrivial R-submodule N ď R˚G bR K.
Consider the length function ℓα : R-Mod Ñ Rě0 Y t8u described in Subsection 3.3 and recall that Kerpℓαq is exactly the class of all left R-modules with Gabriel dimension ď α. Furthermore, Tα`1pKq is a Noetherian module, thus, QαpTα`1pKqq is a Noetherian object in a semi-Artinian category, that is, an object with finite composition length, for this reason ℓαpTα`1pKqq " ℓpQαpTα`1pKqqq ă 8. Using the computations of Example 6.5 and the Addition Theorem, we get ent ℓα pTα`1pR˚GbKqqq " ℓαpTα`1pKqq ă 8 and ent ℓα pTα`1pM qq " ent ℓα pTα`1pM qq`ent ℓα pKerpφqq .
In the above proof we made use of the Addition Theorem for the algebraic entropy, which is quite a deep result. We want to underline that if one is only interested in the second part of the statement, that is, stable finiteness of endomorphism rings, then it is sufficient to use the weaker additivity of the algebraic entropy on direct sums, which can be verified as an easy exercise independently from the Addition Theorem.
Example 7.11. Let G be a free group of rank ě 2 and let K be a field. It is well-known that KrGs is not left (nor right) Noetherian so we can find a left ideal KrGs I ď KrGs which is not finitely generated. Furthermore, by [6,Corollary 7.11.8], KrGs is a free ideal ring, so I is free. This means that I is isomorphic to a coproduct of the form KrGs pNq which is obviously not hopfian.
Let us conclude this subsection with the following problem: Problem 7.12. Study the class of finitely generated groups G such that the group algebra KrGs is hereditarily Hopfian for any skew field K. Is it true that this property characterizes the class of finitely generated amenable groups?
One could also state an analogous problem including all the possible crossed products K˚G, instead of just the group algebras KrGs.
Zero-Divisors
In this last section of the paper we discuss another classical conjecture due to Kaplansky about group rings connecting it to the theory of algebraic entropy: Conjecture 7.13 (Kaplansky). Let K be a field and G be a torsion-free group. Then KrGs is a domain.
Some cases of the above conjecture are known to be true but the conjecture is fairly open in general (for a classical reference on this conjecture see for example [28]). In most of the known cases, the strategy for the proof is to find an immersion of KrGs in some division ring. This is clearly sufficient but, in principle, it is a stronger property. To the best of the author's knowledge, the following question remains open: Is it true that KrGs is a domain if and only if KrGs is a subring of a division ring? The above question is known to have positive answer if G is amenable (see [23,Example 8.16]). In the present section we provide an alternative argument to answer the above question for amenable groups (in the more general setting of crossed group rings) and we translate the amenable case of Conjecture 7.13 into an equivalent statement about algebraic entropy. This approach is inspired to the work of Nhan-Phu Chung and Andreas Thom [5]. Indeed, we can prove the following Theorem 7.14. Let K be a division ring and let G be a finitely generated amenable group. For any fixed crossed product K˚G, the following are equivalent: (1) K˚G is a left (and right) Ore domain; (2) K˚G is a domain; (3) ent dim p K˚G M q " 0, for every proper quotient M of K˚G; (4) Impent dim q " N Y t8u.
Before proving the above theorem we recall some useful properties about Ore domains. We start recalling that a domain D is left Ore if Dx X Dy ‰ t0u for all x, y P Dzt0u. It can be shown that this is equivalent to say that D is a left flat subring of a division ring. Proof. If D is left Ore, then D is a flat subring of a division ring K. Then there is an exact functor K bD´: D-Mod Ñ K-Mod which commutes with direct limits. Thus, we can define the desired length function L simply letting LpDM q " dim K pK bD M q. On the other hand, suppose that there is a length function L : D-Mod Ñ Rě0 Y t8u such that LpDq " 1 and choose x, y P Dzt0u. Since D is a domain, both Dx and Dy contain (and are contained in) a copy of D, thus LpDxq " LpDyq " 1. If, looking for a contradiction Dx X Dy " t0u, then 1 " LpDq ě LpDx`Dyq " LpDx ' Dyq " LpDxq`LpDyq " 2, which is a contradiction.
It is a classical result that any left Noetherian domain is left Ore (see for example [24,Theorem 1.15 in Chapter 2.1]). By the above proposition we can generalize this result as follows: Proof. Let D be a domain with left Gabriel dimension. First of all we verify that G.dimpDDq is not a limit ordinal. Indeed, if G.dimpDDq " λ is a limit ordinal, then D " Ť αăλ TαpDq. This means that, for any non-zero x P D, there exists α ă λ such that Dx P TαpDq. Choose a non-zero x P D, as D is a domain, there is a copy of D inside Dx. Thus, G.dimpDq ď G.dimpDxq ď α for some α ă λ, a contradiction. If G.dimpDDq " α`1 for some ordinal α, then we can consider the length function ℓα : D-Mod Ñ Rě0 Y t8u , ℓαpM q " ℓpQαpM qq .
To conclude one has to show that ℓαpDq " 1, that is, QαpDq is a simple object. Since Cα`1{Cα is semi-Artinian, there is a simple subobject S of QαpDq. Then SαpSq is a sub-module of SαQαpDq. Identify SαpSq, SαQαpDq and D with submodules of EpDq, since D is essential in EpDq, there is 0 ‰ x such that x P SαpSq X D, but then SαpSq contains an isomorphic copy of D. Thus QαSαpSq " S contains an isomorphic copy of QαpDq, which is therefore simple.
We can finally prove our result: Proof of Theorem 7.14. (1)ñ(2) is trivial while (2)ñ(1) follows by Proposition 7.15 and the fact that the algebraic dim-entropy is a length function on K˚G-Mod such that ent dim p K˚G K˚Gq " 1.
(2)ñ(3). Consider a short exact sequence 0 Ñ K˚G I Ñ K˚G K˚G Ñ K˚G M Ñ 0, with I ‰ 0. Choose 0 ‰ x P I, then K˚Gx -K˚G, and so ent dim p K˚G M q " ent dim p K˚G K˚Gq´ent dim p K˚G Iq ď 1´1 " 0.
(3)ñ(4). Let us show first that for any finitely generate left K˚G-module K˚G F , ent dim p K˚G F q P N. In fact, choose a finite set of generators x1, . . . , xn for F and, letting F0 " 0 and Fi " K˚Gx1`¨¨¨`K˚Gxi for all i " 1, . . . , n, consider the filtration 0 Ď F1 Ď F2 Ϩ¨¨Ď Fn " F . By additivity, ent dim pF q " n ÿ i"1 ent dim pFi{Fi´1q .
All the modules F1{F0, . . . , Fn{Fn´1 are cyclics (i.e. quotients of K˚G), thus ent dim pFi{Fi´1q P t0, 1u by hypothesis. Hence, ent dim pF q P N. To conclude one argues by upper continuity that the algebraic dim-entropy of an arbitrary left K˚G-module is the supremum of a subset of N, thus it belongs to N Y t8u. (4)ñ (2). Let x P K˚G and consider the short exact sequence 0 Ñ I Ñ K˚G Ñ K˚Gx Ñ 0 where I " ty P K˚G : yx " 0u. Suppose that x is a zero-divisor, that is, I ‰ 0 or, equivalently, dimp K Iq ‰ 0. By Proposition 6.8, ent dim pIq ą 0 and, by our assumption, entLpIq ě 1. Hence, using additivity, ent dim pK˚Gxq " 0. Again by Proposition 6.8, this implies dimpK˚Gxq " 0 and consequently K˚Gx " 0, that is, x " 0. Thus, the unique zero-divisor in K˚G is 0. | 2017-10-23T16:08:45.000Z | 2014-10-30T00:00:00.000 | {
"year": 2019,
"sha1": "fe0d8bd2f135b103ad3304b9253bbe6e12eb0b21",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1410.8306",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "fe0d8bd2f135b103ad3304b9253bbe6e12eb0b21",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
53580013 | pes2o/s2orc | v3-fos-license | Heterotic Model Building: 16 Special Manifolds
We study heterotic model building on 16 specific Calabi-Yau manifolds constructed as hypersurfaces in toric four-folds. These 16 manifolds are the only ones among the more than half a billion manifolds in the Kreuzer-Skarke list with a non-trivial first fundamental group. We classify the line bundle models on these manifolds, both for SU(5) and SO(10) GUTs, which lead to consistent supersymmetric string vacua and have three chiral families. A total of about 29000 models is found, most of them corresponding to SO(10) GUTs. These models constitute a starting point for detailed heterotic model building on Calabi-Yau manifolds in the Kreuzer-Skarke list. The data for these models can be downloaded here: http://www-thphys.physics.ox.ac.uk/projects/CalabiYau/toricdata/index.html
Introduction
Over the past few years, a programme of algorithmic string compactification has been established where a combination of the latest developments in computer algebra and algebraic geometry have been utilized to study the compactification of the heterotic string on smooth Calabi-Yau threefolds with holomorphic vector bundles satsifying the Hermitian Yang-Mills equations [1,2]. This is very much in the spirit of the recent advances in applications of algorithmic geometry to string and particle phenomenology [3][4][5][6]. Earlier model building programmes which have paved the way for the current systematic approach have led to a relatively small number of models [7][8][9][10] which have the particle content of the minimally supersymmetric standard model (MSSM). In contrast, in the latest scan [2] over 10 40 candidate models on complete intersection Calabi-Yau manifolds (CICYs), around 10 5 heterotic standard models were produced.
Of the databases of Calabi-Yau three-folds created over the last three decades in attempting to answer the original question of [11] whether superstring theory can indeed give the real world of particle physics, the increasingly numerous -and also chronological -sets are the complete intersections (CICY) in products of projective spaces [11], the elliptically fibred [12] and the hypersurfaces in toric four-folds [13] (cf. [14] for a recent review). Such Calabi-Yau datasets provide a vast number of candidate internal three-folds for a realistic model, although many of them may be ruled out even on the grounds of basic phenomenology.
The most impressive list, of course, is the last, due to Kreuzer-Skarke (KS). These total 473,800,776 ambient toric four-folds, each coming from a reflexive polytope in 4-dimensions. Thus there are at least this many Calabi-Yau three-folds. However, since the majority of the toric ambient spaces are singular and need to be resolved the expected number of Calabi-Yau three-folds from this set is even higher. The Hodge numbers are invariant under this resolution and thus have been extracted to produce the famous plot (which we will exhibit later in the text) of a total of 30,108 distinct Hodge number pairs. To establish stable vector bundles over this largest known set of Calabi-Yau three-folds is of obvious importance. To truly probe the "heterotic landscape" of compactifications which give rise to universes with particle physics akin to ours, one must systematically go beyond the set thus far probed, which had been focused on the CICYs [1,2,10,[15][16][17][18][19][20] and the elliptic [21,22] sets.
The study of bundles for model building on the KS dataset was initiated in [23] where the Calabi-Yau manifolds with smooth ambient toric four-folds were isolated and studied in detail. Interestingly, of the some half-billion manifolds, only 124 have smooth ambient spaces. Bundles which give 3 net generations upon quotienting some potential discrete symmetry and which satisfy all constraints including, notably, Green-Schwarz anomaly cancellation, were classified.
Subsequently, a bench-mark study was performed by going up in h 1,1 of the KS list [24]. Now, the largest Hodge pairs of any smooth Calabi-Yau three-fold is (h 1,1 , h 2,1 ) = (491, 11) (with the mirror having (h 1,1 , h 2,1 ) = (11,491)), giving the experimental bound of 960 on the absolute value of the Euler number. In [24], we studied the manifolds up to h 1,1 = 3, which already has some 300 manifolds. The space of positive bundles of monad type were constructed on these spaces.
In any event, the procedure of heterotic compactification is well understood. Given a generically simply connected Calabi-Yau three-fold X, we need to find a freely-acting discrete symmetry group Γ, so that X/Γ is a smooth quotient. We then need to construct stable Γ-equivariant bundles V on the cover X so that on the quotient X = X/Γ, V descends to a bona fide bundle V . It is the cohomology of V , coupled with Wilson lines valued in the group Γ, that gives us the particle content which we need to compute. In other words, we need to find Calabi-Yau manifolds X with non-trivial fundamental group π 1 (X) Γ. Often, the manifolds X and X are referred to as "upstairs" and the "downstairs" manifolds, to emphasize their quotienting relation.
The simplest set of vector bundles to construct and analyze is that of line bundle sums [2,19]. Hence, an important step is to classify heterotic line bundle models on Calabi-Yau manifolds in the KS list and extract the ones capable of leading to realistic particle physics. Of course, the existence of freely acting groups Γ on the Calabi-Yau manifolds is crucial in order to complete this programme. Unfortunately, these freely-acting symmetries are not systematically known for the KS manifolds. Indeed, even for the CICY dataset, which had been in existence since the early 1990s, the symmetry groups were only recently classified using the latest computer algebra [25]. Are there any manifolds in the KS list with known discrete symmetries? A related but simpler question is the following: Are there any manifolds in the KS list already possessing a non-trivial fundamental group? This latter question was already addressed in Ref. [26] and the answer is remarkable: Of the some 500 million manifolds in the KS list, only 16 have non-trivial fundamental group.
In fact, the 16 covering spaces for these are also in the KS list, and the discrete symmetries Γ thereof are known; in particular, their order |Γ| is simply the ratio of the Euler numbers of the "upstairs" and the "downstairs" manifolds. On these 16 special "downstairs" manifolds one can then directly build stable bundles or, equivalently, stable equivariant bundles can be built on the corresponding 16 "upstairs" manifolds. This is the undertaking of our present paper and constitutes an important scan over a distinguished subset of the KS database.
We emphasize that we expect many more than the aforementioned 16 manifolds in the KS list to have freely acting symmetries. However, the quotients of those manifolds do not have a description as a hypersurface in a toric four-fold and can, therefore, not be found by searching for non-trivial first fundamental groups in the KS list. Systematic heterotic model building on this full set of KS manifolds with freely-acting symmetries is the challenging task ahead but this will have to await a full classification of freely-acting symmetries.
The paper is organized as follows. We start in Section 2 by describing the 16 special base three-folds in detail. In Section 3, we consider heterotic line bundle models subject to some phenomenological constraints on these manifolds and the algorithm for a systematic scan over all such models is laid out. The result of this scan follows in Section 4 and we conclude with discussion and prospects in Section 5.
Nomenclature Unless stated otherwise, we adhere to the following notations in this paper:
N
The 4-dimensional lattice space of ∆ As mentioned above, the largest known class to date of smooth, compact Calabi-Yau threefolds is constructed as hypersurfaces in a toric ambient four-fold and is often called Kreuzer-Skarke (KS) data set [13,28]. The huge database consists of the toric ambient varieties A ∆ as well as the Calabi-Yau hypersurfaces X ∆ therein, both of which are combinatorially described by a "reflexive" polytope ∆ living in an auxiliary four-dimensional lattice. The classification of reflexive four-polytopes had been undertaken and resulted in the data set of 473, 800, 766 polytopes, each of which gives rise to one or more Calabi-Yau three-fold geometries.
Only 16 spaces in KS data set carry non-trivial first fundamental groups, which are all of the cyclic form, π 1 ∼ = Z/p Z, for p = 2, 3, 5 [26]. For the heterotic model-building purposes, one is in need of Wilson lines, so these 16 Calabi-Yau three-folds form a natural starting point.
More common in heterotic model building is to start from a simply-connected Calabi-Yau three-fold X with freely-acting discrete symmetry group Γ and then form the quotient X = X/Γ which represents a Calabi-Yau manifold with first fundamental group equal to Γ. Indeed, for the CICY data set [11], all the 7890 Calabi-Yau three-folds turn out to be simply-connected and a heavy computer search had to be performed to classify the freely-acting discrete symmetries [25]. Typical heterotic models have thus been built firstly on the upstairs CICY X and have then been descended to the downstairs Calabi-Yau X. A similar approach has also been taken for the model building based on the KS list carried out in Ref. [24].
In this paper, we attempt to construct heterotic models outright from the downstairs geometry. We shall start in this section by describing some basic geometry of the sixteen toric Calabi-Yau three-folds X with π 1 (X) = ∅. This includes Hodge numbers, Chern classes, intersection rings and Kähler cones. The precise quotient relationship with the corresponding upstairs three-folds X, as well as the full list of relevant geometries, can be found in Appendix B.
The construction
Let us label the sixteen Calabi-Yau three-folds and their ambient toric four-folds by X i=1,··· , 16 and A i=1,··· ,16 , respectively. They come from the corresponding (reflexive) polytopes ∆ i in an auxiliary rank-four lattice N , whose vertex information [26] is summarised in Appendix A. Before describing their geometry in section 2.2, partly to set the scene up, we illustrate the general procedure for the toric construction of Calabi-Yau three-fold, by the explicit example, X 3 ⊂ A 3 and ∆ 3 . For a more detailed introduction, interested readers are kindly referred, e.g., to Ref. [23] and references therein.
Let us first extract the lattice polytope ∆ 3 from Appendix A: It has n = 8 vertices in N Z 4 leading to 8 homogeneous coordinates x ρ=1,··· ,8 for the ambient toric four-fold A 3 ; the 4 rows of the above matrix describe the 4 projectivisations that reduce the complex dimension from 8 down to 4. Next, the dual polytope ∆ • 3 in the dual lattice M is constructed as and one can easily check that ∆ • 3 is also a lattice polytope. Then it so turns out that each of the lattice points in ∆ • 3 is mapped to a global section of the normal bundle for the the embedding, X 3 ⊂ A 3 , of the Calabi-Yau three-fold (see Eq. (45) of Ref. [23] for the explicit map). Here, ∆ • (1) which, when linearly combined, give the defining equation for X 3 .
Note that as the non-trivial fundamental group is torically realised, it is natural to expect that the KS list also contains the sixteen upstairs geometries, which we denote by X i ⊂ A i . By construction, the upstairs three-folds X i should admit a freely-acting discrete symmetry Γ i so that X i = X i /Γ i with π 1 (X i ) = Γ i . We have indeed found the corresponding upstairs polytopes ∆ i associated with the sixteen downstairs (see Appendix A for their vertex lists). It turns out that three of the sixteen upstairs Calabi-Yau three-folds X i ⊂ A i belong to the CICY list [11]: X 1 is the quintic three-fold in P 4 , X 2 the bi-cubic in P 2 × P 2 and X 3 the tetra-quadric in P 1 ×4 . Although the models in this paper are constructed over the downstairs manifolds, one can compare, as a cross-check, the models over X 1 , X 2 and X 3 with the known results over the CICYs [17,18].
We finally remark that the ambient toric varieties A ∆ constructed by the standard toric procedure might in general involve singularities. In order to obtain smooth Calabi-Yau hypersurfaces X, one must resolve the singularities of the ambient space to a point-like level via "triangulation" of the polytope ∆ in a certain manner [29]. The triangulation splits ∆ maximally and leads to a partial desingularisation of the toric variety A ∆ . In principle, there may arise several different desingularisations for a single toric variety A ∆ , in which case the number of geometries increases. Indeed, X 6 and X 14 turn out to have two and three desingularisations, respectively, while the other fourteen Calabi-Yau manifolds only have one each.
Some geometrical properties
Having constructed the Calabi-Yau three-folds in the previous subsection, we now move on to study their geometrical properties relevant to the heterotic model-building. Instead of describing all the details in an abstract manner, we continue with the example X 3 ; the Z 2 -quotient of the tetra-quadric X 3 in P 1 ×4 . The detailed prescription for computing the geometric properties can be found from Appendix B of [23]. Alternatively, one could also make use of the computer package PALP [30] to extract all the information. The resulting geometry can be summarised as follows.
Firstly, we have k ≡ rk(Pic(A 3 )) = 4 and hence, the Picard group is generated by four elements J r=1,··· ,4 . One can then choose the basis elements appropriately so that the toric divisors D ρ=1,··· ,8 defined as the vanishing locus of the homogeneous coordinate x ρ have the following expressions: where, by abuse of notation, the harmonic (1, 1)-forms J r are also used to denote the basis of Picard group. Furthermore, unless ambiguities arise, we shall not attempt to carefully distinguish the harmonic forms of the ambient space from their pullbacks to the hypersurface. Next, the intersection polynomial of X 3 is: which means that the only non-vanishing triple intersections are and those obtained by the permutations of the indices above. The Hodge numbers can also be easily computed: leading to the Euler character χ(X 3 ) = −64. The second Chern character for the tangent bundle, which is crucial for the anomaly check, is given by in the dual 4-form basis ν r=1,··· ,4 defined such that X 3 J r ∧ ν s = δ s r . Finally, the Kähler cone matrix K = [K rs ], describing the Kähler cone as the set of all Kähler parameters t r satysfying K rs t s ≥ 0 for all r = 1, . . . , h 1,1 (X), takes the form thus representing the part of t space with t r=1,··· ,4 > 0.
The reader might have notice that h 1,1 (X 3 ) = 4 = h 1,1 (A 3 ) in this example. In general, however, h 1,1 (X) can be larger than h 1,1 (A) and a hypersurface of this type is called "nonfavourable," as we do not have a complete control over all the Kähler forms of X through the simple toric description of the ambient space A. The notion of favourability means that the Kähler structure of the Calabi-Yau hypersurface is entirely descended down from that of the ambient space; namely, the integral cohomology group of the hypersurface can be realised by a toric morphism from the ambient space. Amongst the sixteen downstairs geometries X i , only the two, X 15 and X 16 , turn out to be non-favourable. As we do not completely understand their Kähler structure, we will not attempt to build models on either of these two manifolds.
In Appendix B.2, the geometrical properties summarised so far for X 3 ⊂ A 3 are tabulated for all the downstairs manifolds X i ⊂ A i , as well as their upstairs covers X i ⊂ A i , i = 1, . . . , 16. Another illustration for how to read off the geometry from the table is given in Appendix B.1 for X 1 ⊂ A 1 and X 1 ⊂ A 1 .
Let us close this subsection by touching upon an issue with multiple triangulations. As mentioned in section 2.1, the Calabi-Yau three-folds X 6 and X 14 turn out to admit two and three triangulations, respectively. Here we take the former as an example. Its toric data is encoded in the polytope ∆ 6 : this polytope turns out to admit the following two different star triangulations * , where triangulations of the polytope ∆ 6 are described as a list of four-dimensional cones. For instance, the first element {1, 2, 5, 6} ∈ T 1 represents the four-dimensional cone spanned by the corresponding four vertices: It also turns out that the two smooth hypersurfaces, associated with the two triangulations T 1 and T 2 , have the same intersection structure and the same second Chern class. It is expected in such a case that the two Calabi-Yau hypersurfaces are connected in the Kähler moduli space. In other words, the two Kähler cones adjoin along a common facet. Thus, the pair can be thought of as leading to a single Calabi-Yau three-fold X 6 , whose Kähler cone is the union of the two sub-cones, K j , * A triangulation is star if all maximal simplices contain a common point, in this case reduced to be cones expanded by four vertices and the origin point. In our notation the origin point is omitted, leaving only the four indices labeling the vertices.
where K 1 and K 2 are the Kähler cones of the two hypersurfaces associated with T 1 and T 2 , respectively (see Ref. [24] for the details). The Kähler cone matrices for the two sub-cones turn out to be and therefore, the Kähler cone matrix for the union can be computed as: One can similarly play with ∆ 14 . For this geometry as well it turns out that the three triangulations lead to a single Calabi-Yau three-fold, X 14 . As for the Kähler cone, the three sub-cones are and via the simple joining one obtains the Kähler cone of X 14 : In summary, although there are different triangulations for ∆ 6 and ∆ 14 , one ends up obtaining a single geometry each, X 6 and X 14 , respectively.
Location in the Calabi-Yau landscape
Since a very special corner in the landscape of Calabi-Yau three-folds has been chosen, it might be interesting to see the location of these sixteen, say, in the famous Hodge number plot [31]. Figure 1 shows the Hodge number plot of all the Calabi-Yau three-folds known to date, together with that of the sixteen manifolds X i and of their mirrors. Some basic topological data for both downstairs X i and upstairs X i is also summarized in Table 1 Figure 1: The Hodge number plot: The left figure is for all the Calabi-Yau three-folds known to date and the right is for the sixteen non-simply-connected Calabi-Yau three-folds X i as well as their mirrors; the blue round dots are for the original sixteen and the purple squares are for the mirrors. Table 1: Picard numbers and Euler characters of the downstairs Calabi-Yau three-folds X i and their upstairs covers X i , for i = 1, . . . , 16. In the last row is also shown the π 1 of the downstairs manifolds X i . The subscript "nf " for Picard number indicates that the geometry is nonfavourable.
Physical constraints and search algorithm
As indicated in Table 1, of the sixteen downstairs three-folds, the first fourteen, X i=1,··· ,14 , turn out to be favourable and, in this paper, we shall take the initial step towards the construction of heterotic line bundle standard models on them. The main difficulty with the two non-favourable geometries arises from the Kähler forms which do not descend from the ambient space; the corresponding components of the Kähler matrix and the triple intersection numbers are difficult to obtain from the ambient toric data, since the line bundles could be safely descended down to CY manifolds are coming only from toric divisors, with a smaller number than the dimension of CY manifold. While for the extra line bundles of CY manifolds, it is not straight forward to write them out and not possible to compute the triple intersection numbers since the calculation is essentially done over the toric variety. Therefore, the missing info makes it impossible to fully check certain consistency conditions of the bundle, notably the poly-stability condition discussed below.
Choice of bundles and gauge group
Let us begin by discussing the choice of gauge bundle and the resulting four-dimensional gauge group. First of all, we need to choose a bundle V with structure group G which embeds into the visible E 8 gauge group. The resulting low-energy gauge group, H, is the commutant of G within E 8 . As discussed earlier, for V we would like to consider Whitney sums of line bundles of the form where the line bundles are labeled by integer vectors k a with h 1,1 (X) components k r a such that their first Chern classes can be written as c 1 (L a ) = k r a J r . The structure group of this line bundle sum should have an embedding into E 8 . For this reason, we will demand that c 1 (V ) = 0 or, equivalently, n a=1 k a = 0 , (10), are attractive grand unification groups and they can be further broken to the standard model group after the inclusion of Wilson lines. Hence, constructing such SU (5) and SO(10) models, subject to further constraints discussed below, is the first step in the standard heterotic model building programme. The additional U (1) symmetries turn out to be typically Green-Schwarz anomalous. Hence, the associated gauge bosons are super massive and of no phenomenological concern.
Anomaly cancelation
In general, anomaly cancelation can be expressed as the topological condition where V is the bundle in the observable E 8 sector, as discussed,Ṽ is its hidden counterpart and [C] is the homology class of a holomorphic curve, C, wrapped by a five-brane. A simple way to guarantee that this condition can be satisfied is to require that where Mori(X) is the cone of effective classes of X. Here, we have used that ch 2 (T X) = −c 2 (T X) and that ch 2 (V ) = −c 2 (V ) for bundles V with c 1 (V ) = 0. Provided condition (10) holds the model can indeed always be completed in an anomaly-free way so that Eq. (9) is satisfied. Concretely, Eq. (10) guarantees that there exists a complex curve C with [C] = c 2 (T X) − c 2 (V ), so that wrapping a five brane on this curve and choosing the hidden bundle to be trivial will do the job (although other choices involving a non-trivial hidden bundle are usually possible as well).
To compute the the second Chern class c 2 (V ) = c 2r (V )ν r of line bundle sums (7) we can use the result where d rst are the triple intersection numbers. For the 16 manifolds under consideration these numbers, as well as the second Chern classes, c 2 (T X), of the tangent bundle are provided in Appendix B.
Poly-stability
The Donaldson-Uhlenbeck-Yau theorem states that for a "poly-stable" holomorphic vector bundle V over a Kähler manifold X, there exists a unique connection satisfying the Hermitian Yang-Mills equations. Thus, in order to make the models consistent with supersymmetry, we need to verify that the sum of holomorphic line bundles is poly-stable.
Poly-stability of a bundle (coherent sheaf) F is defined by means of the slope where J is the Kähler form of the Calabi-Yau three-fold X. The bundle F is called poly-stable if it decomposes as a direct sum of stable pieces, of equal slope µ(F a ) = µ(F), for a = 1, · · · , m. In our case, the bundle V splits into the line bundles L a as in Eq. (7). Line bundles, however, are trivially stable as they do not have a proper subsheaf. This feature is one of the reasons why heterotic line bundle models are technically much easier to deal with than models with non-Abelian structure groups. All that remains from poly-stability is the conditions on the slopes. Since c 1 (V ) = 0, we have µ(V ) = 0 and, hence, the slopes of all constituent line bundles L a must vanish. This translates into the conditions for a = 1, . . . , n which must be satisfied simultaneously for Kähler parameters t r in the interior of the Kähler cone. The intersection numbers and the data describing the Kähler cone for our 16 manifolds is provided in Appendix B.
SU (5) GUT theory
A model with a (rank four or five) line bundle sum (7) in the observable sector that satisfies the constraints (8), (10) and (14) can be completed to a consistent supersymmetric heterotic string compactification leading to a four-dimensional N = 1 supergravity with gauge group SU (5) or SO(10) (times anomalous U (1) factors). Subsequent conditions, which we will impose shortly, are physical in nature and are intended to single out models with a phenomenologically attractive particle spectrum. The details of how this is done somewhat depend on the grand unified group under consideration and we will discuss the two cases in turn, starting with SU (5).
In this case we start with a line bundle sum (7) of rank five (n = 5) and associated structure group G = S(U (1) 5 ). This leads to a four-dimensional gauge group H = SU (5) × S(U (1) 5 ). The four-dimensional spectrum consists of the following SU (5) × S(U (1) 5 ) multiplets: Here, the subscripts a, b, · · · = 1, . . . The multiplicity of these various multiplets is computed by the dimension of associated cohomology groups as given in Table 2. The most basic phenomenological constraint to impose SU (5) × S(U (1) 5 ) repr. associated cohomology contained in Of course, a similar constraint on the chiral asymmetry should hold in the 5-5 sector. In general, for a rank m bundle V , we have the relation So for the rank five bundles presently considered it follows that ind(∧ 2 V ) = ind(V ). Hence the requirement (16) on the chiral asymmetry in the 10-10 sector already implies the correct chiral asymmetry for the 5-5 multiplets, ind(∧ 2 V ) = −3, and no additional constraint is required.
The index constraints imposed so far are necessary but of course not sufficient for a realistic spectrum. For example, one obvious additional phenomenological requirement would be the absence of 10 multiplets which amounts to the vanishing of the associated cohomology group, that is, h 1 (X, V * ) = 0. However, cohomology calculations are much more involved than index calculations and currently there is no complete algorithm for calculating line bundle cohomology on Calabi-Yau hypersurfaces in toric four-folds. For this reason, we will not impose cohomology constraints on our models in the present paper, although this will have to be done at a later stage.
However, working with line bundle sums allows us to impose slightly stronger constraints which are based on the indices of the individual line bundles. Of course we can express the indices of V and ∧ 2 V in terms of the indices of their constituent line bundles as where, by the index theorem, the index of an individual line bundle L = O X (k) is given by Suppose that ind(L a ) > 0 for one of the line bundles L a . Then, in this sector, there is a chiral netsurplus of 10 multiplets which is protected by the index and will survive the inclusion of a Wilson line. Since such 10 multiplets and their standard-model descendants are phenomenologically unwanted we should impose † that ind(L a ) ≤ 0 for all a. Combining this with the overall † The caveat is that line bundle models frequently represent special loci in a larger moduli space of non-Abelian bundles. Line bundle models with exotic states -vector-like under the GUT group/standard model group but chiral under the U (1) symmetries -may become realistic when continued into the non-Abelian part of the moduli space where some or all of the U (1) symmetries are broken. In this case, the exotic states may become fully vector-like, acquire a mass and are removed from the low-energy spectrum. While this is an entirely plausible model building route, here we prefer a "cleaner" approach where the spectrum at the Abelian locus can already lead to a realistic spectrum. constraint (16) on the chiral asymmetry and Eq. (18) this implies that for all a = 1, . . . , 5. A similar argument can be made for the 5-5 multiplets. A positive index, ind(L a ⊗ L b ) > 0, would imply chiral 5 multiplets in this sector. They would survive the Wilson line breaking and lead to unwanted Higgs triplets. Hence, we should require that ind(L a ⊗L b ) ≤ 0 for all a < b which implies that for all a < b. Table 3 summarizes both the consistency constraints explained earlier and the phenomenological constraints discussed in this subsection. This set of constraints will be used to classify rank five line bundle models on our 16 Calabi-Yau manifolds.
SO(10) GUT theory
In this case, we start with a line bundle sum (7) In analogy to the SU (5) The multiplicity of each of the above multiplets is computed from associate cohomology groups as indicated in Table 4. The three generation condition on the 16-16 multiplets remains associated cohomology contained in Table 4: The spectrum of SO(10) models and associated cohomology groups.
the same: For rank four bundles Eq. (17) implies that ind(∧ 2 V ) = 0 so no further constraint needs to be imposed. In analogy with the SU (5) case, in order to avoid 16 exotics, we should impose that for all a = 1, . . . , 4. The line bundle indices can be explicitly computed from Eq. (19). The 10 sector is automatically vector-like so no further constraint analogous to Eq. (21) is required. Table 5 summarizes the consistency constraints explained earlier and the phenomenological constraints discussed above. These constraints will be used to classify rank four line bundle sums on our 16 manifolds.
Search algorithm
In principle, the scanning procedure is straight-forward now. We firstly generate all the single line bundles, L = O X (k) with entries k r in a certain range and with their index between −3 and 0. Then we compose these line bundles into rank four or five sums imposing the constraints detailed in Table 3 and 5, respectively, as we go along and at the earliest possible stage.
Which range of line bundle entries k r a should we consider in this process? Unfortunately, we are not aware of a finiteness proof for line bundle sums which satisfy the constraints in Table 3 and 5, nor do we know how to derive a concrete theoretical bound on the maximal size of the entries k r a from those constraints. Lacking such a bound we proceed computationally. For a given positive integer k max we can find all line bundle models with k r a ∈ [−k max , k max ]. We do this for Table 5: Consistency and phenomenological constraints on rank four line bundles of the form (7).
increasing values k max = 1, 2, 3, . . . and find the viable models for each value. If the number of these models does not increase for three consecutive k max values, the search is considered complete. In this way, we are able to verify finiteness and find the complete set of viable models for rank five bundles. For rank four, we find the complete set for some of the manifolds but are limited by computational power for the others.
Finally, there is a practical step for simplifying the bundle search. If the Kähler cone, in the form given by the original toric data, does not coincide with the positive region where all t r > 0 it is useful to arrange this by a suitable basis transformation. This makes checking certain properties, such as the effectiveness of a given curve class, easier. We refer to Ref. [23] for details.
Results
In this section, we describe the results of our scans for phenomenologically attractive SU (5) and SO(10) line bundle GUT models on the 14 favourable Calabi-Yau three-folds out of our 16 special ones.
SU (5) GUT theory
For the rank five line bundle sums we are able to verify finiteness computationally for each manifold, using the method based on scanning over entries k r a with −k max ≤ k r a ≤ k max for increasing k max , as explained above. As an illustration, we have plotted the number of viable models on X 9 as a function of k max in Fig. 2. As is evident from the figure, the number saturates at k max = 4 and stays constant thereafter. A similar behaviour is observed for all other spaces. Recall from Table 1 that amongst the favourable base manifolds X i=1,··· ,14 , only X 1 has Picard number 1, X 2 and X 4 have Picard number 2, X 5 , X 6 , X 7 , X 8 , X 14 have Picard number 3, and X 3 , X 9 , X 10 , X 11 , X 12 , X 13 have Picard number 4. It turns out that viable models arise on all the six manifolds with Picard number 4 and on two out of the five manifolds with Picard number 3, namely X 6 and X 14 , in total 122 models. The number of models for each manifold is summarized in Table 6 and the explicit line bundle sums are given in Appendix C. A line bundle data set can be downloaded from Ref. [32]. Table 6: Numbers of viable rank five (SU (5)) and rank four (SO(10)) line bundle models and maximal value of |k r a | for each base manifold. For the SO(10) cases marked with a star numbers are converging but have not quite saturated despite the large entries.
SO(10) GUT theory
As in the SU (5) cases, viable models only arise on base manifolds with Picard number greater than 2. It turns out that amongst the five Picard number 3 manifolds, X 7 does not admit any viable models, and the other four, X 5 , X 6 , X 8 , X 14 admit 5, 13, 9, 28 bundles, respectively. For all those cases, the scan has saturated according to our criterion and the complete set of viable models has been found. In total this is 55 models which are listed in Appendix C. For the other six manifolds X 3 , X 9 , X 10 , X 11 , X 12 , X 13 , all with Picard number four, only X 9 is complete and admits 2207 bundles. For the others, the number of viable bundles is converging but still growing slowly despite the large range of integer entries. The number of models found in each case is summarized in Table 6 and the complete data sets can be downloaded from Ref. [32].
An SU (5) example
To illustrate our results we would like to present explicitly one example from our data set, a three generation SU (5) GUT theory on the Calabi-Yau manifold X 9 . We recall that X 9 is a Picard number four manifold, constructed from eight homogeneous coordinates (see Appendix A for details). From Table 6 we can see that there are 12 viable SU (5) models on this manifold, with line bundle entries in the range −4 ≤ k r a ≤ 4. Let us consider the first of these models from the table in Appendix C which is specified by a line bundle sum V of the five line bundles Evidently, c 1 (V ) = 0 and, since three of the line bundles are the same, only two slope-zero conditions (14) have to be satisfied in the four-dimensional Kähler cone. With the intersection numbers and Kähler cone given in Appendix B, we find that this can indeed be achieved.
Hence, the all three chiral 10 multiplets are charged under the first U (1) symmetry and uncharged under the others. Although, at this stage, we do not know the charge of the Higgs multiplet 5H it is clear that all up Yukawa couplings 5H10 10 are forbidden (perturbatively and at the Abelian locus). Indeed, for those terms to be S(U (1) 5 ) invariant we require a Higgs multiplet with charge −2 under the first U (1) and uncharged otherwise, a charge pattern which is not available at the Abelian locus.
We also note from Eq. (25) that the matrix (k r a ) of line bundle entries has rank two. This means that two of the four U (1) symmetries are Green-Schwarz anomalous with corresponding super heavy gauge bosons while the other two are non-anomalous with massless gauge bosons. Those latter two U (1) symmetries can be spontaneously broken, and their gauge bosons removed from the low-energy spectrum, by moving away from the line bundle locus (see Ref. [18] for details).
Conclusion and outlook
In this paper, we have studied heterotic model building on the sixteen families of torically generated Calabi-Yau three-folds with non-trivial first fundamental group [26]. From those 16 manifolds, we have selected the 14 favourable three-folds and we have classified phenomenologically attractive SU (5) and SO(10) line bundle GUT models thereon. Concretely, we have searched for SU (5) and SO(10) GUT models which are supersymmetric, anomaly free and have the correct values of the chiral asymmetries to produce a three-family standard model spectrum (after subsequent inclusion of a Wilson line). For SU (5) we have succeeded in finding all such line bundle models on the 14 base spaces, thereby proving finiteness of the class computationally. The result is a total of 122 SU (5) GUT models.
For SO(10) we have obtained a complete classification for all spaces up to Picard number three, resulting in a total of 55 SO(10) GUT models. For the other six manifolds, all with Picard number four, only one (X 9 ) was amenable to a complete classification. For the other five manifolds, although the number of models were converging with increasing line bundle entries, they had not quite saturated even at fairly high values of about k max = 20. We expect that we have found the vast majority of models on these manifolds with a small fraction containing some large line bundle entries still missing. Altogether we find 28870 viable SO(10) models. All models, both for SU (5) and SO(10), can be download from the website [32].
The main technical obstacle to determine the full spectrum of these models -before and after Wilson line breaking -is the computation of line bundle cohomology on torically defined Calabi-Yau manifolds. We hope to address this problem in the future.
We consider the present work as the first step in a programme of classifying all line bundle standard models on the Calabi-Yau manifolds in the Kreuzer-Skarke list. A number of technical challenges have to be overcome in order to complete this programme, including a classification of freely-acting symmetries for these Calabi-Yau manifolds and the aforementioned computation of line bundle cohomology. Table 7: The sixteen pairs ( ∆ i , ∆ i ) of reflexive four-polytopes, for i = 1, · · · , 16, each pair leading to the upstairs Calabi-Yau geometry X i ⊂ A i and the downstairs geometry X i ⊂ A i with π 1 (X i ) = ∅. The polytopes are described in terms of their integral vertices.
B Base Geometries: Upstairs and Downstairs
In this Appendix, we analyse the quotient relationship between the 16 upstairs manifolds X i ⊂ A i and the corresponding 16 downstairs manifolds X i ⊂ A i whose defining polytopes were given in the previous Appendix. In addition, some geometrical properties of these manifolds relevant to model building will also be discussed.
B.1 An Illustrative Example: the Quintic three-fold
Amongst the sixteen pairs is the quintic manifold X 1 and its Z 5 quotient X 1 , which we take as an illustrative example. The corresponding two polytopes ∆ 1 and ∆ 1 have 5 vertices each.
Firstly, the vertices of ∆ 1 for the quintic three-fold X 1 can be read off from Table 7: wherex ρ=1,··· ,5 are the homogeneous coordinates on the ambient space P 4 . The polytope ∆ 1 naturally leads to the usual 126 quintic monomials inx ρ ; these generate the defining polynomial of the quintic Calabi-Yau three-fold X 1 .
Similarly, the vertices of ∆ 1 for the quotiented quintic X 1 = X 1 /Z 5 are given as follows: where x ρ=1,··· ,5 are again the homogeneous coordinates on the corresponding toric ambient space.
As for the generators of the defining polynomial, the polytope ∆ 1 leads to the following 26 monomials in x ρ : Now, by demanding that the 26 monomials be invariant, we find the following phase rotation rule which links the two sets of homogeneous coordinates.
This phase rotation relates the two manifolds X 1 and X 1 tightly. Not only the Laurant polynomials are explicitly connected, it turns out that the integral cohomology groups are also very much similar under the phase rotation.
As an example illustrating the precise relation between upstairs and downstairs space, consider one of the 126 monomials, x 1 x 3 2 x 3 , defining the upstairs ambient space of the quintic X 1 . If we transform this monomial using the rules in Eq. (30) we obtain The phase independence of the result means that this is one of the 26 monomials which define the downstairs manifold X 1 = X 1 /Z 5 . The remaining 25 downstairs monomials can be obtained by applying this procedure systematically to all upstairs monomials ‡ . ‡ In some cases, an additional permutation of the downstairs homogeneous coordinate has to be included, as in some of the examples in Table. 8. This is to ensure that the linear relationships between divisors and integral basis are literally the same for both the upstairs and the downstairs manifolds.
We next turn to some relevant base geometries, most of which can be easily extracted from PALP [30]. Let us start from upstairs. Firstly, the Picard group of X 1 is generated by a single elementJ 1 and all the toric divisors are rationally equivalent toJ 1 : Note that we do not carefully distinguish harmonic (1, 1)-forms from divisors unless ambiguities arise. The intersection polynomial is: 5J 3 1 , which means that d 111 ( X 1 ) = 5. In general, the coefficient of the monomial termJ rJsJt in the intersection polynomial is the value of d rst ( X), without any symmetry factors. Finally, the Hodge numbers are: leading to the Euler character χ( X 1 ) = −200.
As for the downstairs manifold X 1 , the Z 5 -quotient of the quintic X 1 , the Picard group is again spanned by a single element J 1 and the toric divisors are all equivalent: The intersection polynomial is given as: J 3 1 , and hence, d 111 (X 1 ) = 1. Finally, the Hodge numbers are: h 1,1 (X 1 ) = 1, h 1,2 (X 1 ) = 21 and the Euler character χ(X 1 ) = −40.
Note that the intersection polynomial of X 1 is equal to that of X 1 divided by 5, the order of the discrete group Z 5 . This remains true for all the fourteen favorable manifolds X i=1,··· ,14 in an appropriate basis of H 1,1 .
B.2 Summary of the Base Geometries
For the remaining fifteen cases, the phase rotations of the homogeneous coordinates are not as straight-forward as in the quintic example. One needs to make use of some combinatorial tricks to figure out the explicit results. In some cases, permutations are also required to make the upstairs and the downstairs intersection polynomials proportional to each other. In Table 8, we summarise the complete results for all the sixteen pairs of geometries. For each pair, we first present the phase rotation map between upstairs and downstairs coordinates (and the permutation of the coordinates if required). The base geometries of X i and X i then follow in order: number of generating monomials, § toric divisors in terms of the (1, 1)-form basis elements, intersection polynomial. The Hodge numbers h 1,1 and h 2,1 , as well as the Euler character χ of the manifold X are presented using the notation [X] h 1,1 ,h 2,1 χ .
. In addition, the second Chern class c 2 (T X) and Kähler cone matrix K for the downstairs manifolds are also being listed. The Kähler cone is then given by all Kähler parameters satisfying K rs t s ≥ 0 for all r.
C GUT Models
Technology Facilities Council, UK, for grant ST/J00037X/1, the Chinese Ministry of Education, for a Chang-Jiang Chair Professorship at NanKai University as well as the City of Tian-Jin for a Qian-Ren Scholarlship, the US NSF for grant CCF-1048082, as well as City University, London and Merton College, Oxford, for their enduring support. Chuang Sun would like to thank Andrei Constantin for helpful discussion. S.-J. L. thanks the University of Oxford for hospitality while part of this work was completed. | 2014-05-13T13:08:22.000Z | 2013-09-01T00:00:00.000 | {
"year": 2014,
"sha1": "2c050c7552c6c34468b22ca250add8a98b7f221c",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/JHEP06(2014)077.pdf",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "a8eb5efc460ec3dfea55698c3635e6315071580d",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
93787189 | pes2o/s2orc | v3-fos-license | Growth Morphology of Boron Doped Single Crystal Diamond
Boron-doped single crystal diamond films were grown homoepitaxially on synthetic (100) Type Ib diamond substrates using microwave plasma assisted chemical vapor deposition. A modification in surface morphology of the film with increasing boron concentration in the plasma has been observed using atomic force microscopy. Use of nitrogen during boron doping has been found to improve the surface morphology and the growth rate of films but it lowers the electrical conductivity of the film. The Raman spectra indicated a zone center optical phonon mode along with a few additional bands at the lower wavenumber regions. The change in the peak profile of the zone center optical phonon mode and its downshift were observed with the increasing boron content in the film. However, shrinkage and upshift of Raman line was observed in the film that was grown in presence of nitrogen along with diborane in process gas.
Introduction
Doped diamond films are of interest because of their potential use in the semiconductor industry as well as other applications involving their unique thermal, mechanical and chemical properties. Doping of diamond with boron during its growth process is a widely investigated field of research [1][2][3][4]. Boron provides diamond a wide range of dopant concentrations (1016 up to 1021 atom cm-3) from a wide band gap semiconductor to a metal and to a superconductor [5][6][7]. However, the large scale production of doped diamond film can be possible only if high growth rate and high crystalline quality can be achieved simultaneously. It has been reported that the use of nitrogen during deposition improves the growth rate of diamond films [8][9][10]. In this study we have investigated the effect of nitrogen on growth morphology and growth rate of diamond films during boron doping and also studied their electrical conductivity as a function of temperature.
Experimental
Synthetic (100) oriented Type Ib diamond substrates (size 3.5 × 3.5 × 1.5 mm 3 ) were chosen as seed crystals for this study. The acetone cleaned seed crystal was inserted into 2.45 GHz microwave reactor chemical vapor deposition (CVD) chamber. The microwave power was adjusted to 1.4-1.5 kW to set a deposition temperature of 1100 ± 20 o C at a chamber pressure of 100 Torr. The 10% B2H6 diluted in H2 with 6% of CH4 / H2 mixture was used for the deposition with a total of 400 standard cubic centimeters per minute (sccm) of gas. The details of deposition parameters and growth rate of samples have been summarized in Table 1. Each sample was ultrasonicated in acetone after deposition to remove any residual boron-carbon soot. The quality and surface morphology of as deposited films were determined by Raman spectroscopy, X-ray rocking curve experiment, optical microscopy (OM) and atomic force microscopy (AFM). Raman spectra were recorded using a 514 nm laser excitation wavelength at room temperature. In the X-ray rocking curve experiment, omega scans were obtained by rotating sample with 0.02 o angular step with a detector fixed at 2θ position corresponding to (400) Bragg diffraction peak. Type of doping, doping level and electrical conductivity was determined by room temperature Hall measurement and four point probe measurements respectively. In a four point probe measurement, the samples were first heated to 600K and the electrical resistance was measured during the cool down process. This procedure was followed to remove any adsorbates from the diamond films that could have arisen from the CVD shut down procedure in hydrogen atmosphere. The I-V characteristics were maintained in the linear regime during experiment by limiting the current supply within 1mA at zeromagnetic field. The value of sheet resistance was measured in magnetic field of 0.55 Tesla. The measurements were taken for both positive and negative current with both polarities of magnetic field.
Results and discussion
In order to study the effect of nitrogen on growth morphology of boron doped diamond, a few samples were grown at varying diborane and nitrogen contents in the feed gas for 5 hours deposition. After doping the yellow color of the seed changed into pale blue, then to dark blue and finally opaque to visible light due to the increasing concentration of boron in the films. Optical transmission spectra of the samples in the visible range were taken in order to measure the effect of doping on the color of the films as shown in Fig. 1.
The spectra indicate small downshift of peak and decrease in intensity as the doping level of the films increases. The peak of low doped film (sample BD5) appeared at 454 nm in the optical transmission spectra (Fig. 1). The surface morphology of films was analysed using OM and AFM images taken over scanned areas of 10 × 10 µm 2 as shown in Fig. 2 (a, b and c). The untreated substrate samples (seeds) are macroscopically flat. The OM image of undoped diamond film (HD2) shows some hillocks on its surface as shown in Fig. 2(a). The growth rate and surface morphology are observed to be dependent on B/C ratio in the gas phase. Increasing B/C ratio up to a certain limit in feed gas decreases the growth rate as well as the surface roughness. The surface smoothness of the film was improved with an additional supply of 1000 ppm of nitrogen in feed gas as shown by a comparison of sample BD7 (16000 ppm of B2H6 and no N2) and BD8 (addition of 1000 ppm of N2) in Fig. 2(b) and Fig. 2(c). However, hillocks were observed on the surface of film with the addition of 2000 ppm of nitrogen (BD9), but it improved the growth rate of diamond films by a factor two compared to that using 1000 ppm of nitrogen as shown in Table 1 of samples BD8 and BD9. The root mean square (rms) roughness of the samples measured from contact mode AFM images of the surface are listed in parenthesis for each sample Seed (1.6 nm), HD2 (2nm), BD5 (4nm), BD6 (1.6nm), BD7 (1nm), BD8 (0.8nm) and BD9 (0.9 nm) and was found to decrease with increasing boron concentration. Introduction of 1000 ppm of nitrogen during boron doping improved the surface morphology and increased the growth rate of film (sample BD8). However, when higher amount of nitrogen (2000 ppm) along with the same B/C ratio (16000 ppm) was used, the film growth rate was shown to increase by twice than that using 1000 ppm of nitrogen, but the surface quality degraded. When very high B/C ratio (25000 ppm) or very high methane 10% of total feed gas was introduced in the deposition chamber, amorphous carbon and polycrystalline diamond film grew on the seed crystal. Hence, it can be argued that an optimal amount of diborane and nitrogen is needed to improve the structural quality and the growth rate of boron doped single crystal diamond. Fig. 1. Optical transmission spectra of seed crystal (Seed) and doped diamond films after applying correction for the absorption in the diamond seeed crystal. Photographs of seed crystal (Seed) and doped diamond (BD6) are embedded inside the graph. Yellow color is avidence of substitutional nitrogen atom in seed crystal and bluish color in BD6 represents presence of boron in diamond lattice. The quality of the epitaxial diamond film determined using X-ray rocking curve and Raman spectroscopy is shown in Fig. 3. A single intense peak of (400) was observed in the rocking curve experiment from 30-62 o omega scans indicates high quality film as shown in the inset of Fig. 3. The measured full width at half maximum (FWHM) of the Bragg peak varies from 0.07 o to 0.11 o as the boron content in the films is increased indicating the decrease in structural quality. An intense zone-center optical phonon mode of diamond is visible at 1333 cm -1 along with additional bands at 580, 900, 1042, 1233 cm -1 as shown in Fig. 3. Significant modification and asymmetry in zone center optical phonon line was observed in the presence of such additional bands. In low boron doped films, these bands were absent and no asymmetry in Raman line was observed. Those additional bands were previously reported on both the boron doped polycrystalline and single crystal diamond [6,[11][12][13]. The asymmetry in optical phonon Raman line increased with increasing doping level. The downshift of optical phonon line and broadening of FWHM with doping level were also observed in the spectrum. The Raman line of boron doped samples is observed to be sharpened and shifted towards higher wavenumber region when the sample was grown in presence of nitrogen. The downshift of optical phonon line with increasing boron content has already been described by Gheeraert et al [14]. The bands around 1233 and 580 cm -1 are assigned due to large cluster of boron atoms present in diamond lattice [7,11,15]. The asymmetric broadening and downshift of Raman peak of borondoped diamond films could be explained by . The variation in conductivity, (σ) of the films with temperature (T) was determined from four point probe measurement as shown in Fig. 4. Activation energies of the samples have been obtained by the best fit of the Arrhenius plot of the conductivity data in two different temperature regions. The activation Fig. 1.
Conclusions
Boron-doped single crystal diamonds have been synthesized using MPCVD method and effect of nitrogen during the growth process of boron doped diamond has also been studied. Supply of 1000 ppm of nitrogen in feed gas improves the surface morphology and increases the growth rate of diamond deposition. However, when 2000 ppm of nitrogen was introduced in feed gas, growth rate was double than that using 1000 ppm of nitrogen but some non-epitaxial growth was found on the surface of doped film. The color of seed crystal changed from yellow to pale blue then to dark blue as the boron concentration in the film increased. A few additional bands along with first order Raman line were visible in the lower wavenumber region in Raman spectrum of borondoped diamond films. The downshift and broadening of Raman line was also observed with increasing boron content in the crystal. The decrease in width and upward shift of Raman line was also observed in the samples that grew in presence of nitrogen. The growth rate of diamond was observed to be decreasing with increasing boron content in the film. Temperature dependent resistivity measurements showed that the current conduction mechanism depends upon the doping level and obeys semiconductor behavior in the experimental temperature range 140 to 600 K. | 2019-04-04T13:09:54.153Z | 2013-01-31T00:00:00.000 | {
"year": 2013,
"sha1": "bea0db65f644407c0005fb1b850d13f3f63dfc0e",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "a9196845923da5dc7edf5d594cbf64ee20edf53e",
"s2fieldsofstudy": [
"Materials Science",
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Materials Science"
]
} |
17434756 | pes2o/s2orc | v3-fos-license | Theory of high bias Coulomb Blockade in ultrashort molecules
We point out that single electron charging effects such as Coulomb Blockade (CB) and high-bias staircases play a crucial role in transport through single ultrashort molecules. A treatment of Coulomb Blockade through a prototypical molecule, benzene, is developed using a master-equation in its complete many-electron Fock space, evaluated through exact diagonalization or full Configuration Interaction (CI). This approach can explain a whole class of non-trivial experimental features including vanishing zero bias conductances, sharp current onsets followed by ohmic current rises, and gateable current levels and conductance structures, most of which cannot be captured even qualitatively within the traditional Self Consistent Field (SCF) approach coupled with perturbative transport theories. By comparing the two approaches, namely SCF and CB, in the limit of weak coupling to the electrode, we establish that the inclusion of strong-correlations within the molecule becomes critical in addressing the above experiments. Our approach includes on-bridge-correlations fully, and is therefore well-suited for describing transport through short molecules in the limit of weak coupling to electrodes.
I. INTRODUCTION
Theoretical calculations on single molecule conduction have typically employed coherent Non-Equilibrium Green's function (NEGF) theories ("Landauer limit") [1], [2] coupled with Self Consistent Fields (SCF) to describe charging effects. Though fairly successful in describing many aspects of single molecule conduction [3]- [7], there have been important discrepancies between theory and experiment [8]. The most common ones include poor match between theoretical and experimental current levels and zero-bias currents [3], [4], [8]. It was also pointed out in [9] that a whole class of experimental I-V's show features, which cannot be captured even qualitatively using an SCF theory. Charging energies of short molecules ( 3 eV for benzene) are often larger than their electrode coupling (< 0.2 eV for benzene di-thiol on gold), and thus could be in the Coulomb Blockade (CB) regime where single electron charging effects could dominate. It is thus debatable whether it is better described as a quantum wire in the SCF regime, or as a quantum dot array in the Coulomb Blockade (CB) regime. Nevertheless the wisdom of SCF approaches must be scrutinized especially for conduction through shorter molecules. The purpose of this paper is to present a Coulomb Blockade approach to molecular conduction using a benzene molecule as prototype, and establish it as a different viewpoint from the conventional NEGF-SCF treatment.
Furthermore features obtained via the CB approach can semi-quantitatively explain several nontrivial features commonly observed [10]- [14] in experiments.
It is common to distinguish between two regimes of transport: a) an SCF regime where the dominant energy scale is the contact coupling, allowing for fractional charge transfer through the system; and b) a Coulomb Blockade (CB) regime where the dominant energy scale is the single electron charging, leading to integral charge transfer. In the SCF regime the description of transport via non-interacting single particle energy levels can be justified. In this limit, it is common to use the SCF-NEGF scheme that takes charging effects into account. Here the molecular Hamiltonian is described by a set of single particle levels, which are coupled to reservoirs through their self energies. The electron interactions are taken into account using SCF schemes as shown in the block diagram in Fig. 1a. All quantities in the NEGF formalism are matrices of dimension N × N, N being the number of single particle basis functions used. This allows for an accurate description of quantum chemistry of both the isolated molecule and its bonding to the contacts [15]. In the CB regime with weak contact coupling, charging effects dominate, and the use of single particle basis sets may be questionable. In such cases, it may be preferential to employ a multi-electron description or Configuration Interaction (CI) where feasible. The central quantities in this CI method are now matrices of dimension 2 N × 2 N , thereby accounting for strong interaction accurately. The weakly coupled contacts are treated perturbatively using transition rates between states differing by a single electron [5]. It is interesting to note that most theoretical efforts in molecular conduction have been in the SCF March 23, 2022 DRAFT regime, while energy scales favor the CB regime. Our paper is thus a concrete attempt towards CI based transport.
This paper is organized as follows: we begin by defining an appropriate many-body Hamiltonian for Benzene whose parameters are benchmarked based on well-established mean-field techniques. We then illustrate how a CB treatment is conceptually different from the standard SCF treatment in the weak coupling limit, not only under non-equilibrium conditions, but even under equilibrium conditions. We then point out the importance of inclusion of excited states in transport, that naturally arise within our CI approach. The progressive access of these excited states leads to transport signatures under various non-equilibrium conditions. Before we conclude, a few CB fits to experimental data are presented in support of our analysis.
II. THE MODEL HAMILTONIAN AND EQUILIBRIUM PROPERTIES
An appropriate model Hamiltonian is usually described with an adequate basis set. In this paper, we use a tight binding Hamiltonian with one p z orbital per site to describe our CI based scheme. Although this generates just a minimal 6 × 6 single particle basis set, its many-electron space is 2 12 × 2 12 in size. Besides, our objective here is to describe the CI approach for transport and compare it with the SCF approach for the same Hamiltonian. Better quantum chemical descriptions within the CI approach can be achieved by starting with a reduced but more accurate one-particle Hamiltonian, but we leave these for future work.
One begins with the model Hamiltonian in second quantized notation: where α, β correspond to the orbital indices of the frozen p z orbitals for carbon sites on the Benzene ring,and σ,σ represent a particular spin and its reverse. In connection to its equilibrium configuration, it is more convenient to work with onsite energiesǫ defined as: whereǫ α 's denote the mean-field on-site energies in the equilibrium charge neutral configuration of the molecule and n represents its mean-field value. Now the model Hamiltonian is simply March 23, 2022 DRAFT re-written as:Ĥ The mean-field Hamiltonian derived from the above Hamiltonian, is: where is the Self Consistent Field, the calculation of n α performed self consistently with the one electron Hamiltonianĥ. In the following sections, we derive appropriate parametersǫ, t and U for benzene, to describe the two different approaches i.e., the CI (Eq. 3) and the SCF approaches (Eq. 5), and compare them in parallel in the case of both equilibrium and non-equilibrium conditions.
A. Equilibrium Electron Number v/s chemical potential: Choosing Charging parameters
A distinguishing aspect of CB is the abrupt charge addition as opposed to a gradual one in an SCF calculation shown in Fig.2(b). This fact is readily seen in the figure, in which the SCF and CB calculations are presented using the one electron and many-electron Hamiltonians described by Eq. 3 and 4 respectively. In the weak coupling limit, the CB result is more physical, and (shown dotted line) as well as RSCF calculation is consistent with Gaussian based calculation [22]. c) One particle spectral function shows peaks at the energy levels of the single particle Hamiltonianh. d) Lehmann spectral function evaluated via manyelectron spectrum yields many more spectral peaks corresponging to removal (addition) of electrons from the neutral ground state into various charge configurations (excitations) of singly charged species. Notice that the IP-EA and HOMO-LUMO gaps are equal to the corresponding charge-stability plateaus N = N0 for many-body and SCF calculations shown in b).
March 23, 2022 DRAFT the SCF calculation does not do justice to this integer charge transfer. However, schemes such as self interaction correction could be introduced within the one-electron Hamiltonian [17]- [19] to incorporate this. But it turns out that even such schemes may not capture non-equilibrium correctly [9]. It is however expected that as coupling strength to the electrodes is increased, the electron transfer resembles the SCF result. There is as yet no clear formalism that addresses [20] this crossover, even in the equilibrium case although the two opposite limits namely SCF and CB are well understood. While the two limits can individually be handled by perturbative expansion in the small parameters U/Γ and Γ/U, U being the single electron charging energy and Γ being the level broadening, the intermediate regime is hard to handle owing to the non-existence of a suitable small parameter or 'fine structure constant' for transport.
The plateau in the charge addition diagram N versus µ, in which the electron number is stabilized, spans the HOMO-LUMO gap in the SCF case and the Ionization potential-Electron Affinity (IP-EA gap) in the CB calculation. The IP (EA) is defined as the energy when an electron can be removed (added) to the neutral molecule carrying N 0 electrons. This occurs when the chemical potential µ equals the energy difference between ground states differing by an electron The situation is however different in the case of SCF. Here the charge transfer dictated by a self consistent potential (Eq. 5) is gradual, in which two electrons are transferred adiabatically over a span of 2U corresponding to the removal of two electrons. This is usually referred to as the restricted SCF (RSCF mentioned in Fig. 2b). Most SCF calculations in the literature [21] employ different variants of this scheme.
There are also spin unrestricted SCF techniques [17]- [19] which take into account the abrupt charge transfer in a weakly coupled system, due to self-interaction correction, but it is not yet clear whether they work out of equilibrium [9], [23].
One expects that the IP occurs roughly midway during the gradual charge removal in the RSCF scheme [20], [21]. We use this fact to estimate our charging parameters U αβ , with the aid of a Gaussian-98 based calculation for the equilibrium electron number v/s chemical potential (N − µ), published elsewhere [22]. The calculation corresponding to the equilibrium number of N 0 = 42 maps onto our model calculations for N 0 = 6, focussing thus on the frontier orbitals and ignoring the inner core that is frozen in our estimate forǫ α . By implementing a Restricted SCF scheme using Eq. 5 within in our model Hamiltonian, we obtain a close match of the N −µ plots in the range between N 0 and N 0 −1 in comparison with the Gaussian-98 calculation in [22]. March 23, 2022 DRAFT Using an estimate of the onsite charging U αα , we calculate U αβ using the Matago-Nishimoto approximation: where r αβ is the inter carbon distance in benzene. In each case, evaluation of n α is done selfconsistently using an equilibrium value N 0 = 6, and n ασ = 1 2 . Using exact eigen-energies of the many-electron Hamiltonian Eq. 3 with the above parameters, an N −µ calculation using these total energies (shown dotted red in Fig.2b) is in excellent agreement with respect to Gaussian calculations in [22]. Note that the µ = IP in Fig.2b occurs midway between N = N 0 and N = N 0 −2 in the RSCF charging diagram. It is worth mentioning that the many-body calculation presented in this figure takes all correlation energies into account and is the exact ground state energy within our defined model Hamiltonian.
B. Equilibrium Spectral function
Conduction through molecules via molecular orbitals is well understood in the SCF picture [24]. In the strongly coupled regime (most appropriate for an SCF treatment), fractional charge transfer occurs, and Density of States (DOS) is evaluated at equilibrium [1] in order to capture the effect of the strong coupling with contact. An interplay of molecular DOS and charging treated self-consistently determines the non-equilibrium response (current-voltage or I-V characteristics).
The density of states calculated from the one-electron Green's function [1] in Fig. 2c shows peaks at the single electron eigen spectrum. As the coupling to electrodes gets stronger, the single electron DOS will show signatures and artifacts of contact bondings [15], [24].
In the weak coupling (CB) limit however, integer charge addition is favored, and transitions between states that differ by a single electron appear as spectral signatures [26]. At equilibrium, it is convenient to introduce the Ground State Spectral Function by defining the Green's function in the Lehmann representation [21]: where α, β correspond to the orbital index, which in our case are the sites of the benzene molecule, and |N, i denotes the i th excited state of a charge configuration of N electrons. The poles of this spectral function represent various transition energies for addition (removal) of electrons from the neutral ground state: whose spectral strengths are given by: The first (addition) term adds an electron to orbital β, taking the system from an N electron ground state to the j th (N+1) electron excited state, and then removes it from orbital α, bringing it back to ground state. The second (removal) equation first removes an electron from α and then adds it to β.
One can re-write the expression in terms of diagonal terms only, replacing the recurring index α with a single index, in a more convenient form as: The spectral function shown in Fig. 2d represents the removal and addition strength of various transitions at their energies given by Eq. 10. Notice that there are numerous peaks in this spectrum calculated from the many-electron transitions, due to the possible transfer to various excited states of charged species shown in Fig. 2d. It is important to note that although each transition has a non-trivial spectral weight given by Eq. 9, they satisfy an overall sum rule that amounts to the total electron number in the system. We will see in subsequent sections that these and what the SCF theory completely misses in this regime. Throughout this paper we describe the electrodes (contacts) using corresponding electrochemical potentials µ L and µ R and coupling strengths γ L and γ R .
A. Coulomb Blockade approach: Rate equation model
Transport in the CB limit [25]- [27] is often modeled with a rate equation approach, in which the steady state addition and removal of electrons is described with a rate equation for the nonequilibrium probability P N i of each N electron many-body state |N, i with total energy E N i . The master equation involves transition rates R (N,i)→(N ±1,j) between states differing by a single electron, leading to a set of independent equations defined by the size of the Fock space [26] dP along with the normalization equation i,N P N i = 1. We define rate constants where γ α represents lead molecule broadening or coupling via the end atoms, described using Fermi's Golden rule. These constants represent the partial probability for the electron to be injected by the end atom into a given many-electron ground or excited state. The transition rates are now given by for the removal levels (N, i ↔ N − 1, j), and replacing (r → a, f → 1 − f ) for the addition levels (N, i ↔ N + 1, j). µ α are the contact electrochemical potentials, f is the corresponding Fermi function, with single particle removal and addition energies ǫ N r Finally, the steady-state solution to Eq.(11) is used to get the left terminal current as March 23, 2022 DRAFT where states corresponding to a removal of electrons by the left electrode involve a negative sign.
B. STM Limit:
We briefly elucidate the relationship between spectral functions defined earlier and STM conduction spectra. Electrical conduction depends on the measurement geometry [24] and charging determined by capacitive voltage-division ratio η between leads in the Laplace solution as opposed to the resistive voltage-division ratio γ = γ R /γ L which determines the extent to which the levels are filled or emptied by the leads. The source and drain potentials are then given by Consider a simple picture in which one contact is very weakly coupled (γ << 1, η = 0), equivalent to the molecule being in equilibrium with left contact µ L . In the η = 0 limit the molecular energy levels are pinned to this contact implying that for a positive voltage µ R < µ L , and µ L remains at the equilibrium position. This picture is analogous to STM shell tunneling experiments [30], in which the weakly coupled STM tip acts as a voltage probe, thereby generating the single particle spectrum, the molecule/dot held in equilibrium with the more strongly coupled contact, in this case, the substrate.
• Ground State Spectral function: It is expected, with a more strongly coupled contact, that the right contact voltage probe, such as an STM tip can add or withdraw into or out of the dot at energies corresponding to addition or removal energies defined in Eq. 17. The stronger coupling to the left contact ensures that an electron be added or removed as soon as the tip removes or adds an electron thus maintaining overall charge neutrality. In this case conductance spectrum proportional to the equilibrium spectral function is obtained as shown in Fig.3b.
• Excited Spectral Functions: In the previous case, the chemical potential of the left contact is fixed above the transition level ǫ N r 00 but below ǫ N r 10 , thus maintaining the molecule's charge neutrality in its ground state (i.e., |N, 0 ), and hence only the Ground State spectral signature is observed. However, in a general non-equilibrium scenario, access to excited states of the neutral and charged molecule becomes feasible and hence description in terms of spectral functions corresponding to addition/removal from the i th excited state of the neutral molecule is required: where α now corresponds to the two sites that are coupled to the left (L)/ right (R) contacts.
For example, let the equilibrium chemical potential be situated at a position above ǫ N r Fig.3a. Given, a positive bias (µ L > µ R ) the above transition is energetically feasible only if the ground state of the cation (|N −1, 0 ) is accessed, which occurs for a tip voltage corresponding to µ R below ǫ N r 00 . Once this transition is accessed, spectral function A N 1,L/R involving the first excited state gets involved due to the initial condiction µ L > ǫ N r 10 , due to which the neutral excited state |N, 1 can be accessed. This results in additional satellite peaks S1 and S2 in Fig. 3d. A schematic of transitions that consititute the satellite peaks S1 and S2 due to A N 1,α (E) is shown in Fig.3c. In this figure, we have only shown the removal levels for brevity, and extension of the argument by including addition levels is trivial. In general, in the STM regime, one can write a simple expression to evaluate the conductance formula as a weighted average over various excited state spectral functions: We have thus shown a simple signature that indicates the access of excitations in the many-body spectrum of the neutral molecule. In general, one may view near-equilibrium conduction in the CB regime using single particle energy levels: and their corresponding spectral weights.
C. Break Junction limit
The break junction limit is achieved by setting η = 0.5, γ = 1, implying that both contacts are equally coupled to the molecular dot and half the applied voltage appears across the molecular levels which in our case transition energies ǫ N ij . The many-body configuration of the molecule consists of its ground state |N, 0 and the first excited state |N, 1 separated by a gap similar to the HOMO-LUMO gap ∆, followed by a set of closely spaced excitations denoted by |N, i , i > 1.
The I-V characteristics in this limit show certain key signatures which result from how these excitations are accessed.
The onset of conduction is established by the offset between the equilibrium Fermi energy E F and the first accessible transition energy ǫ N r 00 . The qualitative shape of the I-Vs depends on how the excitations are accessed. Recall that η = 0.5 implies that the molecular levels are displaced with respect to the contact electrochemical potentials by the applied voltage. If the excited states are not accessed simultaneously or prior to the threshold transition ǫ N r 00 , as shown in Fig.4b, the I-V has a brief staircase of plateaus before a quasi linear rise in current. This quasilinear current rise occurs due to a huge number of closely spaced transport channels that are triggered only when transitions involving an excitation appear within the bias window. However, the quasilinear current can also appear prematurely without an intervening plateau, if a feasible transition to an excited state appears in the bias window at or before the threshold transition.
This situation is shown in Fig.4b, where ǫ N r 10 also appears at threshold, resulting in a quasilinear regime immediately following the onset. The two distinct I-Vs have been observed experimentally [10]- [13], [29] and depend merely on the position of the equilibrium electrochemical potential (Fermi energy) with respect to the transition energies. In the meanwhile, similar SCF based I-V characteristics show adiabatically smeared out currents whose onsets get postponed by the changing position of the equilibrium E F , as shown in Fig. 4c. The SCF potential from Eq. 5, determines how levels float with respect to their non-equilibrium occupation [1]. It is readily seen by comparing Fig.3a,b with Fig.3c that any self consistent potential cannot change the qualitative features of the I-Vs in order to resemble the CB features.
D. Connection to experiments: Fitting Data using Coulomb Blockade model
We consider matching the I-V shapes using our CB model using consistent fitting parameters.
The experiments conducted on conjugated phenylenes [11], [12] at low temperatures (T = 30K) suggest strong Coulomb Blockade effects. It is worth noting that an 'orthodox' theory simply involving junction resistors and capacitors would also manage to capture the zero-bias suppressed conductance, the subsequent sharp onset and the linear current rise; however, it would not capture the intervening plateaus, fine structures in the I-Vs, and the gateability of the current levels and their asymmetry features that arise due to discrete transitions in the molecular configuration space [10], [14]. In contrast with metallic islands, a molecular dot shows significant size quantization that leads to quantum corrections to the junction capacitance, and gets further modified at high bias to involve nonlinear corrections to it arising from partial densities of states filled separately by the two contacts.
While our model explains salient features of a lot of Coulomb Blockade experiments [10], [14], [29] it is interesting to note that in some cases the same molecule showed CB behavior at low temperature and SCF behavior at higher temperatures [11], [12]. A possible explanation is that at low temperature the molecule could be frozen into a configuration where the plane of the middle ring is oriented perpendicular to the side rings, while room temperature structures sample other configurations and are rotated on average. This is supported by the fact that current levels at room temperature are an order of magnitude greater, which can be attributed to an increased average degree of conjugation along the molecular backbone. In contrast at low temperature the rotated central ring has a weaker coupling with the rest of the backbone, which could reduce its broadening while increasing electron localization and charging, leading to CB behavior. In fact some of the experiments feature bulky middle groups like antracence. Steric side groups that are deliberately inserted to facilitate this rotation of the central rings and enforce CB [12]. While doing exact calculations on these molecular structures is beyond the scope of the present paper, we consider making simple fits by considering the following facts: • Current Levels: Using the fitting parameters γ 1 = γ 2 ≈ 5 − 10meV we obtain current levels similar to experimental data. It is important to note that changing γ does not affect the Fig. 5. Experimental fits for data [11], [12]. a) T = 30K, γ ≈ 5meV . b) T = 295K, γ ≈ 5meV c)T = 30K, γ ≈ 0.25meV . conductance before the threshold voltage which shows a vanishingly small pre-threshold current.
• Threshold Voltage: We noticed in the last section with that the gap ∆ between ground and first excited states of the neutral molecule is important in determining the qualitative shape of the I-V. When the equilibrium electrochemical potential E F lies above mid-gap between ǫ N r 10 and ǫ N r 00 , the first excited state becomes voltage-accessible before the ground state of the charged species is accessed and populated simultaneously via µ R = ǫ N r 00 , giving rise to the quasi linear I-V immediately following the very first current onset. In all the experimental data, we observe a threshold voltage between 0.5 − 0.7 V thus tuning the gap ∆ ≈ 0.6 − 0.8 V. Fig. 5(a) (b) and (c) are fits obtained for experiments [11], [12]. In case of molecular asymmetries [11], only positive bias is considered [9], the I-V asymmetries themselves being attributed to polarization effects [7]. In obtaining an experimental fit for [12], in Fig. 5(b), we used T = 295 K consistent with experiment. Notice that the first peak has broadened significantly. The higher temperature Coulomb Blockade could possibly be attibuted to the fact that molecule involved has an anthracene based middle ring that is much bulkier, thus leading to a higher temperature frozen configuration stabilized by steric interactions.
The molecular system we consider is a simple prototypical molecule (benzene in our case) with calculations based on simple parameters that are associated with this minimal system. Performing calculations on a real molecule-electrode system will be needed to yield a quantitative fit in terms of threshold voltage, current levels and positions of peaks. However, the conduction mechanism remains the same. The exponentially larger configuration space of even a minimal Coulomb Blockaded molecule makes a first-principles calculation of its transport properties inordinately challenging compared to SCF treatments in the literature. However, the SCF calculations do not capture the non-equilibrium transition rates between the many-body states, which as we argued earlier carry crucial correlation signatures that are experimentally observable for ultrashort molecules. Such a "real" calculation involving the quantum chemistry of larger molecules and contact bondings within this nonequilibrium full CI treatment is still at a very early stage [31].
Furthermore, it needs to be supplemented with the broadening of the many-particle states that could affect the interference between nearby levels, an issue that has received relatively little attention [32]- [36] and requires further work.
IV. CONCLUSION
In this paper, we have developed a Coulomb Blockade approach for molecular conduction through short molecules using Benzene as a prototype. We have shown how equilibrium and non-equilibrium signatures are very different from the traditional NEGF-SCF viewpoint, and that the CB approach is appropriate in the weak coupling limit. Many I-V features distinct to the CB regime are often seen in experiments. These features that are easily obtained using a full Configuration Interaction master equation approach are potentially very hard to obtain within any effective one-electron potential, even for a minimal model. A particular challenge therefore lies in bridging the SCF and CB regimes while paying close attention to coherent level broadening and associated interferences. The emergence of many recent experiments on molecular dots, exploring the interplay between charging, quantization and level-broadening, should prove invaluable in further theoretical developments in this regard. | 2014-10-01T00:00:00.000Z | 2006-10-09T00:00:00.000 | {
"year": 2006,
"sha1": "f67daea2d6e5200ac23dca78299c44570035964b",
"oa_license": null,
"oa_url": "https://docs.lib.purdue.edu/cgi/viewcontent.cgi?article=1011&context=nanodocs",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "f67daea2d6e5200ac23dca78299c44570035964b",
"s2fieldsofstudy": [
"Physics",
"Chemistry"
],
"extfieldsofstudy": [
"Physics",
"Chemistry"
]
} |
131760188 | pes2o/s2orc | v3-fos-license | Licensing of Crude Hydrocarbons in Russia – the Current System , Its Problems and Disadvantages
In this article the authors give a brief assessment of more than 20-year oil and gas subsoil use (since 1993) in the country, set out the negative aspects of subsoil licensing in relation to exploration activities, creation of new reserves and involvement of hydrocarbon reserves in the development; some of the problems are associated with the existing interaction of state authorities and organizations – subsoil users with different forms of ownership. The authors suggest basic ways to improve legislative and regulatory laws, arrangement of subsoil licensing and its monitoring, methodology to manage works on oil and gas subsoil use from the government’s side. The above measures, which can be entered at the same time with the Order of V.V. Putin (until 01.01.2017) on the updating of already issued 3290 licenses, will eliminate the concept of ‘loss of government benefits’ associated with failure to reach efficiency of annual investments in the oil and gas industry for exploration and development of fields.
With the adoption in 1993 of the "Subsoil Law of the Russian Federation" as amended over the years, "Regulations on the subsoil licensing" and other regulatory permits and regulations, the country has developed and operates a modern structure of the study and use of oil and gas, allowing the government to provide licensed right to work on the subsoil in order to study, prospecting, exploration and production of hydrocarbons.
\It is provided to potential users under certain conditions and for a specific time frame on the tender, auction or competitive basis.Note that in accordance with pr.Rosnedra 835 of December 21, 2015 the conditions for issuing licenses and introducing changes to the licensing agreements are the responsibility of Rosgeolekspertiza.20-year period of the licensed right to use subsoil in the Russian Federation has revealed both positive and negative aspects of such an innovation in the oil and gas industry.
On the one hand, for the first time in the country under market conditions the institute of subsoil licensing appeared.According to it any individual or legal person, regardless of their form of ownership could be a subsoil user under certain conditions and requirements.
With the introduction in 2004 of the primary licensed right to use subsoil of oil and gas in the form of auction, the government budget for the first time began obtaining multimillion dollar payouts every year from granting the right to use subsoil on an auction basis.
Excessive granting of subsoil in the mid 90-ies of the last century created a contingent of subsoil users: large (on the basis of the post-Soviet geological and mining production organizations), some of which now form a 'vertically integrated companies' (VICs), and quite a large number of medium and small firms, companies, organizations.
Their main difference is the scale of operating funds, both debt and coming from profits, mostly from hydrocarbon production.
In general VICs problems in this analysis are not considered; mainly VICs inherent problems are associated with deficiencies in the management of mineral resources, provided for the exploration of oil and gas from the government.
In particular, VICs own a license "NR" (Entrepreneurial risk-taking license).To the date of the license issue, as a rule, 1 to 2 fields were located in the area.VICs like small companies do not involve the development of part of the fields on the area.The reason is trivial: VICs have money, but they do not put them into the development of assets (hydrocarbon reserves).Small companies do not have enough money for the full field development.
However, the expected result is the same -the government loses the benefit from non-involvement in the development of known hydrocarbon reserves with the apparent benefit of VICs holding known (recorded in the state balance) hydrocarbon reserves on their balance sheet.At the same time, a number of licenses "NR" have areas from 1 to 2 thousand km 2 .
The license type "NP" (License for Prospecting) is characterized by the opposite situation.When small companies perform even minimal seismic survey and deep drilling for the last, even their own money, the VICs predominantly 'denote' prospecting, evaluation and exploration of new fields.In the best case they will perform 2D seismic survey, often incompletely (the minimum volume under the terms of the license agreements).At the same time for the majority of "NP" licenses, which are owned by small companies, the excess of actually performed seismic work comparing to VICs in different regions of Russia may be 10-30%, and deep exploration and exploration and appraisal drilling up to 50% or higher (according to the authors).
Another problem associated with VICs is non-involvement ("downtime") of hydrocarbon reserves of category C 2 , located on separate deposits, areas, sites, etc., from the major reserves of developed deposits, even on long-developed fields.
The number of licenses already issued to small businesses of all kinds is large enough.In the future due to low demand for licensed areas offered by VICs tenders and auctions, the government support of small enterprises is an urgent task.
Hydrocarbons licensing problems
The prevailing realities in Russia are such that actually the main activities of the government in subsoil field for oil and gas have been reduced to the issue of licensed rights (licenses) for subsoil use with the purpose of prospecting, exploration and production of hydrocarbons.For the period 2010-2012 activity peak of the government in the subsoil licensing ended; in the beginning of 2012 3260 licenses for hydrocarbon were issued; more than 3450 fields were involved in licensing.
However, carried monitoring is not valid in terms of the actual volume of work and the results, does not contain analytic calculations and possible linkages of various performance indicators and work results of subsoil users, as not all subsoil users can honestly report and confirm the amounts and types of work performed.For example, passing volumes and types of work can be displayed in the year they start, the following year or in the year of their completion.Reliable analytics should be performed in existing institutes and centers (named after Shpilman in KhMAD and others), which have a capable staff.
Prompt service of managing and implementing staff in the management structure is not provided with all the necessary and sufficient content of constantly ongoing different bases on the actual data information in the oil and gas industry, partly difficult or impossible.
Information on the activities of subsoil users contained in their information reports, documents, materials, officially entering the Rosnedra, often have the character of formal replies, declarations of intent, which are in fact are not met partly or fully.The presentation of information is arbitrary, graphics materials is minimum (it is the basis of the oil and gas industry).
Official requests to Rosnedra for postponement, types and volumes of work are often not professionally justified (from the point of difficulties encountered in the geological study and exploration, reliability or revaluation of reserves, exploration and development of deposits, etc.); statement of facts must comply with the regulations, but often not respected, and fake.Receipt of additional information, references, documents and materials from the subsoil user is tedious task (even in direct connection with it) that disrupts a reasonable time of conclusions preparation on the address of subsoil users.
Financial crises in the world and in the country occur since the beginning of the introduction of licensing.They are directly related to the financing of capital investments in production activities of any subsoil user.Applications from the user for change of terms, types and volumes of work are not taken into consideration.All this leads to the accumulated debt of almost all subsoil users in the physical volume of works (seismic survey, drilling) and the need for extension or transfer of terms and conditions of license agreements for comprehensible number of years.
The deadline for "NP" licenses is extended for 6-9 months and rarely for 1 year or more.Taking into account the geography of the licensed areas and the seasonal work, especially in the north of the country, as well global warming, 6-and 9-month extension of works may lie in the months of spring, summer and autumn.It is not enough, for example, to test the well in 2-6 column objects in addition to conditional reports or calculation of hydrocarbon reserves.
One of the main subsoil licensing objectives is the recovery of the mineral resource base for hydrocarbons.We can assert that this objective is failed in terms of reserves growth, which the government expected from the subsoil users in prospecting for new deposits or resolved/ It is partly resolved in terms of actual amount of new hydrocarbon reserves, passed the state examination in the National Reserves Committee.The reserves are far from having optimal geography, low (often inflated) average value of reserves for one field, reserves structure, quality of oil, etc.
After the abolition of the government financing, geological exploration (GE) is completely fall on the balance sheet of the company that performs this activity.However, exploration results appear in a few years after the costs incurred.
There has been a disengagement of the government from the immediate solution goals and objectives for the management and monitoring of the recovery of mineral resources in terms of all types of prospecting, evaluation and exploration of new hydrocarbon deposits at own expense of subsoil users.In recent years, the govermnet has reserved only the goals, objectives and scope of state budget financing and management of regional exploration, which are priority in steps and stages on the search for new deposits of hydrocarbons, but crucially cannot be consistent with the recovery of mineral resources.
In addition, there is a substitution of concepts even by leading experts of major oil and gas industry and subsoil users.Today the concept of exploration (regional seismic surveys, drilling of appraisal wells) is mixed with search and assessment works, the basis of which is 2D seismic survey and less 3D survey, drilling of prospecting and exploration wells.
As a result, the main reserves of hydrocarbons have already been transferred to subsoil users for the development and production or retained (for the license "NR") by almost all the subsoil users.Therefore, from the beginning of subsoil licensing in Russia institute there is a negative concept of 'lost profits of the government', in physical and monetary terms (non-additional exploration of known deposits, missing production of hydrocarbons, shortfall of taxes, etc.).The cumulative loss of the government benefits is comparable to the tangible part of the government budget for several years.
Despite that the government declare legislative provisions that a license can not be a guarantee, subject to purchase and sale, etc., de facto licenses for oil and gas are the subject of non-transparent sales transactions, speculative transactions, the outflow of money from the country, feeding of criminal organizations, etc.
Contingent of subsoil users in Russia is presented by natural monopolies, organizations and companies of different forms of ownership, including government-owned, and the professional staff with different skills, which are not appropriate to the technologies of the modern level to solve production challenges.
Many service organizations and companies often derived from the large companies, but affiliated with the parent company, offer their tenants poor quality and unreliable results of services at inflated prices, meaning solely to maximize profits from the performance of the contractual relationship.It leads to unjustified delay of works, extortion of additional costs with lack of interest in the final result in terms of their quality and efficiency.
Employees of subsoil users, especially small structures, often use the incompetence of license holders, are not interested in the success of the work.Their main goal is to stay as long as possible on the payroll in organization.However, the owners of licenses, attracting funding for the works, have no place (in state bodies) to seek help, even in especially important periods of production activities.
Auction subsoil lately is predominant with a few exceptions.However, there is a persistent negative trend -there are no valid members for trading sites.
In accordance with the basic laws of the Russian Federation, the government is the owner and manager of mineral resources.This involves the provision of the right to work in the areas of subsoil in accordance with the agreements stipulated in the license conditions, requirements, terms, types and minimum amounts of the necessary prospecting, evaluation, exploration and development.
However, it also imposes on the public authorities and the use of problems in licensing agreements are not formal, «a blueprint» for almost all areas of mineral resources, and the most realistically attainable conditions, requirements, terms, forms and the minimum volume of work required for each license area.
From the point of view of the legislative and regulatory permits, the government and subsoil users have certain rights and responsibilities, the neglect of which by the subsoil user is unacceptable and punishable in one form or another during their non-compliance (withdrawal of license, early termination, temporary termination of works until corrections performed, etc.) However, the subsoil user draws and invests in the production significant funds.In this respect, the prohibition norms of the government should not be formal but objective, aimed at the creation of prescribed purpose to attract financing (search, evaluation, exploration and development of hydrocarbons), taking into account the imperfections of the legislative, regulatory and permitting rights, characteristics of the geography for oil and gas works, the relationship between customers and contractors, landowners, the emergence of protected and conservation areas in the period of validity of the license, encroachments on the results from other business entities on the territory of the license area, etc.
Initial payments for subsoil use on an auction basis, containing hydrocarbon reserves often suffer exorbitant quantities, not fitting into the existing ranges of rates per unit of reserves in the world, and even in Russia on average.
Terms to input hydrocarbon deposits on fields preparing for auctions and license agreements applicable to licenses issued in the past 8-10 years, contain formulaic wording of the form "after 6, 7, 8 years to enter the field into development".They don't take into account any difference in the geography (regions of the north, southern areas, etc.) or the complexity of geological structure or types of hydrocarbons (oil, natural gas, condensate), no known reserves, nor objective accuracy of the reserves, nor their exploration degree (the ratio of categories C 1 and C 2 ) or the number of deposits, etc.
Paperwork for licensing subsoil also requires perfection.
For example, the winner of the auction was given a photocopy of the Protocol of the auction commission.State authorities believe that it is enough with such 'paper' of potential subsoil user, the organization of which is referred to in this paper, to apply to bank to remove from the account many millions or even billions of dollars for the transfer to the goverment as payment for auction trades.Currently, even the banks in Russia, not to mention the western banks refuse to provide such funds to their customers.
From the above perspective of oil and gas industry in the field of labor management of mineral resources we propose the following directions and recommendations for improvement of the entire oil and gas industry operating in the country.
In the field of legislative and regulatory permitting law: -Continuous improvement of the existing legislative and regulatory law to maximum approach the peculiarities of conducting oil and gas activities; -Strengthening the staff of the entire vertical power structure in administration and management of subsoil licensing; -Granting of direct federal structure (the Federal Agency for Subsoil Use or its equivalent) of at least partly or full management (under the license obligations under existing licenses) and control of the main oil and gas activities (completeness performance, quality and terms of field geophysical works, well drilling for oil and gas, hydrocarbon production, etc.); -Introduction for all federal structures mutually inverse (government-subsoil user) duties by mutual goals and objectives within the framework of licenses and the consistency in the examination of the situation with the observance of the basic principle -effective prevention and resolution of emerging issues and difficulties in anytime action of specific license for subsoil use; -Creation of an effective annual operating system for handling industrial activity of subsoil users not on notification principle (where the subsoil user sends via communication channels or internally reports the results of work without a proper discussion by qualified professionals).The peer principle is to be entered in the protocol part, which contains disadvantages that are required to eliminate; subsoil users should also attend annual or quarterly work plan meetings.
In the field of subsoil licensing arrangement: -Improving the work effectiveness of territorial authorities responsible for subsurface oil and gas; -Improving the paperwork for licensing subsoil, for example, to increase the validity and reality of physical types, volumes and terms of works (especially seismic and prospecting and exploration deep drilling, the transition of hydrocarbon reserves from C 2 to C 1 category); -Conducting a systematic consulting, legal and specialized works (in all matters of oil and gas activities) through seminars, organized by the federal or territorial bodies for investors and licensees (for reference, such seminars have become the prerogative of many business schools, business centers, academies, consultation basis of which are retellings of the existing laws, regulations, regulatory permits and the like, but without the practical applications to the needs of specific license areas); similar seminars advisable for specialists employed by mining companies to work (especially of small states subsoil, where there is only the main experts, and are totally dependent on the quality of service organizations); -To introduce the declarative principle for licenses of "NP" type, when a potential user of mineral resources request the government to give the license right for part of unallocated subsoil fund, not included in the annual realized list of subsoil sites or program of licensing for the next year, for the search and evaluation of new deposits of hydrocarbons; -The public authorities should evaluate the costeffectiveness of actually incurred in the exploration licenses "NP" and "NR" (at least for seismic survey and search and appraisal deep drilling) over the past 10 years; -To maintain a permanent list of contractors, service, research, etc. organizations and enterprises.Based on feedback about the quality of their work from the subsoil user to accredit them under the bodies of the Federal Service for subsoil use as the preferred for mining companies to recruit; the rest of the organization will be the objects of increased attention from the government in terms of the quality of their works and services; -To develop arrangements to consistent fulfillment of license obligations of certain subsoil users (or groups) by sending them official letters, requests, reminders, etc. on correction of technical errors in the licenses, license agreements, additions to them, land arrangement schemes (for example, in the coordinate description of the license areas, overlay areas, location of angular plot points in the neighboring regions of Russia, and so on), with specific recommendations (for example, preparation and signing of the Protocol by all parties in interest on the recognition of line coordinates dividing the area in varying proportions for subsequent application for changes and additions to licenses, etc.).
In the field of methodical approach to manage mineral resources of oil and gas: -To minimize the issue of the license areas with crossborders fields (between countries, the Russian Federation subjects, administrative boundaries, neighboring licenses), prepared for auctions, which include reserves registered on the State balance, divided into two or more parts (with the exception of large fields) because of organizational and technical impossibility of effective management and development of such deposits; -On previously issued licenses containing cross-border fields, to take measures to clearly consolidate such concepts such as "Operator of work types on the cross-border field", "Separation of hydrocarbon production from the field", "Part of the project work on the operation of cross-border fields relating to features of hydrocarbon production near the common borders of two or more licenses", etc.; -Only in 2012 (20 years after the adoption of the Federal Law "On Subsoil Law") we succeeded to adopt amendments.They concern the areas with hydrocarbon reserves from the undistributed fund, but adjacent to the developed deposits (areas called 'incisions').
Legislatively or through regulations on licensing subsoil users, who develop the main part of the reserves of the field, can be granted with the license right for such an area in a prespecified period of preparation for licensing auctions.They have to place on record their desire to join such 'incisions', on the application with mandatory payment for the reserves, cost of auction, etc., and in spite of the presence of others who wish to.In the absence of the officially registered desire for the rights, such 'incision' will be provided to potential users on the general auction basis; -To provide for subsoil users (with the conditions mentioned below) all reserves subject to that there are no neighbors with discovered hydrocarbon reserves and providing the fact that these reserves are of categories C 1 +C 2 .The conditions and requirements for license type "NP" are as follows.The subsoil user according to own seismic survey discovered field.Subsequent drilling of more than 2-3 appraisal wells is required to assess the scale of the industrial reserves.However, the subsoil user did not complete the works by the expiry date of the license "NP".With all unfinished works he is given a license of type "NE" (License for Operation) for the actual part of discovered reserves.The rest of the reserves is transferred into undistributed fund.
-To provide in new licensing agreements of "NP" and "NR" types for the period of granting the right to conduct prospecting and evaluation operations in accordance with the "Regulation on the stages and phases of works…", when there are certain features in the mandatory order to enter at least the minimum requirements regarding the implementation of specific geological trends, ideas, problems, etc. for search and evaluation; -The National Reserves Committee, Central Committee for Reserves effectively work in the Rosnedra structure.At the expert level they provide official conclusions on the studies submitted by the subsoil user reports, records, documents, etc.. Taking into account the above, to provide Rosnedra some geological control functions in terms of completeness, quality and effectiveness of works performed by subsoil user and service organizations, directly affect the efficiency of oil and gas works (without prejudice to the activities of the geological control of Rosprirodnadzor); -To constantly conduct an analysis of organizationssubsoil users under the current license in order to avoid bad situations where even 'big' (by the number of licenses issued) subsoil users over the years violate licensing laws: work or produce hydrocarbons outside the license areas, do not have actual calculations (recalculations) of reserves, existing projects, programs, regulations, etc.
-A special attention should be paid to small (management and senior staff) organizations-subsoil users, which depend entirely on the arbitrary working designers and service organizations (mainly suppliers of seismic and drilling services); such subsoil users have no one to appeal in cases of poor work performance and even consideration of disputes and industrial relations in the courts; -For licenses "NR" to replace the wording of the form "subsoil area is granted with status of mining allotment" (it is obvious that the entire area can not have the status of this kind, but the only area of the deposit) to "subsoil area within the deposit (deposits) is granted with status of a preliminary mining allotment, the rest of the territorystatus of geological allotment".It is more appropriate to the nature and objectives of the works within the entire area.In addition, every new field will have its pre-mining allotment made by surveying service, the coordinates of which can be checked and applied on the map of hydrocarbon deposits and licensing of Russian subsoil.Otherwise, the subsoil user does not provide reliably identified and certified coordinates of new field location; -A number of major and leading specialists in the oil and gas industry do not have a proper mandatory reporting of operating subsoil licensing in Russia; permit in such a situation is given by holding open meetings, open days, daily works with subsoil users in the Russian regions; -Initial payment for the use of new area on an auction basis, containing hydrocarbon reserves should be aligned with the established indicators of oil and gas projects.For example, the price should not be higher than the existing average prices for the particular auction at the date per hydrocarbons unit, for already held auction or a group of related indicators in the technical and economic calculations of hydrocarbon production; -Service companies that give extremely inflated prices primarily on seismic and deep drilling, with inadequate quality or completeness of the required works, should bear responsibility.In case of confirmed by the state examination correctness and conclusions of the subsoil user of the shortcomings and loss of profits, subsoil users should have effective support from the state bodies, including judicial proceedings and recovery of unconditional cash penalty; -To establish an effective subsoil monitoring system, allowing the use of reliable geological and geophysical, other information on subsoil for statistical and analytical purposes, based on the ordering of the actual exploration data.To implement the above, all participants should collect and clearly understand tabular description of the indicators, parameters, etc., their linking to a discrete reporting date, excluding submission of non-existent or unreliable information by introducing various types of sanctions for fake data; -To enter the address supervision for activities of subsoil users on the most important licensed areas from the entire structure of the subsoil, but within the designated authority and without intervention in economic activities of the subsoil user.
Each of the proposals and recommendations derived from the set of descriptions of the typical examples existing in the practice of oil and gas industry.It is assumed that only by such means we can objectively evaluate the effectiveness of inserted in the oil and gas industry huge money of both Russian and foreign investors and own funds of subsoil users.There are cases where foreign investors (especially partners in joint business with the Russian side) are convinced of requirements, regulations, etc. allegedly existing in Russia and demanding the unconditional appreciation of the works.
Such foreign partners who understand the rules of business and partnerships have to take advice from experts, which deny the legitimacy of such requirements in Russia and, most importantly, the effectiveness of their funding.This undermines the confidence of foreign investors and partners to the Russian side, while the country's leadership makes a lot of effort to attract investment in Russia.
The recent Order (to implement until 01/01/2017) of V.V. Putin on the updating of already issued 3290 licenses for oil and gas will eliminate the diversity of form and content of the license agreement text, redundancy of specifications for terms of use.However, it will not improve the investment climate of oil and gas industry and indirectly the economy as a whole.
The execution of the Order will lead to a one-time elimination of the violations in the field of interpretation of the existing legal and licensing documents in accordance with Rosprirodnadzor regulations, the availability of appropriate and approved Projects of works, Calculation of hydrocarbon reserves and elimination of reporting debts to the government.But it will not lead to the objective (enforceable) justification of types, volumes and the timing of exploration for oil and gas, the input of hydrocarbon fields in operation, compliance with the design levels of hydrocarbon production, the implementation of commitments for the rational use of associated petroleum gas, etc.
Most of all license agreements texts imply only a formal approach.There is no question on rational volumes and types of work with real and consistent implementation of them in time.
In other words, with proper and effective cooperation of the government and subsoil users, these tools will work on the recovery of mineral resources base of hydrocarbons, rather than their misuse with declaration of purpose allocations as 'for oil and gas'.It can be argued that the volume of investments, thrown in the oil and gas industry each year, does not match the achieved efficiency of their use (for the discovery of new fields and hydrocarbon reserves, the number of producing wells, the quality and completeness of works). | 2018-12-12T03:44:47.449Z | 2016-01-01T00:00:00.000 | {
"year": 2016,
"sha1": "40641b60af17eaf38032e9ec16dc209b88c6ee61",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.18599/grs.18.1.11",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "40641b60af17eaf38032e9ec16dc209b88c6ee61",
"s2fieldsofstudy": [
"Geology"
],
"extfieldsofstudy": [
"Engineering"
]
} |
202904861 | pes2o/s2orc | v3-fos-license | Regional Development Level Based on Principal Component Analysis: Case Study Uzbekistan
This study aims to define the interstate disparities in provisions of level of socio-economic development in Uzbekistan, compare regions’ level of development, and identify backward parts units of Uzbekistan.
Introduction
At the present time, achieving sustainable development of regions through effective and optimal utilization of the existing potential of social and economic development of the regions, cities and towns, as well as ensuring the rational use of all available resources is one of the pressing issues.
It's a fact that differences between development of regions and close areasare inevitable. Disproportions in development of regional areas, appeared as result of different factors: • the difference in the level of knowledge and talent of human factor; • unequal usage of technical-technological progress; • diversity of natural resources, availability of financial capital; • unequal level of investment; • thenature of the existing economy structure and etc.
In "On Uzbekistan's Development Strategy ")focuses on Integrated and balanced socio-economic development of provinces, districts and cities, optimum and efficient use of their potential as one of the priorities. In addition, a decree "On priority measures to ensure the accelerated socio-economic development of the regions" (signed by President of the Republic of Uzbekistan Sh. M. Mirziyoyev on August 8, 2017). The document was adopted to analyzes the socioeconomic development of the regions, radically improve the forms and methods of organizing work and the quality and living standards of the population, and other important tasks. A major objective of the development programmes lunched in Uzbekistan is to bring the balanced regional development. In order to achieve the goal, the economic planning in the country has traditionally been focused upon the need to provide special support to the disadvantaged areas. Although the country remains on course to achieve its socio-economic development goals, related challenges such as inequality and rural-urban and regional disparities persist.
Realizing the seriousness and importance of the problem of regional socio-economic disparities, the study measures and compares the levels of socioeconomic development of different regional administrative division of Uzbekistan (i.e., 12 regions (Andijan, Bukhara, Jizzakh, Kashkadarya, Navoi, Namangan, Samarkand, Surkhandarya, Syrdarya, Tashkent, Fergana, Khorezm), one autonomous republic (Republic of Karakalpakstan) and one independent city-the capital of Uzbekistan(Tashkent city) ) based on the levels of their development. It is hoped that the results of the study would be useful for regional planning in Uzbekistan.
METHODOLOGY-MATERIALS AND METHODS OF ANALYSIS
The present investigation is exclusively based on secondary data sources. The data is extracted from the database of the State committee of Uzbekistan on statistics. For this study, the regional administrative division has been taken as the unit of analysis. Fourteen regional administrative divisions of Uzbekistan have been included in the analysis.
Socio-economic development is a multidimensional process and it cannot be fully evaluated by a single indicator. Moreover, a number of indicators when analyzed individually do not provide an integrated and easily comprehensible picture of the reality. It necessitates for construction of a composite index of socio-economic development based upon optimal combination of different developmental indicators. There are several methods (e.g., principal component analysis, multiple factor analysis, aggregation method, monetary index, ratio index and ranking method) for combining the effect of various indicators. While one cannot deny usefulness of these methods but most of these methods are having their own limitations.
The following table outlines the advantages and disadvantages of most widely used methods of analysis of developmental level analysis.
Ranking Method
Sum of ranks for all the socio-economic indicators of the unit is taken as the composite index of development.
Ranking procedure does not take into account the magnitude of differences between indicators and units.
Source: Author's research.
A survey of literature on measurement of the level of socio-economic development indicates that the majority of studies has been used principal component analysis approach.
Principal component analysis (PCA) is a mathematical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components (Davis, 1986).The number of principal components is less than or equal to the number of original variables. This transformation is defined in such a way that the first principal component has the largest possible variance (that is, accounts for as much of the variability in the data as possible), and each succeeding component in turn has the highest variance possible under the constraint that it be orthogonal to the preceding components (Cattel, 1966). PCA was invented in 1909 by Karl Pearson and H.Hotelling (1993) has been advanced this method. Now it is mostly used as a tool in exploratory data analysis and formatting predictive models.
According to the principal component featuresand the component index value we use this formula to calculate composite score: Composite Z-score = principal components variance contributionrate *principal component coefficients Assign each of the Z-Scores to a score in the range 0 to 1 by mapping to the cumulative normal distribution.
The inter-district variations are grouped into four categories of less developed region, moderate developed region, developed region and highly developed region on the basis of natural break (Jenks) method.
Statistical data processing was conducted using SPSS software.
1.Preliminary data analysis using PCA
Principal components analysis (PCA) is justified by data set dimension (12 characteristicsfor the 14regional administrative divisions), all the 12 variables being quantitative continuous. Using PCA thedimensionality of data is reduced by creating principal components from the original variablesin the context of this study, principal components analysis is used in order to explorethe original data set and to select the appropriate variables used to identify a regional profileof economic development in Uzbekistan.
In order to verify the adequacy of data for a factorial analysis, the Barlett's test ofsphericity (to test the null hypothesis that the variables in the correlation matrix of the populationare uncorrelated), and the indicator MSA (Measure of Sampling Adequacy) of Kaiser-Meyer-Olkin (to evaluate in which degree each variable may be predicted by all the othervariables) were used.The results obtained by data processing with SPSS are presented in Table 1. The significance level associated to Barlett's test of sphericity, Sig 0.000 , is smaller than 0.05(conventional value), which means the null hypothesis of variables' uncorrelation is rejected.
Table2. KMO and Bartlett's Test
Therefore one can conclude that the considered variables are adequate for a PCA. The value of the indicator MSA of KMO(0.71), greater than 0.5, also indicate the suitability of the considered data for factor analysis (Richarme, 2001). Table 3 represented the varimax rotated factor structure and majority of the variables under study have been appropriately focused on the structure exposes by this factor matrix. The socio-economic communalities value varied from 0.55 for the percentage of household by availability of gas to 0.92 for the services per capita. Others remaining social and economic indicators are suitably represented in the form of two extracted factors. Two factors meet not only the eigenvalue criterion, but also the variance The first factor has a high positive factor loading on variables: Industrial products per capita, total per capita income, the number of enterprises per 1,000 population, percentage of household by availability of gas, services per capita, export per capita. This means that it positively correlates to the respective characteristics oflocal government units. The first factor has a high negative factor loading on "Unemployment rate".
The second factor has a high positive factor loading on the "Number of students per teacher with higher education in secondary school", "Persons employed in trade, transportation and storage, housing and food services as percentage of total employment", "Percentage of household having vehicles", a negative factor loading on the "Poverty rate" and "Persons employed in agriculture, hunting and fisheries as percentage of total".
2.Regional disparities in the level of development in Uzbekistan
By using the method of the composite index (CI) the level of development in terms of socio-economic development in Uzbekistan has been evaluated. The relation between the value of composite score and the level of development is direct for instance regional administrative divisions. with the greater value of composite score have the advanced level of development and the state having lesser value recognize the lower level of improvement.
Table 4. Composite index value matrix of socio-economic indicators
Composite Level of Name of Index development regional administrative divisions 0.33 -0.
Very high developed Region:
Tashkent city, since the capital of the country presents very high characteristics of socio economic development compared to other administrativeterritorial units. Tashkent explains mostly the main characteristics of this county, basically urban. Poverty rate, total per capita income, export per capita, industrial products per capita, services per capita, the number of enterprises per 1,000 population, percentage of household by availability of gas, percentage of household having vehicles, unemployment rate and so on is experienced very high level of development in this area. Tashkent has the highest number and the most important medical and educational centers.
High developed Region:
The region having the composite index between 0.52 and 0.64have been included in this category. They are Tashkent and Navoi regions. An important role must play the neighboring of the Tashkent region, as it is near the capital city-Tashkent. But this influence is not as much as strong and it explains the lower socio economic development rate than Navoi region and lower rate of some socio economic indicators than other regions. With the superior circumstance in terms of industrial products per capita, high level of Export per capita, total per capita income, Services per capita, percentage of household by availability of gas, lower level of unemployment rate,number of students per teacher with higher education in secondary school, Navoi measured as developed region. However in terms of certain socioeconomical factors like poverty rate, the morbidity of the population by main classes of diseases Navoi has justified as a moderate region.
Moderate developed Region:
The underlying region of this zone is Fergana(0,42), Khorezm(0,43), Samarkand(0,45), Andijan(0,53), Bukhara(0,53). Samarkand and Fergana have a lower rate of poverty, but with low level of industrial products per capita and high level of unemployment rate have restraint the development. Khorezmis based on agricultural productivity where percentage of population employed in agriculture, hunting and fisheries is higher and total per capita income is higher even though there is low level of industrial products per capita, export per capita and the number of enterprises per 1,000 population. Other socio economic indicators like poverty rate, unemployment rate, services per capita, percentage of household by availability of gas, percentage of household having vehicle is moderate. At the same time, Andijan and Bukhara also practiced high level of total per capita income, services per capita and lower level of poverty rate. On the other hand Andijanis based on industrial and as well as agricultural activities where the number of enterprises per 1,000 population and persons employed in agriculture, hunting and fisheries as percentage of total, is very high that help in the economic development of the region. But high level of unemployment rate, morbidity of the population by main classes of diseases, number of students per teacher with higher education in secondary school are the reason behind to this moderate development. Republic of Karakalpakstan, Surkhandarya, Kashkadarya, Namangan, Jizzakhand Syrdaryabelong to this low developed region. All the selected socio economic indicators experienced lower development in Republic of Karakalpakstan, Surkhandarya, Kashkadarya, Namanganlike percapita income, poverty rate, services per capita, unemployment rate, number of enterprises per 1,000 population etc. But Republic of Karakalpakstanhas a high percentage of household by availability of gas and lower number of students per teacher with higher education in secondary school. Syrdarya has higher level of poverty and unemployment rate, and lower level of persons employed in trade, transportation and storage, housing and food services as percentage of total employment even though there are a large number of enterprises per 1,000 population. On the other hand in Jizzakhgood indicators of socio economic development is prevailed like lower level of unemployment rate, large number of enterprises per 1,000 population, high percentage of household by availability of gas, less number of Students per Teacher with higher education in secondary school, the lowest level of morbidity of the population by main classes of diseases. But higher level of poverty rate, less industrial products per capita and export per capita are the reason behind to this low development.
CONCLUSION
In the study, we have measured the development levels of different regional administrative division of Uzbekistan applying the composite index based upon optimum combination of selected socio-economic development indicators. The association between developments of different sectors of the economy is assessed and the regional administrative divisions are ranked precisely according to their levels of socioeconomic development. The level of development is assessed overall socio-economic fields. All 14regional administrative divisions have been included in the study and classified into four development categories according to the values of the composite indices.
The results show that wide disparities in the level of socio-economic development exist among different regions of Uzbekistan. The level of development in trade and services is found to be positively and statistically significantly associated with the overall socio-economic development indicating that the growth and progress of these sectors have been going hand in hand in the country. The results show that t in low development regions of Uzbekistan the level of industrial development does not significantly influence the agricultural and overall socio-economic development while agricultural development influences overall socio-economic development. It is noticed that both industrial and agricultural development have a significant bearing on overall socio-economic development in the region. Low developed regions are poorly developed in agriculture and service as well. | 2019-09-17T02:40:21.806Z | 2019-08-30T00:00:00.000 | {
"year": 2020,
"sha1": "1310150d5f19734cc0f53e78ef40d1cfeed55f0b",
"oa_license": null,
"oa_url": "http://www.t-science.org/arxivDOI/2019/08-76/PDF/08-76-15.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "a859ba0fc91cfc8587c48033b0cfe782043b88fd",
"s2fieldsofstudy": [
"Economics",
"Geography"
],
"extfieldsofstudy": [
"Geography"
]
} |
7246133 | pes2o/s2orc | v3-fos-license | Geographic correlation between mortality from primary hepatic carcinoma and prevalence of hepatitis B surface antigen in Greece.
Average annual age-adjusted mortality rates per 100,000 from primary hepatic carcinoma (PHC) among males for 1971-1973 in the urban and rural areas of the 9 geographical regions of Greece were estimated. Hepatitis-B surface antigen (HBsAg) prevalence by region and area was evaluated in a sample of 22,844 Greek Air Force recruits from all parts of the country. Mortality from PHC was found significantly higher in urban areas (28-30 vs. 18-81) whereas prevalence of HBsAg was higher in rural areas (5-3% vs. 3-90%). Nevertheless further statistical analysis showed that there is a strong correlation between HBsAg prevalence and mortality from PHC, which is higher in rural (r = + 0-88) than in urban (+ 0-57) areas. The latter findings indicate that hepatitis B infection and PHC may be causally related.
showed that there is a strong correlation between HBsAg prevalence and mortality from PHC, which is higher in rural (r --+ 0.88) than in urban (+ 0.57) areas. The latter findings indicate that hepatitis B infection and PHC may be causally related.
THE association between hepatitis-B virus (HBV) infection and primary hepatic carcinoma (PHC) is supported by observations on the natural history of chronic liver disease (Sherlock et al., 1970;Hadziyannis, Merikas and Afroudakis, 1970) as well as by findings of studies showing a greater than expected prevalence of hepatitis-B surface antigen (HBsAg) (Vogel et al., 1972;Blumberg et al., 1975) and antibody to hepatitis-B core antigen in patients with PHC (Maupas et at., 1975).
Although this association seems well documented, at least in areas of the world with high incidence of PHC, there is some doubt about the causal nature of the relationship. Indeed the association could be secondary, or the PHC and the frequently related chronic liver disease could facilitate the establishment of persistent antigenaemia (Kumar and Taylor, 1973;Linsell, 1975;Coady, 1975).
Population correlation between pre-valence of HBsAg and incidence of PHC could be considered as further evidence of the aetiological significance of HBV infection in the development of PHC. Indeed it was found that the prevalence of HBsAg is high in the general population of areas such as Africa and Greece where PHC is very common (British Medical Journal, 1975;Trichopoulos et al., 1975). However, the validity of this international population correlation is questionable, since the compared populations are very different in many aspects and the data concerning the prevalence of HBsAg and the incidence of PHC are not strictly comparable. Clearly geographical correlations based on homogeneous population groups and comparable data would have greater value. It is of interest that there has been only one study of this nature, in the Muranga district of Kenya, based on a relatively small sample, and this has shown no significant difference of HBsAg prevalence between areas of high and low incidence of PHC (Bagshawe et al., 1975). The present study was designed to investigate the geographical correlation between HBsAg prevalence and PHC mortality in the various regions of Greece.
SUBJECTS AND METHODS
Prevalence of HBsAg.-Between 1971 and1975 22,844 Greek Air Force recruits were screened for HBsAg at the time of their enlistment. Their age varied between 18 and 22 years. The recruits were classified according to their permanent residence in urban (cities or towns of more than 10,000 inhabitants) or rural (villages or towns of less than 10,000 inhabitants) areas of the nine main geographical regions of Greece.
The average annual mortality rate from PHC per 100,000 male population was 23-12. It was significantly higher in n, Average Annual Mortality from PHC among Males (1971of HBsAg among Greek Air Force Recruits (1971)(1972)(1973)(1974)(1975) The Table shows the average ageadjusted annual mortality rates from PHR among males in the urban and rural areas of the 9 geographic regions of Greece. It shows also the prevalence of HBsAg among Greek Air Force recruits by place of permanent residence as above.
The correlation coefficient between the two variables is +0-88 in rural and +0 57 in urban areas with a common estimate of +0-76, which is again significantly different from zero (P < 0-001).
If the urban-rural difference in the prevalence of HBsAg is allowed for, the difference between the mean mortality from PHC in urban and rural areas is further increased. The adjusted difference (Armitage, 1971) in mortality from PHC between urban and rural areas is 10*99 ± 1-42 and differs significantly from zero (P < 0.001).
DISCUSSION
The high prevalence of HBsAg in Greece has been noted by several authors (Hadziyannis et at., 1973). It is of interest that this prevalence is higher in rural than in urban areas, a fact which could be attributed to the lower hygiene standards of the rural population (Cherubin, 1971).
The estimated annual mortality rate from PHC is much higher than that prevailing in other European countries (I.U. A.C., 1966).
There are certain limitations in the mortality data. In Greece the majority of PHC cases are diagnosed on clinical grounds, since autopsies are rarely performed and alpha-foetoprotein determinations are not routinely performed in most hospitals. It seems likely however that the possible errors are not related to the regional distribution of HBsAg and this substantially decreases the possibility that the association is not real. Therefore, the present study helps to establish that there is a geographical correlation between HBsAg prevalence and mortality from (and hence incidence of) PHC.
Many problems complicate the interpretation of correlation studies. Difficulties arise from the use of populations as sampling units, the long latent interval for most human cancers, and the presence of multiple aetiologic agents. They have recently been discussed by Breslow and Enstrom (1974). However, the demonstration of a population correlation between HBsAg and PHC helps to eliminate a number of alternative explanations for the association of these two factors in individuals.
The population correlation cannot be explained in terms of a tumour effect on the establishment of persistent hepatitis-B antigenaemia. The incidence of PHC and of the frequently associated cirrhosis are in general too low to affect the HBsAg prevalence. Also these diseases develop late in life, while in the present study HBsAg prevalence was estimated in 18-22-year-old males.
It is also less likely (although still possible) that confounding factors would explain the association between HBsAg and PHC in individuals as well as the international correlation and the geographical correlation within Greece. Furthermore, various infections which may affect the incidence of PHC as well as the immunological response to HBV (Dudley, Fox and Sherlock, 1972;Zuckerman, 1972;Coady, 1975) are not encountered in Greece as frequently as in Africa and other parts of the world (WHO Annual, 1975). This is also probably true for aflatoxin (Peers and Linsell, 1973), although sufficient data are not available for Greece. Peers and Linsell (1973) found a near-linear relation between PHC incidence and log aflatoxin intake in the Muranga district of Kenya, whereas Bagshawe et al. (1975) failed to detect a significant difference in HiBsAg prevalence between areas with contrasting incidence of PHC. These findings are not incompatible with the results of the present study. PHC is probably a disease of multifactorial origin (Higginson and Svoboda, 1970) and it is conceivable that a particular pattern of the disease is attributable to one factor, whereas another pattern is explained in terms of another factor. It may also be noted that Bagshawe et al. (1975)
did find a higher
HBsAg prevalence in the areas with higher incidence of PHC although the difference was not statistically significant (3.6% VS. 27 %).
In Greece since the last war a strong wave of internal migration from the villages towards the cities has increased the proportion of urban population from 32.8o% in 194032.8o% in to 53.2% in 197232.8o% in (Trichopoulos et al., 1974. This fact and the probably long latent period of PHC may help to explain the higher correlation found in rural areas, where the population is more stable and to a higher proportion locally born. Regression lines of mortality from PHC on prevalence of HBsAg in urban and rural areas of Greece are parallel, but they differ by about 11 deaths annually per 100,000 population. This implies that irrespective of HBsAg prevalence the mortality from PHC in urban areas is higher than in rural by about 60%. This difference cannot be entirely explained in terms of the better diagnostic facilities available in the towns, since, as already mentioned, most cancer patients are hospitalized in major cities. It would appear, therefore, that in Greece and particularly in urban areas additional factors of aetiologic importance are present and they are partly responsible for the excessively high incidence of PHC in this country. This study was supported by a grant from the Greek Ministry of Social Services. | 2017-11-08T17:45:43.453Z | 1976-07-01T00:00:00.000 | {
"year": 1976,
"sha1": "a35af6ac9d6b330a796efd5488955c21f29ce55a",
"oa_license": "implied-oa",
"oa_url": "https://europepmc.org/articles/pmc2025123?pdf=render",
"oa_status": "GREEN",
"pdf_src": "PubMedCentral",
"pdf_hash": "a35af6ac9d6b330a796efd5488955c21f29ce55a",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
212825757 | pes2o/s2orc | v3-fos-license | Assessing Relativistic E ff ects and Electron Correlation in the Actinide Metals Th to Pu
: Density functional theory (DFT) calculations are employed to explore and assess the e ff ects of the relativistic spin–orbit interaction and electron correlations in the actinide elements. Specifically, we address electron correlations in terms of an intra-atomic Coulomb interaction with a Hubbard U parameter (DFT + U ). Contrary to recent beliefs, we show that for the ground-state properties of the light actinide elements Th to Pu, the DFT + U makes its best predictions for U = 0. Actually, our modeling suggests that the most popular DFT + U formulation leads to the wrong ground-state phase for plutonium. Instead, extending DFT and the generalized gradient approximation (GGA) with orbital–orbital interaction (orbital polarization; OP) is the most accurate approach. We believe the confusion in the literature on the subject mostly originates from incorrectly accounting for the spin–orbit (SO) interaction for the p 1 / 2 state, which is not treated in any of the widely used pseudopotential plane-wave codes. Here, we show that for the actinides it su ffi ces to simply discard the SO coupling for the p states for excellent accuracy. We thus describe a formalism within the projector-augmented-wave (PAW) scheme that allows for spin–orbit coupling, orbital polarization, and non-collinear magnetism, while retaining an e ffi cient calculation of Hellmann–Feynman forces. We present results of the ground-state phases of all the light actinide metals (Th to Pu). Furthermore, we conclude that the contribution from OP is generally small, but substantial in plutonium.
Introduction
In the last several decades, various computational techniques to calculate the electronic structure and ground-state properties of actinides have been suggested in the literature . In these studies, a comparison of the calculated equilibrium volumes with experiments has often served the purpose to evaluate the quality of the modeling and the underlying physics of the 5f electrons. From these works a consistent picture emerges for thorium through neptunium where the 5f electrons can be described as bonding and delocalized (band) electrons. The implication is that the actinide volumes decrease in a parabolic fashion with atomic number for Th to Pu, reminiscent of the non-magnetic d-transition metals. The physical interpretation, given by Friedel [27], is that for the first half of the series, bonding electrons contract the volumes while for the second half, anti-bonding electrons expand the lattice. The parabolic behavior is readily verified by assuming a rectangular electron density of states and the corresponding electron-energy dependence on the volume [27].
Plutonium, on the other hand, deviates slightly from this simple trend. In this case, relativistic effects [4], crystal structure [8,10], and possible electron correlation effects beyond the local density approximation [9] are suggested as explanations for some of the discrepancy noted between theory and experiments. However, a much-improved description of plutonium is obtained, still within the 5f-band picture, when magnetic effects are addressed [13]. Acknowledging these magnetic interactions
Computational Methods
We show results derived from density functional theory that was developed in the mid-1960s [42,43]. Density functional theory is actually correct, but it relies on practical approximations. The central assumption relates to the electrons exchange and correlation interactions. For the actinide metals, we employ the generalized gradient approximation that is more proper than the previously suggested local density approximation [6]. Specifically, we are applying the Perdew-Burke-Ernzerhof (PBE) parameterization of the generalized gradient approximation (GGA) [44]. To our knowledge, there is no established, parameter-free, functional that results in larger atomic volumes in the actinides than that of the GGA. Therefore, there is no choice to replace the GGA to expand the volumes, while a +U approach can accomplish this objective and has often been applied for the actinides, as discussed in the introduction.
The actinide metals exhibit interesting crystal structures, beginning with an innocent face-centered cubic for Th with a gradual escalation of complexity to tetragonal (Pa), orthorhombic (U and Np), and finally culminating with a monoclinic structure (α-plutonium). These structures are described in the compilation by Donohue [45]. For best possible accuracy, when calculating the equilibrium volumes, we have optimized these crystal structures (not the cubic phase) so that the model gives the lowest total energy (structural relaxation). This relaxation has generally a rather weak effect on the actinide volumes, except for α-plutonium where it is significant. In this instance, a structural relaxation contracts the volume substantially relative to a calculation where the experimental (room-temperature) structure is assumed [40,46].
It has been reported that orbital-orbital interactions, absent in conventional DFT, is important for plutonium [14,40,47]. This interaction is efficiently included in an orbital-polarization scheme for some of the calculations presented in this study. In terms of the technical details of the computations, we are applying three techniques. The bulk of the calculations are performed with the pseudopotential plane-wave Vienna ab initio simulation package (VASP) code.
In VASP, the computational scheme relies on the GGA with PBE exchange and correlation functional [44] and the projector-augmented-wave (PAW) method with plane-wave basis set as implemented in VASP [48][49][50]. The computational parameters consist of an energy cut-off of 450 eV, energy convergence of 100 eV, and a Monkhorst-Pack scheme k-point mesh of typically 12 × 12 × 12 for a 1-atom cell and correspondingly less for multi-atom cells. In the cases where we perform DFT + U computations, we employ the technique suggested by Dudarev et al. [51] for the 5f electrons with an effective U = 0.4 eV.
The spin-orbit interaction and orbital polarization are included, allowing for non-collinear spins. For the spin-orbit coupling, we have introduced a correction that retains the scalar-relativistic treatment for the p states. Orbital polarization has been implemented similarly to the scheme proposed by Eriksson et al. [52]. In this approach, the DFT total-energy functional, E DFT , is amended by a quadratic term favoring orbital polarization, where E 3 I is the Racah parameter that is a linear combination of the radial Slater integrals F 2 , F 4 , and F 6 over f-electron orbitals within a muffin-tin sphere centered on site I. For consistency and simplicity, we use a constant Racah parameter that is an average of spin-decomposed values obtained from all-electron calculations (see below). L f I,z is the z component of the angular-momentum operator within the same sphere operating on the f states (5f in this case). This formulation of the orbital-polarization Hamiltonian has mainly been implemented in conjunction with spin-orbit coupling in all-electron codes within the second-variation approximation assuming collinear spin polarization.
Analogous to the conventional formulation of the orbital polarization (OP) method, while having the computational expediency of plane-wave codes and accurate Hellman-Feynman forces, we make use of a variational total-energy formulation of the OP Hamiltonian in the context of non-collinear magnetism. It is easily constructed within the PAW method following the DFT + U implementation by Bengone et al. [53]. See Appendix A for a brief technical description of the PAW method for the purpose of implementation of the orbital-polarization Hamiltonian.
As mentioned, Equation (1) is proposed for collinear magnetism with the spin moment pointing in the z direction. This formalism can be generalized to non-collinear magnetism with arbitrary spin orientations by modifying the quadratic energy functional within each muffin-tin radius to favor orbital polarization in the direction of the expectation value of the local self-consistent spin vector. It is then straightforward to propose the following total-energy functional, consistent with a non-collinear generalization of Equation (1), where L In addition to our plane-wave calculations, we apply two all-electron methods to compare our pseudopotential results with. First, a full-potential linear muffin-tin orbital (FPLMTO) method [54] that has been detailed [55]. Here, no approximations are applied for the core electrons, unlike the pseudopotential method. The core-electron approximation made in the latter technique is computationally expedient but will cause inaccuracies that are not present in all-electron calculations.
The present implementation does not make any assumptions beyond the GGA. Basis functions, electron densities, and potentials are calculated without any geometrical approximation, and these are expanded in spherical harmonics (with a cut-off L max = 8) inside non-overlapping (muffin-tin, MT) spheres surrounding each atom and in Fourier series in the region between these muffin-tin spheres. One has to define an MT sphere with a radius, s MT , and here it is chosen so that s MT /s WS~0 .8, where s WS is the Wigner-Seitz (atomic sphere) radius. The radial parts of the basis functions inside the MT spheres are calculated from a wave equation for the L = 0 component of the potential that includes all relativistic corrections including spin-orbit coupling for d and f states, but not for the p states [20]. It should be noted that excluding the p states also removes the strong and artificial dependence of the calculated volumes on the s MT /s WS ratio [2] that is also not present in the exact relativistic treatment (see below).
The orbital polarization is only considered for d and f states. In the actinides, however, the OP contribution from the d states is minimal and ignored in the VASP calculations. The formulation is self-consistent and parameter-free and given in [52]. The self-consistent FPLMTO Racah parameters for the 5f states are employed in the VASP calculations described above. All crystal-structure geometries were relaxed and for neptunium and plutonium, that have more complex phases, we apply a scheme [56] that avoids the necessity of Hellman-Feynman forces that are troublesome to compute.
The number of k points included in the calculations depends on the particular crystal structure, but we generally use~1000 k points or more for one atom/cell calculations and fewer for cells with many atoms. Each energy eigenvalue is broadened with a Gaussian having a width of 20 mRy.
Second, we employ a full-potential linear augmented plane-wave (FPLAPW) method as implemented in FlapwMBPT (many body perturbation theory) [57]. This approach has many similarities with the FPLMTO method introduced above. As in the FPLMTO method, the space is divided into non-overlapping MT spheres and an interstitial region. The electronic density and effective potential are expanded in spherical harmonics in the MT spheres and in plane waves in the interstitial region (exactly as in the FPLMTO method).
One conceptional difference between FLAPW and the other two methods is that orbital polarization is not applied in the FLAPW calculations. For this reason, Th, Pa, and U were studied with FLAPW, but not Np and Pu where orbital polarization is more important. Another conceptional difference is that FlapwMBPT allows for a complete inclusion of relativistic effects through the four-spinor Dirac equation. Particularly important is that the spin-orbit interaction is here formulated exactly.
Other technical differences between FLAPW and FPLMTO are mostly related to the use of different basis functions in the interstitial region (plane-waves and Hankel functions, respectively) and in the augmentation of them with the functions inside the MT spheres. Details of the implementation of the FLAPW method can be found in [58]. Specific values of the different cut-offs used in the FLAPW calculations are summarized in Table 1. The crystal-structure parameters given in Table 1 are obtained from structural relaxations using the FPLMTO method.
Results
The formally correct way to include spin-orbit interaction in an electronic-structure calculation is to solve the four-spinor Dirac equation. Here, we have applied a FPLAPW method that indeed does this (FlapwMBPT [57]), and we regard the corresponding results as the gold standard for our comparisons. We mentioned [20] that utilizing a scalar-relativistic basis set causes inaccuracies for the p states, but it has been suggested that this problem can be circumvented for the actinides by ignoring spin-orbit interaction for these states [20]. This realization has not been widely recognized in studies using plane-wave pseudopotential techniques, and we therefore reiterate this fact by showing our results in Table 2. Table 2 presents the effect of spin-orbit coupling on the volumes of the actinide elements. For Th, Pa, and U, FPLAW-Dirac calculations provide the reference, while for Np and Pu, that have substantially more complex and large crystal structures, we show results from VASP and FPLMTO. For VASP and FPLMTO the perturbative form of SO is included but for the p states. All three methods give essentially identical results for Th, Pa, and U, suggesting that ignoring spin-orbit interaction for the p states is a very good approximation. The same conclusion was reached in previous studies [59,60] where comparisons were made with the exact muffin-tin orbital method that also solves the four-spinor Dirac equation.
The trend is that spin-orbit coupling is of increasing importance proceeding from thorium to neptunium, causing an expansion of the volumes ( Table 2). These metals are all non-magnetic and the explanation for this behavior is straightforward. As mentioned above, the atomic volumes for Th to Np decrease with atomic number in roughly a parabolic fashion that is due to a gradual population of bonding 5f states. No anti-bonding 5f electrons exist because Th to Np all have less than half-filled 5f band. Inclusion of spin-orbit interaction, however, causes some anti-bonding 5f states to be occupied, and these tend to expand the atomic volumes. The reason is that anti-bonding states mix with bonding states as the 5f band is divided into overlapping 5f 5/2 and 5f 7/2 sub-bands that comprise both bonding and anti-bonding states. The result of this is a reduction of the effective number of bonding electrons and an increase in the atomic volumes (Table 2).
For plutonium, on the other hand, the situation becomes more complicated because here both spin and orbital polarization play important roles as well. In addition, for α-plutonium the crystal and electronic structure correlate sensitively so that the atomic volume cannot be obtained simply from extrapolating the trend established by Th to Np. For plutonium, one must thus carry out accurate electronic-structure calculations, recognizing the magnetic interactions and relaxing the crystal structure, to predict the atomic volume accurately.
Next, we discuss the theory relative to experimental data. In order to do this, we need to establish the experimental volumes at zero temperature to enable a proper comparison with the zero-temperature modeling. Thankfully, Söderlind and Young [61] carefully determined these experimental zero-temperature volumes (V 0 ) for Th, Pa, and U (32.55, 24.72, and 20.47 Å 3 , respectively). For neptunium we use the experimental linear coefficient of thermal expansion (27 × 10 −6 K −1 ) [62] combined with the quasi-harmonic Debye-Grüneisen model [4] to back out V 0 (18.9 Å 3 ). For α-plutonium the Debye-Grüneisen procedure is inaccurate because the temperature dependence of the volume is highly anomalous at low temperatures. At temperatures close to 50 K, there is negative thermal expansion observed in dilatation experiments [63]. This anomaly was proposed to be due to magnetism [63], and this speculation is plausible because the negative thermal expansion observed in δ-plutonium is explained from first-principles theory by magnetic interactions [40]. Here, we extract V 0 (19.53 Å 3 ) for α-plutonium from a linear extrapolation, from the three lowest temperatures, to zero temperature of the dilatation data shown in Figure 1. We compare our VASP results, that include spin-orbit interaction and orbital polarization, with the experimental V 0 values in Figure 2. The agreement is very good and almost as good as all-electron (FPLMTO) data that also include spin-orbit coupling and orbital polarization [24]. The small differences are likely due to the PAW method used in VASP.
Appl. Sci. 2019, 9, x FOR PEER REVIEW 6 of 14 of the volume is highly anomalous at low temperatures. At temperatures close to 50 K, there is negative thermal expansion observed in dilatation experiments [63]. This anomaly was proposed to be due to magnetism [63], and this speculation is plausible because the negative thermal expansion observed in δ-plutonium is explained from first-principles theory by magnetic interactions [40]. Here, we extract V0 (19.53 Å 3 ) for α-plutonium from a linear extrapolation, from the three lowest temperatures, to zero temperature of the dilatation data shown in Figure 1. We compare our VASP results, that include spin-orbit interaction and orbital polarization, with the experimental V0 values in Figure 2. The agreement is very good and almost as good as all-electron (FPLMTO) data that also include spin-orbit coupling and orbital polarization [24]. The small differences are likely due to the PAW method used in VASP. Figure 1. Experimental temperature dependence of the atomic volume for α-plutonium. Redrawn from [63] and assuming that the room-temperature (300 K) atomic volume is 20.0 Å 3 . The 0 K value is obtained from a linear extrapolation from the three lowest temperatures (red line).
Figure 1.
Experimental temperature dependence of the atomic volume for α-plutonium. Redrawn from [63] and assuming that the room-temperature (300 K) atomic volume is 20.0 Å 3 . The 0 K value is obtained from a linear extrapolation from the three lowest temperatures (red line). We assess the pseudopotential approximation in Figure 3 where scalar-relativistic VASP and FPLMTO atomic volumes are compiled. Notice that the all-electron FPLMTO volumes are slightly larger, as are the FPLAPW volumes (not shown), but overall the agreement is quite good. For αplutonium the difference is more significant but, in this case, when spin-orbit and orbital polarization are included, the agreement between the methods improves. In our judgement, the VASP pseudopotential is very reasonable for Th to Pu. We assess the pseudopotential approximation in Figure 3 where scalar-relativistic VASP and FPLMTO atomic volumes are compiled. Notice that the all-electron FPLMTO volumes are slightly larger, as are the FPLAPW volumes (not shown), but overall the agreement is quite good. For α-plutonium the difference is more significant but, in this case, when spin-orbit and orbital polarization are included, the agreement between the methods improves. In our judgement, the VASP pseudopotential is very reasonable for Th to Pu.
We thus have a model (DFT + SO + OP), that reproduces the equilibrium volumes for the actinides very accurately. The parabolic behavior of the volumes is duplicated as well as the upturn observed between Np and Pu. The upturn for plutonium is due to several effects: Crystal structure, spin-orbit coupling, orbital polarization, and formation of magnetic moments. The present model does not depend on an ad hoc Hubbard U parameter. Actually, introducing such a parameter in a DFT + U approach invokes physics that is inappropriate for these metals, unless the parameter is very small and inconsequential. To illustrate this, we performed VASP DFT + U calculations that include spin-orbit interaction with the proper exclusion of the p states, as discussed. In Figure 4 we show these results that suggest an over-estimation of all the volumes, except for thorium. Accounting for the fact that the pseudopotential approximation tends to slightly under-estimate the volume (Figure 3), the conclusion from Figure 4 is that the DFT + U volumes are decidedly too large and worse than those of the treatment without a U (Figure 2). The overall effect of the applied U is here relatively small, cf. Figures 2 and 4, because we investigate a very small U (0.4 eV). Using the more common and larger value of U (~1 eV) exacerbates this over-estimation problem for the volumes, while adopting a larger U (>4 eV), often applied for δ-plutonium, causes wholly unrealistic volumes for all the actinides (including δ-plutonium). We thus have a model (DFT + SO + OP), that reproduces the equilibrium volumes for the actinides very accurately. The parabolic behavior of the volumes is duplicated as well as the upturn observed between Np and Pu. The upturn for plutonium is due to several effects: Crystal structure, spin-orbit coupling, orbital polarization, and formation of magnetic moments. The present model does not depend on an ad hoc Hubbard U parameter. Actually, introducing such a parameter in a DFT + U approach invokes physics that is inappropriate for these metals, unless the parameter is very small and inconsequential. To illustrate this, we performed VASP DFT + U calculations that include spin-orbit interaction with the proper exclusion of the p states, as discussed. In Figure 4 we show these results that suggest an over-estimation of all the volumes, except for thorium. Accounting for the fact that the pseudopotential approximation tends to slightly under-estimate the volume ( Figure 3), the conclusion from Figure 4 is that the DFT + U volumes are decidedly too large and worse than those of the treatment without a U (Figure 2). The overall effect of the applied U is here relatively small, cf. Figures 2 and 4, because we investigate a very small U (0.4 eV). Using the more common and larger value of U (~ 1 eV) exacerbates this over-estimation problem for the volumes, while adopting a larger U (>4 eV), often applied for δ-plutonium, causes wholly unrealistic volumes for all the actinides (including δ-plutonium) . Another very obvious sign of the improper physics that the Hubbard U parameter enforces is found in the prediction of the ground-state phases. For plutonium, DFT + U predicts the wrong ground state, even for the very small U we have adopted in this study (0.4 eV). In Figure 5 we show our fully (both structural and magnetic-configurational) relaxed DFT + U calculations for α-and δplutonium. Notice that the δ-phase energy is about 7 mRy (per atom) lower than the α phase. The fact that the δ phase is predicted to be the ground state when a large U (~ 4 eV) is employed has been discussed [40], but it is remarkable that even a very small U promotes the wrong ground state. Actually, the two phases are energetically degenerate for a small U ~ 0.2 eV. Here, both the α and δ volumes are significantly under-estimated (18.74 and 24.25 Å 3 , respectively). We find that increasing U aggravates the problem and further increases the stability of δ over α (not shown). We emphasize that a theory assuming no ad hoc Coulomb repulsion (U = 0) but orbital polarization and careful Another very obvious sign of the improper physics that the Hubbard U parameter enforces is found in the prediction of the ground-state phases. For plutonium, DFT + U predicts the wrong ground state, even for the very small U we have adopted in this study (0.4 eV). In Figure 5 we show our fully (both structural and magnetic-configurational) relaxed DFT + U calculations for αand δ-plutonium. Notice that the δ-phase energy is about 7 mRy (per atom) lower than the α phase. The fact that the δ phase is predicted to be the ground state when a large U (~4 eV) is employed has been discussed [40], but it is remarkable that even a very small U promotes the wrong ground state. Actually, the two phases are energetically degenerate for a small U~0.2 eV. Here, both the α and δ volumes are significantly under-estimated (18.74 and 24.25 Å 3 , respectively). We find that increasing U aggravates the problem and further increases the stability of δ over α (not shown). We emphasize that a theory assuming no ad hoc Coulomb repulsion (U = 0) but orbital polarization and careful relaxation of all degrees of freedom reproduces the correct ground state in accordance with previous conclusions [18,40].
Summary and Conclusions
We conducted DFT and DFT + U calculations for the purpose of evaluating the potential need for an explicit Hubbard U intra-atomic Coulomb repulsion for the actinide metals Th to Pu. We limit our study to the α phases of these metals (except for δ-plutonium) and find that one carefully needs to address several key approximations in the theory in order to reach a definite conclusion.
First, the spin-orbit interaction has to be introduced in an appropriate fashion. Based on comparisons with Dirac-FPLAPW calculations that include spin-orbit interaction exactly, we recommend a way to accurately account for SO interaction without having to solve the spin-polarized Dirac equation. Namely, by excluding this interaction on the p states, one can accurately apply the usual scalar-relativistic Schrödinger-type equation with perturbative spin-orbit coupling [64] as implemented in VASP and other codes.
Second, one has to make sure that the pseudopotential employed in plane-wave methods is a good representation of the core electrons. We have confirmed that this is the case here by comparisons with results from all-electron codes. The most accurate result is obviously obtained from all-electron approaches, however.
Third, the crystal structures need to be relaxed (i.e., optimized so that the lowest energy is obtained). This is particularly important for α-plutonium because of the sensitive correlation between its crystal and electronic structure. The plutonium computations we have performed are thus nontrivial because forces are required for structural relaxations and these are troublesome to evaluate in
Summary and Conclusions
We conducted DFT and DFT + U calculations for the purpose of evaluating the potential need for an explicit Hubbard U intra-atomic Coulomb repulsion for the actinide metals Th to Pu. We limit our study to the α phases of these metals (except for δ-plutonium) and find that one carefully needs to address several key approximations in the theory in order to reach a definite conclusion.
First, the spin-orbit interaction has to be introduced in an appropriate fashion. Based on comparisons with Dirac-FPLAPW calculations that include spin-orbit interaction exactly, we recommend a way to accurately account for SO interaction without having to solve the spin-polarized Dirac equation. Namely, by excluding this interaction on the p states, one can accurately apply the usual scalar-relativistic Schrödinger-type equation with perturbative spin-orbit coupling [64] as implemented in VASP and other codes.
Second, one has to make sure that the pseudopotential employed in plane-wave methods is a good representation of the core electrons. We have confirmed that this is the case here by comparisons with results from all-electron codes. The most accurate result is obviously obtained from all-electron approaches, however.
Third, the crystal structures need to be relaxed (i.e., optimized so that the lowest energy is obtained). This is particularly important for α-plutonium because of the sensitive correlation between its crystal and electronic structure. The plutonium computations we have performed are thus non-trivial because forces are required for structural relaxations and these are troublesome to evaluate in the presence of strong spin-orbit and orbital-orbital coupling.
The conclusion is that DFT, with the GGA for the electron exchange and correlations, is sufficient for all studied actinide metals (no Hubbard U needed). Recently, numerous others came to the same conclusion [65][66][67][68] for uranium. For plutonium, however, treating the orbital-orbital interaction is necessary for a good electronic structure, total energy, and magnetic form factor [69]. Importantly, we show that DFT + U (U > 0.2 eV) for plutonium predicts the wrong ground-state phase. Because DFT + U defines the static limit of dynamical-mean field theory, that incorporates similar electron correlations, it seems unavoidable that it is also inappropriate for plutonium.
Author Contributions: B.S. and P.S. conceived the study. B.S., P.S., and A.K. performed electronic-structure calculations. All authors helped writing the paper.
Nevertheless, the PAW equations are derived under the assumption of Equation (A4). As a result, the expectation value of any local operatorB confined to the muffin-tin sphere around a nucleus can be simply expressed as follows Ψ kn |B|Ψ kn = ilm,i l m Ψ kn | p ilm φ ilm |B|φ i l m p i l m | Ψ kn . (A5) Finally, when dealing with non-collinear magnetism, the crystal wave functions acquire an extra dimension and become spinor-wave functions Ψ τ kn , with τ now indexing the two spinor components. We are now in a position to derive the expression for the expectation value of any component (α = x, y, z) of the local angular-momentum operator L f α that acts on the f electrons within an atomic sphere. It can be written as where σ ττ α is the Pauli spin matrix in the α direction. The corresponding self-consistent equations are constructed from functional differentiations of Equation (2) with respect to the pseudo spinor wave functions Ψ τ kn and occupations f k,n . | 2019-11-28T12:46:22.562Z | 2019-11-21T00:00:00.000 | {
"year": 2019,
"sha1": "ba2a86369db7bfe4bd2f6dcce14ad887e51dd783",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2076-3417/9/23/5020/pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "a49918e9c90fdc578a03fc76bce5a2e457a1fad3",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
80857277 | pes2o/s2orc | v3-fos-license | Reducing Lumbar Discogenic Back Pain and Disability with Intradiscal Injection of Bone Marrow Concentrate: 5-year Follow-up
Background Context: Surgical treatments of discogenic lumbar back pain are fusion or lumbar artificial disc replacement. Studies comparing lumbar fusion with nonsurgical treatment found no difference in clinical results. Studies comparing lumbar fusion with lumbar artificial disc replacement have had mixed results. In a study with 12-month follow-up, our colleagues reported that intradiscal injections of autologous bone marrow concentrated cells resulted in substantial reductions in pain and disability without treatment complications or other adverse events. Purpose: This article reports a 5-year follow-up of treating lumbar discogenic back pain and disability with bone marrow concentrate. Study Design/Setting: Prospective, open-label, single-center case series. Patient Sample: The initial 26 participants were all surgical candidates according to their history of low back pain, nonsurgical treatment, and measurements of pain, disability, and disc degeneration. Outcome Measures: Visual Analog Scale of pain, and Oswestry Disability Index. Methods: Study design and clinical protocol, bone marrow collection and processing, and intradiscal injection were as previously described in the initial report. The study did not receive any outside funding. The disposable aspiration kits for the BMC injections, which cost about $20 each, were provided without charge by Celling Biosciences, Austin, Texas. Results: Of the initial 26 participants, six proceeded to surgery within 3 years of follow up. Of the remaining 20, 19 were available for follow-up at 5 years. Absolute and percentage reductions in pain and disability scores were sustained through the 5-year follow-up. No adverse events were reported through the 5 years. Conclusions: It may be reasonable to consider injecting participants who have discogenic back pain at one or two levels with bone marrow concentrate before they proceed to surgery.
Introduction
A fact sheet from the United States Bone and Joint Initiative summarizing the burden of musculoskeletal diseases states that one in four adults in the United States suffers from chronic low back pain, with a cost of more than $250 billion in treatment and lost wages [1]. Surgical treatments of lumbar discogenic back pain are fusion or lumbar artificial disc replacement (LADR) [2,3,4]. A 2-year follow-up study [5] and a 4-year follow-up study [6] compared lumbar fusion with nonsurgical care. They reported no differences in clinical results. One study with a 2-year follow-up [7] concluded that LADR is as least as good as fusion. Another two studies with a 2-year follow-up [2,3] reported that LADR is superior to fusion. However, one of these studies continued to a 7-year follow-up [7,8]. This study concluded that outcomes after LADR were similar to those after fusion. A 5 to 10-year follow-up study [9] of LADR clinical results showed substantial improvements in VAS and ODI from baseline at all follow-ups. However, this study also reported that these improvements deteriorated from 48 months after LADR. Furthermore, complications occurred in 26 (14.4%) of the 181 participants who were followed-up.
In 2015, we [10] reported a 1-year follow-up about the use of intradiscal injections of autologous bone marrow concentrated cells (BMC) to treat moderate to severe discogenic back pain in one or two levels in 26 participants. In a single, 45-minute outpatient procedure done under conscious sedation, bone marrow cells were aspirated from the iliac crest, concentrated by centrifugation, and injected into the nucleus pulposus of one or two symptomatic, degenerated discs. The participants had had a history of least 6 months of low back pain and at least 3 months of nonsurgical treatment without resolution. The mean pretreatment pain score as measured by the 0 mm to 100 mm Visual Analog Scale (VAS) [11] was 79.3 mm, and the mean pretreatment disability score as measured by the 0% to 100% Oswestry Disability Index (ODI) [12,13] was 56.5%. They also had a grade of 4 or greater on the 8-grade modified Pfirrmann MRI [14] classification, indicating moderate to severe disc dehydration and degeneration. All were surgical candidates.
By 3 months, the mean VAS and ODI scores were 29.2 mm and 22.8%, statistically significant decreases sustained through 1 year. By the end of the year, 24 (92%) of the 26 participants had not proceed to surgery. Of these, 20 had an MRI to assess the change in Pfirrmann grade. Eight (40%) of these 20 had decreased degeneration in their treated discs as shown by a decrease of at least one Pfirrmann grade. The other 12 had no change.
Although participant age and gender had little effect on reductions in pain and disability, the concentration of mesenchymal stem cells did. At 3 and 6 months after injection, participants receiving of at least 2,000 colony forming units of mesenchymal stem cells per milliliter experienced substantially faster and greater reductions in VAS and ODI scores than those receiving fewer cells. However, the differences between these two groups were not statistically significantly different at 1 year.
We then reported on 2 and 3-year follow-ups [15,16]. By the end of the second and third years, the decreases in VAS and ODI scores that had been observed at the end of the first year were sustained in the participants who had not undergone surgery. No adverse events were reported through 3 years.
Between the first and second years, three participants underwent surgery. Between the second and third years, another participant underwent surgery. Thus, through 3 years of follow-up, 20 (77%) participants of the starting group of 26 avoided surgery.
Here we report on the findings of a 5-year follow-up.
Materials and Methods
Study materials and methods were as described initially [10]. All participants provided written informed consent according to the form provided by Western Institutional Review Board, Puyallup, Washington (clinical study protocol approval number 20120085). The $5,000 per participant cost of the BMC treatment and costs of subsequent follow-ups were provided at no charge. The study did not receive any outside funding. The disposable aspiration kits for the BMC injections, which cost about $20 each, were provided without charge by Celling Biosciences, Austin, Texas.
For the 5-year follow-up, the remaining participants underwent VAS and ODI scoring, and they were asked how they felt and about how often they took opioid pain relievers. The six participants who progressed to surgery after BMC injection were contacted about they felt and how often they took opioid pain relievers. The study did not receive any outside funding.
Participant Retention
Of the 20 participants who underwent BMC injection and had not proceeded to surgery, one could not be located and was lost to follow-up after the 3-year report, for 95% follow-up.
Participant-reported Outcome Measures
Absolute and percentage reductions in VAS and ODI scores were sustained through the 5-year follow-up ( Fig. 1 and Table 1, respectively). The 5-year mean VAS and ODI scores were both statistically significantly less than the 3-year scores. At the 5-year follow-up, eight (42%) of the 19 participants who had not proceeded to surgery reported that their back pain was either 90% reduced or to within pain and disability scores typically reported by adults ages 40 years and older unaffected by back pain, a VAS of 2 mm or less and an ODI of 10% or less [17]. From 3 years to 5 years, 5 of the 19 participants had increased disability as shown by increases in ODI scores from 3 to 5 years: 0% to 6%; 4% to 10%; 6% to 10%; 12% to 14%; and 14% to 22%. Mean (SD) VAS and ODI scores at pretreatment [10] and at 3-year [16] and 5-year (this article) follow-ups (n = 19). SD, standard deviation; VAS, Visual Analog Scale; ODI, Oswestry Disability Index [1]. [12,13]. The 100% values used to compute these percentage decreases were taken from data to calculate the mean values on Fig. 1 At the end of 5 years, four of the six participants who had proceeded to surgery reported taking opioid pain relievers frequently (three to four times weekly), and two of the 19 who did not proceed to surgery reported taking opioids occasionally (three to four times monthly). The other participants who did not proceed to surgery reported no use of opioids.
Adverse Events
As previously reported [16], there were no adverse events through 3-year follow-up. For the 19 participants who underwent BMC injection and were evaluated at the 5-year follow-up, no adverse events were reported between the 3-year and 5-year follow-ups.
Proceeding to Surgery
As previously reported [16], six of the original 26 participants proceeded to surgery through 3 years of follow-up. With the possible exception of the one participant who was lost to 5-year follow-up, no additional participants proceeded to surgery after the 3-year follow-up. Through 5 years, two of the six participants who proceeded to surgery after BMC injection had good surgical outcomes in that they felt better than they did pretreatment.
Discussion
Of the 20 participants who initially underwent BMC injection and did not proceed to surgery, 19 were followed up for 5 years. Overall, pain (VAS scores) and disability (ODI scores) continued to decrease from scores reported in the 3-year follow-up [16]. The increases in disability reported at 5 years in 5 of these 19 participants were only slight. In contrast, only two of the six participants who proceeded to surgery within 3 years after the BMC injection had good surgical outcomes. At 5 years, opioid use was substantially less in the BMC-only participants compared with those six who proceeded to surgery.
Meanwhile, no participant proceeded to surgery after the 3-year follow-up. Compared with pretreatment pain and disability, no participant was made worse from the BMC injections. No adverse events were reported since the 3-year follow-up. Thus, through 5 years, no serious complications or adverse events were associated with the BMC injection.
The results in this study of using bone marrow concentrate to treat discogenic back pain with 5-year follow-up were superior to the 2-year results of lumbar fusion or LADR [2,3,7]. Inasmuch as studies [5,6] comparing lumbar fusion with nonsurgical care reported no difference in clinical results, BMC injection appears to be superior to lumbar fusion surgery in cases of chronic discogenic back pain. The continuing success for most of the 26 initial participants indicates that BMC injection is sustained for as long as 5 years with minimal safety concerns. And as noted, the cost of the BMC treatment was $5,000. This cost was considerably less than the cost of fusion, $50,000 to $75,000, or of lumbar artificial disk replacement, $35,000 to $45,000.
Conclusions
To our knowledge, our previous reports [10,15,16] and this report are the only ones about use of BMC to treat discogenic lumbar back pain and disability. It remains to be determined if the favorable results of this small, preliminary study will continue to be sustained. The study lacked a randomized control group, and MRIs were not done long-term. Nevertheless, we conclude that intradiscal treatment with bone marrow concentrate offers a promising approach to treating lumbar back discogenic pain. It may be reasonable to consider treating patients with who have discogenic back pain at one or two levels with bone marrow concentrate before they proceed to fusion surgery or to artificial disc replacement. | 2019-03-18T14:03:56.013Z | 2018-10-01T00:00:00.000 | {
"year": 2018,
"sha1": "197cd9be13b87ff21ea6b8a90f384111de39c9a3",
"oa_license": "CCBY",
"oa_url": "http://www.sapub.org/global/showpaperpdf.aspx?doi=10.5923/j.ajscr.20180201.01",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "7ca501075d74d11618d408f00f658d5d2faf4191",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
261535054 | pes2o/s2orc | v3-fos-license | Transformation of primary myelofibrosis into acute myeloblastic leukemia: clinical case
Primary myelofibrosis (PMF) is a disease from the group of Ph – negative myeloproliferative tumors, which is characterized by bone marrow fibrosis, splenomegaly and extramedullary hematopoiesis. The mean life in PMF ranges from 7.6 to 10 years and varies widely depending on the appearance of additional mutations and a higher degree of malignancy. Most patients have an indolent flow, and in some patients with PMF, the disease progresses rapidly with the development of acute myeloblastic leukemia. The international stratification system DIPSS-plus allows predicting the occurrence time of the PMF blast transformation phase. The article presents a case of early PMF transformation into acute myeloblastic leukemia in a patient with a heterozygous mutation in the JAK2 gene and an intermediate risk-2 established according to the International Prognostic Scoring DIPSS-plus. The patient received standard cytoreductive therapy.
Introduction
Primary myelofibrosis (PMF) according to the WHO classification (2016) is one of the benign Ph-negative myeloproliferative diseases. The article reports a rare case of early PMF transformation in a 53-year-old patient who ended fatally, despite the timely diagnosis of PMF with the detection of a heterozygous mutation in the JAK2 gene and the establishment of an intermediate risk-2 on the International Prognostic Scoring DIPSS-plus.
The aim of the study is to present a clinical observation of the early transformation of primary myelofibrosis into acute myeloblastic leukemia.
Materials and methods
Description of the clinical case.
Patient T., 53 years old, in December 2013, for the first time noticed heaviness and pain in the left hypochondrium, an increase in the volume of the abdomen, sweating and general weakness.
Patient appealed to the polyclinic at the place of residence, where doctors suspected a myeloproliferative disease, after which the patient was sent to a hematology hospital.
Anamnesis reveals that the patient served in the missile forces in Afghanistan (1979)(1980), suffered from Botkin's disease (1985), abused alcohol, and was registered at a gastroenterologist's dispensary with a diagnosis of Cirrhosis of the liver, class "B" according to Child-Pugh.
On examination, one could see the patient's condition of moderate severity. The skin and visible mucous membranes were pale, there were no hemorrhages. No pathology was detected from the cardiovascular and respiratory systems. The abdomen was enlarged in volume due to hepatosplenomegaly. There was no peripheral edema.
The ultrasound of the abdominal cavity organs dated 15.01.2014 shows: the height of the right lobe of the liver is 176 mm, the thickness is 135 mm. The height of the left lobe of the liver is 103 mm, the thickness is 66 mm. The contours of the liver are smooth, the echostructure is homogeneous, and the echo density is increased. The portal vein is 13 mm. The gallbladder is oval. The bubble wall has no features. There are echogenic contents in the bladder cavity. The common bile duct is 4 mm. The pancreas is not enlarged, the contours are smooth, the echostructure is homogeneous. The echo density is average. The splenic vein is 9 mm. The spleen is enlarged, 200x78 mm. The contour is smooth. The echostructure is homogeneous. In the area of the spleen gate, an additional lobe with a diameter of 24 mm is detected. Conclusion: diffuse changes in the liver parenchyma. Hepatosplenomegaly. An additional portion of the spleen.
Myelogram data dated 15.01.2014: blasts -3%; neutrophilic myelocytes -12.7%; neutrophilic metamyelocytes -4%; banded neutrophils -26.7%; segmented neutrophils -42%; neutrophil maturation index 0.1 RU; eosinophils -0.8%; basophils -2.0%; monocytes -0.2%; lymphocytes -6.3%; all leukocyte cells -94.7%; basophilic normocytes -0.3%; polychromatophilic normocytes -0.2%; oxyphilic normocytes -1.8%; all erythroid cells -2.3%; erythrokaryocyte maturation index -1.1 RU; leukoerythroblastic ratio is 41:1 RU. Conclusion: bone marrow is cellular, polymorphic. The hematopoiesis type is normoblastic. The erythroid lineage is sharply narrowed. The myeloid lineage is hyperplastic due to mature forms of neutrophils. The basophils number has increased. Megakaryocytes are not detected. Histological study of the iliac trepan-biopsy No. 9695 dated 17.01.2014: There is no fat component in the biopsy. There is a significant narrowing of all lineages due to the growth of connective tissue (both in the form of fibrosis fields and in the form of cicatricial connective tissue). Single immature bodies of osteoid. A lot of megakaryocytes are mediumsized and smaller, with acceptable activity. Among the "white row" cells, granulocytes significantly predominate, metamyelocytes are found, there are many segmented cells; there are few lymphocytes. Conclusion: the morphological picture is characteristic of myeloid fibrosis ( Fig. 1-4). Hematoxylin-eosin. Magnification x100. Cytogenetic study of the bone marrow dated 26.01.14: no chromosomal pathology was detected in the examined bone marrow cells. However, a genetic study using the allelespecific polymerase chain reaction in real time to detect the V617F mutation in the JAK2 gene showed the presence of a heterozygous GT mutation (JAK2 gene, rs 77375493). This mutation is associated with an increased risk of thrombosis in the presence of myeloproliferative disease [29,30].
Based on the obtained examination data, doctors drew the following diagnosis: "Primary myelofibrosis, chronic phase; intermediate risk-2 on the DIPSS-plus". The therapy with hydroxyurea at a dose of 1500 mg/s was started. The patient was discharged in a satisfactory condition to continue treatment on an outpatient basis.
Histological study of the iliac trepan-biopsy No. 3381-83 dated 07.12.2015 shows: in the preparations, bone marrow with uneven growth of the fibroreticular stroma, in some places up to complete obliteration of the bone marrow space. The trabeculae of the bone are thickened, with uneven lines, glued together. Myeloid elements are at different stages of maturation, there are relatively many blast cells. The content of lymphoid elements is reduced. The number of megakaryocytes is increased, reaches 30-35 in the field of vision. They are visible both in sclerosed areas and in areas of low cell content. Their size is small, their shape is mostly rounded, the platelet lacing is small. Erythropoiesis is suppressed, but all intermediate growth cells are traced (Fig. 5, 6). Against the background of taking hydroxyurea at a dose of 1500 mg/day, disaggregant therapy (Trombo ASS 100 mg/s), the size of the spleen decreased to 14x10 cm, the hemoglobin level increased to 95 g/l; the number of leukocytes was 4,7x10 9 /l; thrombocytosis remained 408,0x10 9 /l. After 9 months, patient was hospitalized again due to anemia (hemoglobin -62 g/l, erythrocytes -2,1x1012/l) and thrombocytopenia (141,0x10 9 /l). The liver and spleen protrude from under the costal edge by 4 cm and 10 cm, respectively. A course of chemotherapy with small doses of cytosine-arabinoside (20 mg/s) s/c for 21 days, and hemotransfusion were started. Patient was discharged with a slight improvement and a recommendation to continue treatment with hydroxyurea at a dose of 1500 mg/day. Subsequently, frequent hospitalizations due to the steady progression of the disease, accompanied by severe anemia syndrome and dependence on blood transfusions.
At this stage of observation, the doctors established the blast phase of primary myelofibrosis, the high-risk group 3 according to DIPSS-plus. Condition after splenectomy. Complications: Post-transfusion iron overload. Parietal thrombus of the portal and splenic veins.
Since December 2017, alpha-interferon has been added to hydroxyurea therapy for 3 million units 3 times a week.
USDG of the liver vessels dated 05.04.18: Hepatomegaly. The blood flow in the portal vein is not disturbed. No hemodynamic stenoses or thrombosis were detected. They performed transfusions of erythrocyte mass, alpha-interferon million units 3 times per week. Patient was discharged with recommendations to continue therapy with hydroxyurea and intron A.
Since July 2018 until January 2019, the patient was on outpatient observation and received combination therapy with hydroxyurea and alpha interferon. Due to the deterioration of condition, patient was repeatedly hospitalized in the hematology department of a multidisciplinary hospital. The CBC dated 08.01.2019 shows: hemoglobin -69 g/l; leukocytes -13,0x10 9 /l; platelets -67,0x10 9 /l, ESR -39 mm/h. Microscopy: blasts -30%; myelocytes -34%; banded neutrophils -3%; segmented neutrophils -13%; eosinophils -1%; basophils-1%; monocytes-8%; lymphocytes-10%; normocytes 476:100. In connection with the transformation of primary myelofibrosis into acute myeloblastic leukemia, treatment courses with decitabine (dacogen 34,5 mg) were started monthly against the background of hemotransfusion and chelation therapy with exjade 500 mg 2 times per day. A total of 12 courses of decitabine chemotherapy were performed. Only clinical improvement was achieved: the size of the liver, sweating and weakness decreased slightly. During a break from the treatment course, they performed supportive chemotherapy with hydroxyurea (500 mg 1 time per day).
The last hospitalization due to the deterioration of the condition occured on 10.05.2019, the symptoms included the pain in the bones, dyspnea, palpitations and pronounced weakness. The ambulance team was taken to the emergency room of a multidisciplinary hospital. The CBC shows: hemoglobin -60 g/l; platelets -42,0x10 9 /l; leukocytes -37,7x10 9 /l; blasts -66%; banded neutrophils -5%; segmented neutrophils -21%; monocytes -2%; lymphocytes -6%; ESR -52 mm/h. Due to the severity of the condition, patient was urgently hospitalized to the hematology department. On examination, the condition is extremely serious. The skin is pale earthy in color. Respiration rate -24 per min. Auscultation in the lungs is vesicular breathing, wet wheezing in the posterior-lower parts of the chest. Saturation without oxygen is 85%, on oxygen -97%. Heart tones of weakened sonority, tachycardia heart rate -120 per minute, blood pressure 80/50 mm Hg. The abdomen is increased in volume due to hepatomegaly and ascites. On 11.05.2019 at 18:55, due to increasing cardiovascular and respiratory insufficiency, patient was transferred to the intensive care unit. Fibrobronchoscopy dated 11.05.2019 shows indirect signs of pulmonary edema. The condition after the rehabilitation of the bronchial tree.
The patient's condition progressively worsened, and, despite intensive resuscitation manipulations, cardiac arrest occurred, biological death was pronounced at 19:40.
Results
The interest of the presented clinical case is that the transformation into acute myeloblastic leukemia in patient T., 53 years old, occurred quite early (from the date of PMF diagnosis within four years), despite the timely diagnosis of the disease with the detection of a heterozygous mutation in the JAK2 gene and the establishment of an intermediate risk-2 in accordance with the DIPSS-plus scoring.
Discussion
Primary myelofibrosis (PMF) refers to Ph-negative myeloproliferative diseases in accordance with the WHO classification (2016). Myeloproliferative diseases (MPD) are clonal tumors that occur at the level of a hematopoietic stem cell and are characterized by the proliferation of one or more myelopoiesis cell lines in the bone marrow with signs of preserved terminal differentiation [1,3]. The discovery of the JAK2 V617F mutation (2005) contributed to the revision of the diagnostic criteria for MPD. 90% of patients with PMF have a point mutation in the januskinase gene (JAK)a rearrangement of the erythropoietin receptor JAK2 V617F [17,22,25].
The clinical picture of PMF is characterized by a variety of manifestations. In most patients, the initial period of the disease may be asymptomatic for many years [1, 2, 4-6, 8, 9, 21, 24]. During the first years of the disease, the main manifestation is an increased risk of developing thrombosis and thromboembolism against the background of existing cardiovascular pathology and atherosclerosis.
With a prolonged course of the disease and the development of myelofibrosis, symptoms of tumor proliferation and intoxication may appear associated with the secretion of cytokines, the release of myeloid precursors into the peripheral blood with the appearance of extramedullary foci of hematopoiesis in the liver and spleen, the development of cachexia [2,16,21,23]. Prolonged proliferation of the tumor clone leads to the acquisition of additional mutations and a higher degree of malignancy. This process ends with a blast transformation with the development of the terminal stage of the diseasethe blast phase (BF), which is observed in 1-2,5% of patients during the first 10 years of PMF and in 5-8% of patients with a longer duration of the disease [29][30][31].
BF, along with tumor proliferation and intoxication, is characterized by the presence of 20% or more blast cells in the peripheral blood or in the bone marrow [22][23][24].
The most characteristic BF indicators are changes in the hemogram (leukoerythroblastemia, a shift in the neutrophil and erythroid series to young cell elements with the presence of intermediate maturation forms), an increase in the size of the liver and spleen, the presence of tumor intoxication symptoms (fever, weight loss, profuse sweats).
Depending on the severity of bone marrow fibrosis, according to the biopsy, there are prefibrotic and fibrotic stages of the disease [10][11][12][13][14]. The transformation of the prefibrotic stage into the fibrotic one is observed in 65% of patients within 4.2 years, the transformation into acute leukemia is noted in 5-30% of cases [8]. The mean life expectancy is 5 years, although younger patients may live longer [1,[17][18][19][20].
Conclusion
Primary myelofibrosis (PMF) is one of the benign Ph-negative myeloproliferative diseases [1,2,7,12]. The presented clinical case of early PMF transformation in a 53-year-old patient who ended fatally indicates that the current treatment methods of patients with this pathology are aimed only at curbing the progression of the disease, preventing complications, and relieving the symptoms of the disease. The only radical method of therapy for patients with PMF is allogeneic hematopoietic stem cell transplantation, the use of which is associated with a high risk of mortality and complications [15,[25][26][27][28].
To choose the optimal treatment tactics and improve the survival rate of patients with primary myelofibrosis, molecular genetic verification of the disease and stratification of patients using the International Prognosis Scoring DIPSS-plus is necessary. | 2023-09-06T15:18:59.324Z | 2023-01-01T00:00:00.000 | {
"year": 2023,
"sha1": "b415a681d150fab0b8c5d39fdd7c382aaadffb4f",
"oa_license": "CCBY",
"oa_url": "https://www.bio-conferences.org/articles/bioconf/pdf/2023/10/bioconf_ebwff2023_05001.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "5b790fff7df9e18003597529b4d2edbc2a1729aa",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
15227365 | pes2o/s2orc | v3-fos-license | What matters to infrequent customers: a pragmatic approach to understanding perceived value and intention to revisit trendy coffee café
Notwithstanding the rise of trendy coffee café, little is done to investigate revisit intention towards the café in the context of developing markets. In particular, there is a lack of study which provides theoretical and practical explanation to the perceptions and behaviours of infrequent customers. Hence, the study aims to look into the subject matter by using the theory of reasoned action and social exchange theory as the underpinning basis. The framework proposed by Pine and Gilmore (Strat Leadersh 28:18–23, 2000), which asserts the importance of product quality, service quality and experience quality in a progressive manner, is used to decompose perceived value in the model so as to determine their effects on intention to revisit the café. Given the importance to gain practical insights into revisit intention of infrequent customers, pragmatism stance is assumed. Explanatory sequential mixed-method design is thus adopted whereby qualitative approach is used to confirm and complement quantitative findings. Self-administered questionnaire-based survey is first administered before personal interview is carried out at various cafés. Partial least squares structural equation modelling and content analysis are appropriated successively. In the quantitative findings, although product quality, service quality and experience quality are found to have positive effect on perceived value and revisit intention towards trendy coffee café, experience quality is found to have the greater effect than the others among the infrequent customers. The qualitative findings not only confirm their importance, but most importantly explain the favourable impressions they have at trendy coffee café based on their last in-store experience. While product and service quality might not necessary stimulate them to revisit trendy coffee café, experience quality driven by purposes of visit would likely affect their intention to revisit. As retaining customers is of utmost importance to businesses, and infrequent customers more than often make up the majority of the customers, the study provides meaningful and practical understanding of revisit intention.
Background
Creating customer value has become an essential component that contributes to business competitive advantage (Woodruff 1997). The ability of any businesses to deliver superior value is pivotal to customers' perception towards the quality of their offerings. The added value offered to and appreciated by customers, be it product or service, is believed to have consistently generated favourable behavioural outcomes, such as loyalty, consumption behaviour and repeated visit (Wakefield and Blodgett 1996). Hence, in food and beverage businesses, it is pivotal to know customers' perceptions so as to come out with effective operation designs to stimulate their revisit intention.
Although trendy coffee, such as posh coffee, may have once seemed high-priced, people today consider it as an affordable and even indispensable luxury. Despite being Open Access *Correspondence: hiramparousia@gmail.com 1 Institute of Borneo Studies, Universiti Malaysia Sarawak, Kota Samarahan, Sarawak, Malaysia Full list of author information is available at the end of the article a developing country with a population of 30 million, Malaysia is one of the emerging markets which shows promising association between Gross Domestic Product per capital and the amount of coffee consumed (International Coffee Council 2014). Since Malaysian coffee market is in the increasing trend, various business organizations are willing to invest in coffee-related businesses. As a result, the growth of retail coffee franchises, such as Starbucks, Coffee Bean & Tea Leaf, Earthlings, and Dômeand, can be seen in Malaysia in the past decade. Trendy coffee café has become one of the fastest growing business in food and beverage industry, and its businesses is estimated to grow more rapidly in the coming years (Agriculture and Agri-Food Canada 2014).
Competitions between trendy coffee cafés and conventional coffee shops are predictably inevitable and intense. It compels the managers to find cutting-edge strategy to improve their overall quality to enhance customers' perception and their repeat consumption in store. In addition to attracting new customers, retaining existing customers becomes as important to business's long term sustainability (Hoekstra et al. 1999). Despite the prospects of trendy coffee café business and the extant of literature on customer behaviour in food and beverage industry, it is somewhat surprising that little study has been done to delve into customer perception and revisit intention towards trendy coffee café in the context of developing markets. Since existing customers are made up by frequent and infrequent ones, past findings pertaining to revisit intention of the latter are apparently lacking. While perceived value generates transactions between providers and recipients, it is relatively unknown if the same can be said about infrequent customers. Hence, this study aims to look into the relationship between perceived value, revisit intention and their antecedents of the infrequent customers towards trendy coffee café in Malaysia. The framework proposed by Pine and Gilmore (2000), which asserts the importance of product quality, service quality and experience quality in a progressive manner, is used as antecedent variables to provide more explanation to perceived value. The theory of reasoned action (TRA) and social exchange theory (SET) are appropriated to provide theoretical basis to postulated relationships in the study. Instead of addressing theoretical gap, mixed-method approach is used to connect the underpinning theories to the context of the study so as to usher in more relevant and pragmatic discussions (Granek and Nakash 2015;Seddon and Scheepers 2011). As such the understanding of revisit intention of infrequent customers towards trendy coffee café in Malaysia would be extended. Managerial implications are highlighted to articulate the practical contribution of the study.
Trendy coffee café
Trendy coffee café, which is also known as coffee house or coffee concept shop, is a place where coffee is served as the main beverage but food and other beverages are also available for consumption. Unlike conventional coffee house and local food court, it is designed to provide customers with a pleasant in-store experience. Even though each trendy coffee café has its own characteristics, they share the concept of quality and comfort (Ganea 2012). As a result, serving the best coffee is not always the only objective to satisfy customers, but also creating a coffeedrinking experience to leave them with good impression (Stark 2013). Wi-Fi connection and seating arrangements are some of the augmented products to stimulate their interest and enhance in-store experience.
The economic and social development in Malaysia as well as the increasing exposure to the western culture has seen the dynamic change of people's values and lifestyles (Verma 2002). Although the quality of coffee, such as its ingredients and taste, used to be the prime reason for dining at a coffee café, other factors have begun to play significant part in their decision making and revisit intention. Notwithstanding a developing country, Malaysian customers are found to be prepared to pay premium price for a cup of coffee when they see additional values in it. Consequently, trendy coffee cafés, especially the retail coffee franchises, have evolved to become not only a dining place, but also a venue for personal leisure and collective socialization. It even becomes a place for teenagers to hangout (de Luca and Pegan 2013). Ever since Starbucks began its maiden business operation in the capital city of Malaysia in late nineties, trendy coffee cafés have grown rapidly and can now be seen in many populated areas, such as shopping complexes, airports, city centres and town areas throughout the country (Balakrishnan et al. 2009).
Perceived value
Perceived value is one of the most important components in marketing and has become a driving factor to behavioural outcomes in marketplace (Hwang and Kandampully 2015). Value conception is closely associated with the perception of quality (Alex and Thomas 2011). Past findings have shown that product quality has quantifiable impact on customer intention (Woodruff 1997;Alex and Thomas 2011;Asmayadi and Hartini 2015). In the same vein, service quality and performance has long been regarded as an important ingredient that business uses to generate favourable expectation and measure satisfaction, thus ushering in locus of value creation (Parasuraman et al. 1988;Chathoth et al. 2014). Moreover, recent studies have also found that customers make transactions not only for functional reasons, but also emotional satisfaction, which includes fun or enjoyment. Such experience is found to have substantial impact on customer behavioural intention (Ali et al. 2015;Yu and Fang 2013).
The framework proposed by Pine and Gilmore (2000) demonstrates the progression of economic value and provides a comprehensive understanding between the capacity of business and customer behaviour. Apart from the offering of products, it reveals the need and importance of progression in stages: from product to service, and from service to experience. The rationale of the framework is that achieving customer expectation with product might not be sufficient for business to create value with customers in contemporary and future markets. Given the fact that trendy coffee cafés offer not only exquisite food and beverages, but also fine services, and pleasant atmosphere which enhance customers' overall experience, it is interesting to know the perception of infrequent customers towards product quality, service quality and experience quality towards trendy coffee cafés, and how such perception affects their perception of value and revisit intention.
Perceived product quality
Quality has been traditionally defined as a zero error rate, and that is the ability to produce a perfect product at the first try (Parasuraman et al. 1985). As far as customer perception is concerned, the superiority of the product is often regarded as the primary assessment of quality. Past studies have come out with various dimensions to define product quality (Madu et al. 1995;Eduardo et al. 2008;Garvin 1984), and contradicting views are nevertheless found. However, product quality can be generally described as the collection of characteristics and features of a product that contribute to its ability to meet or go beyond given requirement (Parasuraman and Grewal 2000;Ngoc and Uyen 2015). Therefore, perceived product quality can be described as the perceived ability of a product to provide satisfaction relative to the alternatives. It is claimed that product offering value for money not only influences customer behaviour at pre-purchase stage but also their satisfaction as well as post-purchase behaviour (Alex and Thomas 2011). This implies that when customers perceive value in the products they buy or consume due to its superior quality, they are willing to pay more for it, visit the store again and even tell others about it.
In light of the aforementioned, the conception of product quality is of relevance to the present study as it refers to customers' perception towards the products offered by trendy coffee cafés, which in turn infers their evaluation on how well the product specification fits the expectation (Jakpar et al. 2012). Since coffee drinks are the main products and their prices are noticeably higher than those offered in conventional eateries, it is interesting to know the effect of perceived product quality on revisit intention towards trendy coffee café in a developing market.
Perceived service quality
Similar to product quality, various descriptions have been given to explain service quality, including the widely adopted SERVQUAL model (Parasuraman et al. 1988). Service quality has been described as a function of differences between customer expectation and business services (Parasuraman et al. 1985). Customers are often found to be judging the quality of service by comparing what they want from the service with what they eventually receive (Gronroos 1983;Lau et al. 2013). Given the practical gap between what customers want and what they receive, service quality is largely accepted as the perception towards service experience (Ryu et al. 2012). The assumption behind this understanding is that customers perceive quality of service based on service performance or what they actually receive from the deliverance. This includes the behaviour of service employees which has been found to have significant effect on purchase behaviour (Abebe 2014). It has become a significant differentiator and the most powerful competitive weapon to business for attaining efficiency and business performance (Bharadwaj et al. 2015).
In any competitive environment, it is evident that a well-designed process of transaction during which service is purveyed and received is pivotal to business performance and customer satisfaction. When service is perceived as of quality, it generates positive word-ofmouth and continuous visit (Ryu et al. 2012). Since services are variable, and they are produced and consumed at the same time, service providers are often found to be under scrutiny (Kotler and Keller 2007). Hence, perceived service quality is adopted in the present study to investigate its relationship with perceived value and revisit intention of infrequent customers towards trendy coffee café.
Perceived experience quality
The overall level of customer satisfaction and the impression a customer has through interacting with a specific product or service are understood as contextual experience (Alex and Thomas 2011). Hence, customer experience can be described as an engaging act of value co-creation between product or service providers and customers (Schmitt 2010). It can further be defined as cognitive and affective assessment in the process of direct and indirect interaction which eventually affects and moulds perception and subsequent behaviours (Meyer et al. 2007;Page and Conell 2006). Albeit related, experience differs from product and service. While a product is a tangible object, and a service is an intangible activity or process, experience is a personal attachment to the service process in the journey with a series of touch point (Johnston and Kong 2010;Poulsson and Kale 2004). This indicates that the creation of customer experience is embedded in the performance of the business ability to personalize specific prerequisite (Mascarenhas et al. 2006). By instilling memorable experience into the customers across various touch points, it reinforces higher level of loyalty and repeat purchase (Stuart 2006).
It is well documented that an effective creation of customer positive experience is essential to constructing customer loyalty, retention, and subsequently financial success (Klaus and Maklan 2012;Lemke et al. 2011). When customers are impressed in their experience, they tend to remember it and repeat their behaviours. As it is known that trendy coffee cafés offer not only products and services, but also cosy and conducive environment to facilitate various activities, such as reading, socializing and group discussions, they create and offer memorable experience for customers as well. Hence, perceived experience quality is also included in the present study to examine its effect on infrequent customers' perception towards value and revisit intention.
Revisit intention
Intention refers to subjective likelihood of performing a behaviour (Ajzen and Fishbein 1980). It is deemed as the direct determinants of the corresponding behaviour. Past research has supported that beliefs and attitudes are precursors of customers' visible response, such as their purchase and repurchase intention (Huang et al. 2004;Wang et al. 2009). In like manner, past studies in food and beverage sector have also indicated that favourable perception and experience, thus overall satisfaction, are important predictors of customer intent to return (Nadiri and Gunay 2013;Oh 2000;Pettijohn et al. 1997;Qu 1997;Weiss et al. 2005). Therefore, intention is often used to better understand how perception can affect subsequent behaviour.
It is well documented that intention to purchase a product is highly related with the product quality (Wells et al. 2011). In other words, customers are willing to pay more if they perceive value in the product (Michaud et al. 2012). However, studies have also suggested the relevance of other influences, such as internal impulse and external environment (Puccinelli et al. 2009). This ushers in the importance of service quality and experience quality. It is no surprise that perceived service quality is also found to affect behavioural intention (Kim and Moon 2009). Service outcomes, such as revisit intention, come from the integration of customers' perceived performance, expectation, and normative evaluation to the physical environment in which a service process takes place (Wakefield and Blodgett 1999). Past studies not only offer evidence that service quality perceptions positively affect behavioural intentions (Zeithaml et al. 1996), they also highlight the effect of experiential value on positive word of mouth and intention to revisit (Nadiri and Gunay 2013). Hence, revisit intention is adopted as the dependent variable or outcome construct in the study.
Frequency of visit
Frequency of visit is found to be associated with perception towards restaurant performance (Kim and Kim 2004). It is also found to be related to loyalty to preferred restaurant and willingness to pay more for the product and service (Clark and Wood 1998). The underlying principle lies with the expectations of quality. The more frequently customers visited the restaurant, the higher their expectation on quality dimensions (Johnson and Mathews 1997). Hence, frequent customers are likely to have higher expectations of product and service quality than infrequent ones as they anticipate the deliverance of quality during their visits (Bornstein 1989). Moreover, frequent customers are described as ones who will place more reliance on information from internal sources, such as their memories, than external sources, such as advertisements and friends' opinions than infrequent ones (Foulkes 1984). However, in a competitive business environment, the perceptions of infrequent customers cannot be overlooked and they might represent the larger segments in the market. Given the exploratory component of this part of the study, the perception of infrequent customers towards product quality, service quality and experience quality, and revisit intention towards trendy coffee café is looked into. Since past literature reviews little to define frequency of visit in the context of trendy coffee café, a preliminary study is conducted to explore and differentiate infrequent customers from the frequent ones.
Theoretical consideration
Two theories are utilized as the basis to postulate the relationships of variables for the present study. Firstly, the theory of reasoned action (TRA), developed by Ajzen and Fishbein (1980), serves as the primary theoretical foundation to investigate the effect of perceived quality on revisit intention towards trendy coffee café. While belief is generally described as specific descriptions of the object's attributes, attitude is defined as a learned predisposition which ushers in response in a consistent manner (Aaker et al. 2001;Fishbein 1967). As such, the perceptions towards product quality, service quality and experience quality are constructed as beliefs towards trendy coffee café, and perceived value as attitude in the study. Since TRA elucidates direct relationship between attitude and behavioural intention, it establishes the basis for the relationship between perceived value and revisit intention.
Secondly, the social exchange theory (SET) is also utilized as supporting basis to the postulated relationships. SET is among the most influential conceptual paradigms for understanding interdependent transactions (Cropanzano and Mitchell 2005). One of the most fundamental tenets is the expectation of reciprocity (Gouldner 1960). It is also known as reciprocal exchange whereby one party's actions are contingent on the other's behaviour (Molm 2003). Once the transaction is in motion, each outcome will create a self-reinforcing cycle. Such transaction is most often studied in workplace relationships (Shore et al. 2004), and marketplace (Houston et al. 1992;De Ruyter and Wetzels 2000). Given the fact that exchange of value is what drives the development of trendy coffee café, SET provides theoretical and practical basis to the reciprocal relationship between perceived qualities and revisit intention of infrequent customers.
Framework and hypotheses
In light of the underlying theories and literature pertaining to perceived product quality, service quality, experience quality, perceived value and revisit intention, a research framework is developed as shown in Fig. 1.
As past studies generally show the positive effect of perceived product quality, service quality and experience quality on perceived value and various behavioural outcomes (Yu and Fang 2013;Pine and Gilmore 2000), positive-directed hypotheses are developed to test the effect of the said perceptions towards trendy coffee café on revisit intention for infrequent customers. The hypotheses are formulated as follows: There is positive effect of perceived product quality on perceived value of infrequent customers towards trendy coffee café. H 2 There is positive effect of perceived service quality on perceived value of infrequent customers towards trendy coffee café. H 3 There is positive effect of perceived experience quality on perceived value of infrequent customers towards trendy coffee café.
H 4 There is positive effect of perceived value on revisit intention of infrequent customers towards trendy coffee café.
Methods
Due to the lack of literature related to perception of infrequent customers towards trendy coffee café in developing markets and also the potential deficiency of quantitative findings in providing more insights into the investigation (Creswell and Clark 2007), mixed-method design, which assumes pragmatism in its philosophical stance, was adopted in the present study (Teddlie and Tashakkori 2009). As shown in Fig. 2, sequential explanatory design is deemed the most appropriate design as it uses qualitative approach to confirm and complement the findings from the quantitative study (Venkatesh et al. 2013). The combination or "mix" of two approaches would be able to offer more pragmatic and meaningful findings and discussions, thus be of better relevance to the context of the study (Granek and Nakash 2015). A preliminary study using qualitative interview was necessary prior to the main investigation to define infrequent customers based on frequency and purpose of visit. The interview was conducted at the trendy coffee cafés, such as Starbucks and Coffee Beans, rather than anywhere else for two reasons. Firstly, the target respondents had to be customers of any trendy coffee café. Secondly, it was more practical to interview them at the natural setting to find out why they visited trendy coffee café infrequently (Marshall 1996). As such, it is evident that purposive sampling technique was most suited to sample the respondents for the purpose of the study (Onwuegbuzie and Collins 2007). Thirty customers were approached and asked about their frequency and purpose of visit to trendy coffee cafés and sixteen of them were found to be infrequent customers. Hence, frequency of visit was included in the quantitative study as qualifying question so as to set apart the infrequent customers from the frequent ones in the dataset. In the quantitative stage, self-administered questionnaire was used. In addition to questions about respondents' demographic characteristics, all statements related to perceived product quality, service quality, experience quality, perceived value and revisit intention were adapted from past studies (Yu and Fang 2013;Pine and Gilmore 2000). Multiple statements were used to facilitate reliability and validity check in measurement analysis. A seven-point Likert scale was used, where 1 indicates strongly disagree to 7 indicates strongly agree in the questionnaire. After content validity was ascertained by a marketing professor, a pre-test via debriefing method was administered on three individuals to determine the usability and comprehensibility of the questionnaire so as to eliminate potential problems associated with the wordings, instructions and questionnaire design (Bazera 1996;Hunt et al. 1982).
Using purposive sampling technique, 350 copies of questionnaire were distributed at trendy coffee cafés in Malaysia. 250 usable copies were subsequently collected on the spot, with the acceptable response rate at 71 percent (Richardson 2005;Nulty 2008). Statistical Package for Social Sciences (SPSS) was then used for data entry and cleaning purpose. After setting apart data from frequent customers and removing influential outliers and cases with serious missing values, 162 cases were retained in the dataset for data analysis using structural equation modelling (SEM). The sample size was deemed adequate as it was able to detect medium effect size at a statistical power of 80 % using power analysis (Cohen 1988;Hair et al. 2014). Since the emphasis of the study is on prediction, structural equation modelling of partial least squares (PLS-SEM), which is a variance-based SEM, is regarded more appropriate in analysis (Hair et al. 2014).
Once the quantitative findings were obtained, an interview protocol was designed accordingly. The purpose was to find out why infrequent customers would or would not revisit trendy coffee café. Semi-structured interview was administered to allow probes so as to elicit more detailed information pertaining to perceived quality, value and revisit intention from the customers on the basis of data richness (Tong et al. 2007). Fifteen interviews were predetermined with reference to past literature (Guest et al. 2006), and like the preliminary study, they were conducted at various trendy coffee cafés. This was to ensure that all sampled respondents were the actual customers and they could justify their responses with subjective experiences. All interviews were audio-recorded, transcribed and analyzed using content analysis (Kurasaki 2000). Utilizing inter-coder agreement in analysis, data saturation was achieved in the tenth transcript (Carey et al. 1996;Morse 1995). Given the purpose of sequential explanatory mixed-method design, the qualitative findings would then be useful to confirm and complement the quantitative findings pertaining to revisit intention towards trendy coffee café (Onwuegbuzie and Collins 2007).
Preliminary study
In the preliminary study, the interview findings show that the infrequent customers are those who seldom visit trendy coffee café. Either they visit the café not more than once a month or when there is a specific purpose for them to be there. Generally, they neither take the initiative to visit nor suggest to having coffee at trendy coffee café. More than often they are asked to meet up friends there or they need a place to rest, such as at the airport or shopping complex.
Quantitative stage
As sequential explanatory mixed-methods design is adopted, the findings are presented in quantitative stage and qualitative stage. Out of the 162 respondents, the numbers of male and female respondents are about equal. Malay and Chinese respondents make up the majority of the sample as they are the two most populous ethnic groups in Malaysia, and they are more likely to visit trendy coffee cafés which are located in urban areas. Moreover, 110 respondents who are aged 21-25 are found in the sample, showing the popularity of the café among university students and young adults.
Measurement model
Before moving into hypothesis testing, construct validity was first determined. Table 1 indicates the assessment of reliability and validity in the data of the study. The composite reliability (CR) and Cronbach alpha values of more than 0.70 demonstrate that these constructs possess adequate level of internal consistency (Gefen et al. 2000). Similarly, the constructs under investigation demonstrate good convergent validity as they all achieve the minimum threshold value of 0.5 for average variance extracted (AVE) (Bagozzi and Yi 1988). It explains that the items explain more than 50 % of the construct's variances (Hair et al. 2014). Hence, no item was removed from the observation.
Assessment of discriminant validity using Henseler et al. 's heterotrait-monotrait (HTMT) ratio of correlations (2015) criterion is determined as shown in Table 2. Since the correlation values corresponding to the respective constructs do not violate HTMT.90 criterion (Henseler et al. 2015), it is suffice to conclude that construct validity is established in the measurement model.
Structural model
Before performing latent variable analysis in structural model, it is important to ensure that there are no collinearity issues between the constructs under investigation. Table 3 presents that the variance inflation factor (VIF) value for each construct is lower than the offending value of 3.3 (Diamantopoulos and Siguaw 2006). Hence, it suggests that collinearity is of no concern in the study.
In order to test the hypotheses, bootstrapping procedure is used to generate results for each path relationship in the model as shown in Table 4. Bootstrap sub-samples with 5000 cases are created to allow the procedure to estimate the model for each subsample (Hair et al. 2014). All path relationships are found to be significant at 99 and 95 % confidence interval (PROD → VAL, ß = 0.226, p < 0.01; SERV → VAL, ß = 0.201, p < 0.05; EXP → VAL, ß = 0.394, p < 0.01; VAL → INT, ß = 0.623, p < 0.01). Given that all have positive relationship, it is surmised that the four hypotheses in the study are supported.
Despite having decisions for the hypotheses, findings of co-efficient of determination (R 2 ), the effect size (f 2 ), as well as the cross-validated redundancy (Q 2 ) of exogenous constructs on endogenous constructs are also presented as shown in Table 5. The values for co-efficient of determination (R 2 ) of 0.388 and 0.523 suggest that the exogenous variables in this study explain 38.8 % of variances in perceived value and 52.3 % of variances in revisit intention; an indication of substantial explanatory capacity (Cohen 1988). Since the focus of PLS-SEM is on prediction, blindfolding procedure is used (Chin 1998). The Q 2 (Fornell and Cha 1994).
Notwithstanding significant, f 2 values disclose the importance of each exogenous construct to endogenous constructs. It is interesting that effect of perceived experience quality on perceived value (EXP, f 2 = 0.157) is larger than that of perceived product quality and service quality(PROD, f 2 = 0.058; SERV, f 2 = 0.036) for infrequent customers. Evidently, perceived value has large effect on revisit intention towards trendy coffee café (VAL, f 2 = 0.633).
Qualitative stage
Apart from questions pertaining to purposes or reasons to visit trendy coffee café, it has become necessary to find out if infrequent customers are indeed appreciating experience quality more than product and service quality, thus making them willing to revisit. The responses elicited from interviews not only confirm the role of perceived product quality, service quality and experience quality, they also complement the findings on the magnitude of experience quality as shown in Table 6.
In spite of being infrequent customers, the findings show that they believe there is generally good quality in the product and service offered at the trendy coffee café, and they also have had good in-store experience. Even though they think the products are expensive, and the coffee does not necessarily taste better than that of the local coffee shops, they accept that the products are usually good. In like manner, the overall perception about the service quality and performance is also positive. Note worthily, some do not think the staffs are actually doing much. The aforementioned could well explain why the effect size of the relationship between perceived product quality and perceived value, as well as the relationship between perceived service quality and perceived value is found to be minimal. This infers that product quality and service quality at trendy coffee café contribute little to their revisit intention. In other words, infrequent customers would not visit trendy coffee café because of the craving for coffee and the services.
Nevertheless, the findings show the magnitude of experience quality in value perception and revisit intention. Although they rarely consume coffee there, they have fond memory about their last in-store experience. Having good impressions and taking the venue of trendy coffee café into consideration for a specific purpose, such as meeting friends or having discussions there, justify why perceived experience quality carries the most effect on perceived value as shown in quantitative findings. Hence, it can be concluded that the experience quality of infrequent customers at trendy coffee café is mainly constituted by the impressions they have based on their previous visits and their attachment in various contact points. Infrequent customers are not there mainly for the product or service, rather they are there and might visit again for specific purposes which they have experienced before.
Implications and conclusion
This study has provided some empirical insights into the revisit intention of infrequent customers towards trendy coffee café in a developing market with theoretical foundation. Both quantitative and qualitative findings support the notions that revisit intention towards trendy coffee café is affected by perceived value, which, in turn, is predicted by perceived product quality, service quality and experience quality. The findings also support what Pine and Gilmore (2000) advocate about the progression of economic value whereby experience quality is believed to be the cutting edge approach in contemporary market. In general, it is understood that experience quality helps augment the product or service, thus reinforcing customer's revisit intention towards product or service providers (Clark and Wood 1998;Walls et al. 2011). It certainly holds true for infrequent customers as found in the present study.
In light of the findings, it is imperative that managers of trendy coffee café come out with comprehensive business strategies and practices to exemplify in-store experience quality so as to stimulate the interest of existing customers who visit the café infrequently. Over-emphasizing on product quality and frequent customers would likely lead to marketing myopia (Smith et al. 2010). The numbers of frequent and infrequent customers sampled in quantitative stage imply the latter belongs to the majority; and this could well reflect the actual situations of most food and beverage sectors in developing markets. As these markets are emerging as lucrative business regions, the managers cannot afford to lose infrequent ones. Therefore, the contemporary challenge is about keeping existing customers, and converting the infrequent ones to frequent ones by rewarding them with memorable experience when they are at the café.
The pragmatic approach using mixed-method design has indeed elucidated reasons behind the phenomenon under investigation. Such approach is gaining popularity due to its usefulness in providing meaningful and practical interpretations of findings (Granek and Nakash 2015). There are, however, caveats to the study which are worth investigating. Firstly, as different segments are more likely to have different perceptions (Clark and Wood 1998), the sample of the study, which is predominantly made up by university students and young adults, might have come short in extrapolating the findings to wider populations. From the perspectives of developing and multi-cultural country like Malaysia, factors, such as income level, generations and ethnicity, could prove to be vital in divulging more insights about revisit intention towards trendy coffee café. Secondly, the unclassified findings in qualitative stage also suggest that there are more to revisit intention towards trendy coffee café than perceptions about value and about product, service and experience quality. It is worth noting that normative influence could be crucial to improving revisit intention of infrequent customers. Thirdly, the re-specification of perceived value as mediator and the inclusion of moderators and multiple outcome variables could well extend TRA and SET, and thus provide more holistic understanding to revisit intention towards trendy coffee café. | 2018-04-03T01:20:56.371Z | 2016-05-17T00:00:00.000 | {
"year": 2016,
"sha1": "e25a1d6244675a76a5ff34768fc3eb94a8dd0375",
"oa_license": "CCBY",
"oa_url": "https://springerplus.springeropen.com/track/pdf/10.1186/s40064-016-2259-5",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "e25a1d6244675a76a5ff34768fc3eb94a8dd0375",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
232099311 | pes2o/s2orc | v3-fos-license | Penetration Enhancers for Topical Drug Delivery to the Ocular Posterior Segment—A Systematic Review
There is an unmet clinical need for eye drop formulations to efficiently treat the diseases of the posterior ocular segment by non-invasive topical administration. Here, we systematically reviewed the literature on ocular penetration enhancers and their ability to transfer drugs to the posterior segment of the eye in experimental studies. Our aim was to assess which penetration enhancer is the most efficient at delivering drugs to the posterior segment of the eye, when topically applied. We conducted a comprehensive search in three electronic databases (Ovid Embase, Ovid MEDLINE, and PubMed) to identify all the relevant manuscripts reported on ocular penetration enhancers based on the PRISMA guidelines. We identified 6540 records from our primary database search and filtered them per our inclusion/exclusion criteria to select a final list of 14 articles for qualitative synthesis. Of these, 11 studies used cell penetrating peptides (CPPs), 2 used chitosan, and 1 used benzalkonium chloride (BAC) as the penetration enhancer. Cationic and amphipathic CPPs, transactivator of transcription (TAT), and penetratin can be inferred to be the best among all the identified penetration enhancers for drug delivery to the fundus oculi via topical eye drop instillation. Further high-quality experimental studies are required to ascertain their quantitative efficacy.
Introduction
Eye diseases originating in the posterior segment ocular tissues, mainly in the retinachoroid, are a major cause of visual impairment across the globe. It is estimated that worldwide, around 2.2 billion people have some sort of visual impairment and out of these, people affected by diseases of the ocular posterior segment including age-related macular degeneration (AMD) (196 million) and diabetic eye disease (146 million) constitute a large proportion [1]. Specifically, in Europe, AMD and diabetic eye disease lead to a critically high incidence of blindness in older populations and together constitute approximately 35% of the causes of blindness, thus putting a huge burden on public health expenditure [2]. AMD is the reason for 8.7% blindness globally with projected estimates at 196 million people having this disease in 2020 and expanding to 288 million by 2040 and is also the single and most widely recognized cause of elderly blindness in developed nations, especially in individuals over 60 years of age [1][2][3]. Studies also provide strong evidence that AMD is more prevalent in people of European ancestry as compared to other ethnicities, with
Literature Search
Published reports on PEs were identified by running exhaustive searches in three electronic databases: Ovid Embase, Ovid MEDLINE, and PubMed from the time of inception to 30 December 2020. We formulated a common search string for all the databases containing the terms ((penetration enhancers) AND (retina OR retinal ganglion cells OR retinal pigment epithelium)). The search was restricted to English language reports only. Duplicates were removed and manuscripts were initially shortlisted by referencing the titles and abstracts. Final selection was done by reviewing full texts against the inclusion and exclusion criteria by the two authors (A.T. and Z.A.). Additional publications were identified upon reviewing the references of shortlisted articles.
Inclusion and Exclusion Criteria
To be eligible for review, the screened articles had to fulfil certain criteria: (a) only manuscripts of completed experimental studies published in a journal were included; conference abstracts and in-process or non-indexed citations were not included; (b) only studies that investigated topical route of administration of PEs were included; (c) only animal studies were included; clinical or human trials or randomised controlled trials (RCTs) were not; (d) rat, mouse, and rabbit models only for live animal testing; (e) most importantly, only those studies were included where PEs were able to reach the tissues of the ocular posterior segment mainly retina-choroid in-vivo, and studies must demonstrate and compare in-vivo biodistribution and efficacy in the posterior segment of eyes with and without PEs, by standard techniques like immunohistochemistry assays and/or microscopy. Eligibility assessment of studies was performed in an unblinded standardized manner.
Risk of Publication Bias
Risk of bias (RoB) was evaluated using a modified version [17] of SYRCLE's risk of bias tool for animal studies [18], which itself is adapted from the Cochrane Collaboration's tool for assessing risk of bias in RCTs [19]. Briefly, each article was assessed on 13 different risk parameters and the responses were recorded as: (a) yes, if the parameter was positively reported in the article with sufficient information on its methodology; (b) yes, if the parameter was reported positively but with insufficient information or without a methodology; or (c) no, if the parameter was negatively reported or not reported at all. In all of the final shortlisted studies, we surveyed the various critical aspects of bias that play a vital role in animal experiments, to determine the methodological quality of the included studies.
Data Extraction
Data was collected and summarized in a predesigned electronic data extraction form with multiple sections to address all of the key aspects of the study including author(s), journal and year of publication, type of PE used, in-vivo animal model, and efficacy and biodistribution in the posterior segment tissues in-vivo. Since efficacy of drug penetration to the posterior chamber was mostly analysed by immunohistochemistry, metaanalysis was not possible and hence we qualitatively reviewed the reports to answer our review aims.
Study Selection
We retrieved 6540 records from our primary literature search on the databases and after filtering duplicates, 5559 unique original published articles were identified. These were then subjected to a preliminary screening of titles and abstracts, and a list of 28 articles was found to be relevant for review. The whole focus was on identifying PEs that can efficiently access or facilitate penetration of cargo to the fundus oculi by topical eye drop instillation, and to the best of our knowledge we obtained all-inclusive search results that correspond to our eligibility criteria. Full texts of these 28 shortlisted manuscripts were then examined, any relevant references were also analysed and a fi-nal selection of 14 eligible studies was made for qualitative synthesis in our review after application of inclusion and exclusion criteria. The literature identification strategy is demonstrated by means of a PRISMA flow diagram in Figure 1. instillation, and to the best of our knowledge we obtained all-inclusive search results that correspond to our eligibility criteria. Full texts of these 28 shortlisted manuscripts were then examined, any relevant references were also analysed and a final selection of 14 eligible studies was made for qualitative synthesis in our review after application of inclusion and exclusion criteria. The literature identification strategy is demonstrated by means of a PRISMA flow diagram in Figure 1. Of the 14 selected studies, 11 or around 78.6% were exclusively dealing with CPPs as PEs for delivery to the ocular posterior segment and the rest (n = 3 or 21.4%) were based on other compounds. Thirteen studies (92.9%) employed rat or mouse models for in-vivo efficacy and biodistribution tests and one study (7.1%) used rabbits. Since none of the studies quantitatively analysed penetration efficiency to the posterior segment, a metaanalysis of the data was not possible and hence what follows is a qualitative analysis of the results presented for the different PEs. The characteristics of the included studies are defined in Table 1. Of the 14 selected studies, 11 or around 78.6% were exclusively dealing with CPPs as PEs for delivery to the ocular posterior segment and the rest (n = 3 or 21.4%) were based on other compounds. Thirteen studies (92.9%) employed rat or mouse models for in-vivo efficacy and biodistribution tests and one study (7.1%) used rabbits. Since none of the studies quantitatively analysed penetration efficiency to the posterior segment, a meta-analysis of the data was not possible and hence what follows is a qualitative analysis of the results presented for the different PEs. The characteristics of the included studies are defined in Table 1. CPPs can be defined as short chain peptides, often consisting of 30 or less amino acid residues, that exhibit a well-defined property of internalization across cellular membranes with negligible cytotoxicity, through an energy dependent and/or independent mechanism, without the need of chiral interactions with surface receptors. The most commonly studied CPPs are of cationic nature, however a number of anionic or hydrophobic peptides have also been reported. CPPs are able to transport a multitude of covalently or non-covalently linked cargoes of different molecular weights into living cells [34]. Different CPPs have been reported for drug delivery across the skin topically, nasal delivery for pulmonary diseases, and even for crossing the blood-brain barrier (BBB), though the mechanism of their cellular internalization is still debated and is a contentious topic [34,35]. Lately, there has been a rigorous use of CPPs for ocular drug delivery, mostly to the anterior segment tissues of eye and here we report the use of CPPs as penetration enhancers for drug delivery to the ocular posterior segment. Of the total 11 CPP candidates in our review, 5 reports were on transactivator of transcription (TAT), 4 on penetratin, and 2 on other peptides.
TAT
TAT is a protein transduction domain (PTD) with general sequence GRKKRRQRRRPQ of the HIV Tat protein and is one of the most highly investigated CPP for drug delivery. Wang et al. (2010) [21] first demonstrated its potential to deliver a therapeutic growth factor across the posterior segment tissues by eye drop instillation. They investigated the penetration ability of TAT conjugated human acidic fibroblast growth factor (aFGF) in the retina and its therapeutic efficiency in-vivo in a retinal ischemia-reperfusion (IR) injury model in 18 week-old male Sprague-Dawley rats. A primary immunohistochemical analysis was carried out to assess tissue penetration of TAT-aFGF compared with nonconjugated aFGF, where TAT-aFGF was readily detected at 30 min post administration in the retinas whereas non-modified aFGF was not detected. TAT-aFGF was localized mainly in retinal ganglion cells (RGCs) in the ganglion cell layer (GCL), its concentration peaked around 30 min to 1 h, and gradually declined thereafter, but could still be detected up to 8 h post administration.
These results were reinforced with the histological examination of retinal injury from IR after application of TAT-aFGF and aFGF where retinas from TAT-aFGF treated eyes showed significantly less severe pathological changes, lesser damage to tissue morphology, and lower depletion in RGCs than the untreated groups, whereas aFGF only treated groups were no different from untreated eyes. Furthermore, TUNEL staining of diseased retinas affirmed a lower instance of RGC apoptosis induced by IR upon treatment with TAT-aFGF. Apoptotic RGCs in the GCL of TAT-aFGF treated retinas were significantly less in numbers than in the groups treated with aFGF only. In addition, TAT-aFGF treated groups showed accelerated recovery of retinal function compared to those treated with aFGF as measured by electroretinography. The authors suggested a non-corneal route of transport to the posterior segment as TAT-aFGF could not be detected in the deeper layers of the cornea like the stroma and endothelium and were only found in the superficial corneal epithelium cells, meaning it was not fully able to penetrate and pass through the cornea. The mechanism of cellular uptake of TAT is expected to be cell specific because of selective uptake by RGCs. Zhang et al. (2015) [23] demonstrated delivery of endostatin (Es), a specific inhibitor of endothelial cell proliferation and angiogenesis, with TAT via eye drop instillation, tested their ocular penetration system in Kunming mice, and evaluated their inhibitory effects in a choroidal neovascularization (CNV) model in 5-week-old male C57BL/6 mice by immunohistochemistry (IHC). Retinal distribution, as observed by IHC, revealed a good distribution profile of TAT-Es in the retina after eye drop application whereas retinas treated with unconjugated Es showed no detection. The results were supported by the inhibitory effects of TAT-Es on CNV areas observed in choroidal flatmounts. There was a significant reduction in CNV areas in mice treated with TAT-Es eye drops (1378.4 ± 154.3 µm 2 ) compared with negative controls (2623.6 ± 240.9 µm 2 ), while Es only eye drops showed no such protective effects and were similar to negative controls (2514.7 ± 260.7 µm 2 ). They also observed that TAT-Es eye drops showed almost similar inhibitory effects as with the positive control group receiving intravitreal injections of Avastin for 2 weeks (863.5 ± 50.1 µm 2 ) and intravitreally injected TAT-Es (961.2 ± 86.9 µm 2 ), which means that eye drop treatment was as efficient as intravitreal injections and could easily replace it. There have been multiple mechanisms suggested in the study for cellular internalization including energy independent direct translocation for a small portion, but majorly endocytosis dominated by macropinocytosis and relatively smaller instances of clathrin and caveolae-mediated endocytosis. It was speculated that since TAT bears a high net positive charge and could easily interact with negatively charged proteoglycans on the cell membrane like heparan sulphate that work as receptors for positively charged molecules, it may initiate a receptor mediated uptake via macropinocytosis. Li et al. (2016) [25] used TAT combined with a tripeptide of arginine-glycine-aspartic (RGD) as a novel fusion protein for cell-specific Es therapy for anti-angiogenesis effects and inhibition of retinal neovascularization (RNV) in DR in an oxygen-induced retinopathy (OIR) model in C57BL/6 mice via eye drops. ELISA detected 14-fold higher concentrations (9.42 ± 0.77 ng/mg) of TAT-Es-RGD in the retina at 1 h after topical application compared to that detected with unconjugated Es (0.68 ± 0.10 ng/mg). The concentration of TAT-Es-RGD was slightly higher than TAT-Es (8.43 ± 0.85 ng/mg) owing to the endothelial cell targeting ability imparted by RGD that binds preferentially to the α v β 3 integrin expressed on angiogenic endothelial cells. Concentrations of TAT-Es and TAT-Es-RGD peaked at 1.00 ± 0.00 h and 1.125 ± 0.35 h, respectively, and their half-lives were 9.64 ± 1.31 h and 11.03 ± 1.95 h, respectively. Further, fluorescence microscopy analysis of retinal flat mounts of OIR to check the inhibitory effect of TAT-Es-RGD on RNV manifested a significant reduction in avascular areas and vessel tufts in the groups treated with TAT-Es and TAT-Es-RGD via eye drops compared with those of the negative control group. While the Es and Es-RGD treated groups were no different from the negative control group. Breaking of endothelial cells from the inner limiting membrane (ILM) of the retina was also quantified as an evaluation index for hyperoxia-induced neovascularization. A histological analysis of the ILM of retinas proved there was a significant decrease in the number of endothelial cell nuclei, which broke through the ILM in areas of neovascularization in the TAT-Es and TAT-Es-RGD groups as compared with negative controls, whereas Es and Es-RGD groups had no significant difference when compared with negative controls. In addition to these results, levels of VEGF that induced angiogenesis in OIR mice were detected using IHC. The levels of VEGF in TAT-Es and TAT-Es-RGD groups were much lower than that of the negative control groups, whereas in the Es and Es-RGD groups the levels were similar to that of the negative control group. Eye drop delivery of TAT-Es-RGD reduced VEGF expression by a similar amount as intravitreal injection, signifying how eye drops can easily replace invasive treatments. Chu et al. (2017) [27] reported topical ocular delivery of PEG-PLGA nanoparticles (NP) modified with TAT-RGD dual peptides for penetration enhancement and targeting in laser induced CNV models in male Brown Norway rats. Choroid-sclera flatmounts of CNV regions were examined using confocal microscopy to determine the extent of penetration and distribution of NPs. Fluorescent intensity in the CNV region was maximum in the groups treated with RGD and TAT dual-modified NPs and was almost 11 times higher than in the groups treated with non-modified NPs, which showed negligible emission after eye drop administration in the CNV area. Fluorescence intensity in different treatment groups was measured in the order of RGD-TAT-NP > TAT-NP > RGD-NP > NP. TAT modified NPs were able to penetrate the ocular barriers and internalize into retina-choroid, however the internalization was scattered and not focused in the CNV sites due to the absence of selectivity in TAT. RGD and TAT dual-modified NPs exhibited maximum localization and accumulation surrounding and particularly within the CNV region. This is attributed to the complementary targeting property of RGD peptide by binding with integrin α v β 3 .
Penetratin
Penetratin is another commonly used CPP derived from a nonviral antennapedia homeodomain protein in Drosophila melanogaster, with a general sequence of RQIKIWFQN-RRMKWKKK and molecular weight 2.733 kDa. Liu et al. (2014) [22] first reported use of penetratin as a penetration enhancer for drug delivery to the fundus oculi via eye drop instillation. Penetration efficiency and biodistribution of penetratin was evaluated in 17-week-old male Sprague−Dawley rats using fluorescence microscopy. Within 10 min of administration, penetratin was able to internalize in both anterior and posterior segments of the eye. In the posterior segments it was distributed widely in almost all layers of retina-choroid with maximum uptake by the photoreceptor segments and retinal pigment epithelium (RPE) in the subretinal space displaying intense fluorescence in rods and cones, a moderate and uniform distribution in the GCL, inner plexiform layers (IPL), outer plexiform layer (OPL), and mild uptake into the inner nuclear layer (INL) and outer nuclear layer (ONL). Accumulation peaked at 30 min as evidenced by fluorescence intensity and was detectable for as long as 6 h after application suggesting lasting retention in these tissues. An in-vitro cytotoxicity assessment of penetratin by MTT assay was performed on human conjunctival epithelial cells (NHCs) and was found to be the least toxic among a group of CPPs, with a high IC 50 value of approximately 2.5 mM, even higher than that of TAT (2 mM). However, even at a concentration as high as 30 mM of penetratin, the cell viability was still close to 100%. Both transcorneal and conjunctival-scleral transport pathways contribute to penetratin getting into the posterior segment and its penetration ability can be attributed to the high positive charge, its amphipathicity and polyproline type II (PPII) helix structure as observed by circular dichroism. [29] compared the in-vivo penetration ability of different derivatives of penetratin with that of wild-type penetratin via topical eye drop administration in male mice. Pharmacokinetics and biodistribution of hydrophobic and hydrophilic derivatives in comparison with wild-type penetratin were estimated by fluorescence microscopy. Initially all of the peptides were able to internalize in both anterior and posterior segments within 10 min as evident by bright fluorescence and their concentrations peaked around 1-h post eye drop instillation. However, the retinas treated with the hydrophobic derivatives emitted overwhelming levels of fluorescence, revealing that the hydrophobic derivatives internalized more efficiently than did the wild-type penetratin. In contrast, the retinas treated with the hydrophilic derivatives showed weaker fluorescence compared with wildtype penetratin implying impaired penetration ability. To back the results, it has been suggested that higher hydrophobicity induces more helix content in the peptide structure as seen in circular dichroism studies, which in turn helps in stronger affinity interactions with the biological membranes thus facilitating higher permeability. Toxicity analyses performed in-vitro on human corneal and conjunctival epithelial cells via MTT assay revealed no cytotoxicity, and ex-vivo in excised rabbit corneas and sclera by measuring the hydration values also showed good biocompatibility and no leakage during translocation.
Yang et al. (2019) [32] studied penetratin and RGD peptide modified nanocarriers (NCs) of PEG-PAMAM dendrimers for enhanced penetration and complementary targeting to the ocular posterior segment in male ICR mice via topical application. Ocular permeability and posterior segment distribution as estimated by fluorescence microscopy exhibited that although all the NCs were able to penetrate the posterior segment, the mouse retinas treated with NCs conjugated with penetratin or RGD-penetratin showed almost 3 times stronger fluorescence intensity than the retinas treated with non-conjugated nanocarriers, at their peak concentrations at 8 h post application. Penetratin-RGD-NCs initially entered the retinas within 1 h, penetrated the RPE within 2 h, and then gradually internalized to the inner retinal layers, and the accumulation reached a maximum at 8 h post administration. In addition, these retinas kept emitting measurable fluorescence for up to 12 h and were still detectable 24 h after instillation, indicating that the penetratin-RGD modified NCs resided in the retinas persistently, which was not the case with retinas treated with nonmodified NCs as their fluorescence died out within 12 h. MTT viability assays to establish in-vitro cytotoxicity of the peptide-NC complexes in human conjunctival epithelial cells (NHCs), human corneal epithelial cells (HCECs), and human umbilical vein endothelial cells (HUVECs) revealed that these complexes had no detectable cytotoxicity, and the percentage cell survival was 90-110% after incubation with 0.2-20 µM NCs. PEG conjugation significantly attenuated the cytotoxicity of PAMAM dendrimers because of reduced amino groups on the surface and surface passivation, thus preventing contact between cells and primary amines. Johnson et al. (2008) [20] reported a novel CPP called peptide for ocular delivery (POD) with a general sequence GGG(ARKKAAKA) 4 and molecular weight 3.5 kDa that could efficiently reach and deliver cargo to the fundus oculi in-vivo in C57BL/6J mice by topical eye drop instillation. Pharmacokinetics and ocular distribution in the posterior segment were determined by microscopic analysis, which revealed that in the eyes treated with POD-drug complexes the sclera/choroid and dura of the optic nerve were transduced by these complexes in 45 min after application and it was eliminated within 24 h, whereas no such staining was observed in the groups with non-conjugated drug. They have also suggested that cellular uptake of POD is temperature dependant and does not happen by endocytosis as no reduction in peptide uptake was observed in-vitro in the presence of endocytosis inhibitors. Instead, a direct entry mechanism without plasma membrane fixation is supported as also cell-surface proteoglycans inhibited the uptake of POD in-vitro.
Other CPPs
de Cogan et al. (2017) [28] tested delivery of anti-VEGF drugs using another novel poly-arginine based CPP: (5[6]-carboxyfluorescein-RRRRRR-COOH) for their penetration and pharmacokinetics in male Sprague-Dawley rats, and inhibitory effects in a CNV model in wild-type (WT) C57BL/6J mice, via eye drop application. Penetration and biodistribution was evaluated by ELISA. Mice treated with CPP-drug conjugates showed a peak drug concentration about 10-11 times higher in contrast to the groups treated with non-conjugated anti-VEGF drugs. The levels of drug accumulation in CPP-drug conjugates in the retina peaked at 40 min after topical administration representing an upwards of 0.2% of the applied pay load, followed by elimination in 24 h. To estimate the inhibitory effects of CPP-drug conjugates on CNV lesions, fluorescein angiography (FA), infrared imaging, and IHC of disease biomarkers (collagen IV and isolectin B4) in RPE/choroid flatmounts showed that CPP-drug complexes caused considerable reduction to two-thirds extent in the area of scar and neovascularization, whereas CNVs treated with non-complexed drugs had no apparent differences from the untreated or negative control groups. Additionally, the inhibitory effects of eye drop instilled CPP-drug conjugates were similar to those achieved by intravitreal injection of drugs or CPP-drug conjugates, implying that non-invasive topical administration may be a preferred mode for treating posterior segment diseases. No cytotoxicity was detected for CPPs in in-vitro MTT assays in rat retinal cultures, human ARPE-19 cells, and human corneal fibroblasts.
Chitosan
Chitosan is the second most abundant natural polysaccharide derived from deacetylation of chitin and has a sugar backbone of β-1,4-linked glucosamine. It is a polycationic biopolymer and a structural component of the exoskeletons of crustaceans and insects, but also found in some fungi. It bears many favourable properties, like non-cytotoxicity, biocompatibility, and biodegradability and has been used in pharmaceutical and biomedical fields for decades, playing a tremendous role in advancement of drug delivery systems [36]. Chitosan and its derivatives has been long used as a penetration enhancer for transdermal, oral, nasal, buccal, and vaginal drug delivery [37]. In ocular drug delivery it finds wide use as an excipient as penetration enhancer owing to its highly mucoadhesive nature that allows prolonging drug residence in the precorneal area and thus resulting in higher bioavailability.
Li et al. (2019) [31] used chitosan as a penetration enhancer on liposomes for ocular delivery of triamcinolone acetonide via eye drop instillation to the posterior segment of the eye in C57BL/6 mice. Optical coherence tomography (OCT) signal intensities to evaluate penetration and distribution in the eyes showed that eyes treated with chitosan coated liposomes emitted relatively higher levels of signal from the posterior segment of the eye when compared with non-coated liposome treated eyes, though it was not significantly different. The higher levels of the relative signal intensity in the posterior segment of the eye started at 10 min and peaked at 6 h after topical administration. Higher signal intensities from eyes treated with chitosan coated liposomes lasted longer (12 h) than that of non-chitosan groups (10 h) indicating that apart from slightly higher penetration, chitosan also enabled longer retention in the posterior segment. Wang et al. (2020) [33] demonstrated the penetration enhancing ability of carboxymethyl chitosan (CMCS) coating over nanocomposites to deliver dexamethasone (DEX) to the posterior segment of the eye in New Zealand albino rabbits through topical eye drops. In-vivo activity and distribution of the CMCS coated nanocomposites was estimated by measuring the DEX concentrations in retina-choroid tissue homogenates using HPLC. Eyes treated with CMCS conjugated nanocomposites showed the highest DEX concentration in retina-choroid as compared to the groups treated with commercially available DEX eye drops and non-CMCS nanocarriers. In the commercial eye drop groups, DEX was not detectable in the tissues at any time point and its concentration was below detection limits. Even with the non-CMCS nanocarriers, DEX was cleared from the choroid-retina within 1 h post application suggesting impaired retention of the nanocomposites. At 30 min however, DEX tissue concentration in CMCS-nanocomposite groups was 2.5 times higher than that in the non-CMCS nanocomposite groups, which peaked at 1 h and was still detectable up to 3 h post administration, implying greater penetration and longer residence time in retina-choroid. The in-vivo transport pathway for internalization of CMCS was visualized using fluorescence imaging of rabbit eyes after eye drop instillation which measured the fluorescence intensity of all ocular tissues to track the CMCS coated nanocomposites revealed an overwhelmingly stronger fluorescence in the conjunctivalscleral pathway than in the corneal pathway, indicating that CMCS is transported to the posterior segment via the conjunctival-scleral route. The primary mechanism of CMCS uptake in the human conjunctival epithelial cells is via energy-dependent and clathrin-mediated endocytosis as evidenced in-vitro by about 54% reduction in cellular uptake when inhibitors of clathrin-mediated endocytosis were used.
Benzalkonium Chloride (BAC)
BAC is an age-old antimicrobial preservative widely used in ophthalmic formulations and may act as a penetration enhancer, generally to the anterior segment tissues. It facilitates penetration by breaking down the physiological and anatomical diffusion barriers for solute and solvent molecules located in the outer layer of the epithelium.
Mahaling and Katti (2016) [24] used BAC for enhanced penetration of nanoparticles into ocular tissues via topical eye drop administration in pigmented mice. Bioactivity and distribution of BAC coated nanoparticles was determined by fluorescence microscopy. Although the effect of BAC on ocular penetration of nanoparticles and localisation in the retina showed promising results, the enhancement effect was not very pronounced. However, there was some vaguely reported significant increase in bioavailability of BAC coated nanoparticles in the retina-choroid-sclera, as compared to non-coated nanoparticles, which also varied depending on the surface properties of nanoparticles.
Risk of Bias (RoB)
RoB analysis carried out on the methodological quality of the 14 included studies is depicted as a stacked bar-chart in Figure 2. In the included studies, there is a high risk of bias since none of the studies reported treatment concealment, random housing of animals, blinding of interventions, nor assessed outcomes randomly. More than 55% of studies did not include information on random sequence generation for animals, specified the correct timing of randomisation, nor provided sample size calculations. Only 70% of the studies described baseline characteristics, 85% specified primary outcomes but 100% were deemed free from contamination (i.e., when control groups inadvertently receive the treatment or are exposed to the treatment). The RoB analysis of the 14 articles demonstrates that some of the animal studies possess a high degree of bias that may negate/question the findings of those studies.
Discussion
We systematically reviewed all the available literature on experimental PEs for topical delivery to the posterior segment of the eye. From the primary search result of 6540 articles, we filtered the relevant publications to a final 14 in conformity with our inclusion criteria and qualitatively analysed them since a meta-analysis was not possible due to lack of a common standardized quantitative measurement of penetration efficacy in the selected studies. The search results included all the different ocular PEs that have been
Discussion
We systematically reviewed all the available literature on experimental PEs for topical delivery to the posterior segment of the eye. From the primary search result of 6540 articles, we filtered the relevant publications to a final 14 in conformity with our inclusion criteria and qualitatively analysed them since a meta-analysis was not possible due to lack of a common standardized quantitative measurement of penetration efficacy in the selected studies. The search results included all the different ocular PEs that have been reported, however most of them were only effective for anterior segment delivery. For decades, PEs have been used in topical transdermal drug delivery and by oral delivery to the intestinal mucosa, and it is only recently that they are being employed for ocular delivery. Topical eye drop formulations did not evolve functionally for diseases of the fundus oculi, but now there seems to be a renewed vigour in their development especially when PEs can overcome the biggest hinderance of low uptake and bioavailability in the posterior segment tissues.
This systematic review revealed that CPPs are the most favourable PEs for drug delivery to the fundus oculi with high cellular internalization and broad biodistribution in retina-choroid. CPPs have demonstrated their ability to increase the in-vivo uptake of various cargoes including protein drugs [21,23,25,28], nanoparticles [27], genes and nucleic acids [26], dendrimers [26,32], peptides [30], etc., across the ocular biobarriers and into the posterior segment via eye drop instillation. In some cases, topical PEs have affected a similar uptake and biodistribution as observed with intravitreal or intraocular injections thus providing a safe and effective alternative to invasive treatment procedures [23,25,28]. Pharmacokinetic distribution of all the CPPs are very impressive and they were detected in almost every layer of the retina-choroid like RGCs-GCL, photoreceptor segments, RPE, rods and cones, IPL, OPL, ILM, INL, and ONL. Although their penetration mechanisms and transport pathways in the eye need more detailed investigations, this review does show how they may follow all the different internalization mechanisms from passive direct translocation across cellular membranes [20,23] to active endocytosis including proteoglycan-mediated macropinocytosis and clathrin or caveolaemediated endocytosis [23]. All of these pathways are consistent with their general cellular uptake [38,39] and hence, can access the ocular posterior segment from both transcorneal and conjunctival-scleral pathways [22]. CPP studies have also encompassed almost all of the different animal models of major posterior segment pathophysiology including IR injury [21], CNV [23,27,28], RNV, and OIR [25]. Interestingly the CPPs could also be modified with RGD tripeptide for cell-specific and targeted uptake owing to its binding affinity to the α v β 3 integrin expressed on angiogenic endothelial cells [25,27,32]. Additionally, it is worth noting their negligible or non-cytotoxic nature as demonstrated by in-vitro cell viability assays in different cell lines including human corneal epithelial cells (HCECs) and human conjunctival epithelial cells (HConEpic) [22,29,32], rat retinal cultures, human ARPE-19 cells, and human corneal fibroblasts [28].
BAC is a well-known antimicrobial preservative that has been used in ophthalmic eye drop formulations for a long time. Here we see it does work on penetration into the ocular posterior segment but only to a very limited and negligible extent and thus at best can be used for treatment of anterior segment diseases also by virtue of its internationalization mechanism as it works by breaking down the tight junctions of surface epithelia on cornea [24]. Chitosan is reported to enhance the precorneal retention time of drugs due to its mucoadhesive nature thus effectively preventing elimination by tear flow. Chitosan was used as a PE for posterior segment uptake without any encouraging efficacy but could enhance the residence of liposomes in fundus oculi to some extent [31], so we drew the inference that it could be used in combination with CPPs as to increase the residence time of eye drop at the corneal surface resulting in higher uptake and will also increase therapeutic efficacy of the drug by increased retention in the target tissues. CMCS on the other hand showed very promising results with a 2.5 times enhancement in the cargo uptake into the retina-choroid though their major transport was through conjunctival-scleral pathway [33]. As opposed to chitosan, CMCS is anionic in nature and bears a net negative charge but still retains mucoadhesive properties of chitosan residues and enables an increased precorneal and posterior segment residence time, and its uptake is defined as majorly by clathrin-mediated endocytosis.
The risk of bias analysis on the methodologies of selected studies showed a relatively high tendency of bias. Though more than >70% of the studies reported baseline characteristics of animals, a poor randomization of allocation to experimental and control groups and absolutely no concealment of this gives rise to high selection bias. None of the studies report random housing of animals and blinding of investigators and animal caregivers at the intervention stage, thus resulting in extremely high performance bias. Random housing of animals before and after intervention is important because housing conditions like lighting, humidity, temperature, etc., can influence outcomes of the study and behaviour of animals and thus should be comparable in experimental groups [18]. While assessing the outcome, none of the studies report selecting animals at random for analysis and more than >90% of the studies did not ensure blinding of the investigator who made the assessment, which results in very high detection bias. A moderate attrition bias is observed in more studies addressing the incomplete outcome data, however with insufficient information on the methodology used. A good >85% of studies have specified primary outcomes for each intervention, however more than >57% did not calculate a sample size and the few that did calculate had insufficient information on methodology in terms of reporting bias. On the brighter side, there is absolutely no risk of contamination in any of the studies, i.e., neither did the experimental groups receive any additional treatments besides the primary intervention nor were the controls exposed to inadvertent interventions. A 50:50 bias due to unit of analysis errors is reported where one eye of the same animal was used as experimental and control. The high risk of bias can be avoided by adhering to a set of standardised techniques in animal experiments based on the ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments) [40,41].
Limitations
The leading limitation of this article was that a meta-analysis was not possible due to the lack of quantitative data in the included studies. Another limitation is that almost all of the included studies used IHC to assess penetration of drugs to the posterior segment, which is subjective and is prone to inconsistencies due to antibody characteristics such as relative affinity, specificity, and sensitivity to detect the target antigen. The use of poorly characterised antibodies has given rise to the inability of researchers to replicate published data, especially in the analysis of low abundance proteins such as G protein-coupled receptors, steroid hormones, ion channels, and transporters [42]. A quantitative method such as ELISA or HPLC would be preferable when working out posterior segment PE delivery efficiency. Therefore, future studies should take into account these limitations when designing studies to explore the efficacy of PEs to deliver their payload to the posterior segment of the eye.
Conclusions
This systematic review suggests that cationic and amphipathic CPPs like TAT and penetratin are the most promising PE candidates in the present domain for facilitating an increased passage of cargo across the defensive barriers of the eye and into the fundus oculi via topical eye drop administration. These CPPs not only exhibit greater permeability at ocular barriers but also an efficient biodistribution and cellular uptake in the tissues of the ocular posterior segment. A cell-specific and targeted uptake in diseased sites can be achieved in combination with RGD tripeptide. More in-vivo studies are required however, to determine their quantitative efficacy in terms of the proportional bioavailability in the posterior segments of the eye with that of topically applied payload. Most of the in-vivo uptake into the retina-choroid takes place via the conjunctival-scleral pathway. Bioavailability in these tissues can be further enhanced if CPPs can be used in combination with mucoadhesive polymers to increase precorneal residence time and thus effectively allowing for an augmented transcorneal internalization. Further investigations should be carried out on their in-vivo penetration mechanisms and toxicity and also whether the penetration causes any irreversible alterations in barrier properties of the defensive barriers, which can cause more harm than good. | 2021-03-04T05:46:33.364Z | 2021-02-01T00:00:00.000 | {
"year": 2021,
"sha1": "7cfd555d52474ffd1b8e6ea73eab3e592788b725",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1999-4923/13/2/276/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7cfd555d52474ffd1b8e6ea73eab3e592788b725",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
3824069 | pes2o/s2orc | v3-fos-license | A CNN-Based Method of Vehicle Detection from Aerial Images Using Hard Example Mining
Recently, deep learning techniques have had a practical role in vehicle detection. While much effort has been spent on applying deep learning to vehicle detection, the effective use of training data has not been thoroughly studied, although it has great potential for improving training results, especially in cases where the training data are sparse. In this paper, we proposed using hard example mining (HEM) in the training process of a convolutional neural network (CNN) for vehicle detection in aerial images. We applied HEM to stochastic gradient descent (SGD) to choose the most informative training data by calculating the loss values in each batch and employing the examples with the largest losses. We picked 100 out of both 500 and 1000 examples for training in one iteration, and we tested different ratios of positive to negative examples in the training data to evaluate how the balance of positive and negative examples would affect the performance. In any case, our method always outperformed the plain SGD. The experimental results for images from New York showed improved performance over a CNN trained in plain SGD where the F1 score of our method was 0.02 higher.
Introduction
Recently, vehicle detection methods have achieved very high performance owing to deep learning techniques; moreover, many more sources of high-resolution aerial and satellite images have become available and affordable. Worldview3 by Digital Globe [1] provides images with a resolution of 0.3 m per pixel, and now many startup companies such as Planet Labs [2] and Black Sky [3] plan to launch small satellites and provide images with a resolution typically around one meter per pixel. For aerial images, in Japan, NTT Geospace [4] provides aerial images that cover 83% of Japan and updates them frequently. In this context, vehicle detection is now being applied to practical issues such as traffic volume surveys and the estimation of economic activity on the ground.
Research and development of object detection techniques have significantly progressed in recent years by the advancement of deep learning techniques, in particular, the convolutional neural network (CNN). Region-based CNN (R-CNN) [5] was one of the earliest algorithms to employ CNN for object detection and to demonstrate its great capability. In R-CNN, image regions that possibly contain target objects (called "region proposals") are chosen by a selective search algorithm [6], and then a CNN algorithm is applied to map target objects in the region proposals. Following R-CNN, many descendants have been proposed. Fast R-CNN [7] and the Spatial Pyramid Pooling network (SPP-net) [8] have improved accuracy and runtime over R-CNN by utilizing an RoI pooling layer-a special case of the spatial pyramid pooling (SPP) layer-and a SPP layer, respectively. They compute a feature map from an entire image only once, and by utilizing the RoI pooling layer or SPP layer, and the candidate weak classifiers in the next stage were imposed on well classifying those weighted hard examples. While their attempt succeeded, they did not focus on improving the feature learning part in terms of the effective use of training data, which is more straightforward.
In this paper, we proposed the application of HEM to the feature learning process of a CNN model for vehicle detection from high-resolution aerial images.
Methodology
We applied HEM to the stochastic gradient descent (SGD), a commonly used algorithm in deep learning training. Specifically, we used a large batch size, and in each batch, calculated the loss values and employed only examples with the largest loss values for training. In this way, we could always use the most informative examples for training and to improve accuracy.
The details are as follows. In Section 2.1, we introduce our basic methodology and its drawbacks. We first introduce our vehicle detection steps and then explain the characteristics of SGD where there is room for improvement. In Section 2.2, we briefly introduce the related studies of HEM and explain the details of our method. In Section 2.3, we explain our method of accuracy assessment for the experiments in this paper.
Basic Methodology
In this paper, we used a simple sliding window method for vehicle detection. Candidate bounding boxes were scattered densely over an entire image and then those with no existence of vehicles were screened out. HEM was applied to the training of CNN used for the screening. Employed CNN architecture was also simple. We employed the simple sliding window method and CNN architecture, because we mainly focused on the effectiveness of our HEM method. Our HEM method is easily scaled, for instance, by replacing the CNN architecture with a richer one, such as the model used in [18].
Our HEM method was actually a variant of Online Hard Example Mining (OHEM) [22], which was originally designed for Fast R-CNN, but required modifications to suit our method as our training process and the Fast R-CNN training process were different. (The details are described in Section 2.2.)
Vehicle Detection Methodology
We structured the algorithm based on the method of [12]. Figure 1 shows our CNN architecture.
Remote Sens. 2018, 10, 124 3 of 20 examples. While their attempt succeeded, they did not focus on improving the feature learning part in terms of the effective use of training data, which is more straightforward.
In this paper, we proposed the application of HEM to the feature learning process of a CNN model for vehicle detection from high-resolution aerial images.
Methodology
We applied HEM to the stochastic gradient descent (SGD), a commonly used algorithm in deep learning training. Specifically, we used a large batch size, and in each batch, calculated the loss values and employed only examples with the largest loss values for training. In this way, we could always use the most informative examples for training and to improve accuracy.
The details are as follows. In Section 2.1, we introduce our basic methodology and its drawbacks. We first introduce our vehicle detection steps and then explain the characteristics of SGD where there is room for improvement. In Section 2.2, we briefly introduce the related studies of HEM and explain the details of our method. In Section 2.3, we explain our method of accuracy assessment for the experiments in this paper.
Basic Methodology
In this paper, we used a simple sliding window method for vehicle detection. Candidate bounding boxes were scattered densely over an entire image and then those with no existence of vehicles were screened out. HEM was applied to the training of CNN used for the screening. Employed CNN architecture was also simple. We employed the simple sliding window method and CNN architecture, because we mainly focused on the effectiveness of our HEM method. Our HEM method is easily scaled, for instance, by replacing the CNN architecture with a richer one, such as the model used in [18].
Our HEM method was actually a variant of Online Hard Example Mining (OHEM) [22], which was originally designed for Fast R-CNN, but required modifications to suit our method as our training process and the Fast R-CNN training process were different. (The details are described in Section 2.2.)
Vehicle Detection Methodology
We structured the algorithm based on the method of [12]. Figure 1 shows our CNN architecture. CONV, BN, ReLU, POOL, FC represent the convolutional layer, batch normalization layer, Rectified Linear Unit layer, and fully connected layer, respectively. While we simplified the CNN architecture of [12], we added batch normalization [23] layers to accelerate the learning process.
The window size was set so that it finally became 50 pixels before classification, which was large enough to cover a typical vehicle size (see details of our data in Section 3.1). The detailed vehicle detection steps are as follows: • Threshold a test image by pixel intensity greater than 60 or less than 100 and calculate gradient images, yielding three gradient images ( Figure 2).
•
Generate sliding windows that overlap each other on half of width and height ( Figure 3).
•
Move centers of the windows to geometric centers, which represent possible positions of objects in windows. Geometric centers are calculated as Equation (1): The CNN architecture employed in this study. We introduced batch normalization layers to accelerate learning.
CONV, BN, ReLU, POOL, FC represent the convolutional layer, batch normalization layer, Rectified Linear Unit layer, and fully connected layer, respectively. While we simplified the CNN architecture of [12], we added batch normalization [23] layers to accelerate the learning process.
The window size was set so that it finally became 50 pixels before classification, which was large enough to cover a typical vehicle size (see details of our data in Section 3.1). The detailed vehicle detection steps are as follows: • Threshold a test image by pixel intensity greater than 60 or less than 100 and calculate gradient images, yielding three gradient images ( Figure 2). • Generate sliding windows that overlap each other on half of width and height ( Figure 3). • Move centers of the windows to geometric centers, which represent possible positions of objects in windows. Geometric centers are calculated as Equation (1): where g center is a vector which express a pixel position of a geometric center; W and H are the width and height of a window patch, respectively (both are equal to the window size); vector p i,j is a pixel position (i, j) (1 ≤ i ≤ W, 1 ≤ j ≤ H); I i,j is a gradient intensity value of a pixel (i, j); and S is the sum of the gradient intensity values at all pixels in a window patch (Figure 4a,b). • Enlarge them by a factor of √ 2 , and move them to the new geometric centers (Figure 4c,d).
• Discard unnecessary windows that were close to the others. We regarded the windows whose centers were within a distance of 0.15 of window size as unnecessary ( Figure 5).
•
Apply a CNN to RGB pixels in the windows remaining after the above steps. • Examine if the windows had overlapping windows with more than 0.5 of IoU from the highest probability of vehicle existence to the lowest. If a window had overlapping windows, the overlapping windows were discarded (this is called non-maximum-suppression).
where is a vector which express a pixel position of a geometric center; and are the width and height of a window patch, respectively (both are equal to the window size); vector , is a pixel position , 1 , 1 ; , is a gradient intensity value of a pixel , ; and is the sum of the gradient intensity values at all pixels in a window patch ( Figure 4a,b).
•
Enlarge them by a factor of √2, and move them to the new geometric centers (Figure 4c,d).
•
Discard unnecessary windows that were close to the others. We regarded the windows whose centers were within a distance of 0.15 of window size as unnecessary ( Figure 5).
•
Apply a CNN to RGB pixels in the windows remaining after the above steps.
•
Examine if the windows had overlapping windows with more than 0.5 of IoU from the highest probability of vehicle existence to the lowest. If a window had overlapping windows, the overlapping windows were discarded (this is called non-maximum-suppression).
where is a vector which express a pixel position of a geometric center; and are the width and height of a window patch, respectively (both are equal to the window size); vector , is a pixel position , 1 , 1 ; , is a gradient intensity value of a pixel , ; and is the sum of the gradient intensity values at all pixels in a window patch ( Figure 4a,b).
•
Enlarge them by a factor of √2, and move them to the new geometric centers (Figure 4c,d).
•
Discard unnecessary windows that were close to the others. We regarded the windows whose centers were within a distance of 0.15 of window size as unnecessary ( Figure 5).
•
Apply a CNN to RGB pixels in the windows remaining after the above steps.
•
Examine if the windows had overlapping windows with more than 0.5 of IoU from the highest probability of vehicle existence to the lowest. If a window had overlapping windows, the overlapping windows were discarded (this is called non-maximum-suppression). We did not use any meta information such as shadow directions. In terms of sliding window accuracy, we evaluated the negative impact of clutter and shadows in the Appendix A.
Stochastic Gradient Descent (SGD) and Room for Improvement
SGD is an algorithm for optimizing parameters in machine learning that is commonly used in deep learning. First, we explain gradient descent (also called batch gradient descent) on which SGD is based. In machine learning, parameters are optimized by minimizing the objective function (also often called the loss function). In gradient descent, a parameter is updated by where θ is the parameter, α is the learning rate, J is the objective function, and its derivative θ J θ is called the gradient. In gradient descent, gradients are calculated over all examples in the training data and used to update θ [24,25]. This is repeated until there is convergence. However, this becomes inefficient or infeasible when the number of training data is huge [24,25]. Hence in SGD, a small number of examples-called a minibatch-are sampled from the entire training dataset and used for training. Sampling a minibatch is random as giving training data in some meaningful order can bias gradients and lead to poor convergence [25]. Specifically, all of the training data are first shuffled [25] and partitioned (usually equally) into minibatches, then each minibatch is processed for optimization in order. Strictly speaking, this should be called minibatch gradient descent, and SGD originally meant using only a single training example [24]; however, we use this term as it is commonly used in a deep learning context. This is based on the assumption that each minibatch approximates the entire training dataset well [24]. One minibatch process is called an iteration, and processing the entire dataset is called an epoch. Training continues over epochs until convergence. Now let the weight variables of our model be W, minibatch input data be X, labels (the numbers which express the classes) of X be T, and loss function be L(W, X, T). Note that if x and t are single examples of X and T, respectively, L(W, X, T) must be the summation of all L(W, x, t) [26]. Given concrete input X and labels T, we can regard the X and T as the coefficients of L. Therefore, L is regarded as a function of W. We can interpret Equation (2) as follows: This equation updates W so that the loss function becomes smaller. As a consequence, the model becomes able to classify input well. The gradient of each weight variable is calculated by propagating derivatives from the tail to the head of the model based on the chain rule, which is called back propagation [26]. Conversely, calculating output or loss function of a model when given an input is called forward propagation. We did not use any meta information such as shadow directions. In terms of sliding window accuracy, we evaluated the negative impact of clutter and shadows in the Appendix A.
Stochastic Gradient Descent (SGD) and Room for Improvement
SGD is an algorithm for optimizing parameters in machine learning that is commonly used in deep learning. First, we explain gradient descent (also called batch gradient descent) on which SGD is based. In machine learning, parameters are optimized by minimizing the objective function (also often called the loss function). In gradient descent, a parameter is updated by where θ is the parameter, α is the learning rate, J is the objective function, and its derivative ∇ θ J(θ) is called the gradient. In gradient descent, gradients are calculated over all examples in the training data and used to update θ [24,25]. This is repeated until there is convergence. However, this becomes inefficient or infeasible when the number of training data is huge [24,25]. Hence in SGD, a small number of examples-called a minibatch-are sampled from the entire training dataset and used for training. Sampling a minibatch is random as giving training data in some meaningful order can bias gradients and lead to poor convergence [25]. Specifically, all of the training data are first shuffled [25] and partitioned (usually equally) into minibatches, then each minibatch is processed for optimization in order. Strictly speaking, this should be called minibatch gradient descent, and SGD originally meant using only a single training example [24]; however, we use this term as it is commonly used in a deep learning context. This is based on the assumption that each minibatch approximates the entire training dataset well [24]. One minibatch process is called an iteration, and processing the entire dataset is called an epoch. Training continues over epochs until convergence. Now let the weight variables of our model be W, minibatch input data be X, labels (the numbers which express the classes) of X be T, and loss function be L(W, X, T). Note that if x and t are single examples of X and T, respectively, L(W, X, T) must be the summation of all L(W, x, t) [26]. Given concrete input X and labels T, we can regard the X and T as the coefficients of L. Therefore, L is regarded as a function of W. We can interpret Equation (2) as follows: This equation updates W so that the loss function becomes smaller. As a consequence, the model becomes able to classify input well. The gradient of each weight variable is calculated by propagating derivatives from the tail to the head of the model based on the chain rule, which is called back propagation [26]. Conversely, calculating output or loss function of a model when given an input is called forward propagation.
As training progresses, most of the loss values in a minibatch become very small. However, there are still some examples where the loss values are relatively large. We can find analogs of these in the test results. When we conduct vehicle detection with a trained classifier, many of the bounding boxes are classified correctly, but still there can be some that are misclassified. These are sometimes called hard examples. For instance, they may have vehicle-like features that are difficult to discriminate (see Figure 6 for examples). Such examples are likely to give clues to discriminating confusing features; therefore, utilizing them in training processes seems to yield better accuracy. However, they would not sufficiently contribute to learning in an ordinary SGD. As described above, most loss values in a minibatch become very small as training progresses, and gradients calculated over a minibatch are aggregated and averaged. This means the few informative examples are diluted by another large part of the minibatch that do not contribute to improving accuracy. In this way, hard examples contribute little to learning. To address this, we needed to choose the informative examples and preferentially use them for training, which is called hard example mining.
Hard Example Mining (HEM) in SGD Training
In HEM, hard examples, which are difficult to classify correctly, are weighted more than other examples for training. Typically, hard examples are selected if they are difficult to correctly classify for a current classifier. HEM has been conventionally used in machine learning, e.g., for SVM training. For pedestrian detection, Dalal and Triggs [13] searched hard examples with a preliminarily trained detector and additionally used them for training a final detector. Felzenszwalb et al. [27] iteratively updated the training data subset by discarding easy examples that were correctly classified beyond the current classifier's margin and adding hard examples that violated the current classifier's margin. Using a non-SVM method, Tang et al. [16] adopted a cascade of boosted classifiers of shallow decision trees as the final classification part of their vehicle detection method. In each stage of their Real AdaBoost [21] training, a weak classifier that best classified the training data was selected as part of the final classifier. The misclassified examples were weighted, and the candidate weak classifiers in the next stage were imposed on well classifying those weighted hard examples.
In object detection by deep learning, a heuristic method has been previously used. In Fast R-CNN [7] and SPP-net [8], when sampling reference background patches for training data, if the IoU between a background patch and a foreground patch is lower than 0.1, the sampled background patch is excluded from training data, because the patch is not a hard example given that the patch is easily classified to the background patch. If a background patch overlaps a foreground patch in a much portion, such as cases where the IoU is much higher than 0.1, the patch is chosen as training data because the patch is useful as a hard example as the background patch is likely to be confused Such examples are likely to give clues to discriminating confusing features; therefore, utilizing them in training processes seems to yield better accuracy. However, they would not sufficiently contribute to learning in an ordinary SGD. As described above, most loss values in a minibatch become very small as training progresses, and gradients calculated over a minibatch are aggregated and averaged. This means the few informative examples are diluted by another large part of the minibatch that do not contribute to improving accuracy. In this way, hard examples contribute little to learning. To address this, we needed to choose the informative examples and preferentially use them for training, which is called hard example mining.
Hard Example Mining (HEM) in SGD Training
In HEM, hard examples, which are difficult to classify correctly, are weighted more than other examples for training. Typically, hard examples are selected if they are difficult to correctly classify for a current classifier. HEM has been conventionally used in machine learning, e.g., for SVM training. For pedestrian detection, Dalal and Triggs [13] searched hard examples with a preliminarily trained detector and additionally used them for training a final detector. Felzenszwalb et al. [27] iteratively updated the training data subset by discarding easy examples that were correctly classified beyond the current classifier's margin and adding hard examples that violated the current classifier's margin. Using a non-SVM method, Tang et al. [16] adopted a cascade of boosted classifiers of shallow decision trees as the final classification part of their vehicle detection method. In each stage of their Real AdaBoost [21] training, a weak classifier that best classified the training data was selected as part of the final classifier. The misclassified examples were weighted, and the candidate weak classifiers in the next stage were imposed on well classifying those weighted hard examples.
In object detection by deep learning, a heuristic method has been previously used. In Fast R-CNN [7] and SPP-net [8], when sampling reference background patches for training data, if the IoU between a background patch and a foreground patch is lower than 0.1, the sampled background patch is excluded from training data, because the patch is not a hard example given that the patch is easily classified to the background patch. If a background patch overlaps a foreground patch in a much portion, such as cases where the IoU is much higher than 0.1, the patch is chosen as training data because the patch is useful as a hard example as the background patch is likely to be confused with the foreground. This improves accuracy to some extent but is suboptimal, as there could be some hard examples in the excluded patches.
To address this, Shrivastava et al. [22] proposed Online Hard Example Mining (OHEM). In OHEM, the loss values of all region proposals in an image are calculated by the current classifier and only examples with the largest losses are picked for a minibatch. OHEM further improved accuracy over the heuristic method.
However, we could not directly apply OHEM to our method because OHEM is designed for Fast R-CNN, a training process that is different from ours. In Fast R-CNN training, an image is randomly selected from all training images, region proposals are calculated in the image, and 64 of them are selected for a minibatch (in practice, their minibatch consists of 128 examples from two images). As Fast R-CNN employs RoI pooling-in which the feature map is calculated from an entire image only once and region proposals are classified by projecting each of them onto the feature map-this image-wise training is effective. OHEM replaces the selection of region proposals for a minibatch and also benefits from RoI pooling in terms of effective computation. Meanwhile, in our algorithm proposed in Section 2.1.1., we preliminarily extracted patches from all training images, and sampled minibatches randomly from them. We needed to modify OHEM to our training procedure.
Here we explain our method in detail. Figure 7 shows an overview of the algorithm.
Remote Sens. 2018, 10, 124 7 of 20 minibatch and also benefits from RoI pooling in terms of effective computation. Meanwhile, in our algorithm proposed in Section 2.1.1., we preliminarily extracted patches from all training images, and sampled minibatches randomly from them. We needed to modify OHEM to our training procedure.
Here we explain our method in detail. Figure 7 shows an overview of the algorithm. In this way, for training, only the examples with the largest losses are always used, which are the most informative ones. We expect our method to promote the learning of finer features, and it should also find the optimal balance of positive and negative examples in the training examples, the same as in [22]. Recall that SGD is based on the assumption that each minibatch approximates the entire training dataset well, as described in Section 2.1.2. From this viewpoint, we can say that our proposed method approximates checking the loss values of all the training data and only selected the most informative examples for training in one iteration.
In plain SGD, the entire training dataset is split into minibatches and each minibatch is used for training, which means all the examples in the training data are used for training. However, in the proposed method, we only used a part of the examples for training in one epoch because we selected only the examples with the largest loss values. Therefore, we compared plain SGD and the proposed method in the same iteration, not in the same literal epoch.
Here, we also explain the implementation details. As is common practice, we adopted softmax cross entropy as the loss function as defined as follows: where y is the softmax output; N is minibatch size; C is the number of classes; and t is a label. Only one true label among C is 1 and the others are 0. According to this equation, when we want to sort examples of a minibatch by loss values, we only need to check the prediction results, i.e., the last activation of the model corresponding to the true class. There are two ways to implement the proposed method. The first is to calculate all of the loss values in a checkbatch, set the loss values to zeros (except the largest ones), and train. The second is to preliminarily select examples with the worst prediction results in a checkbatch, calculate their loss values, and train. We adopted the second procedure in this paper, mainly because we introduced batch normalization layers [23] into our CNN, as mentioned in Section 2.1.1. Batch normalization normalizes a minibatch so that the mean and the variance of the minibatch become 0 and 1, respectively. If the first implementation is adopted, after examples with the largest losses in a checkbatch are selected, the batch normalization needs to be re-calculated, because the minibatch statistics are generally changed. However, we cannot recalculate the batch normalization by linear transformation of the previous forward propagation result because our CNN also has non-linear ReLU layers as mentioned in Section 2.1.1. Therefore, there is a need to recalculate the loss values of the selected examples after all. For this reason, we adopted the second implementation.
Batch normalization has a training mode and a testing mode, and different statistics are used to normalize a minibatch in each mode. In the training mode, it uses the statistics of the current minibatch, and in testing mode it uses the statistics of all the data that have been used for training. We used the testing mode in the loss checking process, because our idea was to use examples that were difficult to classify by the current classifier.
Vehicle Detection Criteria
We adopted the same criteria as [12]. When a window was detected as containing vehicles, if the distance of centers between it and any groundtruth was smaller than 0.45 of the window size, it was judged as true positive (TP), otherwise it was a false positive (FP). A groundtruth was judged to be detected if it had at least one corresponding TP. In these criteria, it is possible to have multiple TPs for one vehicle, which are redundant except for one TP. One TP is allowed to detect only one vehicle.
Quantitative Measure
We calculated the recall rate (RR), precision rate (PR), and false alarm rate (FAR) as per [12,14]. We also calculated the F1 scores by using the obtained RR and PR scores.
Experiment and Results
We evaluated the performance of our method by comparing it with plain SGD training (hereinafter called the normal method). First, we trained CNNs by the normal and proposed methods, and then conducted vehicle detection with those classifiers. An overview of the experiment is shown in Figure 8. We used sparse training data as explained in Sections 3.1 and 3.2. We conducted preliminary experiments using the normal method and found that it still had room for improvement, because the FAR in the result was high. We aimed to reduce false positives and improve the accuracy by using the proposed method, which was the first motivation of this paper.
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Training Images
Test Images We used sparse training data as explained in Sections 3.1 and 3.2. We conducted preliminary experiments using the normal method and found that it still had room for improvement, because the FAR in the result was high. We aimed to reduce false positives and improve the accuracy by using the proposed method, which was the first motivation of this paper.
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Training Images
Test Images
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Data Preparation
We prepared groundtruth maps of all the images described in Section 3.1 by choosing a pixel in the center of each vehicle by hand. Then we generated the training dataset by extracting patches from the training images. For positive examples, we first extracted bounding boxes around the dots in the groundtruth maps as groundtruth patches. The window size was 50 pixels, which was designed to well cover the typical size of vehicles. To increase the variance of the positive examples, we generated 10 rotated duplications of each groundtruth patch at rotation angles from 9° to 90° in increments of 9°. This is called data augmentation. Then, for negative examples, we extracted background patches randomly where the IoU between a candidate patch and any groundtruth was lower than 0.4. These types of methods are commonly used. The authors in [14] used similar methods for data preparation. Finally, all patches were resized to 48 by 48 pixels, which was the input size of our CNN.
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Data Preparation
We prepared groundtruth maps of all the images described in Section 3.1 by choosing a pixel in the center of each vehicle by hand. Then we generated the training dataset by extracting patches from the training images. For positive examples, we first extracted bounding boxes around the dots in the groundtruth maps as groundtruth patches. The window size was 50 pixels, which was designed to well cover the typical size of vehicles. To increase the variance of the positive examples, we generated 10 rotated duplications of each groundtruth patch at rotation angles from 9° to 90° in increments of 9°. This is called data augmentation. Then, for negative examples, we extracted background patches randomly where the IoU between a candidate patch and any groundtruth was lower than 0.4. These types of methods are commonly used. The authors in [14] used similar methods for data preparation. Finally, all patches were resized to 48 by 48 pixels, which was the input size of our CNN.
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Data Preparation
We prepared groundtruth maps of all the images described in Section 3.1 by choosing a pixel in the center of each vehicle by hand. Then we generated the training dataset by extracting patches from the training images. For positive examples, we first extracted bounding boxes around the dots in the groundtruth maps as groundtruth patches. The window size was 50 pixels, which was designed to well cover the typical size of vehicles. To increase the variance of the positive examples, we generated 10 rotated duplications of each groundtruth patch at rotation angles from 9° to 90° in increments of 9°. This is called data augmentation. Then, for negative examples, we extracted background patches randomly where the IoU between a candidate patch and any groundtruth was lower than 0.4. These types of methods are commonly used. The authors in [14] used similar methods for data preparation. Finally, all patches were resized to 48 by 48 pixels, which was the input size of our CNN.
Training and Test Images
We downloaded aerial ortho images of New York from the U.S. Geological Survey (USGS), cut out areas of harbors and malls, and used them for training and testing. The pixel size of all images was 0.15 m. Table 1 shows the images and their attributes. The train_2 image was taken in the spring of 2013, and the rest was taken in April-May of 2014. These were the only images used in this paper.
Data Preparation
We prepared groundtruth maps of all the images described in Section 3.1 by choosing a pixel in the center of each vehicle by hand. Then we generated the training dataset by extracting patches from the training images. For positive examples, we first extracted bounding boxes around the dots in the groundtruth maps as groundtruth patches. The window size was 50 pixels, which was designed to well cover the typical size of vehicles. To increase the variance of the positive examples, we generated 10 rotated duplications of each groundtruth patch at rotation angles from 9° to 90° in increments of 9°. This is called data augmentation. Then, for negative examples, we extracted background patches randomly where the IoU between a candidate patch and any groundtruth was lower than 0.4. These types of methods are commonly used. The authors in [14] used similar methods for data preparation. Finally, all patches were resized to 48 by 48 pixels, which was the input size of our CNN.
Data Preparation
We prepared groundtruth maps of all the images described in Section 3.1 by choosing a pixel in the center of each vehicle by hand. Then we generated the training dataset by extracting patches from the training images. For positive examples, we first extracted bounding boxes around the dots in the groundtruth maps as groundtruth patches. The window size was 50 pixels, which was designed to well cover the typical size of vehicles. To increase the variance of the positive examples, we generated 10 rotated duplications of each groundtruth patch at rotation angles from 9 • to 90 • in increments of 9 • . This is called data augmentation. Then, for negative examples, we extracted background patches randomly where the IoU between a candidate patch and any groundtruth was lower than 0.4. These types of methods are commonly used. The authors in [14] used similar methods for data preparation. Finally, all patches were resized to 48 by 48 pixels, which was the input size of our CNN.
We generated five different training datasets. The groundtruth patches were always the same, whereas the amounts of sampled background patches were different. The ratios of background patches to groundtruth patches (without the augmented ones) were 100:1, 200:1, 300:1, 400:1, and 500:1 (hereinafter called ×100, ×200, ×300, ×400, and ×500), respectively. For instance, in the case of ×100, the ratio of positive examples (including augmented ones) to negative examples was 11:100. We used them, because the balance of positive and negative examples in the training data generally affects the result, and we aimed to evaluate this effect. As the background area is generally larger than the foreground area in an image, it is common to use more negative examples than positive examples in the training data for better accuracy [1, 7,8,11,14]. This is more conspicuous in the case of vehicle detection, because vehicles are small objects. Taking these into account, we began the ratio of positive to negative examples from 11:100.
In each training dataset, we randomly selected one-tenth of the dataset and used it as a fixed hold-out validation dataset. During training, we calculated the loss and accuracy on this validation dataset in every epoch, which was its only use.
Training Results
We initialized the CNN weight variables at random. We used the Adam solver [28], and the training iterations were equivalent to 100 epochs in the normal method. (e.g., 500 epochs when ncheck was five times larger than nlearn). The batch size for learning (i.e., nlearn) was constantly 100. These were decided empirically based on preliminary experiments. In every epoch during training, the mean loss and mean accuracy were calculated for the training and validation datasets.
As the shapes of all of the graphs were similar for the different conditions, we present only one example. Figure 9 shows the training curves where the background patch amount was ×200 (one of the results of repeated experiments). In Figure 9, all methods seem to have sufficiently converged. For training loss and accuracy, the fluctuations of the proposed method were larger than the normal method. This seems natural because in every iteration, the CNN is updated and used to calculate loss values and select training examples, which means that the criteria of selecting training examples changed in every iteration. Figure 10 shows the moving average of Figure 9d, which aimed to show the convergence trend more clearly. In Figure 10, we averaged about 6350 iterations, which corresponded to two epochs of the normal method.
As Figure 10 shows, while the accuracy in the normal method still improved slightly toward the last epochs, it improved much faster in the proposed method. This means that the proposed method markedly accelerated convergence. In addition, the curve of HEM1000 seems to have converged slightly earlier than that of HEM500. This indicates that the larger ncheck accelerated convergence more.
To evaluate the training results, we compared the final values of validation loss and validation accuracy under different conditions. To mitigate fluctuations, we calculated the moving average of iterations corresponding to 10 epochs of the normal method and then averaged the repeated experiments. Table 3 shows the statistics, which include the standard deviations and standard errors. While validation losses were not necessarily smaller in the proposed method than in the normal method, validation accuracies were always higher in the proposed method than in the normal method, which can be said explicitly according to the standard errors. Although the main purpose of this validation was to check the overfitting occurrence, this result is evidence that the proposed method yielded better generalization. When we compared the validation accuracies of HEM500 and HEM1000, we could not see a significant difference. Note that this accuracy calculation included classifying the background patches. The performance of our system in terms of vehicle detection is evaluated in Section 3.5.
Remote Sens. 2018, 10, 124 11 of 20 the convergence trend more clearly. In Figure 10, we averaged about 6350 iterations, which corresponded to two epochs of the normal method. As Figure 10 shows, while the accuracy in the normal method still improved slightly toward the last epochs, it improved much faster in the proposed method. This means that the proposed method markedly accelerated convergence. In addition, the curve of HEM1000 seems to have converged slightly earlier than that of HEM500. This indicates that the larger ncheck accelerated convergence more.
To evaluate the training results, we compared the final values of validation loss and validation accuracy under different conditions. To mitigate fluctuations, we calculated the moving average of iterations corresponding to 10 epochs of the normal method and then averaged the repeated the convergence trend more clearly. In Figure 10, we averaged about 6350 iterations, which corresponded to two epochs of the normal method. As Figure 10 shows, while the accuracy in the normal method still improved slightly toward the last epochs, it improved much faster in the proposed method. This means that the proposed method markedly accelerated convergence. In addition, the curve of HEM1000 seems to have converged slightly earlier than that of HEM500. This indicates that the larger ncheck accelerated convergence more.
To evaluate the training results, we compared the final values of validation loss and validation accuracy under different conditions. To mitigate fluctuations, we calculated the moving average of iterations corresponding to 10 epochs of the normal method and then averaged the repeated
Vehicle Detection Results
We conducted vehicle detection using the method described in Section 2.1.1 and the test images described in Section 3.1. Results of repeated experiments were averaged. Figure 11 shows the F1-measure results and Table 4 shows the statistics of all quantitative measures, which include the standard deviations and standard errors. method, validation accuracies were always higher in the proposed method than in the normal method, which can be said explicitly according to the standard errors. Although the main purpose of this validation was to check the overfitting occurrence, this result is evidence that the proposed method yielded better generalization. When we compared the validation accuracies of HEM500 and HEM1000, we could not see a significant difference. Note that this accuracy calculation included classifying the background patches. The performance of our system in terms of vehicle detection is evaluated in Section 3.5.
Vehicle Detection Results
We conducted vehicle detection using the method described in Section 2.1.1 and the test images described in Section 3.1. Results of repeated experiments were averaged. Figure 11 shows the F1-measure results and Table 4 shows the statistics of all quantitative measures, which include the standard deviations and standard errors. Figure 11. F1 scores in each condition. Our proposed method improved the scores in all cases over the normal method. Figure 11. F1 scores in each condition. Our proposed method improved the scores in all cases over the normal method.
In terms of F1 scores, while almost all of the standard errors were smaller than 0.01, the proposed method improved the scores by over 0.02 when compared to the normal method in most cases, which proved the effectiveness of our proposed method. Moreover, because the standard deviations of all methods were not very different, we can say our proposed method worked stably. When we compared the results in terms of the background patch amount, the F1 scores in the normal method tended to be higher when the background patch amount was larger, and this seems to also apply in the proposed method.
When we compared the F1 scores of HEM500 and HEM1000, from ×100 to ×500, HEM500 won in two cases and HEM1000 won in the other three cases, which seems almost even. The score differences were less than those between the normal method and the proposed method. Figure 12 plots the FAR versus the RR. As can be seen, the proposed method greatly reduced the FAR while retaining nearly the same RR. This mainly contributed to accuracy improvement, because our training data were sparse and the FARs were relatively large throughout our experiments. Note that the power of the proposed method was not restricted to FAR reduction, because the most informative examples were automatically selected in every checkbatch.
Although the non-maximum-suppression (NMS) was properly applied, there were some duplicated detections (redundant TPs) due to a limitation of NMS. However, this does not affect accuracy assessment according to the definitions of PR and F1. Figure 13 shows an example of good and bad results by HEM500 where the background patch amount was ×400. We chose the classifiers that achieved the best F1 scores from repeated experiments. While many FPs were reduced in the pair of images in Figure 13a, a few vehicles became undetected in Figure 13b. There seems to have been a kind of trade-off, while overall accuracy was improved.
Remote Sens. 2018, 10, 124 14 of 20 In terms of F1 scores, while almost all of the standard errors were smaller than 0.01, the proposed method improved the scores by over 0.02 when compared to the normal method in most cases, which proved the effectiveness of our proposed method. Moreover, because the standard deviations of all methods were not very different, we can say our proposed method worked stably.
When we compared the results in terms of the background patch amount, the F1 scores in the normal method tended to be higher when the background patch amount was larger, and this seems to also apply in the proposed method.
When we compared the F1 scores of HEM500 and HEM1000, from ×100 to ×500, HEM500 won in two cases and HEM1000 won in the other three cases, which seems almost even. The score differences were less than those between the normal method and the proposed method. Figure 12 plots the FAR versus the RR. As can be seen, the proposed method greatly reduced the FAR while retaining nearly the same RR. This mainly contributed to accuracy improvement, because our training data were sparse and the FARs were relatively large throughout our experiments. Note that the power of the proposed method was not restricted to FAR reduction, because the most informative examples were automatically selected in every checkbatch. Although the non-maximum-suppression (NMS) was properly applied, there were some duplicated detections (redundant TPs) due to a limitation of NMS. However, this does not affect accuracy assessment according to the definitions of PR and F1. Figure 13 shows an example of good and bad results by HEM500 where the background patch amount was ×400. We chose the classifiers that achieved the best F1 scores from repeated experiments. While many FPs were reduced in the pair of images in Figure 13a, a few vehicles became undetected in Figure 13b. There seems to have been a kind of trade-off, while overall accuracy was improved.
(a) (b) Figure 13. An example of good and bad cases in the tested images. In each pair of images, the left one shows the result of the normal method and the right one shows the result of HEM500. (a) Good case. On the right, FPs were much reduced; (b) bad case. On the right, some vehicles became undetected. In terms of F1 scores, while almost all of the standard errors were smaller than 0.01, the proposed method improved the scores by over 0.02 when compared to the normal method in most cases, which proved the effectiveness of our proposed method. Moreover, because the standard deviations of all methods were not very different, we can say our proposed method worked stably.
When we compared the results in terms of the background patch amount, the F1 scores in the normal method tended to be higher when the background patch amount was larger, and this seems to also apply in the proposed method.
When we compared the F1 scores of HEM500 and HEM1000, from ×100 to ×500, HEM500 won in two cases and HEM1000 won in the other three cases, which seems almost even. The score differences were less than those between the normal method and the proposed method. Figure 12 plots the FAR versus the RR. As can be seen, the proposed method greatly reduced the FAR while retaining nearly the same RR. This mainly contributed to accuracy improvement, because our training data were sparse and the FARs were relatively large throughout our experiments. Note that the power of the proposed method was not restricted to FAR reduction, because the most informative examples were automatically selected in every checkbatch. Although the non-maximum-suppression (NMS) was properly applied, there were some duplicated detections (redundant TPs) due to a limitation of NMS. However, this does not affect accuracy assessment according to the definitions of PR and F1. Figure 13 shows an example of good and bad results by HEM500 where the background patch amount was ×400. We chose the classifiers that achieved the best F1 scores from repeated experiments. While many FPs were reduced in the pair of images in Figure 13a, a few vehicles became undetected in Figure 13b. There seems to have been a kind of trade-off, while overall accuracy was improved.
Improvement Extent
While we could see that our method improved accuracy, the improvement of F1 did not seem very significant. We investigated the reasons. First, the improvement was different between test_1 image and test_2 image. Table 5 shows the F1 result of each test image in the case of ×400. As can be seen, although the result of test_2 image was much improved, the result of test_1 was originally good, because test_1 image had very similar features to the train_1 image and did not much improve (became slightly worse in this case) using our method. Thus, the improvement of both test images became relatively small.
The second reason was redundant TP. While our method greatly reduced FPs, the improvement of PR was relatively small. In Table 4, in the case of ×400, while FAR reduction was about 0.13 on average, PR improvement was about 0.06 on average. This was because not only FPs, but also redundant TPs hurt PR. As we checked, the ratios of redundant TPs to detected vehicles in normal, HEM500, and HEM1000 were about 0.31, 0.25, and 0.27, respectively. Our method also seems to have reduced redundant TPs, which were because redundant TPs were relatively distant from the vehicles. For example, in one experiment using the normal method with ×400, the average distances between vehicles and TPs that detected vehicles, and between vehicles and redundant TPs, were 6.6 pixels and 11.5 pixels, respectively. As our training data only included positive examples that exactly matched the locations of vehicles, such redundant TPs would have been reduced by HEM. However, the reduction was smaller than that of FPs, because the redundant TPs overlap vehicles to some extent. This diluted the improvement of FAR.
The third reason was RR decrease. In Table 4, in the case of ×400, RR decreased by 0.02 on average. As above-mentioned, our HEM method seems to have reduced TPs that were relatively distant from vehicles, and a small part of them may have detected vehicles before applying HEM. This slightly hurt the RR.
The second and third reasons came from the inaccuracy of sliding windows. By replacing sliding windows with a more accurate method such as RPN, we can see the improvement by HEM more explicitly.
The best average F1 score throughout our experiments was 0.71, which is actually not a state-of-the-art result. For instance, [16] reported an F1 of 0.83, and [18] reported very high F1 of 0.94. Although it is not a fair comparison, because the training data and test settings are totally different, 0.71 is not the best. The primary reason would be the insufficiency of our training data. The authors in [16] used the Munich dataset [29], which has 9433 annotated vehicles, and [18] proposed and used the COWC dataset, which has 32,716 annotated vehicles. Compared to those datasets, our training data were sparse. Although our HEM could improve accuracy, sufficient training data were necessary for high performance. Another reason would be the simplicity of our CNN architecture. When we compared just the numbers of convolutional layers in each method, [16] had five and [18] had 50, whereas our model only had three.
We can combine our HEM method with adequate training data and rich CNN architecture for higher accuracy. Moreover, as discussed above, the accuracy would further become better by improving or replacing the sliding window method. Our HEM method can easily scale with those options.
Training Loss Values and Duration
Shrivastava et al. [22] reported that the loss values during training became smaller by using their OHEM method because they conducted a fair comparison where all region proposals in an image-not just the ones selected for a minibatch-were used to calculate the loss values in every method. Meanwhile, the loss values of our method seemed to have been better than the normal method; however, the difference was very small (Figure 9a). This was because those loss values were calculated from examples that were actually used for training; those examples had relatively large loss values, because we chose such examples for training. If we evaluated the loss values over a checkbatch, the trend would be similar to [22].
Our method took more time to train for the same iterations than the normal method, because of the overhead of calculating loss values in a checkbatch. For instance, training durations of normal, HEM500, and HEM1000 were 2.5, 6, and 10 h, respectively, with Tesla K20X manufactured by NVIDIA (Santa Clara, CA, USA) in Figure 9. In the original OHEM, it took approximately 1.7 times longer in one iteration to select 64 out of around 4000 region proposals in an image. Our method is more time-consuming than theirs, because our method requires extra forward propagation calculations as described in Section 2.2, whereas the original OHEM can calculate the loss values of region proposals with small additional costs due to RoI pooling.
However, our method strongly accelerates convergence. Therefore, the training could be stopped much earlier in the proposed method, which would cancel out the demerit.
Source of Accuracy Improvement
In our experiment, more background patches tended to yield better F1 scores in the normal method. This may be because more background patches contain more "negative" features to be discriminated from vehicles, and it kept improving the accuracy in the range of our experiments. FP reduction seems to have contributed more to the performance improvement, because the FAR was relatively high in our experiments.
To confirm how accuracy improvement continued, we conducted additional experiments. We gradually increased the amount of background patches to 1000 times the number of groundtruths. Figure 14 shows all the F1 scores of vehicle detection tests by the normal method, including the additional experiments. The standard deviations and standard errors of the additional experiments had values similar to the previous experiments. As can be seen, the accuracy does not seem to have improved after ×600, which indicates the best balance was around ×600. Although the best score was at ×800, which was about 0.7, the difference from ×600 was smaller than the standard errors, and was thus negligible.
When we compared these results with those of the proposed method, four out of 10 scores of the proposed method surpassed the best scores of the normal method, which had the best balance of positive and negative examples in the training data. This fact proves that our method improved accuracy by learning finer features, and not only by balancing positive and negative examples.
Continuing to increase negative examples could lead to a worse result. Even in such cases, our method is expected to find the optimal balance between positive and negative examples during training, which we could not confirm explicitly in our experiments. This will be our future work. Figure 14 shows all the F1 scores of vehicle detection tests by the normal method, including the additional experiments. The standard deviations and standard errors of the additional experiments had values similar to the previous experiments. As can be seen, the accuracy does not seem to have improved after ×600, which indicates the best balance was around ×600. Although the best score was at ×800, which was about 0.7, the difference from ×600 was smaller than the standard errors, and was thus negligible. When we compared these results with those of the proposed method, four out of 10 scores of the proposed method surpassed the best scores of the normal method, which had the best balance of
HEM500 vs. HEM1000
The HEM500 and HEM1000 scores were almost even, and the score differences were less than those between the normal and proposed methods. Considering that the validation accuracy results of HEM500 and HEM1000 were not very different, and taking these results into account, we can suppose that the performances of HEM500 and HEM1000 were similar, because an ncheck value of 500 was large enough, and increasing it did not improve accuracy in the range of our experiments. However, our experiments were not enough to prove this conjecture. Moreover, we could test only two values of ncheck due to our limited computing resources. Further verification will be our future work.
Usabilty of Our Method
In our experiments, we used a sparse training dataset. However, our method does not depend on model architecture or training data; therefore, it would be effective even when training data are abundant.
In this paper, we used our CNN as a classifier. However, it could also be used as a feature extractor by removing the last fully connected layer, and it could be combined with other classifiers. For instance, Tang et al. [16] used a cascade of boosted classifiers that was fed features extracted by a CNN. We would be able to further improve accuracy by combining our method with such a method.
Conclusions
We applied HEM to the SGD training of a CNN vehicle classifier, which successfully promoted learning finer features and improved accuracy. It took more time to train for the same number of iterations; however, this could be canceled out by breaking training earlier, which would be acceptable as the proposed method markedly accelerates convergence. Although we used sparse data in our experiments, our method would be effective even when training data are abundant. However, we could not confirm the effect of balancing positive and negative examples in the training data explicitly, and the effect of checkbatch size was not sufficiently determined. These issues will form the basis of our future work.
In this appendix, we present an example of HEM500 with ×400 to further investigate the result.
Effect of Clutters and Shadows
To evaluate how clutter and shadows impeded vehicle detection, we visualized undetected vehicles. Figure A1 shows 100 out of 211 randomly selected undetected vehicle patches of test_1 and all 31 undetected vehicle patches of test_2. As can be seen in Figure A1b, only one undetected vehicle was covered by any clutter or shadows. Although we could not fairly evaluate the robustness of the performance of our sliding window method, because our test images did not originally have many obstacles, there seems to have been only a few vehicles that were not detected due to clutter and shadows.
Furthermore, to evaluate how clutter and shadows confused CNN, we visualized the FP patches. Figure A2 shows all 18 FP patches of test_1 and 100 out of 231 randomly selected FP patches of test_2. As can be seen in Figure A1b, only one undetected vehicle was covered by any clutter or shadows. Although we could not fairly evaluate the robustness of the performance of our sliding window method, because our test images did not originally have many obstacles, there seems to have been only a few vehicles that were not detected due to clutter and shadows.
Furthermore, to evaluate how clutter and shadows confused CNN, we visualized the FP patches. Figure A2 shows all 18 FP patches of test_1 and 100 out of 231 randomly selected FP patches of test_2.
In Figure A2b, several patches with shadows were misclassified as vehicles. This was probably because the shadows looked like straight edge features that are similar to vehicle features, thus confusing CNN. This seems to have been caused by the insufficiency of the classification performance of CNN rather than the inaccuracy of the sliding windows. These misclassifications would be further reduced by using abundant training data.
As can be seen in Figure A1b, only one undetected vehicle was covered by any clutter or shadows. Although we could not fairly evaluate the robustness of the performance of our sliding window method, because our test images did not originally have many obstacles, there seems to have been only a few vehicles that were not detected due to clutter and shadows.
Furthermore, to evaluate how clutter and shadows confused CNN, we visualized the FP patches. Figure A2 shows all 18 FP patches of test_1 and 100 out of 231 randomly selected FP patches of test_2. | 2018-04-03T06:22:52.791Z | 2018-01-18T00:00:00.000 | {
"year": 2018,
"sha1": "f22a8c28a6de723e5451ce577a3ef8dfb26f5e2a",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2072-4292/10/1/124/pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "f22a8c28a6de723e5451ce577a3ef8dfb26f5e2a",
"s2fieldsofstudy": [
"Computer Science",
"Engineering",
"Environmental Science"
],
"extfieldsofstudy": [
"Computer Science",
"Geology"
]
} |
269195383 | pes2o/s2orc | v3-fos-license | The effect of Orem self-care model on the improvement of symptoms and quality of life in patients with diabetes: A scoping review
Abstract Objective: to evaluate the association of Orem self-care model improvement of symptoms and quality of life in patients with diabetes. Methods. A scoping review was carried on bibliographic databases: PubMed-Medline, Scopus, SID and Magiran. The inclusion criteria encompassed studies examining the impact of the Orem self-care model on diabetic patients. Studies considered for inclusion needed to have full-text availability and be written in either English or Persian, with key words including “Models”, “Nursing”, “Quality of Life”, and “Diabetes Mellitus”. CONSORT checklist and STROBE statement were selected for quality assessment. Results. A total of 9 studies were included, all using quantitative methodology and focusing on adults or older adults. The majority of articles focused on quality of life and diabetic symptoms. 8 studies showed positive outcomes after implementation of the model. The findings indicate that this model led to an enhanced level of self-efficacy, improved quality of life, and better self-care practices among diabetic patients. Conclusion. Orem self-care model can reduce the diabetic symptoms and improve the quality of life, self-efficacy and self-care in these patients.
Introduction D
iabetes originates from impairment in the metabolism of proteins, carbohydrates and lipids. (1)It is a chronic metabolic disorder that is defined and recognized by the increase in blood sugar caused by defects in the secretion or action of insulin or both.This disease leads to many complications such as limb amputation, cardiovascular problems, kidney diseases, blindness, and long-term disabilities. (2)Diabetes is one of the most common non-communicable diseases that has the highest prevalence rate among metabolic diseases.More than 90% of diabetic patients suffer from type 2 diabetes.This disease is associated with short-term and long-term complications that are irreversible in many cases. (3)Complications and deaths caused by diabetes are among health problems all over the world.The increase in the number of patients with type 2 diabetes indicates a global epidemic. (4)In addition to direct complications, diabetes also has many costs that affect the lives of the patients.In 2017, 327 billion dollars was the total cost of diagnosed diabetes. (5)e prevalence of type 1 and type 2 diabetes is increasing worldwide, but the rate of type 2 diabetes is higher than that of type 1. (6) Diabetes is a chronic condition that imposes major threats to the mankind and affected 435 million individuals in 2017. (7)It will involve 552 and 624 million patients by 2030 and 2040, respectively. (8,9)According to the International Diabetes Federation (IDF), Iran ranks third among Middle Eastern countries in terms of diabetes (5.4 million cases).It is expected that the rate of diabetes will reach 9.2 million Iranians by 2030. (10)Aging population will be another effective factor in increasing the prevalence of diabetes in the future. (11)The increase in the number of diabetic patients has faced the care provider organizations with increasing financial problems and a decrease in the number of care providers. (12)15-20% of hospital beds are occupied by patients with diabetes and it is difficult to determine which of these patients require more medical attention. (13)Regardless of age, country and economic conditions, diabetes is a major health challenge and has a global impact, so that the prevalence of diabetes is reaching an alarming level. (14)herefore, the prevention and control of this disease is considered one of the health challenges, so that one of the goals of the Ministry of Health, Treatment and Medical Training of Iran is to stop the spread and growth of diabetes until 2025. (15)e of the most important clinical goals of nursing care is to reduce the severity of disease symptoms and the stress caused by it in order to improve the quality of life in these patients.Without patient participation, favorable outcomes cannot be considered.Quality of life is a crucial aspect of healthcare, particularly in the context of nursing care.The primary clinical objective for nurses is to alleviate the severity of disease symptoms and mitigate the associated stress experienced by patients.By doing so, the overarching goal is to enhance the overall quality of life for individuals undergoing medical treatment. (16)Self-care behaviors are one of the most important components which requires special attention.These behaviors in diabetic patients includes appropriate glycemic control, following a healthy diet, taking medicines regularly, monitoring blood sugar, taking care of feet and doing sports activities. (17)Self-care behaviors are naturally multidimensional and include all the activities focused on health maintenance, disease prevention and treatment.These kinds of behaviors are done purposefully and with the consent of the patient. (18)Improving the self-care behaviors is the first step toward helping the diabetic patients.Normally, self-care practices can reduce the costs and prevent acute or chronic complications of diabetes. (19,20)Appropriate behaviors can lead to a decreased level of cardiovascular risks among diabetic patients. (21)Since self-care activities are effective in glycemic control and other outcomes in diabetic patients, it can be claimed that selfcare behaviors can be an underlying reason for mortality among them. (22,23) e of the models that can contribute to the patient's self-care is Orem self-care model. (24)ccording to Orem, humans have the ability to take care of themselves, and if this ability is disturbed, nurses help them to regain this ability. (25)Understanding the impact of the Orem Self-Care Model on symptom alleviation and quality of life is crucial for healthcare professionals and policymakers.Positive findings could lead to the development of targeted interventions and strategies that enhance patient outcomes, potentially improving the overall quality of care for individuals with diabetes.The need for research on factors affecting self-care based on the needs of patients to increase their self-care ability seems necessary.Hence, this study aimed to evaluate the association of Orem self-care model improvement of symptoms and quality of life in patients with diabetes.
Methods
This scoping review adheres to PRISMA guidelines.The effect of Orem self-care model on the improvement of symptoms and quality of life in patients with diabetes: A scoping review patients, participants with physical or mental handicap and patients involved in forensic process.Gray literature which did not publish as articles, such as posters, organizational projects, class presentations, health messages and other similar literature did not meet the eligibility criteria.
Quality Assessment.For the assessment of observational studies, STROBE statement (Strengthening the Reporting of Observational studies in Epidemiology) was used.The statement consists of 22 items. (26)In addition, CONSORT checklist was used for the assessment of interventional studies. (27)ta Extraction.In order to find eligible studies, titles and abstracts from databases and additional sources were assessed by two authors independently.Then, the retrieved full texts were checked by the same two authors.Any disagreement was resolved by consultation with the third author.Unavailable data was requested from study authors.
Results
The initial search yielded 259 records on PubMed, 226 on Scopus, 215 on SID and 189 on Magiran.After removal of duplicates, the final records were 625.By reviewing the titles, the eligible articles reduced to 321.The selection by abstract brought the total to 74.The main purpose of the study was to select the studies have examined the special association between Orem self-care model and diabetic patients.This led to a significant reduction of eligible articles.Eventually, 9 studies met the inclusion criteria and were considered.Figure 1 shows the process of study selection.Two studies were from PubMed, whilst 6 were extracted from SID and Magiran.Additionally, one study gained from snowball method.Almost, all of the studies were interventional.There were 597 participants in this systematic review.(36) 8 out of 9 articles showed a positive statistically significant correlation between Orem self-care model and positive outcomes in diabetic patients.
The details and main findings are given in Table 1.For instance, it was found that implementing Orem's self-care model can lead to a decreased level of diabetic neuropathy. (28)As well, the study showed that the group receiving Orem's model has more appropriate level of self-care. (29)imilarly, as previously mentioned, other factors such as quality of life (31)(32)(33) and self-efficacy (30) had improved in the group receiving Orem's self-care model.
An overview of quality assessment of studies using STROBE statement and CONSORT checklist and is presented in Tables 2. For the quality assessment of observational studies, the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) was used.STROBE consists of a checklist with 22 items.The total scores lower than 17 were indicative of low methodological quality. (37).On the other hand, experimental and quasi-experimental studies were assessed by the CONSORT checklist.This checklist includes 25 items (38).Two authors were responsible for methodological quality of included studies.The third author resolved any disagreement between the reviewers.Seemingly, the included studies possessed acceptable quality.The effect of Orem self-care model on the improvement of symptoms and quality of life in patients with diabetes: A scoping review
Discussion
The results indicated Orem self-care model can lead to an improved level of self-efficacy, quality of life and self-care among the diabetic patients Also, the diabetic symptoms diminished.In this way, similar studies have showed that self-care behaviors can lead to the improvement of quality of life and reduction of costs. (39)It was proved that Orem self-care model is effective in changing the dietary habits among the participants, (35) as other studies have shown the efficacy of educational programs for improving the dietary behaviors and controlling the blood sugar are effective. (40)owever, the seasonal preferences need to be taken into account.For instance, in autumn, the desire for protein and fat increases, whilst patients in summer prefer to consume more carbohydrate. (41)Even, religious beliefs, culture and lifestyle can affect dietary habits.In other words, dietary habits are individualized. (42)Even though, culture can alter the perception of patients regarding healthcare providers. (43)me of the included studies were conducted on patients with diabetic foot ulcer and showed promising results. (29,34)Other similar programs conducted in Iran were indicative of effectiveness of self-care training programs among diabetic patients as they could be helpful in wound healing process. (44)Similarly, studies in other countries highlighted the pivotal role of educating patients for the self-care. (45)It is proved that controlling the diabetic foot ulcer mainly depends on the patient. (46)However, the factors such as marital status, educational level, gender, age and occupation paly important role in the level of self-care. (47)or instance, it is claimed that rate of self-care is higher in females compared with males. (48)ikewise, single patients pay less attention to their self-care, especially regarding diabetic foot ulcer. (48)It is because stronger emotional and social bonds exist among married patients. (49)e study participants were middle-aged and older adults.In this way, a study on older adults with diabetes showed health-related quality of life was moderate. (50)Though, married older adults had higher scores of quality of life. (51)Education was another multi-dimensional factor which can lead to higher contribution in the areas such economy, society and politics.Educational level directly affects the quality of life in the diabetic patients. (52)On the other hand, illiterate individuals are less familiar with coping strategies.Also, these individuals are in an inappropriate economic situation, which can cause more stressful situation and lower level of quality of life. (53)e results indicated that Orem self-care model in female patients with diabetes suffering from neuropathy can be effective in reeducation of symptoms and severity and it led in improvement of knowledge, attitude and other related skills. (28,36)However, their performance and self-care abilities was reported moderate. (36)Other studies conducted in Iran have also shown that training can be effective in communication skills and selfcontrol in patients with diabetic neuropathy. (54)In addition, a study in USA proved that education is effective in regulation of cognitive and emotional states. (55)e psychological effect of self-care programs is also undeniable.For instance, a study showed that self-management were effective in improvement of mood condition in diabetic patients. (56)However, many programs do not consider education in the area of psychological skills, as problems such as depression are common in diabetic patients. (57)A study in Iran showed psychological training can alter the level of stress, anxiety and depression among diabetic patients. (58)Additionally, existential need in patients encourages them to seek meaning in life and this can help to increase their wellbeing. (59)In explaining the role of the spiritual dimension of patients, it can be acknowledged that patients who receive more services have more ability to deal with psychological pressure and psychological trauma. (60)For the success in implementing such programs, the imperative role of economy should not be neglected.Economic well-being and availability of financial services can be effective in medical adherence of patients. (61)Naturally, unemployed individuals or those with a low-income profession are more vulnerable to experiencing adverse effects of the disorders. (62)dditionally, comorbidities play an important part in lack of self-care behaviors in diabetic patients. ( Obstacles such as physical limitations, polypharmacy, adverse symptoms and lack of social and emotional support exist that can have negative impact on self-control and selfcare behaviors. (64)However, it was claimed that increasing comorbidities result in poor diabetes self-care. (65)e of the findings was focused on gestational diabetes, which indicated Orem self-care model can be effective for the patients. (30)Similarly, other studies also proved that self-care training can bring about positive outcome both in Iran and other countries. (66,67)However, one study claimed that education cannot alter the scores related to blood sugar. (68)Specially, after COVID-19 pandemic, the psychological problems have increased, which it may be reduced in gestational diabetes by self-care training. (66,69)Notably, lack of knowledge regarding health is a risk factor for poor self-management in women suffering from gestational diabetes. (70)Likewise, It is believed that lack of social support is another important risk factor for women and has adverse effects on gestational outcomes. (71)One of the important steps toward coping behaviors in gestational diabetes is acceptance.Emotional stability is a factor which may lead to the acceptance. (72) conclusion, looking at the relation found between Orem self-care model and diabetic patients' outcome, it can be stated the model can be effective for the diabetic patients.Positive and supportive implementation of the model can improve patients' quality of life, self-efficacy, and self-care behaviors and reduce diabetic symptoms.The studies analyzed in this review have revealed that Orem self-care has a significant association with positive outcomes among diabetic patients.This shows that Orem self-care model can induce transformational alterations in self-care and quality of life in diabetic patients.However, future studies need to focus on other outcomes related to diabetes as it is a multi-dimensional disease.Specially, psychological outcomes need to be considered.
Limitations.Despite the efforts to present a perfect systematic review, there were some shortcomings.In practice, the studies were included if were published in English or Persian owing to inclusion criteria.Also, the full text of one article was not available.Additionally, there were various tools for the measurement of outcomes.Although the search strategy was designed precisely and snowball search of included studies was considered, there may be some missing data.Absolutely, there is need for conducting more studies to complement the findings of current study.
The effect of Orem self-care model on the improvement of symptoms and quality of life in patients with diabetes: A scoping review Inclusion/Exclusion Criteria.The inclusion criteria were as following: Studies published in English or Persian with full text available, participants aged 10-65 years, participants with the diagnosis of diabetes mellitus at any stage of the disease.Included interventions were those that aimed to implement Orem model.Moreover, the intervention can be emanated from either inpatient or outpatient setting.Additionally, randomization may reduce the risk of bias, it may not be fully possible in health care.Hence, the studies with nonrandomized or one-group study designs were included.
Table 1 . CONSORT and STROBE checklist scores of selected studies Author and year of publication Reference Type of study Score
*: CONSORT score; †: STROBE score Invest Educ Enferm.2024; 42(1): e08
Table 1 . Overview of Included Studies. (Cont.)
Randomized controlled Trial; NR: Not reported; N/A: Not Applicable ; F: Female ; M:Male Invest Educ Enferm.2024;42(1): e08The effect of Orem self-care model on the improvement of symptoms and quality of life in patients with diabetes: A scoping review | 2024-04-18T15:13:05.506Z | 2024-03-01T00:00:00.000 | {
"year": 2024,
"sha1": "b8819f2a7b254684ef819da6b7e9f5f96a7ca584",
"oa_license": "CCBY",
"oa_url": "https://revistas.udea.edu.co/index.php/iee/article/download/356797/20813795",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c505d84442b6d3d1e3bdb8d58e5b79bfced132e1",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
133555011 | pes2o/s2orc | v3-fos-license | NATURAL HERITAGE OF TAIMYR: CHALLENGES FOR ITS CONSERVATION AND SUSTAINABLE USE
The article discusses the natural heritage of Taimyr (northern Siberia) within the perspective of the environmental conditions of the region. The assessment of its significance is based on the results of many years of Russian-Dutch ecological expeditions. Issues discussed are the responsibility for the preservation of heritage, the interests of the local indigenous peoples, and the role of public authorities. Based on the results of a pilot project, the feasibility of developing ecotourism as one of the most effective forms of sustainable use of the heritage is demonstrated. Key words: natural heritage, Taimyr, Arctic, protected areas, indigenous people, Russian-Dutch expeditions, ecotourism
INTRODUCTION
The Taimyr Peninsula is one of the largest socio-economic regions of the Russian Arctic, characterized by a unique combination of environmental features and almost pristine state of natural systems. The uniqueness of Taimyr is due, primarily, to its geographical position: the Peninsula is the northernmost of the continental land masses not only in Eurasia, but in the world, resulting in a singular diversity of its landscape structure.
From north to south within the Taimyr Peninsula ( Fig. 1), there is a transition in natural landscapes ranging from the polar desert in the far north (75° N) to the foresttundra at the latitude of the Norilsk industrial complex. The enormous territory occupied by the Peninsula -about 1,200 km from east to west and over 1,000 km from north to south -resulted in a significant variation in the availability of natural resources: various mineral ores and precious stones, water resources, commercial hunting species and reindeer. Except for the heavily polluted zone around the axis Dudinka -Norilsk in the south, the harsh climatic conditions and remoteness of Taimyr from the developed areas of the country (no highways nor train connections to other parts of Siberia) contributed to the conservation of the fragile natural ecosystems of the Peninsula in an almost intact state. Through a combination of these factors, Taimyr today is one of the last major etalons of pristine nature in the world.
From this perspective, Taimyr is of particular interest in identifying its role in the dynamics of the global ecosystem that currently have a special topicality in relation to global change. This interest is manifest in the growing number of international research projects, among which a cooperation of more than 20 years between Russian and Dutch ecologists and geographers to study high-latitude ecosystems. The results of these studies, conducted mostly in the vast region of the Pyasina River delta, give reason to consider the ecosystems and the natural landscapes of the region as a whole to be an absolute priority category of natural resources to be preserved for the benefit of present and future generations, i.e. natural heritage.
NATURAL HERITAGE OF TAIMYR
The presence of large areas of pristine nature on the Taimyr Peninsula is reflected in the existence of the three largest Russian state nature reserves: the Great Arctic, the Putorana and the Taimyr Strict Nature Reserves. In the global system of nature reserves, they are characterised above all by a high level of biodiversity and occurrence of habitats for rare species. Precisely these reserves constitute the ecological framework of one of the 200 ecoregions of the world (Global 200) identified by WWF with the denotation "coastal tundra of Siberia and the Taimyr Peninsula", which to our opinion is not a very accurate indication.
The Pyasina River delta, among other areas of the Peninsula, has high biodiversity value; in 1995, it was designated as one of the clusters of the "Great Arctic" reserve ( Fig. 2). The Pyasina site covers an area of more than one million hectares of the delta of the Pyasina River and the adjacent large areas, as well as numerous islands of the coastal waters of the Kara Sea. The cluster projected so that its boundaries comprise the entire range of biological and ecological diversity aspects of the Arctic, that need of protection. The wetlands of The delta complexes are the most favorable habitat for water birds. For example, on the right-bank side of the Pyasina delta, in addition to numerous watercourses, rivers, and wetlands, there are over 260 lakes. The developed drainage system, significant reserves of food in the polygonal swamps and wetlands along the lowlands, and the lack of disturbance create optimal conditions for the existence of numerous water birds, which are one of the main objects of protection in the reserve (Fig. 4) (Fig. 5). Here, also red-breasted geese (Branta ruficollis) -a rare species emblemic for Taimyr -nest. In the last decade, there was a marked steady growth of this species and the expansion of its range to the north. At the same time, there was a decrease in the population of brent goose (with almost 100,000 individuals, compared with the beginning of the 1990s), which could not but cause concern of the international community of ornithologists and ecologists [Pakina, 2006]. In particular, this fact alarmed scientists from the Netherlands: the wintering black geese in this country that breed in the "Great Arctic" are symbolic for international biodiversity.
Studies conducted in the Arctic by scientists from different countries have revealed a close relationship between the state of populations of different species of birds and mammals. For example, the breeding success of many migratory birds is influenced by the condition of the population of lemmings (Fig. 6), whose population dynamics in recent years have been subject to serious disturbances. There is evidence of disruption of a three-year cycle of abundance of lemmings in the Scandinavian Arctic and the tundra of Alaska. Many scientists tend to explain this phenomenon of the nature by global warming. The Taimyr Peninsula is the last region on the planet where the cyclical fluctuations in the number of lemmings until 1994 remained normal. The need to trace the full range of complex interactions in arctic ecosystems that determine the success of reproduction of populations of migratory birds was the main reason for the organization of systematic research in the Pyasina delta. The idea was supported by the administration of the "Grand Arctic" reserve as well as by the Russian and the Dutch branches of World Wildlife Fund (WWF) and by the Royal Netherlands Embassy in Moscow.
The results of the international expedition confirmed the existence Based on the results of the recent expeditions, as well as on modern scientific concepts [Mazurov, 2005], we note the following aspects of the scientific value of the nature of Taimyr, exemplified by the Pyasina delta: preserved pristine state of the nature in general; etalon status of the natural complexes; high biological and landscape diversity of the territory and of aquatic systems; the presence of rare, valuable, and endangered species; high aesthetic value of landscapes; -relatively well scientifically studied area, especially in comparison with other regions equally difficult to reach.
The above-mentioned features confirm the importance of the functioning there of a Strict Nature Reserve -the institution designed to preserve the pristine nature and conduct relevant research and scientific observations of changes in the state of natural systems.
The findings on the value of natural systems of the Pyasina cluster can be extended to other parts of Taimyr. First, these are the ranges of existing natural protection areas such as, in particular, the Putorana state nature reserve. The uniqueness of the landscapes and the scientific value of the natural systems of the Putorana reserve provided the basis for its nomination to be included in the List of the World Heritage sites by UNESCO. In 2010, this proposal was adopted by the World Heritage Committee, which reflected the recognition of the global natural heritage values of the Taimyr region by the international community.
The existing nature reserves of Taimyr do not yet form a complete comprehensive ecological network in the region, as would be required in the interest of preserving the region's natural heritage. New components of this system will be the regional wildlife reserves "Agape" and "Gorbita" that were approved in the Taimyr Dolgan-Nenets Region (the authority competent for the Taimyr peninsula). Giving these habitats a conservation status will enhance the preservation of their highly valuable wetlands and aquatic areas of international significance. Further development of the ecological network of Taimyr will increasingly mitigate the growing risk to natural heritage of the region associated, primarily, with the expansion of economic activities in the North.
HUMAN GEOGRAPHY OF THE STUDY AREA
The most important indicator of the economic development of an area is its population or population density. Until now, the vast expanses of the Taimyr tundra have been virtually uninhabitable. The members of the expedition did not observe any traces of settlements, except for some seasonal structures of commercial fishermen. In general, adjacent to the "Great Arctic" reserve vast territories, as well as many other regions of the Far North, are characterized by extremely low population density.
The closest to the base camp at the mouth the Pyasina River is the small urban settlement of Dixon, located at a distance of about 200 km to the west (about an hour by helicopter). One of the oldest Russian hydrometeorological stations (founded in 1916) is located there. In the Soviet period, Dixon was one of the most important ports in the Arctic Sea Route, which predetermined the rapid development of this settlement.
In the post-Soviet period, the intensity of freight traffic in the Arctic Sea has fallen sharply, which resulted in the decline of Dixon.
The distance to the nearest settlement to the northeast and southeast of the base complex of the reserve, i.e., to the border post at Cape Chelyuskin, and to temporary camps of geologists on the islands the Severnaya Zemlya archipelago, as well as to the settlements in the Khatanga valley, is 600-700 km. Much closer to the camp is a small village of Ust-Tareya (about 150 miles in a straight line or three times longer by the meanders of the Pyasina basin). It is difficult to overestimate its importance for aviation operation in the region; however, it has been almost completely depopulated in recent years.
Under current conditions, transport of people and goods to the reserve is almost only possible by helicopter from the airport Valyok, located near the city of Norilsk in more than 500 km to the south. Communication with the developed parts of the continent is also possible by sea and river routes, but due to climatic conditions and the economic situation at the present time, these routes are almost never used.
Thus, the territory of the "Great Arctic" reserve is located in one of the most remote and inaccessable Arctic regions of Russia and the world at large, which makes it especially appealing for studying natural processes in ecosystems.
At the same time, the territory is a historical zone of vital interest of the local native peoples. This is reflected, among other things, in the local toponyms. Despite the predominance of Russian toponyms in the names of geographical features (rivers, lakes, mountains, capes, etc.) there are many native names, including those captured on modern maps, down to the coast of the Kara Sea. An example is the River Khutudabiga (Fig. 7), which means "river rich in life". The Nganasans, an indigenous people of Taimyr, gave this river its name, with reason. The river and its surroundings are known not only for the rich fish stocks but also for the abundance of game species, including many geese and wild reindeer.
The presence of local toponyms in the investigated region is the most obvious indication of this region's cultural heritage, represented, first of all, by the local indigenous population. Precisely these people have preserved to this day the extremely fragile nature of their land and specifically they are vitally interested in its preservation in the future. Therefore, the policy of conservation of natural heritage of the region should be formed in the interests of the native Taimyr peoples and with their participation.
THE POPULATION AND THE INDIGENOUS PEOPLE OF TAIMYR
Today, the population of the Taimyr Dolgan-Nenets region of the Krasnoyarsk Krai (the region corresponds to the former Taimyr Autonomous District) is about 53 thousand people. Within the Taimyr Dolgan-Nenets region, there is Norilsk industrial district (180 thousand population), which is administratively subject to the administrative center of Krasnoyarsk and is not part of the Taimyr Dolgan-Nenets region. The population density of this actual autonomous region, even including Norilsk, has an extremely low value of about four persons per km 2 , which is about 30 times lower than the average in Russia [Chistobaev, 2003].
The ethnic structure of the population is dominated by Russians who came to the Taimyr Peninsula in the XVII th century, when the region was annexed by the Russian Empire. They are followed by Ukrainians and Tatars. A particularly rapid growth of these and of other ethnic groups took place in the XX th century, which was due to economic development of the region. A crucial role in economic development of Taimyr was played by the Norilsk industrial complex and the associated construction of the sea ports in Dixon and Dudinka.
The large-scale economic development of Taimyr that began in the Soviet period, radically changed the lives of the indigenous population. It was almost completely converted from a nomadic to a settled way of life. These people were consistently subjected to a policy of state paternalism, incompatible with fundamental interests of the indigenous people, which often gave rise to social and socio-environmental problems that even grew dramatically in the post-Soviet period. In recent years, there is a steady upward trend in the Taimyr indigenous population, mainly due to the excess of births over deaths. Over the period between the 2002 and 2005 censuses, it increased with 320 people, a relative increase of 1,07%. However, there is a significant variation in these indicators among ethnic groups, which does not allow considering this situation safe from the standpoint of the stability of the native population. Thus, the demographic growth of the Nenets -the second largest indigenous group in the region -is approximately 1,5 times higher than that of Dolgans, who are the largest group [Chistobaev, 2003]. But an even greater contrast is in the reproduction of the population between the smallest groups of indigenous people, the Taimyr -Evenks and Entsys: the former group shows the largest increase in the region (3,67%), while the latter group suffers a threatening decrease of -8,79%, close to the depopulation. However, as evidenced by numerous data, this adaptation took place not without affecting the mentality of the natives. In urban areas, it is more difficult to maintain traditional working practices and cultural customs. The situation was complicated by the arrival of "civilization diseases": social dependency, alcoholism, crime, chronic illness, suicide, etc. The above problems worsened dramatically in the post-Soviet period, superimposed on the explicit and hidden unemployment, which hit the hardest the indigenous people of the North. Almost all of them were unable to adapt to market conditions, to live in a competitive environment that is incompatible with the traditional mentality and the system of spiritual values of these people.
It is obvious that the current difficult socioeconomic situation in the Russian North actually "squeezes out" the indigenous people from urban and rural settlements to the tundra and forces them to return to the traditional economy, which, though not promising prosperity, can, under favorable conditions, ensure survival.
In this situation, it is objectively inevitable that Dolgans and other ethnic groups turn to the previously neglected Taimyr grazing, hunting, and fishing lands. Naturally, this interest will manifest itself in the claims of indigenous communities and/or families in recovery and preservation of property rights to their ancestral lands, covering, apparently, almost the entire territory of Taimyr. In the medium to long term, one should expect the spread of these claims to the most northern regions as well, not excluding the study area of the expedition, the Pyasina delta. One might also assume that in the future, it may be the ethnic interests that will be the decisive factor for the regional policy on the Taimyr Peninsula and the Russian North in general.
The territory of the Taimyr Peninsula is the living environment of the local indigenous people, first of all of Dolgans and Nenets, to which they have the undoubted historical rights. Obviously, in these circumstances, these rights in some cases may compete or even conflict with the federally designated status of certain areas, such as nature reserves. In this regard, improvement of the legislative base for the northern territories should take into account the current situation and prospects of its development.
An important component of the living environment of the local indigenous population, who in the long term will be increasingly involved in the economy, is its natural heritage. This perspective includes both positive opportunities as well as the inevitable risks. Let us illustrate this with the example of ecotourism as the most promising use of natural heritage.
OPPORTUNITIES AND RISKS OF ECOLOGICAL TOURISM DEVELOPMENT
Until now, the fishing and recreational appeal of the Taimyr coast have been known to only a relatively limited number of individual fishermen and rare tourists, whose impact on the nature was narrowly local and relatively small. However, in view of the above-mentioned circumstances, the situation can quickly change toward, for example, the formation of a sufficiently substantial tourism (ecotourism, sport fishing and hunting tourism) and a concurrent increase in tourist numbers associated with environmental risks.
The investigation on the potential for the development of ecological tourism in the "Great Arctic" reserve was conducted considering the fact that tourism, even ecotourism, is also, apart from the obvious advantages over traditional types of tourism, a definite threat to the natural systems. Today, the definition of "ecological" is often used merely as a marketing label to create a more appealing image of this kind of activity. Ecological tourism in reality is a complex interaction of various structures, including regional and local administrations, the management of nature reserves, private businesses, local communities, etc.
In the summer of 2005, a pilot project on ecotourism in the northwest of the Taimyr Peninsula [Henkens et. al., 2005] was implemented. Along with the assessment of conditions for the development of ecotourism at the field camp at the Pyasina Delta, the prospects of an ecological tour there were evaluated. The Khutudabiga River was chosen as an object of investigation, which represents great interest in terms of landscape and biological diversity, as well as being traditionally attractive for the organization of sport fishing. In the course of the project, the conditions for the arrangement of the ecological tours in this region, including all possible types, i.e., educational, sport, adventure, and fishing, were assessed.
The Khutudabiga River is located in the northwestern part of the Taimyr Peninsula and pours directly into the Kara Sea in its southeastern part. The mouth of the river is about 30-40 km north from the Pyasina River mouth. The northern boundary of the Pyasina site of the state nature reserve "Great Arctic" runs along the Khutudabiga River.
The remoteness of the area, poor transport accessibility, and the need to use the most expensive type of transport, i.e., helicopter, resulted in a high degree of conservation of the local natural systems and of their biological and landscape diversity.
A necessary premise for tourism development and for its effective management is to identify issues that are of interest for the potential visitors to the reserve and to the adjacent areas. The development of tourism on the Taimyr Peninsula involves the use of all available there specific characteristics defined by the natural environment.
Opportunities for tourism development in the region are limited, by a short tourist season, harsh conditions, and a poorly developed transport network. However, these aspects are also advantageous for targeting the group of tourists interested in the opportunity to visit extremely remote areas unaffected by human activities.
Comparing the Taimyr region with other regions of tourist attention, its main features can be identified as follows [Ebbinge and Mazurov, 2006 The studies were conducted from July 26 to August 16, in the most favorable period for the organization of tours. In the area, the beginning of mid-July to early September, is characterized by the best mild weather conditions. Snow cover is virtually non-existent, snow patches remaining only in the lowlands on the northern slopes, and the tundra shows all its variety of colors and biological activity. The air temperature ranges from +5-+10°C, occasionally rising to +16-+18°C or more. Mosquitos and midges are rare.
The fauna of the study area is often described as "poor", which is understandable considering harsh natural conditions. At the same time a high degree of preservation of the natural systems, vast open spaces, and the polar day create a unique opportunity to observe much of the animal world, including rare species. Numerous water birds and wader birds can easily be observed at suitable locations. On the ledges of coastal cliffs there are often nests of peregrine falcons, snowy owl, and other rare birds. Different types of waders and quail are common to the tundra. Among the mammals, one is likely to see a fox ( Fig. 8), reindeer, and musk ox, sometimes wolf , and many lemmings in years of peak population. Directly at the bay, i.e., in the point of contact of the river and the sea, there are sightings of seals, as well as of "sea hare" -bearded seal. There are also sightings of whales at the mouth of the Khutudabiga.
In the area of the start of the tour (approximate coordinates: 74°30'N, 90°00'E), a group of four was dropped by helicopter along the Kutudabiga. They continued farther down the river by inflatable boats. Rafting on the river was done in a mixed mode, alternating with sufficient stops to get acquainted with the natural phenomena. In some of the most picturesque places, suitable for placement of the camp, the stays were longer, up to two days. In places most favorable for fishing, camps were set. Fishing was organized based on the "catch and release" principle, which, in general, is typical for fly fishing, the most sporting method of fishing that is also the most gentle and "environmentally friendly. " Experience has shown that tourism development on the Taimyr Peninsula is quite feasible. Decisive for the development of this sphere of activity are such aspects as ensuring the safety of tourists, especially important in the Arctic, and increasing the service level. In addition, transport accessibility remains one of the main factors determining the development of ecotourism in remote regions in the short term.
Sustainable tourism can contribute to the international status of the "Great Arctic" nature reserve and the involvement of the local communities in the management of the reserve. One can assume that this will also help reducing the incidence of unauthorized hunting and fishing on its territory, as well as providing an additional source of income. The benefits of the development of this work may be directed toward increasing the involvement of participation of the local population in the management of the reserve and the development of longterm cooperation with other community agencies engaged in the management of the Taimyr region.
There is no doubt that the organization of environmental and any other types of tourism in the territories of national parks should be accompanied by monitoring changes in the natural environment, ecosystems, etc.
Currently, many areas of the Russian North, characterized by a significant natural heritage, have a protection status. For professionals it is clear that designation of such a status improves the protection of their natural and cultural heritage. In this case, the protected areas actually acquire a new quality: they are universal human heritage subjected to a strict social imperative of their priority preservation as public good. This fact largely determines the relatively high preservation requirements of natural systems within the boundaries of nature reserves. At the same time, economic realities require a more flexible approach to the conservation of natural resources. Ecotourism is one of the few options for meeting the needs of society in preserving their heritage and wise and sustainable use of natural resources in the interests of not only present but also future generations.
CONCLUSION
The foregoing discussion suggests the following conclusions regarding the natural assets of the Taimyr Peninsula: they are of particular natural and cultural heritage value, implying the need for conservation; they show a large spatial variation and characteristic temporal dynamics; they are of interest to different users, whose number may increase and interests may give rise to competing claims; they are vulnerable to a number of risk factors, induced by various uses of which the incidence and intensity in the future may increase significantly; they are within the competence of various administrative bodies that still are in the These features fully allow inclusion of the natural assets discussed above into the category of "natural heritage" [Mazurov, 2005], i.e. having special values as public goods, which suggests first of all their preservation for the benefit of present and future generations. The need to preserve this heritage as a basis for maintaining the ecological balance in the Russian Arctic and harmonious development of indigenous people must be supported by an appropriate legislative framework at both the regional and federal levels.
Until now, the Russian government carried the responsibility for preserving the unique natural heritage of Taimyr, including the establishment and maintenance of nature protection sites and other protected areas, such as the "Great Arctic" reserve. The State represents the interests of the entire population, including the local indigenous population who is objectively interested the most in preserving the heritage. It can be argued that, as awareness of their interests and of their role in preserving their natural heritage grows, the Taimyr indigenous people will take a more active role in the responsibility for it, which fully corresponds to their own cultural traditions. Responsibility for the preservation and for passing on the unique and precious natural heritage of Taimyr to future generations is now with the local indigenous population and the population of Russia as a whole.
A situation is gradually developing now where the natural environment of Taimyr is increasingly being perceived as a heritage, the preservation of which is of equal interest for the entire local population, the Russian State, a number of foreign nations -the countries associated with Russia at the ecosystem level (Fig. 9), and, to some extent, the world community. Largely due to the international nature of research, understanding is being formed that there is no second Taimyr in the world and that this Taimyr plays an indispensable role in ensuring a sustainable and prosperous future for mankind. | 2019-04-26T14:25:58.250Z | 2012-01-01T00:00:00.000 | {
"year": 2012,
"sha1": "61888f3548410f848e605ae1186df2706b5d3a8c",
"oa_license": "CCBY",
"oa_url": "https://ges.rgo.ru/jour/article/download/172/170",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "16c3a3f2375fc35d57f447d535aec12912cbead2",
"s2fieldsofstudy": [
"Sociology"
],
"extfieldsofstudy": [
"Geography"
]
} |
246434208 | pes2o/s2orc | v3-fos-license | Characterizing the Physicochemical Properties of Two Weakly Basic Drugs and the Precipitates Obtained from Biorelevant Media
Generally, some weakly basic insoluble drugs will undergo precipitate and redissolution after emptying from the stomach to the small intestinal, resulting in the limited ability to predict the absorption characteristics of compounds in advance. Absorption is determined by the solubility and permeability of compounds, which are related to physicochemical properties, while knowledge about the absorption of redissolved precipitate is poorly documented. Considering that biorelevant media have been widely used to simulate gastrointestinal fluids, sufficient precipitates can be obtained in biorelevant media in vitro. Herein, the purpose of this manuscript is to evaluate the physicochemical properties of precipitates obtained from biorelevant media and active pharmaceutical ingredients (API), and then to explore the potential absorption difference between API and precipitates. Precipitates can be formed by the interaction between compounds and intestinal fluid contents, leading to changes in the crystal structure, melting point, and melting process. However, the newly formed crystals have some advantageous properties compared with the API, such as the improved dissolved rate and the increased intrinsic dissolution rate. Additionally, the permeability of some precipitates obtained from biorelevant media was different from API. Meanwhile, the permeability of rivaroxaban and Drug-A was decreased by 1.92-fold and 3.53-fold, respectively, when the experiments were performed in a biorelevant medium instead of a traditional medium. Therefore, the absorption of precipitate may differ from that of API, and the permeability assay in traditional medium may be overestimated. Based on the research results, it is crucial to understand the physicochemical properties of precipitates and API, which can be used as the departure point to improve the prediction performance of absorption.
Introduction
Physiologically based pharmacokinetic (PBPK) models have been widely used to predict the pharmacokinetic characteristics of compounds, but they cannot always capture the absorption process, leading to low confidence (greater than 2-fold) in the prediction performance [1,2]. In fact, the oral absorption of some insoluble drugs is susceptible to gastrointestinal (GI) conditions, e.g., the pH in the GI tract, the bile micelle concentration under both fasted and fed states, and so on, which can impact the solubility and permeability of drugs, increasing the difficulty of accurately estimating the absorption characteristic of compounds [3][4][5][6].
Additionally, oral absorption is a complex process in the GI tract, especially concerning salts and weakly basic insoluble drugs. Generally, the weakly basic insoluble compounds show good solubility in the acidic gastric environment through ionization, but the advantage dissipates when the less soluble unionized compounds are formed in the near-neutral intestinal microenvironment. Nucleation will occur when a large number of the unionized compounds are continuously concentrated to form a supersaturation solution, followed by crystallization and precipitate [7,8]. Moreover, the precipitate can be redissolved again and cross the intestinal barrier after the full absorption of the dissolved unionized compound [9]. Therefore, one compound can undergo dissolution, supersaturation, precipitation, and redissolution before transporting across the barrier tissue into the systemic circulation in theory [10]. Many studies have been carried out on the supersaturation and precipitate of the weakly basic insoluble drugs in vivo, and the phenomenon of precipitate does exist in vivo [11][12][13].
To improve the prediction accuracy of absorption for insoluble compounds, the PBPK absorption model focuses on the dynamic process of the compounds in the GI tract. Although many factors, such as supersaturation, precipitation time, and dissolution profiles [14,15], have been considered in the development of the PBPK absorption model for poorly soluble drug candidates, it is unclear whether the absorption process of the new crystals is the same as that of the API in the intestinal microenvironment. Nucleation and crystallization occur in the intestinal environment, which may be affected by the chemical bond or the properties of the intraluminal components in intestinal fluid [16,17]. Some studies have reported that the crystallization structure is different from the API [7,11]. Therefore, it is crucial to characterize the physicochemical properties of crystals to better understand the absorption differences between precipitates and API.
Considering that biorelevant media, simulating physicochemical parameters and components of intestinal fluid, is commonly used in dissolution tests to evaluate the potential dissolution processes of formulations in vivo in advance [15], it is an alternative method for gathering sufficient crystals in biorelevant media in vitro. Therefore, evaluation of physicochemical properties in vitro is a workable plan. Herein, the crystals of model drugs were collected in biorelevant media and their physicochemical properties were evaluated systematically by X-ray diffraction, scanning electron microscopy, Fourier transform infrared spectroscopy, differential scanning colorimetry, permeability assay, and so on. Through these explorations, we hope to shed light on the potential absorption difference between API and the redissolved precipitates and provide some points for reasonably improving the prediction accuracy of absorption.
Materials
Rivaroxaban (pKa 13.6, log p 1.90, purity ≥ 98%) and Drug-A (pKa 10.41, log p 2.102, purity ≥ 98%) were selected as weakly basic model drugs in this manuscript. Both of them were obtained as gifts from sponsors. Biorelevant powder, a commercial product of premixed bile salts and phospholipids, was purchased from Biorelevant.com Ltd. (Surrey, UK). Chemicals were purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China), except for methanol, acetonitrile, and potassium bromide obtained from Merck (Merck KGaA, Darmstadt, Germany). All reagents were analytical grade and were used without further purification. Parallel artificial membrane permeability assay (PAMPA), acceptor sink buffer, and lipid solution were obtained from Pion Inc. (Billerica, MA, USA).
Equilibrium Solubility of API in Different Media
To select an appropriate method for the preparation of precipitates, the equilibrium solubility of model drugs was firstly investigated in various media, including pH 1.0 HCl solution, pH 5.0, pH 6.5, and pH 8.0 phosphate buffer solution (PBS), fasted state simulated intestinal fluid (FaSSIF, Fa), and fed-state simulated intestinal fluid (FeSSIF, Fe) [18]. The excess sample was added to the medium, and the test of equilibrium solubility
Preparation of Precipitates
The biorelevant media, including FaSSIF and FeSSIF media, were prepared based on the instructions from biorelevant.com. The preparation method of double diluted (0.5 × FeSSIF, 0.5 Fe) and double concentrated (2 × FeSSIF, 2 Fe) FeSSIF media was similar to FeSSIF medium, but the amount of biorelevant powder was reduced and increased to twice that of FeSSIF medium, respectively. The composition of the media is summarized in Table 1 [19]. The precipitates of the compound were prepared by the solvent shift method [20]. Rivaroxaban and Drug-A were first dissolved in a minimal amount of glacial acetic acid and acetonitrile, respectively. Then, the concentrated compound solutions were separately added into the biorelevant media (FaSSIF, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF) and aqueous buffer media (FaSSIF buffer (FaB) and FeSSIF buffer (FeB)). The precipitate in each medium was obtained according to the solubility difference of model drugs in organic solvent and the biorelevant media/aqueous buffers. The precipitates were collected by vacuum suction filtration and washed three times with a few milliliters of distilled water. The samples were dried in a vacuum drying oven at 40 • C for 24 h and stored with desiccators until further analysis [21].
Recovery of Precipitates
The content of rivaroxaban and Drug-A in the corresponding crystal was determined with a known concentration solution, which was prepared by dissolving a certain amount of precipitate in a volumetric flack [22]. Then the absolute content of corresponding crystals was determined by high-performance liquid chromatography [23] (HPLC, Agilent Technologies Santa Clara, CA, USA) and compared with that of plain API solutions. All these tests were performed three times.
X-ray Diffraction of Precipitates and API
The diffraction patterns of API and precipitates that were obtained from the organic solvent, FaSSIF buffer, FeSSIF buffer, FaSSIF, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media, respectively, were analyzed by X-ray diffraction (XRD) using Rigaku ultima IV with CuKα radiation (λ = 0.154184 nm) operating at 40 KV and 40 mA. The pattern was recorded for angles (2θ) in the range of 0-45 degrees at a scanning rate of 4 • C/min [24].
Crystal Morphology of Precipitates and API
The surface morphology of different crystal structures was characterized by scanning electron microscopy (SEM, Zeiss SUPRA 55 VP FE-SEM; Carl-Zeiss Jena; Jena, Germany). All samples were coated with a moderate gold-palladium layer and then observed and recorded at an acceleration voltage of 10.0 kV [25].
Fourier Transform Infrared Spectroscopy
A Fourier transform infrared spectroscope (FTIS, Thermo Fisher Scientific Inc., Waltham, MA, USA) was used to evaluate the transition of functional groups of precipitates after the interaction between the compound and molecules in solutions. Samples were ground with spectroscopy grade KBr at the weight ratio of 1:50 and then compressed into a semitransparent pellet under the pressure of 20 tons for 30 s. The spectra signal was collected 16 times in the range of 400-4000 cm −1 with a resolution of 4 cm −1 for each sample [26]. The OMNIC 9.3 was used for FTIR data analysis.
Differential Scanning Calorimetry
Thermal analyses of various samples were carried out by differential scanning calorimetry (DSC Q20, TA Instruments, New Castle, DE, USA). Indium was firstly used to calibrate the instrument. Then, the samples were weighted accurately and placed in aluminum pans. The temperature was programmed at a rate of 10 • C/min with a starting temperature of 100°C. Nitrogen at a flow rate of 40 mL/min was used as purge gas [26].
Equilibrium Solubility of Precipitates and API in pH 6.5 Aqueous Buffer
To estimate the solution properties of different crystal structures in near neutral or slightly acidic environments, the equilibrium solubility of various precipitates and API in the pH 6.5 FaSSIF buffer medium (FaB) was investigated. An excess sample was added to the medium, and the test of equilibrium solubility was performed with the stirring rate of 200 revolutions per minute (rpm) under 37 • C for 24 h. Sample concentration was detected every minute by Pion's in situ fiber-optic UV dissolution monitoring system. All measurements were performed three times.
Intrinsic Dissolution Rate of Precipitates and API
The intrinsic dissolution rate (IDR) of compounds with different crystal structures was measured by the disc method based on the literature [27][28][29][30]. About 10 mg of powder was loaded into the Mini-IDR compression system (Pion Inc., Billerica, MA, USA) and compressed at 100 psi (rivaroxaban) and 200 psi (Drug-A) for 1 min to obtain the miniaturized discs with a surface area of 0.071 cm 2 . The discs were inserted into stirrer-die assembly and placed into flat-bottomed glass vials for dissolution analysis. At the start of the experiments, the magnetic stirrer was immediately switched on with the speed of 200 rpm, and then 10 mL of preheated (37 • C) dissolution media (FaSSIF buffer medium, FeSSIF buffer medium, FaSSIF medium and FeSSIF medium) were added into the vials at a constant temperature. The concentration was determined directly in the dissolution media every 30 s using an in situ UV probe for 2 h. All experiments were performed in triplicate, and the IDR was calculated with the following equation [30]: where V is the volume of medium in the glass vial; dc/dt is the slope of the straight line from the dissolution profile; A is the surface area of the disc.
Permeability Assay of Precipitates and API in Different Media
In order to simulate the permeation of compounds in different crystal structures in the GI tract, the stirring PAMPA method (Pion Inc., Billerica, MA, USA) was used to assess the passive transcellular permeability. The donor and acceptor chambers representing the GI tract and systemic circulation were separated by the artificial membrane (PVDF, polyvinylidenfluoride, 0.45 µm, 0.78 cm 2 ). The membrane surface was coated with 20 µL of lipid, which was used to simulate the microenvironment of lipids in vivo. Then, 16 mL PBS/FaSSIF/FeSSIF medium with pH 5.0 was added into the donor chamber, and 16 mL sink buffer was added into the acceptor chamber. In order to simulate the redissolution progress of precipitate in the GI tract, a slight excess of crystalline powder was added into the donor chamber. Both chambers were stirred at 200 rpm at 37 • C, and the concentration of each chamber was detected by immersed UV-probes at 320 nm and 340 nm for rivaroxaban and Drug-A, respectively. The flux was defined as the amount of compound passing through the unit area perpendicular to its flow direction in unit time and was calculated by the following equation [31]: where J(t) is the flux of a compound; C t is the concentration of compound at t; V is the solution volume of the donor chamber; A is the area of the artificial membrane. The correlation between J and the effective permeability (P e ) can be described as follows [32]: where J is the flux of the compound; C d is the concentration of the compound in the donor chamber; C a is the concentration of the compound in the accepter chamber; P e is the effective permeability of the compound. Since the concentration of the compound in the receptor is very low at the beginning of the experiment, assuming that the concentration of API in the receptor chamber is zero, the equation can be simplified as follows [32]:
Statistical Analyses
The statistical analyses regarding significance levels and correlation analysis were assessed with the GraphPad Prism Version 8.0 (GraphPad Software, San Diego, CA, USA), and a p-value < 0.05 was used to estimate the significance.
Solubility of the API in Different Media
The equilibrium solubility of rivaroxaban and Drug-A in various aqueous buffers and biorelevant media was firstly tested. The detailed results are shown in Table 2. Because the change of pH value of the medium had no significant effect on the solubility of model drugs, the solvent shift method was used to prepare precipitates [20].
Recovery of Precipitates
The content of rivaroxaban and Drug-A in precipitates were accurately determined with HLPC. The absolute quantitative method for rivaroxaban refers to the published literature [23], and Drug-A refers to the declared data of new drug applications (Unpublished). The recoveries of all the test samples are within the range of 95-101%, indicating that there is little loss during the recrystallization. The detailed results are shown in Table 3.
X-ray Diffraction Results of Precipitates and API
XRD analyses are performed to identify crystalline phases, and the detailed results are shown in Figure 1. There is little effect of organic solvent on the characteristic peak of rivaroxaban and Drug-A, respectively. However, the media have different influences on the crystal structure. Aqueous media mainly affect the peak strength rather than the peak position, and the presence of bile salts in the medium can change the crystalline natures, as the characteristic diffraction peaks appear at 2θ equal to 3 and 9 degree for rivaroxaban, and 6, 32, and 41 degree for Drug-A. Moreover, changes in bile salts concentration can impact the relative strength of peaks. Therefore, the crystal lattice structure and lattice energy of compounds have been changed during the formation of precipitates.
X-ray Diffraction Results of Precipitates and API
XRD analyses are performed to identify crystalline phases, and the detailed results are shown in Figure 1. There is little effect of organic solvent on the characteristic peak of rivaroxaban and Drug-A, respectively. However, the media have different influences on the crystal structure. Aqueous media mainly affect the peak strength rather than the peak position, and the presence of bile salts in the medium can change the crystalline natures, as the characteristic diffraction peaks appear at 2θ equal to 3 and 9 degree for rivaroxaban, and 6, 32, and 41 degree for Drug-A. Moreover, changes in bile salts concentration can impact the relative strength of peaks. Therefore, the crystal lattice structure and lattice energy of compounds have been changed during the formation of precipitates.
Scanning Electron Microscopy Analyses
To investigate the possible microscopic morphology of crystals in GI tract, API and precipitates obtained from FaSSIF and FeSSIF media were characterized by SEM. The different rulers were selected to capture the complete crystal morphology based on the size difference of crystals. Figure 2A show the microscopic morphology of rivaroxaban (API). Clearly, there are some nearly spherical particles with the size range of 1-10 µm, and some small particles stick together to form larger particle clusters. Figure 2B present the morphology of precipitate (rivaroxaban) obtained from the FaSSIF medium. Some lamellar blocks with different sizes were recorded, and the thick sheet block consisted of many thin sheets with a thickness of several nanometers. The microscopic morphology of precipitate (rivaroxaban) obtained from the FeSSIF medium is shown in Figure 2C, and there exist some loose blocks with many irregular small particles adhered to the surface. Figure 2D (API of Drug-A) display many small sugar block particles with a clear surface contour, and the particle size is less than 1 µm. The microscopic morphology of precipitate (Drug-A) obtained from FaSSIF medium ( Figure 2E) presents a similar morphology to that of precipitate (rivaroxaban) obtained from the FeSSIF medium ( Figure 2C). Namely, some large block samples have many small particles adhered to the surface. Figure 2F present many long rod-shaped blocks with different sizes, and the largest size is up to dozens of microns. Overall, the precipitates obtained from FaSSIF and FeSSIF media have an obvious cuboid structure with fine particles attached to the surface. Moreover, the particle size of precipitates is larger than that of API. tate (Drug-A) obtained from FaSSIF medium ( Figure 2E) presents a similar morphology to that of precipitate (rivaroxaban) obtained from the FeSSIF medium (Figure2C). Namely, some large block samples have many small particles adhered to the surface. Figure 2F present many long rod-shaped blocks with different sizes, and the largest size is up to dozens of microns. Overall, the precipitates obtained from FaSSIF and FeSSIF media have an obvious cuboid structure with fine particles attached to the surface. Moreover, the particle size of precipitates is larger than that of API.
Fourier Transform Infrared Spectroscopy Analysis
Since the crystalline properties of precipitates were different from those of API, FTIR analysis was used to determine whether the functional groups of compounds had changed in the microenvironment. Figure 3 show the FTIR spectra of APIs and precipitates of rivaroxaban and Drug-A. The characteristic peaks of the rivaroxaban related samples appear at the same absorption position, indicating that precipitates and API are still the same compounds with different crystal structures ( Figure 3A). However, Drug-A
Fourier Transform Infrared Spectroscopy Analysis
Since the crystalline properties of precipitates were different from those of API, FTIR analysis was used to determine whether the functional groups of compounds had changed in the microenvironment. Figure 3 show the FTIR spectra of APIs and precipitates of rivaroxaban and Drug-A. The characteristic peaks of the rivaroxaban related samples appear at the same absorption position, indicating that precipitates and API are still the same compounds with different crystal structures ( Figure 3A). However, Drug-A related precipitates obtained from the media containing acetic acids, such as FeSSIF buffer, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media, have two new absorption peaks at the position of 1750 cm −1 and 1725 cm −1 , which may be caused by the stretching of C=O bonds of saturated fatty acid monomers (acetic acid). In order to confirm the type of carbon-oxygen double bond, API and these precipitates were dissolved in acetonitrile and analyzed by mass spectrometry. Both the parent ions and product ions between API and the precipitates obtained from medium containing acetic acid are the same (Unpublished). Combined with the structural formula of Drug-A, the basic drug may interact with the acetic acid and form carboxylate.
Differential Scanning Calorimetry Analysis
Under the same conditions, there is no significant difference in melting endotherm among the precipitates of rivaroxaban ( Figure 4A). While for Drug-A, the melting points and melting process are changed dramatically with precipitates obtained from FeSSIF buffer, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media. As shown in Figure 4B, the endothermic peak of precipitates obtained from organic solvent, FaSSIF buffer, and FaSSIF media is very sharp but precipitates obtained from FeSSIF buffer, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media display a smooth and shallow peak during the endothermic process, respectively. Changes in thermostability may be related to the newly formed carbon-oxygen double bond after the interaction between the Drug-A and acetic acid in the buffer medium. Hence, the composition of supersaturated solutions can influence the crystal structure and then affect the thermostability of the compound. position of 1750 cm −1 and 1725 cm −1 , which may be caused by the stretching of C=O bonds of saturated fatty acid monomers (acetic acid). In order to confirm the type of carbonoxygen double bond, API and these precipitates were dissolved in acetonitrile and analyzed by mass spectrometry. Both the parent ions and product ions between API and the precipitates obtained from medium containing acetic acid are the same (Unpublished). Combined with the structural formula of Drug-A, the basic drug may interact with the acetic acid and form carboxylate.
Differential Scanning Calorimetry Analysis
Under the same conditions, there is no significant difference in melting endotherm among the precipitates of rivaroxaban ( Figure 4A). While for Drug-A, the melting points and melting process are changed dramatically with precipitates obtained from FeSSIF buffer, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media. As shown in Figure 4B, the endothermic peak of precipitates obtained from organic solvent, FaSSIF buffer, and FaSSIF media is very sharp but precipitates obtained from FeSSIF buffer, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF media display a smooth and shallow peak during the endothermic process, respectively. Changes in thermostability may be related to the newly formed carbonoxygen double bond after the interaction between the Drug-A and acetic acid in the buffer medium. Hence, the composition of supersaturated solutions can influence the crystal structure and then affect the thermostability of the compound.
Equilibrium Solubility of Different Precipitates and API in the Same Medium
Thermodynamic properties, e.g., melting point and fusion enthalpy of crystals, and the interaction between solutes and compounds can jointly influence the equilibrium solubility of a compound in a specific medium. Hence, the equilibrium solubility of these precipitates was measured in FaSSIF buffer medium at pH 6.5. As shown in Figure 5A (rivaroxaban related), the dissolved rates of all the precipitates are higher than that of API, and the extent (achieved in about 4 h and 24 h) is almost the same. Interestingly, the dissolved rates ( Figure 5B) for Drug-A and its precipitates are different. Some precipitates (Fa and 0.5 Fe in Figure 5B) dissolved more slowly and had less solubility than that of API in the first 8 h. Precipitates obtained from FeSSIF and 2 × FeSSIF media have high
Equilibrium Solubility of Different Precipitates and API in the Same Medium
Thermodynamic properties, e.g., melting point and fusion enthalpy of crystals, and the interaction between solutes and compounds can jointly influence the equilibrium solubility of a compound in a specific medium. Hence, the equilibrium solubility of these precipitates was measured in FaSSIF buffer medium at pH 6.5. As shown in Figure 5A (rivaroxaban related), the dissolved rates of all the precipitates are higher than that of API, and the extent (achieved in about 4 h and 24 h) is almost the same. Interestingly, the dissolved rates ( Figure 5B) for Drug-A and its precipitates are different. Some precipitates (Fa and 0.5 Fe in Figure 5B) dissolved more slowly and had less solubility than that of API in the first 8 h. Precipitates obtained from FeSSIF and 2 × FeSSIF media have high solubility at the beginning of dissolution, but then the solubility decreases. This phenomenon may be caused by the formation of supersaturation solution after rapid dissolution in the medium and then trend to the equilibrium solubility. Furthermore, for rivaroxaban and Drug-A, there is no statistical difference in equilibrium solubility between API and precipitates obtained from biorelevant media (FaSSIF, 0.5 × FeSSIF, FeSSIF, and 2 × FeSSIF). Therefore, the biorelevant precipitates mainly have the characteristic of faster dissolution than API, which may be caused by the improvement of wettability.
Pharmaceutics 2022, 14, x FOR PEER REVIEW 9 of 17 solubility at the beginning of dissolution, but then the solubility decreases. This phenomenon may be caused by the formation of supersaturation solution after rapid dissolution in the medium and then trend to the equilibrium solubility. Furthermore, for rivaroxaban and Drug-A, there is no statistical difference in equilibrium solubility between API and precipitates obtained from biorelevant media (FaSSIF, 0.5 × FeSSIF, FeS-SIF, and 2 × FeSSIF). Therefore, the biorelevant precipitates mainly have the characteristic of faster dissolution than API, which may be caused by the improvement of wettability.
Intrinsic Dissolution Rate
Since the dissolved rate of biorelevant precipitates is faster than API, the intrinsic dissolution rate of samples was estimated to explore the impact of crystal structure on dissolution rate [30]. To ensure the consistency of experimental conditions, IDR is governed by the surface area of the compound in contact with the dissolution medium [28]. IDR of different samples was calculated according to Equation (1). Figure 6 exhibit the correlation between concentration and time in the experiment of IDR. The final results are shown in Figure 7. The crystal structure of rivaroxaban has a significant influence on IDR (p < 0.05). However, except for IDR of 0.5 Fe in FaSSIF buffer, there is no statistical difference in the IDR of Drug-A with different crystal structures in the same dissolution medium. In addition, the pH value and composition of the dissolution medium have great influences on the IDR of the same sample (p < 0.05). Therefore, precipitates obtained
Intrinsic Dissolution Rate
Since the dissolved rate of biorelevant precipitates is faster than API, the intrinsic dissolution rate of samples was estimated to explore the impact of crystal structure on dissolution rate [30]. To ensure the consistency of experimental conditions, IDR is governed by the surface area of the compound in contact with the dissolution medium [28]. IDR of different samples was calculated according to Equation (1). Figure 6 exhibit the correlation between concentration and time in the experiment of IDR. The final results are shown in Figure 7. The crystal structure of rivaroxaban has a significant influence on IDR (p < 0.05). However, except for IDR of 0.5 Fe in FaSSIF buffer, there is no statistical difference in the IDR of Drug-A with different crystal structures in the same dissolution medium. In addition, the pH value and composition of the dissolution medium have great influences on the IDR of the same sample (p < 0.05). Therefore, precipitates obtained from biorelevant media may have some improved properties of dissolution, which are the pivotal points affecting the absorption of compounds.
Permeability Assay
Considering that the crystal structure, dissolved rate, and IDR were different among precipitates, permeability tests were performed with the PAMPA. The permeability is calculated according to Equations (2)-(4), and the results are shown in Figure 8. In Figure 8A, the permeability of some rivaroxaban related precipitates was larger than that of API at pH 5.0 PBS (donor chamber) (p < 0.05). When the dissolution medium was FeSSIF in the donor chamber, only the permeability of API (rivaroxaban) was significantly different from that of precipitates obtained from 0.5 × FeSSIF and FeSSIF media (p < 0.05). However, the permeability of API was larger than that of precipitates obtained from bio-
Permeability Assay
Considering that the crystal structure, dissolved rate, and IDR were different among precipitates, permeability tests were performed with the PAMPA. The permeability is calculated according to Equations (2)-(4), and the results are shown in Figure 8. In Figure 8A, the permeability of some rivaroxaban related precipitates was larger than that of API at pH 5.0 PBS (donor chamber) (p < 0.05). When the dissolution medium was FeSSIF in the donor chamber, only the permeability of API (rivaroxaban) was significantly different from that of precipitates obtained from 0.5 × FeSSIF and FeSSIF media (p < 0.05). However, the permeability of API was larger than that of precipitates obtained from biorelevant media. As shown in Figure 8B, only the precipitate obtained from 0.5 × FeSSIF medium had different permeability from other samples when the donor chamber contained pH 5.0 PBS (p < 0.05). Therefore, under the same conditions, the crystal structure has limited influence on permeability. Nevertheless, the dissolution medium in the donor chamber is one of the factors affecting permeability. The permeability of rivaroxaban and Drug-A in the donor chamber containing aqueous medium is about 1.92-fold and 3.53-fold higher than that of samples in the donor chamber containing biorelevant medium, respectively. Actually, solubility and permeability are related to the polarity and non-polarity of the compound, respectively [33]. Although biorelevant media have good solubilization ability, the amount of bile salts is limited in its capacity to dissolve all compounds in the donor chamber. Theoretically, the hydrophobic groups of compounds can be wrapped by bile salts, and the hydrophilic groups are exposed to the medium. After the binding between the compound and the bile salts, the increasing molecular size and the hydrated layer will prevent the compounds from passing through the lipophilic permeable membrane. In addition, a medium with a high bile salts concentration (such as FeSSIF medium) can increase the binding of insoluble compounds to bile salts in the donor chamber, thus further reducing the permeability of the compounds. Therefore, there is a negative correlation between permeability and solubility.
Discussion
There are several ways to form a supersaturated solution in vivo. The supersaturated solution of some insoluble weak basic compounds can be triggered by the pH gradient from the stomach to the small intestine, resulting in the generation of precipitate and reduction of absorbable compounds, which has attracted extensive attention [34]. A supersaturated solution is a thermodynamically unstable system with high chemical potential, which drives the formation of crystallization/precipitate into an energetically favorable crystal structure. Part of the precipitate is excreted from the body with the intestinal peristalsis. When the concentration of the compound in the GI tract falls below the equilibrium solubility, the remaining precipitates are redissolved. Therefore, the complex precipitate kinetics of compounds in vivo play a key role in governing oral absorption. However, the limiting step of precipitate absorption is unknown. Absorption is related to the solubility and permeability of the substrate [35] and is determined by several factors, including physicochemical, formulation-relevant, and physiological factors [36]. Chemical groups of compounds may interact with intestinal fluid components during crystallization, resulting in different physicochemical properties between the newly formed crystals and API. Systematic study of the physicochemical properties of precipitates obtained from biorelevant media in vitro is the prerequisite for understanding the absorption mechanism of redissolved precipitates.
Precipitates obtained from the biorelevant media have many different physicochemical properties from that of API, but precipitates cannot be regarded as new active substances unless they demonstrate different safety and efficacy [37]. Actually, precipitates have some improved properties compared with the API, such as the rapid dissolution and the increase in intrinsic dissolution rate. These favorable changes may be related to the interaction between biorelevant media and compounds in the crystallization process so as to improve the wettability of compounds. Therefore, the absorption of redissolved precipitate may differ from API. Since the absolute amount of unabsorbed precipitate is unknown, this may be an important factor affecting the prediction performance of absorption. In addition, not all weakly basic compounds precipitate in the GI tract and not all precipitate events are equal, which depend on the solubility difference between the stomach and intestinal chambers, chemical groups, lattice energy, and so on. Hence, it is very important to determine whether precipitate can be formed in the GI tract through a biorelevant twostage in vitro test [38] and then characterize the physicochemical properties of precipitates from biorelevant media.
Considering that changes in bile salts composition and content may lead to differences in crystallization trends of compounds [39], physicochemical properties of precipitates obtained from biorelevant media with different bile salts contents were investigated in this manuscript. The changes of dissolved rate, intrinsic dissolution rate, and melting point appear in different crystals of rivaroxaban and Drug-A, indicating that bile salt concentration will affect the absorption of compounds and may also be one of the factors leading to individual absorption differences. Bile secretion is related to food intake and fat content. Hence, the influence of fat content and the difference of absorption mechanism between precipitates and API should be considered when using the PBPK model to predict the absorption of compounds. In addition, bile salts such as endogenous surfactants play an important role in solubilizing and improving the rate of dissolution of insoluble compounds. With the rapid dissolution of compounds into the medium, supersaturated solutions exceeding the equilibrium solubility can be formed [40]. Meanwhile, bile salts can interact with drug molecules, modify the crystallization kinetics, prolong the duration of supersaturated behavior, and finally promote the absorption of compounds [41][42][43]. Therefore, bile salts have complex effects on solubility and permeability, and the permeability of compounds should be estimated in the biorelevant systems.
Many studies have focused on evaluating drug permeation under fed state conditions. However, exogenous compounds, such as intestinal fluid and bile salts under fed state, can induce detrimental changes to epithelial cell monolayer integrity [44], cytotoxic effects [45], cell viability [46], and so on. These disadvantageous changes make it more difficult to evaluate the permeability of compounds by cells experiments under fed states, and the accuracy of experimental results may be limited [47]. In addition, crystal structure and dissolution rate are the pivotal factors determining the solubility of compounds, and supersaturation solution has a high solute activity that can promote the penetration of the compound into the intestinal epithelial cells [10]. Therefore, adding the predissolved solution of compounds into the donor chamber may ignore the generation of supersaturation solution. Here, the crystal structures, bile salts, and other factors affecting drug dissolution are incorporated into the permeability evaluation of the PAMPA method by adding excess compounds into the donor chamber. Meanwhile, the experimental results are meaningful, as the permeability is negatively correlated with bile salt concentration. This phenomenon indicates that compounds may have different permeability under fasted and fed states in vivo, and traditional permeability evaluation under standard conditions cannot represent the permeability of compounds in vivo.
PBPK absorption models have been widely used to predict the effects of food on the exposure of compounds, with 50% of the observed food effects within the predefined boundary of 25% [1,2]. To accurately capture absorption characteristics, factors affecting the absorption of compounds in the model, such as dissolution rate and precipitate rate, are usually evaluated under the biorelevant media to simulate the physiological environment [48,49]. However, the absorption process of redissolved precipitates may differ from that of API due to differences in physicochemical properties, particularly permeability. Moreover, apparent permeability is a ubiquitous descriptor of drug absorption and a widely used input value of the PBPK model based on experimental results. Based on the research results in this manuscript, this seemingly simple parameter requires additional consideration in order to better understand the most appropriate value that should be used in the model. Although PAMPA can evaluate the effects of crystal structure and intestinal fluid on permeability, it is an artificial membrane, and the metabolism and transport of compounds in the GI tract are neglected. The permeability assay of rivaroxaban and Drug-A in this manuscript shows that the medium in the donor chamber is the pivotal factor affecting the apparent permeability. The correlation between apparent permeability in biorelevant media and the effective permeability of humans should be further studied. Hence, it is urgent to establish sophisticated experimental systems to evaluate the permeability of drugs with a high level of bio-relevance, which is also the focus of our next research.
Conclusions
Physicochemical properties are the key points affecting the absorption of compounds in vivo. The reabsorption of precipitates is an important part of the absorption of some insoluble, weakly basic oral drugs. Hence, evaluating the physicochemical properties of precipitates collected from biorelevant media is fundamental for a full understanding of compound absorption. The crystal structure, melting point, melting process, and fusion enthalpy are different between API and precipitates. The interactions between API and biorelevant media can affect the molecular structure of a certain compound, such as Drug-A. Nevertheless, precipitates have many physicochemical properties superior to API, including the rapid dissolution into the medium and the increasing intrinsic dissolution rate. In addition, the permeability of some precipitates is statistically different from that of API. Meanwhile, bile salts can reduce the permeability of rivaroxaban and Drug-A by 1.92-fold and 3.53-fold, respectively. Considering that the dissolved rate of the precipitates is improved, the permeability of rivaroxaban-related precipitates is smaller than that of API, and the permeability assessment in vitro overestimates that in vivo, more attention should pay for absorption of redissolved precipitates and reasonably evaluating the permeability of compounds with a high level of bio-relevance.
Conflicts of Interest:
The authors declare no conflict of interest. | 2022-02-01T16:15:50.601Z | 2022-01-29T00:00:00.000 | {
"year": 2022,
"sha1": "8d49e03240df322d5bedccdf0a82416c3e7cc81b",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1999-4923/14/2/330/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "ade32fad2f538dd81943e857ce3873928032bfa7",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
55822792 | pes2o/s2orc | v3-fos-license | Lipid sources in diets for egg-laying japanese quail: performance and egg quality
This study evaluated the effect of using lipid sources of plant and animal origin in diets for Japanese quail (Coturnix coturnix japonica) on production performance and internal egg quality. Japanese quails (n = 160) were distributed in a completely randomized design with four treatments and eight replications of five birds each. Lipid sources evaluated were: soybean oil, from poultry slaughterhouse, fish waste and grape seed. The characteristics of performance and internal egg quality were measured every 21 days for a total period of 84 days. Data were subjected to analysis of variance and the means were compared by Tukey’s test. There was no effect of treatments on the production performance of birds, except for feed intake, which was higher for birds fed grape seed oil in the diet. Lipid sources evaluated had no influence on the internal egg quality. It can be concluded that the lipid sources evaluated here in can be used as alternative energy in feed for laying quails, without affecting the performance and internal egg quality.
Introduction
Animal origin products have great importance in the human diet and health.Thus, products, such as milk, meat and eggs have been the focus of much research aimed at identifying nutrients or agents that can promote healthy nutrition and, consequently, human health.
Accordingly, the use of certain ingredients in the diet of farm animals has attracted attention in research, in order to evaluate positive and negative points of their use.The increased animal productivity and even the change in the chemical, physical or organoleptic composition of products obtained are taken into account in these researches, which may affect both animal production costs and the gain in human health.
Among all the nutrients in a food, the energy source is one of the ingredients with high contribution in the total cost of animal feed, besides having the ability to change the fatty acid composition of egg and/or meat, especially polyunsaturated, improving its availability for human consumption.
An egg supplies about 13% of the daily requirements of protein for an adult and 25% for children.Thus, the egg is an important component in human diet due to protein input, containing nutrients essential for vital functions, such as, for example, lipids, vitamins and minerals.Moreover, it has been noted the interest of consumers for the quality of animal products (egg, meat or milk), considering characteristics such as color, size and weight, factors influencing the acquisition of a particular product (Grela, Ognik, Czech, & Matras, 2014).According to Bologa, Pop, and Albu (2013), many studies show that the chemical composition, biological value and physical and chemical properties of eggs are results of the effects of numerous factors, including breed, rearing condition, storage and culinary process of eggs in addition to animal nutrition.
In this context, the inclusion of ingredients, more specifically, oils or fats of plant or animal origin to the diet for birds, to change the lipid profile of animal products is a global trend given the demand of critical consumer for high quality products, with appeal to human health.
Thus, this study evaluated the addition of lipid sources to the diet of egg-laying Japanese quails in relation to production performance and internal egg quality.
Material and methods
The study was conducted at the Poultry Sector, Experimental Center of Unicastelo, Campus Descalvado, and previously approved by the Ethics Committee on Animal Use of the institution (003/2013).
Japanese quails (n = 160; Coturnix coturnix japonica), with 16 weeks of age were housed in galvanized wire cages with trough feeder and nipple drinker in conventional shed with galvanized wire mesh and curtains.The experimental design was completely randomized with four treatments (lipid sources) and eight replications per treatment, with five birds per repetition.Lipid sources evaluated were: soybean oil, poultry slaughterhouse waste, fish waste and grape seed.Alternative sources were used to replace soybean oil, considered as the control diet.
Experimental diets were based on corn and soybean meal, prepared according to the nutritional requirements determined by Silva and Costa (2009), Table 1.The composition of food was determined according to Rostagno et al. (2011).
Birds were fed ad libitum throughout the rearing period.The lighting program was 16 hours light daily.Maximum and minimum temperatures, inside the shed, were recorded daily with the aid of a digital thermometer (29±1.5°Cand 22±1.0°C,respectively).Production performance characteristics (% egg laying, feed intake, feed conversion and viability) were measured every 21 days, considered a production cycle, during the 84 experimental days.To quantify the total feed intake, feed was weighed at the beginning and end of each production cycle.Feed conversion ratio was determined from the relationship between the feed intake and the number of dozen eggs produced (kg dz -1 ) and egg mass (kg kg -1 ).Viability of birds, expressed in percentage, considered the mortality during the experimental period.
For the evaluation of the internal egg quality, in the last three days of each cycle, total egg production was used for determining the average egg weight.Three eggs per experimental unit were used to determine the Haugh Unit, yolk index and yolk color.
Eggs were individually weighed, cut in the central region, and placed on a flat surface to measure albumen height, height and diameter of yolk and yolk color.Haugh Unit was determined according to Brant and Shrader (1958), according Equation 1: HU = 100*log (H + 7.57 -1.7 W 0.37 ) (1) where: H = albumen height (mm) and W = egg weight (g).
Yolk index was obtained by the ratio between height and diameter of the yolk (Nesheim, Austic, & Card, 1979).Yolk color was obtained with the aid of a Roche ® color fan with a numerical scale from 1 to 15 (from white to red).
Data were subjected to analysis of variance and means were compared by Tukey's test at 5% significance using the statistical software Statistical Analysis System (SAS, 2002).
Results and discussion
Mean values of the production performance characteristics are listed in Table 2.There was no significant difference (p > 0.05) for the characteristics evaluated, except for daily feed intake, because quails fed grape seed oil in the diet showed higher intake in relation to the other treatments, with no difference only from diet containing soybean oil.
The results above corroborate those obtained by Costa et al. (2008), who evaluated different levels of linseed oil content replacing soybean oil in the diet for laying hens and observed no significant results for the feed intake, % of lay, egg weight and mass and feed conversion ratio (per mass and per dozen eggs).Likewise, Oliveira et al. (2010) analyzed the addition of vegetable oil to the diet for young hens, aged 20 to 28 weeks of age, and verified no significant effect on feed intake, feed conversion and egg production.In contrast, Santos et al. (2009) investigated the addition of vegetable oils in diets for laying hens and detected a significant increase in the percentage of eggs for birds fed diet with 2% linseed oil and soybean oil.However, there was no significant increase in feed intake between the levels of inclusion of vegetable oils.
On the other hand, Al-Daraji, Al-Mashadani, Al-Hayani, Mirza, and Al-Hassani (2010) emphasized that the inclusion of fish oil and linseed oil in the diets of groups with male and female quails improved the production performance (body weight, feed intake, egg weight, % of lay, egg mass and feed conversion ratio) and reproductive performance (fertile eggs and embryo mortality) of these animals.
Regarding characteristics of internal egg quality of quails fed different lipid sources, the mean results are presented in Table 3.There was no significant difference (p > 0.05) for egg weight, yolk color, Haugh Unit (HU) and yolk index between the treatments.Nevertheless, the results characterize a good internal quality of eggs.The HU values above 72 indicate good quality with respect to freshness, classifying eggs into AA, according to the United States Department of Agriculture (USDA, 2000).
These results are in agreement with the findings of Santos et al. (2009) who verified a greater egg weight of laying hens fed diets containing 4% soybean oil, although the diet with 4% cottonseed oil resulted in eggs with lower weight.
As for yolk color, similar results were found by Faitarone et al. (2016), who concluded that the inclusion of vegetable oils in commercial white layer (Lohmann LSL) diets does not significantly change egg yolk pigmentation, as colorimetrically evaluated.The diets evaluated were with 2.5% linseed oil inclusion; or with 2.5% canola oil; or with 2.5% soybean oil; or with 5.0% linseed oil; or diet with 5.0% canola oil; diet with 5.0% soybean oil; diet with 2.5% linseed oil + 2.5% soybean oil; diet with 2.5% canola oil + 2.5% soybean oil; and T1 diet with 2.5% linseed oil + 2.5% canola oil.
According to Stadelman and Cotterill (1995), Lee, Kim, and Lee (2001), Santos-Bocanegra, Ospina-Osorio, and Oviedo-Rondón (2004), usually a darker yolk color in eggs of commercial laying hens is desirable and depends exclusively on the feed supplied, since they are not able to synthesize these color pigments, but can absorb 20-60% of the feed pigments.
The use of alternative lipid sources in feed for quails has been evaluated considering the high cost of the energy component in the feed and hence in the total production of poultry.Certain lipid sources can alter the lipid profile of eggs, so as to increasing the content of polyunsaturated fatty acids.Therefore, assessments in the chemical composition of eggs are required for a greater power of decision making in the use of alternative energy sources in the diets for laying quails.Mean ± standard deviation; Different letters in the same column are significantly different by Tukey's test (p < 0.05).FCR = feed conversion ratio (kg kg -1 and kg dz -1 ); FI = daily feed intake (g bird -1 day -1 ).
Conclusion
In summary, lipid sources evaluated in this study can be used as alternative energy sources in diets for egg-laying quails, without affecting the performance and internal egg quality.
Table 1 .
Proximate and nutritional composition of the control diet for egg-laying quails.
Table 2 .
Production performance characteristics of Japanese quais fed diets containing different lipid sources.
Table 3 .
Characteristics of internal egg quality of quails fed diets containing different lipid sources.Different letters in the same column are significantly different by Tukey's test (p < 0.05). | 2018-12-05T04:39:22.627Z | 2016-08-08T00:00:00.000 | {
"year": 2016,
"sha1": "322a8d3fe29153c3890b3337853c48ca7218a8c3",
"oa_license": "CCBY",
"oa_url": "http://periodicos.uem.br/ojs/index.php/ActaSciAnimSci/article/download/30925/pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "322a8d3fe29153c3890b3337853c48ca7218a8c3",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
101240416 | pes2o/s2orc | v3-fos-license | Sensitive and selective detection of alcohols via fluorescence modulation
Abstract Reported herein is the selective detection of aliphatic alcohols using cyclodextrin-promoted, proximity-induced fluorescence modulation of a high-quantum yield fluorophore. This fluorescence modulation occurred when the analyte was held in close proximity to the fluorophore via non-covalent cyclodextrin–analyte–fluorophore interactions, and led to unique modulation responses for each analyte, fluorophore and cyclodextrin investigated. These changes in fluorescence were used for the generation of an array using linear discriminant analysis that successfully generated unique pattern identifiers for 99% of the investigated analytes, and could detect alcohols at micromolar concentrations. These results represent a fundamentally new detection approach for these challenging analytes, and have significant potential in the development of novel detection schemes.
Introduction
The sensitive and selective detection of alcohols is an important research objective with implications in public health and environmental remediation, as many of these alcohols are known or suspected toxicants (1)(2)(3). Selectivity in distinguishing aliphatic alcohols from similar analogues is particularly crucial, as structurally similar analogues often have widely disparate toxicities. These aliphatic alcohols are found in fuels and fuel additives (4), foods and beverages (5), and other commercial products (6,7). Current methods for aliphatic alcohol detection overwhelmingly rely on mass spectrometry (8,9); however, the high sensitivity associated with mass spectrometry often leads to extraneous signals, which increases the analysis time and limits the ability to conduct high-throughput assays (10,11). Some examples of non-mass spectrometry detection methods have been reported, most of which require a derivatization step to enable accurate detection (12,13).
A January 2014 chemical spill in Elk River, West Virginia of more than 7500 gallons of 4-methylcyclohexanemethanol (4-MCHM) (14,15) highlighted a key problem with the use of mass spectrometry-based detection in complex environments: the need for a priori knowledge of the toxicant of interest to enable high-throughput detection of contaminated samples (16,17). Because every compound generates a mass spectral signal, first responders needed to know the identity of the contaminant prior to detection and clean up. In the case of the Elk River spill, this knowledge was not disclosed by the offending company in a timely fashion, significantly hampering remediation efforts (18,19).
We recently developed a fundamentally new approach for small-molecule detection using γ-cyclodextrin as a supramolecular scaffold to promote proximity-induced energy transfer from aromatic toxicants and toxicant metabolites to high-quantum yield fluorophores (Figure 1) (20)(21)(22)(23). This energy transfer operates successfully in multiple complex not allow for the detection of non-aromatic small molecule analytes such as aliphatic alcohols. Although these alcohols do not participate in proximity-induced energy transfer, they nonetheless are known to bind in or near the cyclodextrin cavity (31). This binding, in turn, can lead to significant fluorescence modulation of a high-quantum yield fluorophore held in close proximity by the cyclodextrin scaffold (Figure 2). This modulation relies on favourable intermolecular interactions between the analyte, fluorophore and cyclodextrin, including hydrogen bonding, hydrophobic interactions, CH-π interactions and a variety of other intermolecular forces (32)(33)(34). Although all of these forces are well-known in the chemical literature, to the best of our knowledge, they have not been applied to the detection of this highly relevant class of analytes to date. Reported herein is the successful realisation of such proximity-induced, cyclodextrin-promoted fluorescence modulation that uses a multitude of intermolecular forces to achieve the sensitive and selective detection of a variety of alcohols, including aliphatic, aromatic and benzylic alcohols. environments, including biological fluids (24) and aqueous extracts from crude oil samples (25)(26)(27), and has been used for the selective array-based detection of 30 aromatic analytes with 99% accuracy in buffer and 92% accuracy in human urine (28).
All previous work on cyclodextrin-based detection relied on the ability of the toxicants to act as energy donors to high quantum yield fluorophores (29,30), which does 14. About 20 μL of an analyte solution (1 mg/mL) was added to the cuvette, and fluorescence measurements were repeated at 460 nm. The experiments were repeated for fluorophore 15 and 16 at excitation wavelengths of 420 nm and 490 nm, respectively. The fluorescence spectra were integrated vs. wavenumber on the X-axis, and fluorescence modulation was determined using Equation (1).
where Fl alcohol is the fluorescence emission of the fluorophore in the presence of the alcohol, and Fl blank is the fluorescence emission of the fluorophore in the absence of the alcohol. Each experiment was repeated four times. The reported values are the average of the results.
Experimental details for array generation experiments
Array analysis was performed using SySTAT 13 statistical computing software with the following settings: (1) Fluorophore Ratio = Fl alcohol ∕Fl blank
Materials and methods
All analytes (except for compound 8) and cyclodextrins were obtained from Sigma-Aldrich chemical company. MCHM 8 was obtained from Tokyo Chemical Company as a mixture of cis-and trans-isomers. Fluorophore 14 was synthesised following literature procedures (35). All fluorescence measurements were performed using a Shimadzu RF 5301 spectrophotometer. Both the excitation slit width and the emission slit width were 1.5 nm. All fluorescence spectra were integrated vs. wavenumber on the X-axis using originpro software. All arrays were generated using SySTAT Version 13.
Experimental details for fluorescence modulation experiments
In a quartz cuvette, 2.5 mL of a cyclodextrin solution and 100 μL of a fluorophore 14 solution (0.1 mg/mL) were combined. The solution was excited at 460 nm, which is the excitation wavelength to directly excite detection (LoD), defined as the lowest concentration of analyte at which a signal can be detected, and the limit of quantification, defined as the lowest concentration of analyte that can be accurately quantified.
Selection of alcohols
The alcohol analytes were selected to include a wide range of analytes (Chart 1), including those that are structurally • Classical Discriminant Analysis • Grouping Variable: Analytes • predictors: Bodipy, Rhodamine 6G, Coumarin 6 • Long-Range Statistics: Mahal
Experimental details for limit of detection experiments
These experiments were conducted following literature-reported procedures (36), to determine the limit of
Selection of cyclodextrins
Cyclodextrins were chosen as supramolecular hosts for this proximity-induced fluorescence modulation, as a result of their high aqueous solubility and known ability to promote non-covalent intermolecular interactions (42). Most of these interactions rely on the ability of cyclodextrin to bind hydrophobic guests in the hydrophobic interior cavity (43); an alternate mode of complexation involves hydrogen bonding between the cyclodextrin hydroxyl groups (hydrogen bond donors) and hydrogen bond acceptor moieties on small molecule guests (44). The alcohol analytes are expected to associate through hydrogen bond formation involving the free hydroxyl groups, as well as hydrophobically driven inclusion involving the hydrocarbon segments of the alcohol structures. We selected four commercially available cyclodextrin hosts: α-cyclodextrin (α-CD), β-cyclodextrin (β-CD), randomly methylated β-cyclodextrin (Me-β-CD, with an average degree of methylation of 1.8 methyl groups per glucose unit) (45) and 2-hydroxypropyl-β-cyclodextrin (2-HpCD) (46).
Fluorescence modulation experiments
These experiments measure the ability of the analyte to induce measurable and unique changes in the fluorescence emission spectra of a high-quantum yield fluorophore in the presence of a cyclodextrin host. The experiments were conducted by adding small amounts of a concentrated solution of the fluorophore in tetrahydrofuran to a 10 mM cyclodextrin solution in aqueous buffer, which led to a 91% aqueous/ 9% organic final solvent composition (volume/ similar but have widely disparate toxicities. Ethylene glycol 5 was chosen due to its widespread availability and known toxicity (37), as well as the desire to assay the ability of the cyclodextrin-based system to identify diol motifs via fluorescence modulation. 1-Hexanol (compound 1) and cyclohexanol (compound 2) were selected to challenge the ability of the fluorescence modulation system to distinguish between analytes with identical molecular weights, which are difficult to distinguish via high-throughput mass spectral analysis (although such compounds have unique fragmentation patterns in the mass spectrometer) (38). 4-MCHM (compound 8) was selected due to its prevalence in the 2014 Elk River contamination event (14), as well as its structural similarity to cyclohexanol 2. Alcohols 9-12 were chosen to assay the ability of the fluorescence modulation technique to detect propargylic alcohol, aromatic alcohols and highly hindered tertiary alcohols, and to distinguish between these analytes and simple primary aliphatic alcohols 1, 3 and 4. Control analytes 6 and 7, which lack free hydroxyl moieties, were also included in these investigations, as was tetrahydrofuran (13), the solvent used to dissolve the fluorophores.
Selection of fluorophores
These compounds include three of the major classes of high-quantum yield organic fluorophores: BoDIpy, Rhodamine, and Coumarin. They were either commercially available or were synthesised following literature-reported procedures (35). The efficacy of these fluorophores in energy transfer-based detection systems has already been established by us (20) and others (39-41). (56), which may be a factor in establishing its limited response to these analytes. (2) Control analytes 6 and 7 generally lead to increases in the fluorophore emission signal, with the exception of the methyl-β-cyclodextrin solutions which demonstrate no significant change. Analyte 6 is likely effecting these increases through acting as a hydrogen bond acceptor with the cyclodextrin hydroxyl groups, in line with analogous literature reports (57). Analyte 7, in contrast, is known to organise into hydrophobic aggregates in a highly polar, aqueous solvent system; these hydrophobic regions, in turn, are attractive regions for localising a hydrophobic fluorophore and facilitating moderate fluorescence increases (58).
(3) Diol 5 shows limited changes in the fluorophore emission signal, with the exception of a significant fluorescence decrease observed in the presence of β-cyclodextrin. Ethylene glycol is known to hydrogen bond to cyclodextrins and lead to the self-assembly of cyclodextrin monomers into supramolecular aggregates (59,60); to the extent that these cyclodextrin aggregates cause fluorophore aggregation, that would explain the observed fluorescence decreases (as many types of fluorophore aggregates are also less fluorescent than their monomeric counterparts) (61). (4) Analytes 2 and 8 exhibit variable changes in the fluorophore emission signal, with no clearly observable trends. Interestingly, the structural differences between these two analytes are relatively minor, as they both contain free hydroxyl moieties and a cyclohexane ring structure. The minor differences (primary vs. secondary alcohol, presence of additional methyl group) are sufficient to lead to clear differences in the fluorophore emission ratios of compound 14 in volume). These solvent conditions were designed to maximise hydrophobic association between the cyclodextrin hosts and small molecule guests (47), while ensuring solubility of all organic components. The fluorescence emission spectrum of the fluorophore under these conditions was measured and compared to the fluorescence emission spectrum of the fluorophore obtained after introduction of 20 μL of the analyte of interest under the same conditions. The difference between these two fluorophore emission spectra was defined as the degree of fluorescence modulation, and quantified according to Equation (1) (above). Selected results of the fluorophore modulation experiments are summarised in Table 1 and in Figures 3-5. Several aspects of these results merit further discussion.
The introduction of an alcohol to a fluorophore-cyclodextrin solution leads to marked changes in the observed fluorophore emission signal. In the case of fluorophore 14, these changes generally show up as an increase in the fluorophore emission signal in the presence of the alcohol (indicated by a ratio higher than 1.00). Aliphatic alcohols have been shown to bind to cyclodextrins under a variety of conditions (48)(49)(50), although in most instances the alcohol addition causes a decrease in the observed fluorophore emission signal. In our system, the observed fluorescence increases are a result of multiple, analyte-specific factors: (1) Linear alcohol analytes (1, 3 and 4) assist in the formation of cyclodextrin-fluorophore-alcohol complexes, leading to higher fluorophore emission signals (51,52). Ethanol 4, in particular, has been shown to form ternary complexes in 2-HpCD and to stabilize complex formation in Me-β-CD (53). The addition of these alcohols also increases the cyclodextrin solubility, although the small amount of alcohol added (20 μL in a 2.5 mL solution) means that the effect on cyclodextrin solubility is likely to be negligible (54,55).
The two exceptions to this trend are analytes 3 and 4 in the presence of β-cyclodextrin, which cause no measurable change in the fluorophore hydrocarbon segments of the molecules and the resultant effects of these segments on cyclodextrin inclusion complexes, and less affected by the free hydroxyl groups in these cases. (9) The introduction of control analyte 13 led to limited fluorescence changes in all hosts investigated, with no change observed in the methyl-β-cyclodextrin solution and limited fluorescence increases observed in other cases. These results indicate that the fluorescence changes observed in the other cases are related to specific analyte-cyclodextrin-fluorophore intermolecular interactions rather than being solely a result of the addition of organic solvent.
The results for fluorophores 15 and 16 are markedly different from those obtained using fluorophore 14, and reflect the fact that intermolecular interactions that lead to fluorescence modulation are exquisitely sensitive to the molecular structures of each component (see ESI for full tables with these fluorophores). For example, in the presence of methyl-β-cyclodextrin, fluorophore 15 undergoes substantial fluorescence decreases upon introduction of all analytes. These decreases are likely a result of the fact that fluorophore 15 binds strongly in the methyl-β-cyclodextrin cavity in the absence of any added alcohol; the addition of alcohols then disrupts the strong fluorophore-cyclodextrin complex and lowers the observed fluorescence. In another example, the fluorescence emission of fluorophore 16 changes very little with the introduction of analytes 1-13 to the cyclodextrin solution for all analytes in α-cyclodextrin and β-cyclodextrin, which indicates a limited interaction between the analytes and the fluorophore in these cases. Steric incompatibilities may prevent the larger fluorophore 16 from binding in the cyclodextrin cavity; as a result, the alcohol-cyclodextrin non-covalent association complexes have no effect on the photophysical properties of fluorophore 16.
These differential responses between analytes in the presence of different cyclodextrin hosts can be seen more clearly in Figures 3-5. As illustrated in Figure 3, in the presence of methyl-β-cyclodextrin, compounds 1-4 effect noticeably different changes in the fluorophore emission spectra of compounds 14-16. Compounds 3 and 4 lead to markedly higher fluorescence emissions of compound 14 compared to compounds 1 and 2 (Figure 3(A)). Fluorophore 15, in contrast, demonstrates relative insensitivity to the presence of the aliphatic alcohols (Figure 3(B)), with minimal differences in the fluorophore emission spectra observed. Finally, fluorophore 16 demonstrates higher emission signals in the presence of compounds 3 and 1, with relatively lower signals observed for compounds 2 and 4 (Figure 3(C)). cyclodextrin solutions with the introduction of these two analytes. (5) Alcohol 9 is the only analyte that led to a decrease in the fluorescence of compound 14 in the presence of an α-cyclodextrin host, which correlates well with the reported ability of the alkyne moiety to bind in the sterically constrained α-cyclodextrin cavity (62), and highlights the fact that the fluorescence modulation results are intimately related to key structural features of the analytes. In the presence of the other cyclodextrin hosts, analyte 9 behaved similarly to most of the other investigated analytes, with an increase in fluorescence in the presence of β-cyclodextrin and 2-hydroxypropyl-β-cyclodextrin, and a moderate fluorescence decrease in the presence of methyl-β-cyclodextrin, which is likely driven by the strong association of fluorophore 14 with the methyl-β-cyclodextrin host. (6) Alcohol 10, the only example of an aromatic alcohol, led to increases in fluorescence for all hosts investigated, which is qualitatively similar to the results obtained with aliphatic alcohols 1 and 2. However, the magnitude of the fluorescence changes were markedly higher for alcohol 10 compared to the other investigated analytes, and indicates that aromatic alcohols demonstrate different intermolecular interactions that translate into easily observable differences in fluorescence modulation behaviours, in line with literature reports of aromatic guest binding in the cyclodextrin cavities (63, 64). (7) The behaviour of fluorophore 14-cyclodextrin complexes with the introduction of alcohol 11 resembles that obtained with introduction of alcohol 9, with the exception of α-cyclodextrin solutions which demonstrated significant changes in the presence of alcohol 9, but virtually no changes in the presence of alcohol 11. These results highlight that the alkyne-α-cyclodextrin intermolecular interactions lead to unique photophysical read-out signals. (8) Alcohol 12 demonstrates behaviours that are qualitatively similar to analyte 8, with moderate fluorescence increases observed in the presence of α-cyclodextrin, β-cyclodextrin, and 2-hydroxypropyl-β-cyclodextrin, and slight fluorescence decreases observed in the presence of methyl-β-cyclodextrin. Interestingly, the behaviour of these analytes is also similar to that of control analytes 6 and 7, which indicates that the fluorescence modulation is driven by the features contribute to different intermolecular interactions with the fluorophores and cyclodextrin hosts, leading in turn to unique spectroscopic signatures. Tertiary alcohol 12, in contrast, generated response patterns that were similar to many of the other aliphatic alcohols investigated, which indicates that the substitution around the free hydroxyl group is less important than the other structural features of the molecule such as the presence or absence of an aromatic ring or alkyne group that lead to grouping of the analytes in similar regions in the LDA-generated plots.
The practical utility of this array would be markedly enhanced by its ability to generate well-separated signals for mixtures of analytes, as nearly all instances of analyte identification in complex environments involve toxicant mixtures (70). To that end, an array was generated using mixtures of analytes 2 and 8, and the results are summarised in Figure 7 for fluorophore 14. These results indicate that well-separated signals are generated for all mixtures of analytes 2 and 8 in all cyclodextrin hosts, indicating the high degree of selectivity. Moreover, the signals for the single component mixtures were relatively far away from those of the analyte mixtures, which points to the possibility of some degree of cooperativity between analytes 2 and 8 that is responsible for the observed non-linearity in array response patterns.
The fundamental basis for the observed high levels of differential selectivity in these alcohol-cyclodextrin-fluorophore systems is that non-covalent interactions between the analyte and fluorophore lead to measurable, analyte-specific fluorescence changes. Control experiments using a cyclodextrin-free phosphate-buffered saline solution indicate that the presence of the cyclodextrin hosts affects the fluorescence modulation results in specific, cyclodextrin-dependent ways.
one phenomenon that is facilitated in the presence of cyclodextrin hosts is the binding of the fluorophores in the cyclodextrin cavity, and concomitant increases in the fluorophore emission signals (Figure 8). Fluorophore emission signals in the absence of cyclodextrin and in the presence of α-cyclodextrin are markedly lower compared to the emission signals observed in the presence of β-cyclodextrin, methyl-β-cyclodextrin, and 2-hydroxypropyl-β-cyclodextrin, which indicates that the fluorophores bind in the cavities of larger cyclodextrin hosts (71). This increase occurs both for systems without an alcohol analyte (Figure 8(A-C)) and systems that contain an alcohol additive (compound 1 used as an example; Figure 8(D-F)). This fluorophore-cyclodextrin binding is one of several intermolecular interactions that exist for each unique analyte-fluorophore-cyclodextrin combination and enables excellent differentiation between the target analytes. Figure 4 illustrates the differential responses of fluorophores 14-16 to the introduction of analytes 1, 2 and 8 in the presence of β-cyclodextrin host. Here, noticeably different response patterns are observed for each analyte, which is striking given the structural similarities between the analytes. Furthermore, it would be difficult to distinguish analytes 1 and 2 via mass spectral analysis in a rapid time frame, due to their identical molecular weights (despite known differences in their fragmentation patterns) (65), which makes the differential fluorescence response patterns even more significant.
Finally, Figure 5 illustrates the limited degree of fluorescence differentiation observed in the presence of analytes 6 and 7. Relatively little changes in the fluorophore emission signals are observed upon introduction of these analytes, or with the introduction of 20 μL of water (instead of an organic analyte). The one exception to this is the marked decrease observed in the fluorescence emission of fluorophore 16 in the presence of 20 μL of analyte 6 ( Figure 5(C)), and current efforts are focused on elucidating the reason for this aberrant behaviour.
overall, substantially different fluorescence response patterns were observed for fluorophores 14-16 in the presence of cyclodextrin hosts with the introduction of analytes 1-13, which enables the development of selective array-based detection systems. Array-based detection has been well-established as a tool for selective analyte detection (66), using statistical analysis such as linear discriminant analysis (LDA) and principal component analysis (pCA) to identify unique response patterns for a variety of target analytes (67-69) For the fluorescence modulation system reported herein, the integrated fluorophore emission in the presence of the target analyte (Fl alcohol , see Equation (1)) was used as the input data for the arrays. These modulation experiments generated well separated signals for 99% of the analyte-cyclodextrin combinations investigated ( Figure 6). Interestingly, the signals for cyclohexanol and 4-MCHM (analytes 2 and 8) are grouped near each other for each cyclodextrin host, but still generate well-separated signals in every case, highlighting their structural similarities that lead to similar response patterns in combination with cyclodextrin hosts and high-quantum yield fluorophores. Similarly, ethanol and ethylene glycol (analytes 4 and 5) generate signals that are close to, but well separated from, each other for each cyclodextrin host. Finally, the signals for methanol and ethanol (analytes 3 and 4) are found far away from each other, indicating that despite their apparent structural similarities, their interactions with cyclodextrin hosts are in fact quite different.
Analytes 9-11 displayed signals that were relatively near each other and clearly distinct from signals generated by analytes 1-8, highlighting that their unique structural alcohol detection systems. Efforts towards the development of such detection systems are currently underway in our laboratory, and results of these and other investigations will be reported in due course.
Disclosure statement
no potential conflict of interest was reported by the authors. The practicality of this modulation was further demonstrated through determining the fluorescence modulation induced by analyte 8 in β-cyclodextrin dissolved in different solvent systems: ultrapure water, phosphate-buffered saline, 1 M naCl, and crude seawater collected directly from the narragansett Bay. The resulting ratios demonstrated solvent dependence that was translated into a highly selective array which distinguishes the analyte in different solvent environments with 100% accuracy ( Table 2). The ability of the cyclodextrin-promoted fluorescence modulation to operate in crude seawater highlights the practicality of this method in real-world environmental contamination scenarios.
Funding
practical applications of this detection method require the ability to detect alcohols at levels at or below literature-reported levels of concern (72). Limits of detection for all analyte-cyclodextrin combinations were calculated following literature-reported procedures, and selected results are highlighted in Table 3. Micromolar detection limits for these analytes are a promising start towards the development of highly sensitive detection systems. one way to further improve sensitivity would be to introduce specific recognition groups to the fluorophores. Such groups would improve binding ability by promoting stronger intermolecular interactions between the analyte and fluorophore (73). Efforts towards improving the sensitivity of this system are currently underway in our research group.
Conclusions
A variety of high-quantum yield fluorophores demonstrate measurable changes in their fluorescence emission spectra with introduction of small amounts of alcohols. These changes, termed 'fluorescence modulation' , are highly specific to each analyte-fluorophore-cyclodextrin combination and can be used to develop an array with 99% success in generating unique pattern identifiers for each analyte. notably, small structural changes in the alcohols lead to markedly different fluorescence response patterns and unique pattern identifiers that can distinguish, for example, cyclohexanol from 1-hexanol, methanol from ethanol, and benzyl alcohol from phenol. More broadly, different classes of alcohols such as aliphatic alcohols displays signals in a region of the plot that is unique from signals obtained from propargylic, aromatic and benzylic alcohols, and point to the potential of developing class-specific signatures for different alcohol classes. Finally, the highly selective array demonstrates significant promise in detecting binary mixtures of analytes with high selectivity, a result that significantly broadens the practical applicability of this technology for the development of fluorescence-based | 2019-04-07T13:06:57.766Z | 2016-02-05T00:00:00.000 | {
"year": 2016,
"sha1": "ef90710915b18cd430b9d44e082af3b1432ec027",
"oa_license": "CCBY",
"oa_url": "https://figshare.com/articles/journal_contribution/Sensitive_and_selective_detection_of_alcohols_via_fluorescence_modulation/2074294/files/3704056.pdf",
"oa_status": "GREEN",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "89fdc4c3f0e6b6a99d156a81c2e8dd74d233df7c",
"s2fieldsofstudy": [
"Chemistry"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
8100408 | pes2o/s2orc | v3-fos-license | Toxicity of Transition Metal Oxide Nanoparticles: Recent Insights from in vitro Studies
Nanotechnology has evolved to play a prominent role in our economy. Increased use of nanomaterials poses potential human health risk. It is therefore critical to understand the nature and origin of the toxicity imposed by nanomaterials (nanotoxicity). In this article we review the toxicity of the transition metal oxides in the 4th period that are widely used in industry and biotechnology. Nanoparticle toxicity is compellingly related to oxidative stress and alteration of calcium homeostasis, gene expression, pro-inflammatory responses, and cellular signaling events. The precise physicochemical properties that dictate the toxicity of nanoparticles have yet to be defined, but may include element-specific surface catalytic activity (e.g., metallic, semiconducting properties), nanoparticle uptake, or nanoparticle dissolution. These in vitro studies substantially advance our understanding in mechanisms of toxicity, which may lead to safer design of nanomaterials.
upon molecular self-assembly, the development of novel materials with dimensions on the nanoscale, and even the direct control of matter on the atomic scale. The application of nanotechnology in biology (nanobiotechnology) encompasses development of nanomaterials for delivering and monitoring biologically active molecules, disease staging, therapeutical planning, surgical guidance, neuro-electronic interfaces, and electronic biosensors.
In 2000 the U.S. National Science Foundation estimated that the market for nanotechnology products will be over one trillion US dollars by 2015. This increase in use will likely lead to unintended exposures to nanomaterials by occupational workers and end product users via inhalation, dermal absorption, or gastrointestinal tract absorption. In particular, the direct use of nanomaterials in humans for medical and cosmetic purposes dictates vigorous safety assessment of toxicity. Presently, the adverse effects of such exposure on human health and the environment are incompletely understood [2].
Nanotoxicology is an emerging field that builds upon previous work on airborne particle toxicity. Given (1) fixed particle mass, (2) unitary density, and (3) particle surface bioreactivity, nanoparticles possess better tissue penetration and higher biological potency than coarse (2.5-10 µm) and fine (<2.5 µm) particles, reflecting their small sizes and large reactive surfaces (Figure 1). Inhaled particles and fibers generate oxidants including reactive oxygen species (ROS) and reactive nitrogen species (RNS) [3][4][5]. Knaapen et al. characterized the generation of cellular ROS/RNS as -primary source‖ if arising from the target cells themselves or -secondary source‖ if arising from inflammatory cells [6]. Oxidative stress (OS) plays important roles in cellular signaling, inflammatory, and genotoxic and proliferative responses [5][6][7][8][9][10][11]. A three-tiered response to oxidative stress model has been outlined [12]: The first tier involves the induction of antioxidant enzymes such as HO-1, NQO1, superoxide dismutase, catalases and glutathione peroxidases. If tier 1 protection fails to restore cellular redox equilibrium, tier 2 responses involving activation of pro-inflammatory signaling pathways such as the JNK and NF-κB cascades are triggered. At higher and more prolonged oxidative stress levels, cellular perturbation and disarray result in a decrease in mitochondrial membrane potential, leading to cell death.
TiO 2 allows osseointegration of artificial medical implants and bone. TiO 2 is also extensively used as a pigment, a thickener, and a UV absorber in cosmetic and skin care products. TiO 2 , particularly in the anatase form, is a photocatalyst under ultraviolet light. It can be employed for solar energy conversion: Dye, polymer, or quantum dot sensitized nanocrystalline TiO 2 solar cells can be manufactured using conjugated polymers as solid electrolytes. TiO 2 is also used as a material in the memristor, a new electronic circuit element.
Nanostructures of ZnO, including particles, rods, wires, belts, tubes, cages, walls, and rings, have found utility due to their unique nanoscaled electronic and optoelectronic properties [24][25][26][27][28][29][30]. ZnO has wide direct band gap (3.37 eV or 375 nm at room temperature), making it suitable for use in laser diodes and light emitting diodes. ZnO has high biocompatibility and fast electron transfer kinetics; these features enable the use of this material as a biomimic membrane to immobilize and modify biomolecules [31]. Other applications are in catalysis, paints, abrasives, wave filters, UV detectors, transparent conductive films, varistors, gas sensing, solar cells, sunscreens, and cosmetic products.
Cupric oxide (CuO) is used as a pigment in ceramics to produce blue, red, and green (and sometimes gray, pink, or black) glazes. CuO is a p-type semiconductor due to its narrow band gap of 1.2 eV. CuO can be used to produce dry cell batteries as well as wet cell batteries as the cathode. CuO is used as an abrasive to polish optical equipment, and it can be used to dispose of hazardous materials such as cyanide, hydrocarbons, halogenated hydrocarbons and dioxins via oxidation processes.
Methodologies to Study Toxicity of Transition Metal Oxides
Both in vitro and in vivo methods provide valuable tools to investigate nanotoxicity. In this article, we review in vitro studies. In vitro toxicity testing is a cost effective tool to evaluate the toxicity of nanomaterials; there are too many nanomaterials to evaluate each in vivo. Some advantages of in vitro studies using various cell lines are that they: (1) reveal effects of target cells in the absence of secondary effects caused by inflammation; (2) permit the identification of primary mechanisms of toxicity in the absence of the physiological and compensatory factors that confound the interpretation of whole animal studies; and (3) are efficient, rapid and cost-effective; and 4) can be used to improve design of subsequent expensive whole animal studies. For instance, we and others have used A549 (human bronchoalveolar carcinoma-derived cells), U87 (astrocytoma cells), U937 (human monoblastoid cells), mouse Leydig TM3 cells, human V79 and L929 fibroblasts, human SCCVII, B16F10 and FsaR tumor cells, and RAW264.7 (mouse peritoneal macrophage) cell lines to characterize oxidative stress-related signaling pathways [32][33][34][35][36][37][38][39][40][41][42]. A nontransformed human lung cell line (BEAS-2B) has been used to decipher expression of alteration of genes pertaining to oxidative stress and cell death pathways [43][44]. Co-culture of cell lines is another valuable technique to manipulate environmental influences in vitro studies [45].
A limitation of in vitro testing is that cells in culture do not experience the range of pathogenic effects observed in vivo, partly related to issues of translocation, toxicokinetics and coordinated tissue responses. A detail discussion of the limitations of in vitro studies can be found in Donaldson et al. [46].
Mechanisms of Cellular Uptake and Intracellular Interactions
Transmission electron microscopy (TEM) and scanning electron microscopy (SEM) have been the most widely used techniques to visualize agglomerated nanoparticles in cells [35,[40][41]47]. To understand mechanisms of cellular uptake of nanoparticles, we recommend live cell imaging techniques, as cell fixation process can alter membrane properties leading to artificial uptake of nanoparticles [48]. Fluorescent probes such as Texas Red and FITC can be conjugated to nanoparticles to ascertain bioavailability. Studies on uptake mechanisms of transition metal oxide nanoparticles are limited; however, the following findings on the uptake of quantum dots, carbon nanotubes, gold nanoparticles, and silicon nanoparticles are likely to be relevant to transition metal oxides.
Cationic nanoparticles may enter cells more easily than anionic nanoparticles since cationic nanoparticles can interact with heparin sulfate proteoglycans on the membrane surface [49][50]. This increased bioavailability may explain why cationic silicon nanoparticles are more toxic than neutral and anionic silicon nanoparticles [51]. This electrostatic interaction is a prelude for a subsequent endocytic process. Exogenous materials can be internalized via multiple pathways, including clathrin-mediated endocytosis, caveolar-mediated endocytosis, macropinocytosis, phagocytosis, flotillin-dependent endocytosis, arf6-dependent endocytosis, IL2Rβ endocytic pathway, CLIC/GEEC endocytic pathway, circular doral ruffles, and trans-endocytosis. A few studies have suggested that the uptake of nanomaterials involves energy-, lipid raft-and actin-dependent macropinocytosis. Phagocytosis is an alternate proposed venue of entry. For instance, single wall carbon nanotubes conjugated with proteins as cargoes enter cells via energy-dependent pathways and become entrapped inside endosome/lysosomes [52]. We have observed that CdSe/ZnS quantum dots non-covalently conjugated with nona-arginines are internalized via lipid raft dependent macropinocytosis, and the complex is confined in lysosomes [53][54]. One study has demonstrated that Tat peptide-conjugated quantum dots (Tat-QD) are internalized via actin-dependent endocytosis, transported by microtubules, and aggregated around microtubule-organizing center; the Tat-QD vesicle are shed from filopodia [55]. On the other hand, gold nanoparticles conjugated with biomolecules are internalized via phagocytosis by mouse macrophages [56], and organic monolayer-coated silicon nanoparticles are phagocytosed by rat alveolar macrophage cells (NR8383) [51]. Clearly, studies on uptake mechanisms of transition metal oxides are needed.
Toxicity of Transition Metal Oxide Particles
Using commercially available raw nanoparticles to test toxicity is essential to understand human health risk and environmental impact, as this addresses unintended exposure. It is equally important to test surface modified-nanoparticles, as these modified nanoparticles have different kinetics and bioavailability than the native nanoparticles, and they are frequently applied directly on human body for purposes such as disease diagnostics, treatment, and prognostics.
The toxicological literature reveals a trend among these transition metal oxides: TiO 2 is less toxic than CuO and ZnO in human cell lines [33,35,43,[57][58]. A similar trend was observed in E. Coli [59] and yeast [60]. Using BEAS-2B cells, we investigated the toxicity of oxides of Cr, Mn, Fe, Co, Ni, Cu, and Zn, each of which is widely used in industry and is in the same period as Ti (V 2 O 3 particles were excluded from this analysis because they are not commercially available in the same size range and morphology as the other compounds). Toxicity increased with atomic number (Figure 2), with the exceptions of Fe 2 O 3 (lower toxicity than expected) and CoO (higher toxicity than expected). Fahmy and Cormier also identified a similar relationship of CuO and Fe 2 O 3 toxicity in airway epithelial cells (HEp-2) [39]. It is difficult to compare the toxicity of TiO 2 and V 2 O 5, as synthesis methods, particle characteristics, and cell lines (Leydig TM3 cells, human V79 and L929 fibroblasts, human SCCVII, B16F10 and FsaR tumor cells) differ among the few published studies; however, it is quite clear that TiO 2 is minimally toxic while V 2 O 5 is toxic [40][41][42]. Data from these studies are consistent with our findings presented in Figure 2. (adapted from [43], with additional data).
While the factors underlying this trend have not been established, several physicochemical properties have been considered, including element-specific catalytic activity on the particle surface, released ions (particle dissolution), and differential cellular uptake. In a study with A549 cells, Limbach et al. [61] demonstrated that 1) among titanium-, manganese-, iron-, or cobalt-doped silica, the addition of Ti or Fe has a much smaller effect on ROS generation than Mn due in part to surface catalytic activity; and 2) particles of Co 3 O 4 and Mn 3 O 4 , but not TiO 2 and Fe 2 O 3 , can significantly dissolve in a cell-free culture system leading to elevated ROS levels. They further demonstrated that much less ROS is formed when an aqueous iron solution contains A549 cells than when the solution contains no cells. The authors attributed this difference to a barrier function of the cell membrane for ions. It is also likely that that adsorption and binding of the metal oxide particles to the matrix of the cell medium may also be involved [62].
The potential role of dissolved ions in toxicity was also highlighted in a study by George et al. [63]. They found remarkably high levels of Zn 2+ released into aqueous solutions from pure and Fe-doped ZnO nanoparticles. In one study, we suspended transition metal oxides in an aqueous solution for a period of 6 hours, and utilized a dialysis membrane with a molecular weight cutoff 500-1,000 to separate ionic species. We observed a trend of increased ion dissolution of metal oxides from TiO 2 (0.0017%-0.0065%, wt based) to ZnO (1.25%-1.67%, wt based) (Huang et al. unpublished data), corresponding to the toxicity trend ( Figure 2). While initial attempts to measure ion dissolution from metal oxide nanoparticles have been made by us and others, it would be desirable to assess ion dissolution in a cell medium relevant setting using a standardized protocol to separate dissolved ions. Moreover, pH maintenance during the experimental protocol is essential to allow comparisons with the cellular environment, and adsorption effects by culture medium can complicate the subsequent ICP-MS analysis and must be taken into consideration.
While Limbach's findings [61] may explain the relatively low toxicity of Fe 2 O 3 , it is not obvious why CoO is more toxic than NiO. We suspect involvement of a variety of factors: surface catalytic activity, released ion, cell membrane as a barrier, adsorption to the matrix of the cell medium, cellular uptake, and possibly point of zero charge. Experiments using oxides of different oxidation states (CoO, Co 3 O 4 , NiO, Ni 2 O 3 ) may be revealing in this regard.
Although we and others have contributed to the understanding of the physico-chemical properties governing cytotoxicity of transition metal oxides, there are still gaps in our knowledge. While studies on nanoparticles of industrial origin have contributed to risk assessment and management, these materials possess considerable heterogeneity in size, morphology, surface defects, etc. This makes it difficult to relate toxicity to specific particle characteristics. Thus, it is desirable to synthesize highly defined and uniform nanoparticles to identify the toxicity-contributing properties of nanoparticles for comparative and predictive nanotoxicology. Not only would this type of particles enrich our knowledge of particle physics, chemistry and biology, it would also assist in the design and production of safer alternative particles.
Exposure to Metal Oxides Tips Off Cellular Redox State-Elevated Oxidative Stress
Oxidative stress is a normal cellular process involved in many aspects of cellular signaling, though excessive oxidative stress can be harmful. Many studies have shown that exposure to nanoparticles elevates cellular oxidative stress [34][35][37][38][39]61]. To cope with elevated oxidative stress, cells mount protective or injurious responses. For instance, cells activate enzymatic and non-enzymatic antioxidant defense mechanisms. Glutathione peroxidases, catalases, superoxide dismutases, and phase II enzymes play essential roles in returning cells to a normal redox state (Figure 3). The transcription factor Nrf-2 has been shown to play an essential role in the antioxidant response element (ARE)-mediated expression of phase II detoxifying and antioxidant enzymes, as well as other stress-inducible genes, in response to oxidative stress [64][65][66]. A recent study with CeO 2 nanoparticles has demonstrated that Nrf-2 can be activated and translocated into nucleus with subsequent induction of heme oxygenase-1 (HO-1) in CeO 2 -exposed BEAS-2B cells [67]. Nrf-2 can therefore be considered as a master switch for antioxidation defense, serving as a functional indicator of oxidative insult caused by nanomaterials.
In our pathway-specific microarray study involving 84 oxidative stress responsive genes [20], exposure of BEAS-2B cells to a ROS-elevated but sublethal concentration of ZnO nanoparticles elevates the expression of genes pertaining to oxidative stress (PRDX3, PRNP, and TXNRD1 genes) and apoptosis (a pro-apoptotic BNIP3 gene). This is consistent with our biochemical and cytotoxicity findings. One caution in assessing gene regulation under oxidative stress is that the response is quite dynamic and dependent on somewhat arbitrary cutoff criteria relative to control levels. For instance, we did not find elevated expressions of several common antioxidant genes, including catalase, glutathione peroxidase, glutathione reductase, glutathione transferase, and superoxide dismutase in cells exposed to ZnO at the ROS-elevated but sublethal level [20]. This finding is consistent with results reported by Sarkar et al. [44] with toxic single-walled carbon nanotubes in BJ Foreskin cells. At different time points or concentrations (i.e., stress levels), gene expression may be alternately up-regulated or down-regulated. Understanding the regulation of enzymatic and non-enzymatic antioxidant defense mechanisms may suggest strategies to mitigate elevated oxidative stress. For example, Gilliland et al. used knockout and genetic polymorphisms of genes that encode phase II enzymes to characterize a susceptibility mechanism that may explain why only some people develop PM-induced injury [68]. Catalase and glutathione peroxidase are two other examples: Catalase has a relatively low constitutive level in cardiomyocytes which predisposes cardiac muscle to oxidative stress damage. Cardiac muscle is also very susceptible to oxidative damage due in part to the rapid inactivation of glutathione peroxidase [69]. Overexpression of glutathione peroxidase in endothelial cells and myocytes significantly decreases oxidative stress-induced NF-kB activation, which leads to apoptosis [70]. Thus, tissue-specific responses to nanoparticle exposure should be considered when risk assessment is performed.
The fundamental factor that governs initiation of cellular oxidative stress by transition metal oxides is still unknown. Oxygen vacancies on metal-oxide surfaces are electrically and chemically active. Once entering circulatory systems or cells, charge accepting molecules, such as NO 2 and O 2 , may be in close proximity at the vacancy sites; de-coupling of charged NO 2 or O 2 is possible and could be sources of initiation of oxidative stress. Interestingly, non-metal nanomaterials such as carbon nanotubes behave like (transition) metal oxides in inducing cellular oxidative stress. We hypothesize that oxidative stress initiated by metal or non-metal nanoparticles may be related to their metallic or semiconducting properties at nano-scale sizes. It remains to be determined how the metallic or semiconducting behaviors and metal dissolution differentially contribute to initiation of cellular oxidative stress and subsequent pathologic outcome.
When oxidative stress overwhelms defense mechanisms, cellular macromolecules such as proteins, lipids, and DNA are subject to damage. DNA damages include deletions, mutations, single-and double-strand breakages, adduct formation, and cross-linking with proteins. Studies have confirmed DNA adducts and oxidation-induced DNA fragmentation following exposure to metal oxide nanoparticles [35,58,[71][72][73]. In response to DNA insult, cells attempt to repair the damaged DNA. Repair failure may lead to cell death (e.g., apoptosis) or cell transformation. It is important to identify types of DNA damages and repair mechanisms involved in nanotoxicity (e.g., nucleotide excision repair, base excision repair, mismatch repair, double strand repair, direct reversal), as this understanding may suggest measures for prevention or intervention, such as the over-expression of DNA repair genes.
In the case of severe damage to DNA, cells may die by either necrosis or apoptosis. Apoptosis is a complex and highly regulated process that is invoked to eliminate irreparably damaged cells. In contrast, necrosis has long been thought of as a disorderly event (although new evidence suggests otherwise), and is considered to be a consequence of extreme physiochemical stress. Although studies have shown that exposure to certain metal oxide nanoparticles induces apoptosis [74], it is unclear which pathways are involved. There are at least four major apoptotic pathways: extrinsic apoptotic (receptor mediated pathway), intrinsic apoptotic (mitochondrial pathway), caspase-2 dependent, and caspase-independent. Again, an understanding of which pathway(s) is involved in nanoparticle induced cell death may suggest measures for prevention or mitigation.
Insofar as multiple genes have been demonstrated to be involved in responses to oxidative stress, DNA damage, and cell death (apoptosis and necrosis), a custom, pathway-specific genetic approach could serve to decipher toxicity mechanisms of nanoparticles. On the other hand, whole genome approaches may serve as a discovery tool to identify unanticipated signaling pathways.
Oxidative Stress and Perturbation of Intracellular Calcium Homeostasis
The intracellular calcium concentration ([Ca 2+ ] in ) plays major regulatory roles in cellular metabolism, signal transduction, and gene expression. For instance, calcium can influence cell cycle by activating protein kinase C (PKC), Ca 2+ /calmodulin-dependent protein kinases (CaMK), and mitogen-activated protein kinase (MAPK). Accordingly, [Ca 2+ ] in is tightly regulated, and alterations of [Ca 2+ ] in are associated with cellular dysfunction, metabolic and energetic imbalance, disease states, and cell death. A host of environmental toxicants elevate [Ca 2+ ] in directly or indirectly by promoting Ca 2+ influx, releasing Ca 2+ from intracellular stores, inhibiting Ca 2+ sequestration, or blocking Ca 2+ efflux from the cell [75]. A study with ultrafine particles demonstrated that calmodulin-dependent signaling pathways are crucial for cytotoxicity and cytoskeletal dysfunctions [76]. In our studies, ZnO caused a concentration-dependent elevation of [Ca 2+ ] in , that could be partially attenuated by the antioxidant N-acetylcysteine (NAC), indicating an effect of oxidative stress on calcium homeostasis [43]. The inverse correlation between [Ca 2+ ] in and cell viability suggests a role for calcium in cell death. The moderation of this increase by nifedipine suggests that a portion of this increase reflects the influx of extracellular calcium. Membrane disruption (e.g., oxidative stress-induced lipid peroxidation) may also play a role in this influx. The involvement of Ca 2+ release from intracellular stores has yet to be evaluated.
In another study, the effects of ZnO nanoparticles on store operated calcium entry was studied in Chinese hamster ovary (CHO) cells stably transfected with a M3 muscarinic receptor. M3 receptors activate Gα q transducer proteins that stimulate phospholipase Cβ (PLCβ) [77]. PLCβ hydrolyzes phosphatidylinositol 4.5-bisphosphate, releasing two second messengers, diacylglycerol and inositol trisphosphate (IP3). IP3 binds to the IP3 receptor on the endoplasmic reticulum (ER) to release calcium into the cytosol. Sensors in STIM1 proteins that are components of ER membranes detect the depletion of calcium in the ER. STIM1 proteins thus activated interact with Orai channel proteins in the plasma membrane to stimulate the entry of extracellular calcium, i.e., the store operated calcium entry (SOCE) (Figure 4) [78][79]. At a non-cytotoxic concentration (10 µg/mL), ZnO increased resting [Ca 2+ ] in of Chinese hamster ovary (CHO) cells expressing M3 muscarinic receptors from 40 to 130 nm without compromising calcium homeostatic mechanisms and the CHO cells had no store operated calcium entry response in the presence of 10 µg/ml ZnO. Hence, ZnO particles had minimal effects on IP3-or thapsigargin-mediated release of intracellular calcium from the endoplasmic reticulum, but strongly inhibited store operated calcium entry. This effect was seen a decrease in Ca 2+ entry upon introduction of calcium to the extracellular medium following thapsigargin-induced depletion of calcium from the endoplasmic reticulum (EC50's ≈ 2 µg/ml). Thus, ZnO nanoparticles interfere the M3 signaling pathway (at least) via disruption of store operated calcium entry. . Store-operated Ca 2+ entry (SOCE). Ligand binding to certain G protein coupled receptors leads to activation of phospholipase Cβ. The IP3 thus released increases Ca 2+ release from the lumen of the endoplasmic reticulum. Depletion of ER calcium leads to a Stim1 (Ca 2+ sensor)-Orai (Ca 2+ channel) interaction and the entry of extracellular Ca 2+ . ZnO nanoparticles inhibit this pathway by blocking the SOCE without affecting proximal receptor signaling events [67]. Nanoparticle-induced cell death by various means is summarized in Figure 5 [34][35][37][38]43]. The increases in intracellular content OS may have multiple sources. We postulate that the elevated OS is a consequence of ZnO surface reactivity and/or defects interacting with intracellular reductants in combination with the effects of dissolved metal ions that catalyze redox reactions. Synergistic interactions between intracellular [Ca 2+ ] and OS are a likely contributing factor. While [Ca 2+ ] in and OS affect the activity of each other, they both induce cell death by distinct pathways. Finally, though calcium-dependent kinase activation pathways leading to cell cycle regulation or cell death have been well documented, studies in this area with transition metal oxides are lacking.
Pro-Inflammatory Response
Inflammation is clinically defined as the presence of redness, swelling and pain. Histologically, it is defined as the presence of edema fluid and the infiltration of tissues by phagocytic cells. Chronic inflammation can lead to diseases such as atherosclerosis, pulmonary diseases, and cancer. Oxidative stress is closely related to inflammatory response by activating the nuclear factor-kappaB (NF-κB) signaling pathway that controls the transcription of pro-inflammatory genes such as IL-1, IL-8, and tumor necrosis factor- (TNF-) (Figure 6).
Figure 6.
A model shows the simplified NF-kB signaling pathway that is activated by oxidative stress. Persistent activation leads to chronic inflammation.
Exposure to nanoparticles has been found to result in oxidative stress-induced activation of pro-inflammatory factors such as IL-1, IL-6, IL-8, and macrophage inflammatory proteins (MIP) at both mRNA and protein levels in vitro [45,[80][81][82][83][84][85][86]. Additional effects such as activation of IL-2, IL-4, IL-5, IL-6, IL-12, TNF-α, increased number of polymorphonuclear leukocyte (PMN) and lymphocyte, increased B cell distribution, and T cell diminish have been observed in vivo ( Table 1). These factors along with histopathological evidence can be included in the assessment of the risks associated with nanoparticle exposure.
Conclusions
Recent studies have increased our understanding of the nanotoxicity of metal oxide particles, particularly with respect to oxidative stress-induced cascade pathways that lead to inflammatory responses. Several important challenges remain. First, inter-laboratory discrepancies contribute to uncertainties in risk assessment. Second, as new nanomaterials continue to emerge, a systematic approach to identifying the physicochemical properties of nanomaterials that determine toxicity is required. This is problematic using nanomaterials of industrial origin since they 1) vary in multiple physical characteristics, 2) are synthesized using different methods, and 3) vary widely in purity. Third, to attribute toxicity to a particular factor, one needs to consider the extent to which the test design is relevant to the cellular environment. It is likely that toxicity is a non-linear function of multi-variables, which highlights the necessity of developing a theoretical computing model for predictive nanotoxicity. | 2014-10-01T00:00:00.000Z | 2010-10-01T00:00:00.000 | {
"year": 2010,
"sha1": "a481c3da96a294488dfdc775a93ab4b4c44a6eb5",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1996-1944/3/10/4842/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a481c3da96a294488dfdc775a93ab4b4c44a6eb5",
"s2fieldsofstudy": [
"Environmental Science",
"Materials Science",
"Medicine"
],
"extfieldsofstudy": [
"Materials Science",
"Medicine"
]
} |
238586366 | pes2o/s2orc | v3-fos-license | A Novel Autophagy-Related Marker for Improved Differential Diagnosis of Rheumatoid Arthritis and Osteoarthritis
Rheumatoid arthritis (RA) and osteoarthritis (OA) are two most common rheumatic diseases in the world. Although there are standard methods for the diagnosis of both RA and OA, the differentials in some cases are poor. With deepening research, the role of autophagy in maintaining cell homeostasis and thus enabling cells adapt to external environments has become increasingly prominent. Both RA and OA, two diseases with inherent differences in pathogenesis, gradually show differences in autophagy levels. Our study therefore aims to further understand differences in pathogenesis of RA and OA through in-depth studies of autophagy in RA and OA. We also define appropriate autophagy-related markers as recognition indicators. Differences in autophagy levels between RA and OA were found based on analysis of the Kyoto Encyclopedia of Genes and Genomes (KEGG) and single-sample gene set enrichment (ssGSEA). These differences were mainly caused by 134 differentially expressed genes (DEGs). In two autophagy-related genes, CXCR4 and SERPINA1, there existed significant statistical difference between RA and OA. An autophagy related index (ARI) was thus successfully constructed based on CXCR4 and SERPINA by binary logistic regression of the generalized linear regression (GLR) algorithm. Pearson analysis indicated that the expression of CXCR4, SERPINA1, and ARI were closely correlated with autophagy scores and immune infiltration. Moreover, ARI showed high disease identification through receiver operating characteristic (ROC) analysis (AUCtesting cohort = 0.956, AUCtraining cohort = 0.867). These results were then verified in GSE12021 independent cohort. In conclusion, ARI associated with autophagy and immune infiltration was successfully constructed for accurately identifying OA and RA. The index, thus, has great potential in clinical applications.
INTRODUCTION
With aging population, musculoskeletal-related chronic disability that is caused by both rheumatoid arthritis (RA) and osteoarthritis (OA) can be further aggravated and this imposes huge economic and social burden (Woolf and Pfleger, 2003). Recent epidemiological data showed that from 0.24 to 1% of the people in the world suffered from RA, while OA affected approximately one third of the population over 65 years (Gibofsky, 2014;Johnson and Hunter, 2014). These two are considered the most common rheumatic diseases in the world (Sangha, 2000). A deep understanding of the diseases reveals that although RA and OA have great differences in pathophysiology, similar and overlapping features are presented in their underlying mechanisms (de Lange-Brokaar et al., 2012;Frangos and Maret, 2020). Standard differential diagnoses for RA and OA in some cases lead to poor outcomes (Toes and van der Woude, 2011;Kiltz et al., 2017). There is, thus, an urgent need to further improve their differential diagnosis for better clinical decisionmaking and early intervention.
Autophagy is a highly conserved intracellular lysosomaldependent catabolism pathway. It is characterized by the ability of autophagosomes to degrade damaged organelles and maintain cellular homeostasis, thus enabling cell adaption to environmental stress (Mizushima and Komatsu, 2011). Recently, a large number of research showed that autophagy is involved in the development of a variety of rheumatic diseases, including RA and OA (Rockel and Kapoor, 2016;ysl et al., 2016;Marta et al., 2018). In RA, synovial hyperplasia, cartilage degeneration, fibroblast-like synoviocytes (FLS) infiltration into cartilage and bone surfaces, and other pathological features are presented (Pope, 2002). Zhu et al. (2017) showed that autophagy in synovial tissue was significantly enhanced in RA patients, and the level reflected the disease severity. Meanwhile, Shin et al. (2010) indicated that the increased autophagy in RA-FLS promoted RA-associated synovitis because of the decreased rate of apoptosis in RA-FLS-increased synovial fibrosis. On the other hand, the main pathogenesis of OA is the irreversible destruction of cartilage and increase of proinflammatory factors and catabolic enzymes, both of which accelerate the progression of inflammation (Kraan and Berg, 2008). Sasaki et al. (2012) proved that increased autophagy is an adaptive response of cells from stress, this protective mechanism inhibits the expression of genes related to cartilage rupture by regulating apoptosis and reactive oxygen species (ROS), thus, affecting the progress of OA. In addition, Gang et al. (1038) found that increased autophagy could protect chondrocytes against additional apoptosis and senescence. In summary, certain differences in the level of autophagy between RA and OA have been found through these studies, providing hope for the establishment of promising markers of differential diagnosis.
In this study, a comprehensive analysis was used to evaluate autophagy differences between RA and OA. The aim was to establish the importance of autophagy from the perspectives of pathogenesis and diagnosis. Results from this study offer a deeper understanding of autophagy in both diseases and provide assistance for clinical differential diagnosis.
Data Download and Pre-processing
First, raw data of OA and RA were downloaded from the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) using GEOquery R package. Next, background correction and normalization with RMA algorithm were processed through the "affy" R package. Probe IDs were transformed to gene symbols by using corresponding R packages (Table 1). Finally, to improve accuracy of analysis, sva algorithm was performed to merge differential datasets and remove batch effect.
Identification of Differentially Expressed
Genes of Osteoarthritis Versus Rheumatoid Arthritis GSE55457, GSE55584, and GSE55235 were merged to screen differentially expressed genes (DEGs) between OA and RA through the limma R package. The genes with p < 0.05 and |log2FC| > 1 were identified as DEGs.
Analysis of the Kyoto Encyclopedia of Genes and Genomes and Single-Sample Gene Set Enrichment
In order to explore differences of functional enrichment in OA and RA, the DEGs were implemented by Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. This was completed with the enrichKEGG function of the clusterProfiler R package. The parameters of the enrichKGG function were used as default (Yu et al., 2012). Autophagy score of each sample was thus calculated by single-sample gene set enrichment analysis (ssGSEA) (Barbie et al., 2009;Hnzelmann et al., 2013). Gene sets for ssGSEA were downloaded from the Human Autophagy Database (HADb, http://www.autophagy.lu/index.html). Differential autophagy levels were identified by Wilcoxon test, with p < 0.05 considered as significant statistical difference.
Acquisition and Verification of Autophagy Related Identified Markers
Overlapped genes among autophagy-related gene set and DEGs were selected as autophagy-related identified markers. Furthermore, an autophagy-related index (ARI) was constructed as identifier model using binary logistic regression of the generalized linear regression (GLR) algorithm. Finally, the GSE12021 dataset was used as a testing cohort to verify differential expression of markers and the accuracy of prediction by ARI. (Smolen et al., 2016), while OA is traditionally classified as non-inflammatory arthritis, and predominantly geriatric and degenerative (Glyn-Jones et al., 2015). In recent years, a large number of research studies have also shown that OA is related to individual immunity, but there is still great difference in the levels of immunity between RA and OA (Haseeb and Haqqi, 2013). Hence, we performed ssGSEA analysis to score immune factors for each of the RA and OA samples, based on 29 immune gene sets. Meanwhile, principal components analysis (PCA) was performed to explore differences in immune microenvironments between RA and OA based on scores of the 29 immune factors. The Wilcoxon test was further used to assess potential differences in immune cell environments of RA and OA. A p < 0.05 was considered as significant statistical difference.
Statistical Analysis
Statistical analyses were performed using R (version 4.0.4) (http:// www.r-project.org/) and its corresponding packages. The relationship between ARI and the 29 immune factors was determined through correlation analysis. The area under the curve (AUC) of the ROC curve was calculated using the ROCR R package. p-Values less than 0.05 indicated statistically significant differences in the analysis.
Differentially Expressed Genes of Osteoarthritis Vs. Rheumatoid Arthritis
The research flow chart is shown in Figure 1. In our study, 134 genes with p < 0.05 and |log2FC| > 1 were identified as differentially expressed genes between OA and RA by differential expression analysis. The number of differentially encoded genes for both upregulated and downregulated genes was 67 ( Figure 2).
Kyoto Encyclopedia of Genes and Genomes and Single-Sample Gene Set Enrichment Analyses
KEGG analysis declared that multiple important autophagyrelated signaling pathways were differentially enriched between RA and OA, such as NF-kappa B, TNF, NOD-like receptor, and PPAR ( Figure 3A, p adjust < 0.05). These signaling pathways were closely associated with autophagy (Jiang et al., 1998;Lin et al., 2013;Xu et al., 2019). Moreover, the ssGSEA algorithm showed that significant statistical differences existed between the autophagy levels of RA and OA ( Figure 3B, p 1.1e−07). In summary, these analyses indicated that differences in autophagy status are common in the two diseases, RA and OA.
Acquisition and Verification of Autophagy-Related Identified Markers
CXCR4 and SERPINA1 associated with autophagy were differentially expressed between OA and RA (Figures 4A-C, p < 0.05). The difference was also verified in a testing cohort ( Figure 4D,E, p < 0.05). Based on these two genes, an ARI was successfully constructed as an identified marker, based on binary logistic regression of GLR: Exp SERPINA1 and Exp CXCR4 represented the expressions of SERPINA1 and CXCR4 genes, respectively, in each sample. The p is the probability of being diagnosed as RA, and 1-p is the probability of being diagnosed as OA. The Pearson correlation analysis demonstrated that ARI was closely positively correlated with autophagy level, an indication that ARI could represent autophagy level ( Figure 5C, R 0.732, p 1e−10). Moreover, the expressions of SERPINA1 and CXCR4 genes were also closely positively correlated with autophagy level and was found consistent with the ARI results ( Figure 5A, R 0.7, p 6.63e−10; Figure 5B, R 0.584, p 1.19e−06).
To confirm the accuracy and potential ability of differential diagnosis of ARI, ROC values of two cohorts were calculated. Results showed CXCR4 and SERPINA1 had high accuracy in the training ( Figure 6A, AUC CXCR4 0.897, AUC SERPINA1 0.923) and testing ( Figure 6B, AUC CXCR4 0.842, AUC SERPINA1 0.833) sets, and the ARI built from CXCR4 and SERPINA1 performed with higher accuracy ( Figure 6A, AUC ARI 0.956, Figure 6B, AUC ARI 0.867). Figure 7 revealed that OA and RA were significantly divided into two groups based on the scores of 29 immune factors in the PCA analysis. This indicated RA and OA existed with difference in immune environments ( Figure 7A). Results of further analysis found that the infiltration score of most immune cells in OA was significantly lower than that in RA, while the infiltration score of mast cells was contrary (Figure 7 B,C, p < 0.05).
Relationship Between Immunity and Autophagy-Related Index
To explore relationship between autophagy and immune microenvironments, Pearson correlation analysis was used. Results showed that CXCR4, SERPINA1, and ARI were positively correlated with most immune factors but negatively correlated with mast cells (Figure 8, p < 0.05). This indicated autophagy was closely associated with immune microenvironments in both diseases.
DISCUSSION
RA is a chronic systemic disease mediated by both inflammatory and autoimmune responses (Smolen et al., 2016; Update on the Pathomechan, 2020; Page et al., 2021). OA which is traditionally classified as non-inflammatory arthritis has predominance as FIGURE 2 | The volcano plot of differentially expressed genes. Red dot represents upregulated genes from comparing OA with RA, while green dot represents downregulated genes. The number of differential encoded genes for both upregulated and downregulated genes was 67. RA, rheumatoid arthritis; OA, osteoarthritis.
Frontiers in Genetics | www.frontiersin.org October 2021 | Volume 12 | Article 743560 FIGURE 4 | (A) The Venn diagram between differentially expressed genes and autophagy-related genes. Results showed autophagy-related genes CXCR4 and SERPINA1 were differentially expressed in RA and OA. (B,C) Training cohort. Autophagy-related genes CXCR4 and SERPINA1 were higher expressed in RA than in OA.
(D,E) Testing cohort. Results of differential expressions of autophagy-related genes CXCR4 and SERPINA1 also verified in validation cohort, higher expressed in RA than in OA. Frontiers in Genetics | www.frontiersin.org October 2021 | Volume 12 | Article 743560 5 geriatric degenerative disease (Glyn-Jones et al., 2015). In recent years, a deepening understanding of these two diseases, RA and OA, showed similar and overlapping characteristics in their underlying mechanisms (Frangos and Maret, 2020). Therefore, although there are standard differential diagnostic methods for RA and OA, poor differential diagnosis still occurs in some cases (Toes and van der Woude, 2011;Kiltz et al., 2017). Prior research has shown that rheumatoid factor serves as an important index of diagnosis for rheumatoid diseases, with sensitivity and specificity being 60-90% and 85%, respectively. This finding indicates that the diagnosis of some patients with joint pain might not be accurate, especially at the early stages (Mun et al., 2021). Our in-depth review of OA cases found the proportion of such patients was still on the increase (Mobasheri, 2012).
Some studies have shown that many signaling pathways were differentially enriched between OA and RA (Charles, 2017;Tateiwa et al., 2019). In our analysis, we found that diverse differential signaling pathways were mainly mediated by 134 factors. From these, factors like NF-kappa B, TNF, NOD-like receptor, and PPAR have been confirmed to be closely related to autophagy of OA and RA. NF-kB, an important regulator of inflammation and immune response (Dell'Accio and Sherwood, 2015), is involved in many biological processes and its dysregulation is often observed in many conditions like arthritis and autoimmune diseases (Courtois and Gilmore, 2006; Herrington et al., 2015). Choi et al. concluded that targeted intervention in the NF-kB signaling pathway could be used as a treatment method for OA . Jiang et al. further demonstrated that dihydroartemisinin could inhibit the NF-kB signaling pathway and thus activate autophagy, inhibiting levels of catabolism and inflammatory factors in the chondrocytes of rats with osteoarthritis (Jiang et al., 2016). TNFα plays a key role in the pathogenesis of RA and it contributes to synovial inflammation and bone degradation (Kobayashi et al., 2000). Lin et al. demonstrated that autophagy was activated in RA in a TNFα-dependent manner and stimulated osteoclast differentiation, thus, promoting the progress of RA (Lin et al., 2013). This affirmed the importance of TNF signaling pathway in RA. Meanwhile, NOD-like receptor signaling pathway was also shown to play an important role in OA. The inflammatory corpuscles, in particular NOD-like receptor protein 3 (NLRP3), are the core of inflammation and a major cause of OA (Zhang et al., 2018). Zhou et al. demonstrated that autophagy could promote the production of IL-1B and inhibit the degradation of mitochondrial components, thus negatively regulating the NLRP3 signaling pathway and preventing its overactivation (Zhou et al., 2011). In addition, Xu et al. showed that the activation of autophagy reduced the activity of NLRP3 in articular cartilage and delayed its degeneration (Xu . In recent years, PPAR was also found to be closely related to energy metabolism, cell differentiation, apoptosis, inflammatory response, and autophagy (Jiang et al., 1998). Studies have shown that PPARγ-mTOR axis played an important role in the pathogenesis of OA which was closely related to autophagy (Dell'Accio and Sherwood, 2015; Vasheghani et al., 2015). These differentially enriched signaling pathways demonstrate the differences between RA and OA in autophagy levels. The importance of autophagy in OA and RA is thus clearly identified by previous studies and ours, and provides basis for further research. Hence, we believe that autophagy-related genes could identify and reflect differences between these two diseases, RA and OA. Our further analysis found that CXCR4 and SERPINA1 played a main role in differential autophagy levels of the two diseases. Relevant studies have shown that CXCR4 activated autophagy by mediating autophagy signaling pathways, and increased the levels of autophagy proteins and autophagosomes (Hashimoto et al., 2008;Wang et al., 2020). Xia et al. found that low-intensity pulsed ultrasound activated autophagy and enhanced the SDF-1/CXCR4 signaling pathway, thereby promoting the migration of mesenchymal stem cells and contributing to the enhancement of cartilage repair effect on OA (Xia et al., 2021). This further suggests that CXCR4 is of significance as an autophagy-related identification marker of OA and RA. Although the role of SERPINA1 in autophagy has not been studied much in OA and RA, it has been shown to be associated with autophagy in a variety of other diseases (Teckman et al., 2004;Jxd et al., 2020).
A likely role of SERPINA1 in RA and OA could thus be explored. An ARI composed of CXCR4 and SERPINA1 and established by us, served as identified markers of RA and OA. The accuracy of the ARI was verified in an independent cohort. The ARI should have great potential in clinical applications.
Rheumatoid arthritis (RA) is believed to be significantly related to autoimmunity, and immunotherapy is one current treatment method for the disease (Firestein and McInnes, 2017). In recent years, greater attention has been paid to the relationship between immunity and osteoarthritis (OA), and the role of immunity in the pathogenesis of OA was considered important (Rollín et al., 2008;Martel-Pelletier et al., 2016). However, the role of immunity in the two diseases was still different (Haseeb and Haqqi, 2013;Smolen et al., 2018). Our analysis gave results consistent with those of previous studies. The infiltration score of most immune-related factors in RA was significantly higher than that in OA (de Lange-Brokaar et al., 2012). Notable also was that the infiltration scores of mast cells were higher in OA than in RA. This was consistent with most immunological studies on OA (Dean et al., 1993;Pu et al., 1998), but some studies still reported lower numbers of mast cells (MCs) in OA compared with RA (Gotis- Graham and Mcneil, 2014). Moreover, the specific mechanism of the influence of mast cells on OA and RA was still not fully understood (Rivellese et al., 2017). The association of mast cells with local inflammation, chondrocyte apoptosis, cartilage breakdown, and positive autoantibodies (Rivellese et al., 2019;Wang et al., 2019) is cause for further study. We could thus confirm that immunity FIGURE 8 | Correlation analysis of autophagy-related genes CXCR4, SERPINA1, and ARI with 29 immune-related factors. Results indicated that autophagyrelated genes CXCR4, SERPINA1, and ARI were positively correlated with most immune-related factors but negatively correlated with mast cells.
Frontiers in Genetics | www.frontiersin.org October 2021 | Volume 12 | Article 743560 8 played an indispensable role in RA and OA, and further research on it ought to provide new directions for improving the diagnosis and treatment of RA and OA in the future.
Interestingly, our analysis found that the ARI was positively correlated with a variety of immune cells, while negatively correlated with mast cells. It could be inferred that autophagy was related to immunity in RA and OA, this being also consistent with results from a large number of studies on autophagy and immunity. Autophagy is closely related to immune functions such as intracellular bacterial clearance, inflammatory cytokine secretion, lymphocyte development, and pro-inflammatory signal transduction (Levine et al., 2011). Relevant studies have proved that autophagy-related gene polymorphisms were associated with a variety of autoimmune diseases, including rheumatoid arthritis, systemic lupus erythematosus, psoriasis, and multiple sclerosis (Wu and Adamopoulos, 2017). Masaru et al. proved that autophagy was associated with the survival of RA fibroblastic synovial cells (RA-FLs), and Ra-FLs are the primary source of pro-inflammatory cytokines responsible for the activation of osteoclasts and subsequent bone destruction (Kato et al., 2014). Wang et al. indicated that monocyte chemoattractant-protein-induced protein 1 (MCPIP-1) could upregulate autophagy-related protein Beclin-1 and microtubule-associated protein 1A/1B-light chain 3 (LC3) in monocytes, thereby promoting an initial differentiation of monocytes into osteoclasts by increasing endoplasmic reticulum stress and oxidative stress, which was closely related to bone resorption in OA (Wang et al., 1093). Thus, autophagy was closely related to immunity in RA and OA, and any correlation between ARI and immunity would further enhance the feasibility and potential of our identification indicators.
In general, the ARI could reflect not only the differences in autophagy but also the differences in immunity in OA and RA. This further increases the potential ability and reliability of ARI in differentiating RA and OA. The ARI composed of two genes was expected to provide new ideas for the diagnosis and treatment of OA and RA and further research on new therapeutic drugs. However, there are some limitations to the study that are worth considering. We included independent datasets for validation but this study was retrospective as the number of datasets was small. Therefore, a well-designed prospective study is needed in the future to further confirm the discriminative ability and accuracy of the ARI.
CONCLUSION
In this study, a comprehensive autophagy-related index (ARI) consisting of autophagy-related genes CXCR4 and SERPINA1 was successfully constructed for distinguishing RA and OA. The ARI showed a high differential diagnostic accuracy and could well reflect the immune and autophagy status of RA and OA. To a certain extent, this provides hope for the difficulties in distinguishing RA from OA. This finding is conducive to prompt implementation of clinical decision-making and intervention measures.
AUTHOR CONTRIBUTIONS
FL conceptualized the study. RH provided the visualization and formulated the methodology of the study. JZ was in charge of the software used in the study. RH, JZ, FL, QL, WH, and DZ wrote the original draft. QW was in charge of the project administration and supervised the study. | 2021-10-12T13:16:49.404Z | 2021-10-12T00:00:00.000 | {
"year": 2021,
"sha1": "3cacfe48f28e5ea2285ac170713e57999bcd8596",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fgene.2021.743560/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3cacfe48f28e5ea2285ac170713e57999bcd8596",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
203581547 | pes2o/s2orc | v3-fos-license | Atomic-resolution crystal structures of the immune protein conglutinin from cow reveal specific interactions of its binding site with N-acetylglucosamine
Bovine conglutinin is an immune protein that is involved in host resistance to microbes and parasites and interacts with complement component iC3b, agglutinates erythrocytes, and neutralizes influenza A virus. Here, we determined the high-resolution (0.97–1.46 Å) crystal structures with and without bound ligand of a recombinant fragment of conglutinin's C-terminal carbohydrate-recognition domain (CRD). The structures disclosed that the high-affinity ligand N-acetyl-d-glucosamine (GlcNAc) binds in the collectin CRD calcium site by interacting with the O3′ and O4′ hydroxyls alongside additional specific interactions of the N-acetyl group oxygen and nitrogen with Lys-343 and Asp-320, respectively. These residues, unique to conglutinin and differing both in sequence and in location from those in other collectins, result in specific, high-affinity binding for GlcNAc. The binding pocket flanking residue Val-339, unlike the equivalent Arg-343 in the homologous human surfactant protein D, is sufficiently small to allow conglutinin Lys-343 access to the bound ligand, whereas Asp-320 lies in an extended loop proximal to the ligand-binding site and bounded at both ends by conserved residues that coordinate to both calcium and ligand. This loop becomes ordered on ligand binding. The electron density revealed both α and β anomers of GlcNAc, consistent with the added α/βGlcNAc mixture. Crystals soaked with α1–2 mannobiose, a putative component of iC3b, reported to bind to conglutinin, failed to reveal bound ligand, suggesting a requirement for presentation of mannobiose as part of an extended physiological ligand. These results reveal a highly specific GlcNAc-binding pocket in conglutinin and a novel collectin mode of carbohydrate recognition.
The collectin family of mammalian innate immune system proteins recognizes pathogens by the repeating pattern of carbohydrates on their surfaces and organizes their disposal through neutralization, agglutination, and opsonization while recruiting immune system cells for their destruction (1,2). Collectins are also involved in clearance of apoptotic cells and modulation of inflammatory and immune responses (3,4). The surfactant proteins SP-A and SP-D are found in the lung where they form the first line of defense against inhaled pathogens and allergens. Mannose-binding lectin (MBL) 5 is a serum protein which, in common with CL-L1, CL-K1, and CL-P1, is able to activate the complement system to dispose of pathogens (5,6). In addition to the collectins found in other mammals, the family Bovidae has three serum collectins: bovine conglutinin, CL-43, and CL-46. Phylogenetic analysis indicates that these bovine collectins evolved from a common ancestral SP-D gene after the divergence of the Bovidae from the mammalian line (7).
The rapid evolution of additional collectins suggests that their presence confers a survival advantage to the animals. The digestive system in ruminants relies on a symbiotic relationship with vast numbers of microorganisms in the rumen, which are able to break down the cellulose in food. The bovine serum collectins provide extra defense against microorganisms which leak into the bloodstream. As well as binding, neutralizing, and opsonizing the microorganisms, it is thought that they lead to reduction in inflammatory responses (7), whereas low conglutinin levels in plasma are associated with increased susceptibility to infection (8). Mice injected with conglutinin followed by Salmonella typhimurium showed significantly reduced blood bacterial levels and greater survival compared with controls (9). In vitro experiments also demonstrated antibacterial activity of conglutinin against S. typhimurium and Escherichia coli (9). In addition to this antimicrobial role, conglutinin levels in the Bovidae vary with diet, season, infections, and stage of the reproductive cycle (10), with a significant presence in fetal serum (11). The significance of these variations remains unclear.
Bovine serum conglutinin has been shown to bind, through the globular carbohydrate recognition domain (CRD), to specific carbohydrate structures on complement effector protein iC3b, which arises following activation and cleavage of complement component C3 (12). This is thought to enhance complement-dependent clearance through agglutination of complement-reacted microorganisms. The iC3b ␣-oligosaccharide is of high-mannose type, being (Man) 9 (GlcNAc) 2 -Asn or (Man) 8 (GlcNAc) 2 -Asn (13,14), and the binding is proposed to be via the terminal ␣1-2 mannobiosyl unit (15). It is suggested that C3b undergoes a conformational change that exposes its glycan when cleaved (iC3b) leading to conglutinin recognition of iC3b (13). The binding of conglutinin to iC3b is inhibited by N-acetylglucosamine (GlcNAc) (12) which, unique to the bovine collectins conglutinin and CL-46, is reported to be the highest affinity ligand, in the case of conglutinin by an order of magnitude relative to other monosaccharide ligands such as N-acetyl mannosamine (ManNAc), mannose and glucose and to the maltose disaccharide (16 -20). Conglutinin has been shown to bind to some complex-type as well as high mannosetype oligosaccharide chains (21). This is thought to be via terminal GlcNAc or mannose residues, and for the high mannosetype there seems to be a preference for the terminal man␣1-2 sequence although there is increased reactivity from the Man 5 to the Man 8 structures. Complement-independent anti-mycobacterial activity has also been demonstrated for a recombinant neck plus CRD fragment of conglutinin which is thought to bind directly to Mycobacterium bovis Bacille Calmett-Guérin (BCG) lipoarabinomannan in a calcium-dependent manner, inhibiting uptake by macrophages (22). Binding to yeast carbohydrate structures including zymosan and mannan has also been reported (23).
Bovine serum conglutinin has been shown to inhibit influenza A virus (IAV) more efficiently than other collectins, and a functionally intact recombinant homotrimer (neck plus CRDs) (24) has been shown to be active against IAV (25). An engineered SP-D/conglutinin (neck/CRD) fusion protein (26) appears to combine the beneficial properties of the two components with enhanced IAV neutralizing activity, whereas targeted mutation (SP-D R343V) enhances SP-D recognition of high mannose viral glycans (27). Reported ligand-bound crystal structures include CRD and trimeric fragments of MBP/MBL, and trimeric fragments of human SP-D, porcine SP-D, rat SP-A, and CL-K1 (27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37). These structures reveal a conserved calcium-binding site within the CRD as part of the significant sequence conservation throughout the ligand-binding domain of the CRD.
We present here high-resolution crystal structures with and without bound ligand of a recombinant fragment of the C-terminal carbohydrate recognition domain of conglutinin, the first crystal structures of a bovine collectin. The results reveal a highly specific binding pocket for GlcNAc, as opposed to other monosaccharide ligands, and a novel collectin mode of carbohydrate recognition which utilizes interactions which extend beyond the conserved collectin calcium-binding pocket and the binding-site flanking residues.
Protein expression and purification
Expression of the trimeric recombinant neck plus CRD conglutinin fragment provided ϳ40 mg protein per liter culture, with almost all the protein in the inclusion bodies as evidenced by the absence of corresponding protein in the sonicate supernatant. The recombinant fragment showed a relative molecular mass of ϳ23 kDa under reducing conditions on a 15% (w/v) acrylamide gel, corresponding to the monomeric recombinant neck plus CRD (rfBC) (Fig. 1a). The recombinant fragment sequence is shown in Fig. 2.
The protein eluted as a single peak from a Q-Sepharose column and the subsequent Mono Q column at salt concentrations of ϳ200 mM and ϳ225 mM, respectively. Following ion exchange purification, the MW of rfBC under nondissociating conditions was determined by gel filtration, the recombinant protein eluting as a single peak corresponding to 64 KDa, which indicated the trimeric nature of the recombinant rfBC (Fig. 1b). The N-terminal sequence obtained following chromatographic purification and characterization was GSAEANALKQRVTIL-DGHLRRFQNAFSQYKKAVLFPDGQA, where the first five residues GSAEA are from the vector. Only a single sequence was observed, with the conglutinin sequence starting at Asn-199 in agreement with the published sequence in the protein
Crystal structures of native and ligand-bound conglutinin
database (Fig. 2). Cysteine residues are present only in the CRD of the recombinant fragment (Fig. 2).
Crystallization of the purified protein resulted in unliganded and ligand-bound crystal structures of the recombinant fragment of bovine conglutinin which revealed a monomeric CRD rather than the expressed trimeric rfBC. The protein used for crystallization, several months after the initial preparation, was recharacterized by N-terminal sequencing and SDS-PAGE (data not shown). The results showed that 80% of the preparation had the N-terminal sequence starting at conglutinin Lys-224 or Ala-225 whereas the remaining 20% was composed almost equally of material starting at conglutinin Arg-214 and Val-205 (Fig. 2). The SDS gel of the protein also showed that the molecular weight of the majority of the recombinant fragment was less than that expected for the recombinant neck plus CRD monomer. This indicates that the majority of the protein used in the crystallization had lost a substantial amount of the neck region between initial preparation and crystallization and was a monomer of 21.3 kDa, which is consistent with the monomeric CRD fragment seen in the crystal structures.
Crystal structure
Both unliganded and ligand-soaked structures have been determined. The initial structure of the conglutinin CRD showed electron density at the CRD ligand-binding site (Ca1) into which the cryoprotectant, ethylene glycol, was subsequently fitted. A further structure was therefore determined, to a higher resolution of 0.97 Å, utilizing 2,4 methanepentanediol (MPD) as an alternative cryoprotectant. This structure did not show any extra electron density in the ligand-binding site, except for water molecules, and is therefore designated as the native structure. To investigate the structural detail of the known binding properties and specificities of conglutinin (16 -20), various ligands which should be able to access the binding site in the crystal were introduced into crystals by adding ligand to cryobuffers. These included GlcNAc, maltoheptaose, inositol 1-monophosphate, ␣-1,2-mannobiose, and ManNAc. Only the highest affinity ligand, GlcNAc, was visualized in the ligandbinding site along with the closely similar N-acetyl-␣-D-glucosamine 1-phosphate (␣GlcNAc-1-P) (Fig. 3). Initial experiments for ␣-1,2-mannobiose, implicated in the binding of conglutinin to iC3b, used a ligand molar excess of three times that was successfully used for GlcNAc but did not show any ligand in the crystal. Extending the soaking time and final ligand concentration in the cryobuffers still did not show the ligand in the binding site, or at any other location in the protein. The GlcNAc and ␣GlcNAc-1-P ligand-bound structures were subsequently refined along with the unliganded and ethylene glycol-bound structures (see Table 1).
The conglutinin CRD monomer has an overall topology similar to that of the known collectin structures (Fig. 4). In all struc-tures, calcium Ca1 is present at the collectin CRD ligand-binding site. Additionally, except for the structure without bound ligand, there is evidence of a calcium ion in the collectin Ca2 site although the electron density and the average B-factors (Ca1 14.0 Å 2 , Ca2 33.0 Å 2 ) suggest this site is not as fully occupied as Ca1, particularly for the ethylene glycol and ␣GlcNAc-1-P structures. In the unliganded structure, the lack of electron density for the calcium ion at the Ca2 site is accompanied by a lack of electron density, or movement away from the site, for some of the side chains of the Ca2 coordinators. There is no electron density in any of the structures to support the presence of the third calcium ion observed in some collectins. Ca1 is coordinated in a similar way to other structurally characterized collectins ( Fig. 5 and Table 2). Protein coordination to Ca1 is by Glu-315, Asn-317, Glu-325, Asn-337, Asp-338, and the carbonyl O of Asp-338. In the ligandbound structures Ca2 is coordinated by Asp-291 (2), Glu-295 (2), Asn-318 (this interaction is not present in the ␣GlcNAc-1-Pbound structure), Asn-326, the carbonyl O of Glu-325, and one water molecule.
Ligand-binding site
Adjacent to the calcium CRD site, the longer of the two loops involved in binding to the calcium ions (see Fig. 4) has an insertion compared with the homologous collectin human surfactant protein D (hSP-D) (Fig. 6). These additional two residues compared with hSP-D result in an extended loop near the CRD ligand-binding site, in a similar manner to pig surfactant protein D (pSP-D). In the unliganded and ␣GlcNAc-1-Pbound structures this Asn-318 -Pro-324 loop is disordered (318 -323). Upon binding of the ligand GlcNAc or ethylene glycol the 318 -324 loop becomes ordered, although the electron density is not as clearly defined as in the majority of the CRD structure. Within the structure without bound ligand there is some ill-
Crystal structures of native and ligand-bound conglutinin
defined electron density for the loop that could correspond to the ligand-bound loop, suggesting a small degree of order. However, it would not have been possible to discern the location of the loop without the prior knowledge of the position of the loop in the ligand-bound structures, supporting that the loop is mobile in the absence of ligand.
The GlcNAc-bound structure shows evidence of both ␣ and  anomers present in the electron density (Fig. 5), reflecting the heterogeneous mixture of the commercially sourced GlcNAc. GlcNAc interacts with the CRD calcium and its coordinators Glu-315 OE2, Asn-317 ND2, Glu-325 OE2, and Asn-337 ND2 via the O3Ј and O4Ј hydroxyls, with Asp-320 from the extended loop interacting with the N group of GlcNAc (␣ 3.04 Å;  3.18 Å) ( Table 2). Additionally, Lys-343 NZ interacts with O7 on the acetyl head group (2.79 -2.83 Å) (Fig. 5). Lys-343 adopts the same conformation in all the structures, being held in place by Glu-329, which in turn is interacting with Asn-337 which coordinates Ca1 and hydroxyl of bound ligand (Fig. 7). There is some evidence in the electron density of an interaction between where F oh and F ch are the observed and calculated structure factor amplitudes, respectively, for the reflection h. c R free is equivalent to R conv for a 5% subset of reflections not used in the refinement. Both ␣GlcNAc and GlcNAc are seen in the binding site with the anomeric carbon C1 indicated. Amino acid interactions are shown for ␣GlcNAc. GlcNAc interactions are the same except for that with Glu-321 which is only present for ␣GlcNAc. The Lys-246 interacting residue from a neighboring subunit via a crystal contact is shown in blue and calcium ion Ca1 is shown as a green sphere.
Crystal structures of native and ligand-bound conglutinin
Glu-321 in the extended loop and O1Ј of the ␣ anomer of GlcNAc. In addition to the CRD interactions with the N-acetyl group, Lys-246 from a symmetry-related molecule interacts with GlcNAc O5 and O6Ј (Fig. 5). This residue is in a similar orientation in all the structures determined, including that without ligand, regardless of interactions between protein and ligand. The ␣GlcNAc-1-Pbound structure shows ligand binding in the calcium site and the additional protein-ligand interaction with Lys-343 as seen for GlcNAc. Residues 318 -323 in the extended loop are disordered in the ␣GlcNAc-1-P structure, the bulky ␣-linked phosphate precluding the interaction of Asp-320 with the ligand nitrogen (Fig. 3) and preventing the loop adopting the orientation seen in the other ligand-bound structures.
The structure of the conglutinin CRD with ethylene glycol from the cryoprotectant bound at the ligand-binding site reveals ligand binding in a similar manner to that seen for all collectin ligands to date. Two hydroxyls from ethylene glycol, which have the same configuration as the O3Ј and O4Ј mannose-type pair, interact with the binding-pocket residues and Ca1. For this small ligand no further interactions with protein are seen. Although the 318 -324 loop in the ethylene glycol-
Crystal structures of native and ligand-bound conglutinin
bound structure makes no direct interactions with the ligand, the loop is ordered and adopts the same overall conformation as that seen in the GlcNAc-bound structure.
Discussion
The trimeric recombinant fragment comprises amino acids 199 -351 of conglutinin, the N-terminal sequence of the preparation being confirmed by sequencing. The crystal structure however reveals only the monomeric CRD, with electron density definition at the N terminus starting at around amino acid 230 (the residue neighboring proline 229 at the neck-CRD interface, Figs. 2 and 3). Subsequent N-terminal sequencing of the protein used for crystallization revealed that the majority (ϳ80%) starts at amino acid 224/225 (KAVLFP/AVLFP; Fig. 2) consistent with the crystal structure. The remaining 20% gave an N-terminal residue of either Arg-214 or Val-205 (Fig. 2). A significant portion of the recombinant chain (25/26 residues for the major component) is believed to have been proteolytically cleaved just prior to the neck-CRD interface because of traces of proteases in the preparation. The time delay of several months between preparation and use, for crystallization, may also be a contributory factor. The absence of the trimerizing neck residues, known to be essential for trimer formation (24,38) results in the dissociation of the three chains of rfBC to give the monomeric CRD observed here.
The conglutinin CRD fold is very similar to the other known collectin structures (Fig. 4). The overall structure is stabilized by the two typical collectin CRD intermolecular disulfide bridges, Cys-255-Cys-349 and Cys-327-Cys-341 here (Fig. 2). No other cysteines are present in the expressed fragment providing no opportunity for intrachain disulfides to be formed either in the full expressed trimeric fragment (Fig. 1b) or the cleaved fragment present during crystallization. Calcium Ca1 at the CRD ligand site is coordinated by conserved collectin residues. The second calcium ion Ca2 is not as tightly bound as Ca1 (as evidenced by higher B-factor compared with Ca1) and is not observed in the structure without bound ligand. This may be because of a number of interdependent factors including the loss of order of 318 -323 in the calcium binding loop, loss of Ca2 coordination by Asn-318, the lack of an interaction between Asp-320 and ligand, and low calcium concentration in the crystallization buffer. A third calcium ion is not present in any of the conglutinin CRD data consistent with the presence of Asn at the Ca2/Ca3 coordinating residue conglutinin 326 rather than the Asp present in rat MBP-A, hSP-D, and human MBL which all contain a third calcium ion (Fig. 6).
Although the structures of the conglutinin CRD reveal ligand binding to calcium in a similar manner to that seen for all collectin ligands to date, with two hydroxyls which have the same configuration as the O3Ј and O4Ј hydroxyls of mannose/glucose-type rings interacting with Ca1, there are additional specific interactions with the high-affinity GlcNAc ligand. The N-acetyl head group interacts with Lys-343 (via O7Ј) and with Asp-320 (via N) in the extended conglutinin loop. Conglutinin does not discriminate between the ␣/ GlcNAc anomers, with the crystal structure suggesting that both are bound with similar occupancy despite an indication of an additional proteinligand interaction involving Asp-321 and the anomeric hydroxyl O1Ј in the ␣ anomer. As this anomer is only partially occupied (as evidenced by the presence of both ␣ and  forms in the structure), this may explain why the Glu-321 density is not as clearly defined as may be expected for an interacting residue.
The head group NZ of conglutinin Lys-343 which interacts with GlcNAc O7Ј is held in position by Glu-329 which is linked to the collectin-conserved Asn-337 which coordinates the CRD calcium Ca1 and hydroxyl of bound ligand (Figs. 6 and 7). The Lys-343 occupies a position between the small, hydrophobic residues Val-339 and Leu-345 (Fig. 7) which are replaced in hSP-D by Arg-349 (hSP-D numbering) and the key flanking residue Arg-343. Thus in conglutinin the flanking residue Val-339, equivalent to SP-D Arg-343, indirectly influences, and enhances, GlcNAc binding by allowing O7Ј of the acetyl head group direct access to Lys-343 NZ (NZ-O7Ј 2.79 -2.87 Å) (Fig. 7). Positioning GlcNAc in hSP-D in a similar manner to that observed here would give a clash between the hSP-D Arg-343 head-group and ligand, specifically the acetyl head group of GlcNAc (39) which is presumably why ManNAc is rotated 180°w hen bound to hSP-D (29). This would also be true for pSP-D which has Lys and Arg in equivalent positions to hSP-D Arg-343 and Arg-349 (Fig. 6).
The specificity of conglutinin for GlcNAc provided by the Lys-343 to N-acetyl binding alongside the standard collectin interactions of the Glc O3Ј and O4Ј hydroxyls with calcium and protein, is enhanced by the extended loop Asn-318 -Pro-324, bounded at both ends by conserved residues which coordinate both ligand and Ca1, which becomes ordered on GlcNAc ligand binding, with Asp-320 OD1 making a specific interaction with the N-acetyl nitrogen (Fig. 5). The structure of the ␣GlcNAc-1-P complex again shows interaction with Lys-343 in addition to Ca1 binding-site residues, but with the extended flexible loop displaced by the phosphate group such that Asp-320 is no longer within binding distance. The extended flexible loop in pSP-D (insertion of GSS; Figs. 4 and 6) also becomes more ordered on sugar binding, although no specific interactions between the loop and bound ligand are reported (40,41), whereas the RAK insertion in CL-43 is reported to enhance specificity for mannose-rich viral glycoconjugates (42).
The combination of a flanking valine (339) with Lys-343 and Asp-320 appears to be a key element in the recognition and binding of GlcNAc by conglutinin with an affinity order of magnitude higher than other monosaccharides (16 -20). The only
Crystal structures of native and ligand-bound conglutinin
other collectin where GlcNAc is reported to bind with high affinity (7) is CL-46 which also has a flanking valine, however the potential GlcNAc N-acetyl coordinating residues, in equivalent positions to conglutinin Lys-343 and Asp-320, are glutamate and asparagine, respectively (Fig. 6). Several other collectins have specificity, albeit usually weak, for GlcNAc ligands (16,17), the known structures with bound GlcNAc or GlcNAc type terminal ligand revealing the orientation of the bound sugar ring as generally reversed to that observed in the conglutinin structure here. This is the case for two of the three subunits in the GlcNAc-rat MBP-A trimeric complex, for GlcNAc-rat MBP-C (33,34), and for ManNAc-hSP-D (29). The exception is the C subunit of rat MBP-A (34) which shows the same GlcNAc orientation as conglutinin, although this is thought to be because of crystal contacts and there are no specific interactions between the protein and the acetyl group. Although the ability of crystal contacts to influence ligand orientation is well-documented (32), the interaction here of Lys-246 from a neighboring subunit with GlcNAc O5 and O6Ј does not appear to influence ligand orientation, protein-GlcNAc binding determinants (Glu-315, Asn-317, Glu-325, Asn-337, Asp-320, Ca1, Lys-343), or the extended loop 318 -324, because each of these for GlcNAc is closely similar to those in both the unliganded and the ethylene glycol-bound structures where the Lys-246 -ligand interaction is absent. The ligand-binding site is formed ready to interact with the incoming ligand, the NZ of the contact residue Lys-246 then shifting by ϳ2 Å to interact with the bound GlcNAc ligand.
The biological and physiological relevance of the high specificity for the N-acetyl group, and for GlcNAc in particular, is far from clear. GlcNAc3 Glc represents the conserved motif hexose-132-hexose in cores of many Enterobacteriaceae, including all Salmonella, which share a common major core structure (that of the Ra mutant), and E. coli R2 and K12 (43)(44)(45). Where the LPS is extended to O antigen the Ra GlcNAc is present as a branch linked through O1Ј to the LPS main chain glucose and is thus available for recognition by conglutinin. The ability of the conglutinin-binding pocket to accommodate the phosphate in GlcNAc-1-P is an important indicator that 1-linked GlcNAc such as that of the Salmonella LPS can be recognized and bound by conglutinin. This is consistent with the in vitro and in vivo studies which demonstrate the antibacterial activity of conglutinin against S. typhimurium and E. coli (8,9).
Conglutinin, which is known to bind to high mannose-type oligosaccharide with a preference for the terminal man␣1-2 sequence (21), binds to the C3-derived glycoprotein iC3b present on pathogens and facilitates clearance (13)(14)(15), with binding reported to be exclusively mediated through the N-linked Man oligosaccharides on the iC3b ␣ chain (23,46). In addition, binding to yeast mannan and complement-activated immune complexes has been shown to be calcium-dependent and inhibited by GlcNAc (12), suggesting a central role for the CRDbinding site. In our studies, soaking of ␣1-2-mannobiose into the unliganded crystals, successfully demonstrated for hSP-D (32), failed to reveal any ligand in the CRD site (or elsewhere) even at high concentrations. Because the binding site appears to be openly accessible in the crystal, the failure to bind this simple mannobiose suggests a requirement for the iC3b ␣ chain oligosaccharide to be presented by the carrier protein iC3b in a conformation which enables high-specificity binding (46,23), perhaps utilizing an extended binding site as for Salmonella enterica Minnesota R5 LPS (36). This is supported by Mizuochi et al. (21), who report that high mannose-type oligosaccharides exhibit increased reactivity from the Man 5 to the Man 8 structures. Calcium-dependent binding of M. bovis Bacille Calmett-Guérin by conglutinin is also thought to be via terminal dimannose units of the bacterial lipoarabinomannan (22), our inability to visualize binding of simple ␣1-2-mannobiose in the openly accessible ligand-binding site further suggesting a requirement for presentation of dimannose as part of extended physiological ligands. The primary biological roles of bovine conglutinin remain to be determined as do the details of the binding interactions with iC3b, bacterial LPS, and viral glycans. The uniquely designed, highly specific GlcNAc-binding pocket and the ability of GlcNAc to abolish binding to iC3b suggest a dual role for conglutinin, directly binding to iC3b via the high mannose oligosaccharide attached to the N-terminal of the ␣ chain and directly binding pathogens via exposed and accessible GlcNAc on the pathogen surface. The competition between these two conglutinin targets for the Ca1-binding site suggests overlap between the two recognition sites. Whether conglutinin can simultaneously bind, via different homotrimeric CRDs, to both iC3b coated microorganisms and to the microorganism itself is not at all clear, but it would be surprising if the exquisitely designed GlcNAc-binding site was not central to the associated biological roles.
Protein expression
Expression of the recombinant fragment (rf) of bovine conglutinin (BC) consisting of the CRD and neck regions (rfBC, residues 199 -351 of native conglutinin) followed the procedure described by Prasad (47) which is based on the initial cDNA clone of Lu et al. (48) and the transformation into competent E. coli DH5␣ cells as described by Wang et al. (24). Briefly, the clone was purified from DH5␣ and used to transform competent NovaBlue cells for maintaining the plasmid. A single colony was picked from a LB agar plate (containing 50 g/ml carbenicillin and 34 g/ml chloramphenicol) streaked with BL21(DE3)pLysS containing the recombinant plasmid and was used to inoculate 10 ml of LB media. The culture was incubated overnight at 37°C with vigorous shaking (225 rpm). Cells were harvested next day at 2000 ϫ g and washed once in fresh LB media. The cell pellet was then resuspended into 5 ml of fresh media and used as inoculation for 1 liter of fresh LB media containing the antibiotics chloramphenicol (34 g/ml) and carbenicillin (50 g/ml) and incubated at 37°C with vigorous (225 rpm) shaking.
The cells were induced at A 600 of 0.6 with isopropyl--Dthiogalactopyranoside (IPTG, NOVA Biochem), to a final concentration of 0.5 mM, and incubated further for 2 h. The culture was then chilled on ice for 30 min with gentle shaking. The cells were harvested at 8000 ϫ g for 10 min at 4°C and washed once with fresh media. Samples were analyzed on SDS-PAGE. Sam-
Crystal structures of native and ligand-bound conglutinin
ples were heated to 100°C for 5 min (10 min for total cell lysates). 20 l of each sample was loaded onto the wells and run on a 15% (w/v) acrylamide gel.
Protein extraction
The cell pellet was resuspended in 25 ml of ice-cold cell-lysis buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM CaCl 2 ) and frozen at Ϫ80°C. The cells were thawed on ice at 4°C and the remaining 25 ml of lysis buffer containing Triton X-100 (final concentration 0.05% v/v), DNase (5 g/ml), MgCl 2 (0.04 M), and proteases inhibitor mixture tablet was added. The thawed cell suspension was shaken at 4°C for 45 min. The lysate was then sonicated with 10 30-s pulses of 16 s amplitude at 15-s intervals. After sonication, the cell soup was centrifuged at 15,000 ϫ g. The protein, pelleted as inclusion bodies, was washed twice with ice-cold deionized water to remove traces of Triton X-100.
The pellet was then subjected to a denaturing and refolding procedure, involving repetitive dialysis against concentrations of urea decreasing from 6 M to zero. Before use, all urea solutions were passed through Duolite MB6113 Indicator Mixed Resin (BDH) to remove traces of cyanate ions, which may irreversibly bind to the protein and affect function. The pellet was resuspended in dialysis buffer (20 mM Tris-HCl, 150 mM NaCl, 5 mM CaCl 2 , and 5% v/v glycerol) containing 6 M urea and shaken for 30 min at 4°C. The denatured protein solution was centrifuged to remove insoluble material. An activated dialysis membrane tubing (Medicell 10 K MWCO) was filled with the cleared supernatant and placed in 2 liters of dialysis buffer containing 4 M urea. The protein was dialyzed against decreasing concentrations of urea with buffer changes every 2 h followed by an additional 24-h dialysis to remove traces of urea. The dialysate was then centrifuged at 15,000 ϫ g to remove contaminating aggregates. Samples were analyzed on SDS-PAGE for purity as described elsewhere.
Protein purification and characterization
Protein samples were passed through Anotope 0.2 m filters (Whatman) prior to loading. Purification was performed on a Pharmacia Biotech AKTA purifier FPLC system.
Ion-exchange purification
20 ml of the dialyzed recombinant protein solution was diluted to 50 ml with water and loaded at 2 ml/min onto a Fast Flow Q-Sepharose (16 ϫ 200 mm) column in 20 mM Tris-HCl buffer containing 5 mM CaCl 2 and 50 mM NaCl. A linear gradient, 0 -50% buffer B (buffer A containing 500 mM NaCl), for 90 min was used to elute the protein in 2-ml volume fractions. The fractions were analyzed by SDS-PAGE on 15% (w/v) and the peak fractions pooled. 10 ml of the sample from the Q-Sepharose, diluted to 50 ml with water to reduce the salt concentration, was loaded onto a 5 ϫ 50 mm Mono Q column at 1 ml/min. A steep linear gradient 0 -100% buffer B was applied over 5 min. 1-ml fractions were collected and analyzed by 15% SDS-PAGE.
Gel-filtration chromatography
The peak fractions from Mono Q were loaded onto a prepacked Superdex-75 (HR 10/300 mm) column, 200-l protein samples were loaded onto the column in Tris buffer, containing 10 mM EDTA. The flow rate was 1 ml/min and 0.5-ml fractions were collected. The column was calibrated with alcohol dehydrogenase (150 kDa), BSA (66 kDa), carbonic anhydrase (29 kDa), and cytochrome C (14.4 kDa).
N-terminal sequencing
The protein samples were run on a 4 -12% (w/v) bis-tris NuPAGE precast gel (Novex UK Ltd) at 200 mA per gel using a Novex XCell II Mini-cell gel apparatus. The gel was electroblotted to ProBlott membrane (PE Applied Biosystems) in a Bio-Rad Trans-Blot electrophoretic transfer cell. The membrane was stained with Coomassie Brilliant Blue. The rfBC band was excised and sequenced using Edman degradation on an Applied Biosystems 494A Procise protein sequencer (PE Applied Biosystems UK). The samples were run on the sequencer by applying to a glass fiber disc pretreated with polybrene to limit sample washout.
Crystallization and data collection
Crystals of conglutinin were grown in sitting drops consisting of an equal volume (2 l) of protein solution (8 mg ml Ϫ1 protein in 13 mM Tris, 100 mM NaCl, 7 mM CaCl 2 , pH 7.5) and precipitant buffer (2.5 M (NH 4 ) 2 SO 4 , 0.1 M Tris, pH 8.0 -8.5). For the initial data, crystals were prepared for cryocooling using ethylene glycol in precipitant buffer. Successive addition of 2-l aliquots of increasing concentrations (5, 10, 15, 20%) of ethylene glycol cryobuffer were added to the well, followed by addition of a further 2-l aliquot of 20% cryobuffer, and an exchange of 10 l of the resulting well buffer with 20% cryobuffer.
Further experiments indicated the crystals lasted longer in 2-4 methanepentanediol and this was then used as the cryoprotectant for a further unliganded data set and for all the ligand-bound data, with ligands introduced into crystals by their inclusion in the cryobuffer. Ligand concentrations in the cryobuffers for those ligand soaks that showed bound ligand in electron density maps were 10 mM for GlcNAc and 0.1 M for ␣GlcNAc-1-P, with CaCl 2 added to give a final concentration of 10 mM in the cryobuffers. All ligands were purchased from Sigma. Where ligand could not be identified in electron density maps, extended soaking times at higher molarity were followed by further data collection. All data were collected either at Daresbury SRS or Diamond Light Source (DLS). Integrated intensities were calculated with the program MOSFLM (49) and data processed using CCP4 suite (50). Data collection and processing statistics are given in Table 1.
Structure solution and refinement
The structure of conglutinin without bound ligand was solved by molecular replacement with AMoRe (51) using the homologous hSP-D structure (PDB ID 1PW9) as a search model. The recombinant homotrimeric fragment of conglutinin consisted of the C-terminal 153 residues (Asn-199 to Phe-351). The size of the unit cell indicated that a trimer would not fit into the unit cell and there was space only for a monomeric unit. The neck region of collectin molecules is required for trimerization (24,38), suggesting that some or all of the Crystal structures of native and ligand-bound conglutinin expressed neck region (Fig. 2) was absent from the crystallized protein. This was confirmed by subsequent recharacterization of the protein used for crystallization by N-terminal sequencing and SDS-PAGE.
Only the CRD portion (residues hSP-D 236 to 355 which correspond to conglutinin residues 230 to 351) of the native recombinant fragment hSP-D structure (PDB ID 1PW9) was thus used as a molecular replacement model. One clear rotation solution was found in AMoRe and translation functions indicated a space group of P4 3 . Initial electron density maps (50) indicated the CRD portion only was present in the structure. Isomorphism was sufficient to allow the coordinates of the unliganded conglutinin CRD structure to be used as a starting model for the ligand-bound structures. The ligand-soaked structures which gave no sign of ligand in the electron density were not refined further. Cycles of iterative model building and refinement were performed; early work on the ethylene glycol and GlcNAc-bound structures used O (52) and CNS (53), however COOT (54) and REFMAC5 (55) were subsequently used throughout for refinement of these and all further structures. Final models were checked via MolProbity (56) and using Protein Data Bank validation software. For all the structures, the first few residues at the N terminus are not visible in the electron density, definition starting at amino acid 230 (Fig. 2). Final details of the refinement are given in Table 1. There are no residues in the outlier region of the Ramachandran plot (Mol-Probity). Figures were generated using CCP4 mg (57). | 2019-09-29T13:01:32.728Z | 2019-09-27T00:00:00.000 | {
"year": 2019,
"sha1": "37e4b2a8c3cba664c6ed41b0d173754672c877cf",
"oa_license": "CCBY",
"oa_url": "http://www.jbc.org/content/294/45/17155.full.pdf",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "f221e033fe746f9f53f647ccf7a28e6cfd729f4d",
"s2fieldsofstudy": [
"Chemistry",
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Chemistry"
]
} |
201115167 | pes2o/s2orc | v3-fos-license | MiR-106a aggravates sepsis-induced acute kidney injury by targeting THBS2 in mice model
Abstract Purpose To investigate the role and related mechanisms of miR-106a in sepsis-induced AKI. Methods Serum from sepsis and healthy patients was collected, sepsis mouse model was established by cecal ligation and puncture (CLP). TCMK-1 cells were treated with lipopolysaccharide (LPS) and transfected with THBS2-small interfering RNA (siTHBS2), miR-106a inhibitor, miR-106a mimics and their negative controls (NCs). The expression of miR-106a, thrombospondin 2 (THBS2), Bax, cleaved caspase-3 and Bcl-2, cell viability, relative caspase-3 activity and TNF-α, IL-1β, IL-6 content were respectively detected by quantitative real-time polymerase chain reaction (qRT-PCR), western blotting, Cell Counting Kit-8 (CCK-8) and enzyme linked immunosorbent assay (ELISA). The relationship between miR-106a and THBS2 was confirmed by dual luciferase reporter assay. Results MiR-106a was up-regulated in serum of sepsis patients, CLP-induced mice models and LPS-induced TCMK-1 cells. LPS reduced cell viability and Bcl-2 expression, and increased caspase-3 activity, Bax expression, the content of TNF-α, IL-1β, IL-6. THBS2 was a target of miR-106a. The decreases of caspase-3 activity, TNF-α, IL-1β, IL-6, Bax expression and the increases of cell viability, Bcl-2 expression caused by miR-106a knockdown were reversed when THBS2 silencing in LPS-stimulated TCMK-1 cells. Conclusion MiR-106a aggravated LPS-induced inflammation and apoptosis of TCMK-1 cells via regulating THBS2 expression.
by cecal ligation and puncture (CLP) and stimulated mouse kidney epithelial TCMK-1 cells with LPS in vitro to investigate the role of miR-106a in sepsis-induced AKI and its related mechanisms, as well as the possible regulatory relationship with THBS2.
■ Methods
Between March, 2018 and January, 2019, a total of 50 patients (34 males and 16 females, mean age: 48±7.21 years) with sepsis who lived in Intensive Care Unit (ICU) and 30 healthy controls (21 males and 9 females, mean age: 42±5.46 years) were recruited in this study from The Affiliated Hospital of Hangzhou Normal University. The diagnosis of sepsis complied with the American College of Chest Physicians (ACCP) and the Society of Critical Care Medicine (SCCM) joint definition of sepsis diagnostic criteria in US 17 .
The venous blood (5 mL) was collected from patients with sepsis and healthy people, and then the whole blood was preserved by anticoagulation with 1 mL of Edathamil (EDTA-K2, Solarbio, Beijing, China). Serum was obtained by centrifugation after 4 mL EDTA-K2 coagulation and stored at -20°C. All patients signed the informed consent, and the experiment was approved by the Ethics Committee of The Affiliated Hospital of Hangzhou Normal University.
The establishment of the sepsis model
All animal experiences were approved by the Laboratory Animal Management Committee of Hangzhou Normal University Laboratory Animal Center.
A total of 12 adult clean Kunming mice aged 6-8 weeks and weighing 18-22g were provided by the Laboratory Animal Center of Chongqing Medical University (Chongqing, China). All mice were raised independently at 20-25°C and 50% humidity, and then were randomly divided into two groups: the sham group and the CLP group. Animals were fasted one day before the experiment and anesthetized with 0.3% sodium pentobarbital (30 mg/kg, Solarbio, Beijing, China). In the midline of the mouse, we cut the mouth about 1cm long, found the cecum and ligatured the root, retained the blood supply, and then used the suture to puncture the ligated cecum through the serosal surface of the blind end of the intestine opposite to the mesentery. Two holes were made once to extrude a little intestinal content, close the abdominal cavity layer by layer with 1-0 silk thread, and immediately subcutaneously inject 1 mL of normal saline to supplement the intraoperative fluid loss 18. Animals in the sham group were treated in the same manner, but the cecum was not perforated and not ligated.
■ Introduction
Sepsis is a systemic inflammatory response syndrome caused by infection or highly suspicious infection, and is also a major cause of multiple organ failure and septic shock until death 1,2 . The uncontrolled inflammatory response and immune dysfunction of the body are the main pathophysiological basis of sepsis 2 . The degree of excessive inflammation and immune dysfunction directly affect the occurrence and development of sepsis 2 . Moreover, acute kidney injury (AKI) is one of the most common organ failure symptoms in sepsis 3 . The pathways of AKI induced by sepsis may include coagulopathy, inflammation, oxidative stress and changes in renal tubular epithelial cells to injury 4 . However, the mechanism of these effects is complex and there is no clear research report at present.
MicroRNAs (miRNAs) are non-coding endogenous RNAs that exert translational inhibition by targeting messenger RNA (mRNA), and have been shown to regulate various stages of sepsis [5][6][7] . A number of studies have shown that miRNAs disorders are associated with clinical symptoms of inflammation and sepsis [6][7][8] . For example, the up-regulation of miR-21-3p promotes sepsisinduced cardiac dysfunction 6 . MiR-15a/16 is up-regulated in serum of neonatal sepsis patients, and inhibits LPSinduced inflammatory signaling pathway activity 7 . MiR-128 expression decreased, while miR-21 levels increased in serum and LPS-induced podocytes of patients with sepsis, leading to podocyte damage in sepsis 8 . The above studies show that miRNAs play a crucial role in regulating immune response and sepsis. Therefore, understanding the potential mechanism of miRNAs in sepsis will help to better treat this disease. In addition, it has been reported that miR-106a expression is up-regulated in the serum of sepsis mice and inflammatory bowel disease mice model, and then the absence of miR-106a can alleviate the inflammatory response 9,10 . However, the role of miRNA-106 in sepsis-induced AKI remains unclear.
Thrombospondins (THBSs) is a protein family with a minimum of five members that participate in a variety of biological processes by combining some target proteins, such as cell motility, apoptosis, cytoskeletal formation, and can also serve as a response platform for extracellular matrices 11,12 . THBS2 is a relatively special member of THBSs family that has anti-angiogenic effects and interacts with various cellular receptors and growth factors to regulate cell proliferation, apoptosis and adhesion 13 . THBS2 has also been reported to act as important downstream targeting gene of miRNAs to form a miRNAs/THBS2 network for various diseases, such as cervical cancer, human myxoid liposarcoma and renal cancer [14][15][16] .
In this study, we collected serum from sepsis and healthy patients, constructed sepsis mouse models Cell culture and transfection TCMK-1 mouse kidney epithelial cell line was purchased from BioVector NTCC Inc. (Beijing, China) and cultured in 90% high-glucose Dulbecco's modified Eagle medium (DMEM, Solarbio, Beijing, China) and 10% fetal bovine serum (FBS, Solarbio, Beijing, China) at 37°C. After incubation for 24 h, cells were divided into two groups: one group was added with 5 mg/L of lipopolysaccharide (LPS, Solarbio, Beijing, China) to mimic inflammation, and the other group was added with equal amount of DMEM for 2 h at 37°C as control group.
HEK293 cell line was purchased from American Type Culture Collection (ATCC; Manassas, USA) and cultured in Eagle's Minimum Essential Medium (EMEM, Solarbio, Beijing, China) with 10% FBS at 37°C for dual luciferase reporter assay.
Cell viability
Cell viability was detected by Cell Counting Kit-8 (CCK-8, MedChemExpress, New Jersey, USA). Briefly, cells (100 μL/well) were seeded in 96-well plate for 24 h, added to 10 μL of CCK-8 solution and incubated in an incubator at 37°C for 3h. The absorbance at 450 nm (OD450) was detected by a microplate reader (Bio-Rad, Hercules, USA).
Enzyme linked immunosorbent assay (ELISA)
Cultured cells were centrifuged for 10 min to obtain the culture supernatant. The contents of tumor necrosis factor (TNF-α), interleukin 1 beta (IL-1β) and interleukin 6 (IL-6) were detected by ELISA Kit (Abcam, Cambridge, USA) following the instructions. The absorbance at 450 nm (OD450) was measured by a microplate reader.
Quantitative real-time polymerase chain reaction (qRT-PCR)
Cells were lysed with Biozol Reagent (Biobee, Beijing, China), extracted with chloroform and isopropanol (Sustgreen, Nanjing, China), precipitated in 75% ethanol and washed to obtain total RNA. The synthesis of cDNA was performed using SYBR primeScript TM RT-PCR Kit (TaKaRa, Dalian, China) according to the Kit instructions. The primers were designed by Primer 5.0 software (PREMIER Biosoft International, California, USA) and synthesized by TaKaRa Biotechnology Co., Ltd (Dalian, China). The specific sequences of primers were shown in Table 1. qRT-PCR was performed by SYBR Green I dye method on the PCR instrument according to the operation instructions of ExScript TM RT-PCR Kit (TaKaRa, Dalian, China). Reaction conditions of PCR were as follow: pre-denaturation 95°C for 10s, denaturation 95°C for 5s, fluorescence detection 72°C for 10s, in total of 35 cycles. The data were calculated by 2 -ΔΔCt method 19.
Statistical analysis
All data were analyzed by SPSS 19.0 software, and students' t-test was used for comparison between the two groups. The difference was significant when p<0.05, and then the results were represented by mean ± standard error (SE).
MiR-106a was up-regulated in serum of sepsis patients and CLP-induced mice models
In order to study the expression level of miR-106a, we collected serum from sepsis patients and healthy patients, established CLP-induced sepsis model in mice and treated TCMK-1 cells with LPS. The results in Figure 1 showed that miR-106a expression level in the sepsis, CLP and LPS group was higher than that in the control or sham group (p<0.001), illustrating that sepsis and sepsis-induced inflammation promoted miR-106a up-regulation in vivo and in vitro.
Knockdown of miR-106a reduced LPS-induced apoptosis in TCMK-1 cells
In order to investigate the mechanism and effect of miR-106a on sepsis-induced AKI, we used miR-106a inhibitor to transfect TCMK-1 cells treated with LPS. In Figure 2A, miR-106a expression level was significantly reduced in the miR-106a inhibitor group compared with the NC inhibitor group, indicating that miR-106a was efficiently transfected into cells (p<0.001). CCK-8 results in Figure 2B showed that cell viability was decreased in cells treated with only LPS, and this decrease was reversed after adding miR-106a (p<0.01 or p<0.001). On the contrary, caspase-3 activity in the LPS group was higher than that in the control group (p<0.001, Fig. 2C). But caspase-3 activity in the LPS+miR-106a inhibitor group was less than that in LPS+NC inhibitor (p<0.001, Fig. 2C). Moreover, western blotting results showed that the expression of Bax and cleaved caspase-3 were clearly increased and Bcl-2 expression was reduced when adding LPS; conversely, miR-106a inhibitor alleviated these increase and reduction (p<0.001, Fig. 2D). Above data indicated that miR-106a knockdown increased cell viability and decreased apoptosis-related protein expression in LPS-induced TCMK-1 cells.
Knockdown of miR-106a reduced LPS-induced inflammatory factor levels in TCMK-1 cells
In order to further investigate the effects of miR-106a on sepsis, we selected three cellular inflammatory factors TNF-α, IL-1β, IL-6 as representatives for research. The results from qRT-PCR and ELISA in Figure 3A and B displayed that TNF-α, IL-1β, IL-6 expression and content were significantly increased in the LPS group compared with the control group (p<0.05, p<0.01 or p<0.001). After adding miR-106a inhibitor to LPS-induced cells, the expression and content of TNF-α, IL-1β, IL-6 were reduced, suggesting that miR-106a knockdown mitigated the increases of TNF-α, IL-1β, IL-6 caused by LPS.
THBS2 was a target of miR-106a
In this study, we used TargetScan website (http://www.targetscan.org) to find a potential target gene THBS2 downstream of miR-106a, and their binding site was shown in Figure 4A. Then, we also used double luciferase reporter assay to identify the targeting relationship between THBS2 and miR-106a. Figure 4B displayed that miR-106a mimics were successfully transfected into cells (p<0.001). In Figure 4C, relative luciferase activity in the miR-106a mimics+THBS2-WT group was less than that in the NC mimics+THBS2-WT group (p<0.001). However, relative luciferase activity did not change in the miR-106a mimics+THBS2-MUT group. The results in Figure 4C suggested that there was a targeting relationship between THBS2 and miR-106a. Furthermore, western blotting results showed that THBS2 expression level was up-regulated in the miR-106a inhibitor group in relation to the NC inhibitor group; conversely, it was clearly decreased when miR-106a overexpression in TCMK-1 cells (p<0.01). Hence, these results illustrated that THBS2 was a target of miR-106a and negatively regulated by miR-106a.
MiR-106a increased LPS-induced inflammation and apoptosis of TCMK-1 cells via regulating THBS2 expression
In order to study how miR-106a and THBS2 together affect sepsis-induced AKI in vitro, we used a complement experiment and simultaneously transfected siTHBS2 and miR-106a inhibitor into the LPS-treated TCMK-1 cells. qRT-PCR results in Figure 5A showed that THBS2 mRNA level was reduced in the miR-106a mimics group, indicating that there was a negative regulatory relationship between miR-106 and THBS2. Then, the increase of cell viability caused by miR-106a knockdown was decreased after adding siTHBS2 (p<0.05, Fig. 5B). On the contrary, siTHBS2 alleviated miR-106a knockdown-induced the reduction of caspase-3 activity (p<0.05, Fig. 5C). Furthermore, the mRNA level and content of TNF-α, IL-1β, IL-6 in the miR-106a inhibitor+siTHBS2 group were significantly more than that in the siNC+miR-106a inhibitor (p<0.01 or p<0.001, Fig. 5 D-E). Finally, western blotting results in Figure 5F showed that the down-regulations of Bax and cleaved caspase-3, and the up-regulation of Bcl-2 were reversed by siTHBS2 (p<0.01 or p<0.001). Sum up, miR-106a decreased cell viability and increased apoptosis by regulating THBS2 expression in LPS-induced TCMK-1 cells.
■ Discussion
Sepsis is a disordered state in which an inflammatory response is out of control due to severe infection 3 . In recent years, it has been found that miRNAs can regulate the imbalanced inflammatory response by down-regulating inflammatory factors at the posttranscriptional level, suggesting that miRNAs play a very important role in inflammatory response and immune regulation of sepsis 5 . The latest findings confirm that the expression of miRNAs in serum of sepsis patients is also different from that of healthy people 6,7,20,21 . Vasilescu et al. 20 used microarray detection technology to compare the differences in miRNA expression profiles in sepsis patients with normal human serum, and then the results showed that the expressions of miR-150, miR-182 and miR-486 were significantly different in patients with sepsis. Wang et al. 21 reported that miR-146a and miR-233 expressions were reduced in serum of patients with sepsis. Moreover, Wu et al. 9 displayed that the expression of miR-106a was up-regulated in the serum of sepsis mice and inflammatory bowel disease mice model. However, our results in this study were consistent with Wu's report, that is, we found that the level of miR-106a was increased in the serum of sepsis patients, CLP-treated mice models and LPS-induced cells. These results suggested that miR-106a might be an ideal indicator for the detection and treatment of sepsis or sepsis-induced AKI. Furthermore, a research showed that the absence of miR-106a could alleviate the inflammatory response 10 . It has also been reported that miR-106a can inhibit cell proliferation and promote apoptosis in astrocytoma 22 . Hence, we detected LPS-stimulated TCMK-1 cell viability and apoptosisrelated proteins expression, and found that miR-106a knockdown decreased apoptosis and increased cell viability, illustrating that miR-106a could involve in the process of sepsis-induced AKI.
During the recurrence and development of sepsis, the release of cytokines shows cascade effect, including the pro-inflammatory factors TNF-α, IL-1β and IL-6 23,24 . Then, low levels of TNF-α, IL-1β, IL-8 and IL-6 could enhance cellular immune responses, activate defensive function and reduce the incidence of infection [24][25][26] .
TNF-α has the characteristics of early rapid synthesis and release, and its content changes can reflect the changes of other inflammatory factors 23 . In addition, IL-6 in serum is a delayed cytokine, which can be used as a marker of cytokine cascade activation, reflecting the relationship between host inflammatory response and severity of disease, and also can be used as a prognostic indicator in sepsis 27 . In this study, the results showed that knockdown of miRNA-106a reduced LPS-induced the levels of inflammatory factor TNF-α, IL-1β and IL-6, suggesting that inhibiting the expression of miRNA-106a might inhibit the occurrence of AKI induced by sepsis.
THBS2 is considered to be a powerful angiogenesis inhibitor and a regulator of tissue remodeling after trauma. In recent years, studies have shown that THBS2 plays an important role in tumors, but no studies related to sepsis have been reported [14][15][16] . However, Wei et al. 14 found that miR-221-3p could increase metastasis of cervical cancer by targeting THBS2. Then, we used TargetScan website to predict the potential target gene of miR-106a and found that there were seven binding sites between miR-106a and THBS2. Furthermore, we also confirmed THBS2 was a target gene of miR-106a and negatively regulated by miR-106a. In addition, our results showed that miR-106a increased LPSinduced inflammation and apoptosis of TCMK-1 cells via regulating THBS2 expression.
In conclusion, miR-106a aggravates sepsisinduced AKI by targeting THBS2 in mice model and cells. MiR-106a might be an ideal test and evaluation index, and further research on the downstream targeting gene THBS2 might provide a broader approach and method for the diagnosis and treatment of AKI caused by sepsis, the development of molecular targeted drugs and prognosis of sepsis patients. | 2019-08-22T13:04:19.159Z | 2019-08-19T00:00:00.000 | {
"year": 2019,
"sha1": "95129a290da8afe6b006495ac4b34620d12b6641",
"oa_license": "CCBY",
"oa_url": "http://www.scielo.br/pdf/acb/v34n6/1678-2674-acb-34-06-e201900602.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "5c39b42bfd6377f3e5a592c681c2a4c0beb7621f",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Chemistry"
]
} |
119028667 | pes2o/s2orc | v3-fos-license | Accessing Sivers gluon distribution via transverse single spin asymmetries in p(transv. polarized) p -->D X processes at RHIC
The production of D mesons in the scattering of transversely polarized protons off unpolarized protons at RHIC offers a clear opportunity to gain information on the Sivers gluon distribution function. D production at intermediate rapidity values is dominated by the elementary g g -->c cbar channel; contributions from q qbar -->c cbar s-channel become important only at very large values of x_F. In both processes there is no single spin transfer, so that the final c or cbar quarks are not polarized. Therefore, any transverse single spin asymmetry observed for D's produced in p(transv. polarized) p interactions cannot originate from the Collins fragmentation mechanism, but only from the Sivers effect in the distribution functions. In particular, any sizeable spin asymmetry measured in p(transv. polarized) p -->D X at mid-rapidity values will be a direct indication of a non zero Sivers gluon distribution function. We study the p(transv. polarized) p -->D X process including intrinsic transverse motion in the parton distribution and fragmentation functions and in the elementary dynamics, and show how results from RHIC could allow a measurement of the Sivers gluon distribution function.
I. INTRODUCTION AND FORMALISM
Within the QCD factorization scheme, the cross section for an inclusive large p T scattering process between hadrons, like p p → h + X, is calculated by convoluting the elementary partonic cross sections with the parton distribution functions (pdf's) and fragmentation functions (ff's). These objects account for the soft non-perturbative part of the scattering process, by giving the probability density of finding partons inside the hadrons (or hadrons inside fragmenting partons) carrying a specific fraction x (or z) of the parent light-cone momentum. The parton intrinsic motion -demanded by uncertainty principle and gluon emission -is usually integrated out in the high energy factorization scheme, and only partonic collinear configurations are considered. However, it is well known that the quark and gluon intrinsic transverse momenta k ⊥ have to be taken into account to improve agreement with data on unpolarized cross sections at intermediate energies [1,2]. Moreover, without intrinsic k ⊥ one would never be able to explain single spin asymmetries within QCD factorization scheme; several large single spin asymmetries have been observed [3,4,5], which in a collinear configuration are predicted to be either zero or negligibly small. Although not rigorously proven in general [6,7], the usual factorized structure of the collinear scheme has been generalized with inclusion of intrinsic k ⊥ , so that the cross section for a generic process A B → C X reads: The pdf's and the ff's are phenomenological quantities which have to be obtained -at least at some scale -from experimental observation and cannot be theoretically predicted. The pdf's of unpolarized nucleons, q( , are now remarkably well known; one measures them in inclusive deep inelastic scattering processes at some scale, and, thanks to their universality and known QCD evolution, can use them in different processes and at different energies. The k ⊥ dependence ofq(x, k ⊥ ) is usually assumed to be of a gaussian form, and the average k ⊥ value can be fixed so that it agrees with experimental data. Notice that, in our notations, a hat over a pdf or a ff signals its dependence on k ⊥ and k ⊥ = |k ⊥ |.
When considering polarized nucleons the number of pdf's involved grows and dedicated polarized experiments have to be performed in order to isolate and measure these functions. We have by now good data on the pdf's of longitudinally polarized protons -the helicity distribution ∆q(x) ≡ g q 1 (x) -but nothing is experimentally known on the transverse spin distribution -the transversity function ∆ T q(x) ≡ δq(x) ≡ h q 1 (x). The situation gets much more intricate when parton intrinsic transverse momenta are taken into account. Many more distribution and fragmentation functions arise, like the Sivers function [8,9], which describes the probability density of finding unpolarized partons inside a transversely polarized proton; similarly, the Collins fragmentation function [6] gives the number density of unpolarized hadrons emerging in the fragmentation of a transversely polarized quark. These are the functions which could explain single spin asymmetries in terms of parton dynamics [10,11].
One of the difficulties in gathering experimental information on these new spin and k ⊥ dependent pdf's and ff's is that most often two or more of them contribute to the same physical observable, making it impossible to estimate each single one separately.
In Ref. [12] it was shown how properly defined single spin asymmetries in Drell-Yan processes depend only on the Sivers distribution function ∆ N f (x, k ⊥ ) of quarks (apart from the usual known unpolarized quark distributions). In Ref. [13] it has been suggested to look at back-to-back correlations in azimuthal angles of jets produced in p ↑ p RHIC interactions in order to access the gluon Sivers function. We consider here another case which, again, would isolate the gluon Sivers effect, making it possible to reach direct independent information on ∆ N f g/p ↑ (x, k ⊥ ).
Let us consider the usual single spin asymmetry for p ↑ p → DX processes at RHIC energy, √ s = 200 GeV. These D mesons originate from c orc quarks, which at LO can be created either via a qq annihilation, qq → cc, or via a gluon fusion process, gg → cc. The elementary cross section for the fusion process includes contributions from s, t and u-channels, and turns out to be much larger than the qq → cc cross section, which receives contribution from the s-channel alone. Therefore, the gluon fusion dominates the whole p ↑ p → DX process up to x F ≃ 0.6. Beyond this the qq → cc contribution to the total cross section becomes slightly larger than the gg → cc contribution, due to the much smaller values, at large x, of the gluon pdf, as compared to the quark ones (see Fig. 1).
As the gluons cannot carry any transverse spin the elementary process gg → cc results in unpolarized final quarks. In the qq → cc process one of the initial partons (that inside the transversely polarized proton) can be polarized; however, there is no single spin transfer in this s-channel interaction so that the final c andc are again not polarized. One might invoke the possibility that also the quark inside the unpolarized proton is polarized [14], so that both initial q andq are polarized: even in this case the s-channel annihilation does not create a polarized final c orc. Consequently, the charmed quarks fragmenting into the observed D mesons cannot be polarized, and there cannot be any Collins fragmentation effect [see further comments after Eq. (9)].
Therefore, transverse single spin asymmetries in p ↑ p → DX can only be generated by the Sivers mechanism, namely a spin-k ⊥ asymmetry in the distribution of the unpolarized quarks and gluons inside the polarized proton, coupled respectively to the unpolarized interaction process qq → cc and gg → cc, and the unpolarized fragmentation function of either the c or thec quark into the final observed D meson. That is [6,15]: where q = u,ū, d,d, s,s and Q = c orc, according to whether Notice that z is the light-cone momentum fraction along the fragmenting parton direction, identified byp c , z = p + D /p + c . Throughout the paper we choose XZ as the D production plane, with the polarized proton moving along the positive Z-axis and the proton polarization ↑ along the positive Y -axis. In such a frame k ⊥a and k ⊥b have only X and Y components, while k D has all three components; the function δ(k D ·p c ) ensures that the integral over k D is only performed along the appropriate transverse direction, k ⊥D , that is the transverse momentum of the produced D with respect to the fragmenting quark direction. The factor C contains the flux and relevant Jacobian factors for the usual transformation from partonic to observed meson phase space, which, accounting for the transverse motion, reads [15,16]: Notice that for collinear and massless particles this factor reduces to the familiarŝ/πz 2 . The Sivers distribution functions [8] for quarks and gluons are defined by where a can either be a light quark or a gluon. Similarly,D D/Q (z, k ⊥D ) is the probability density for a quark Q to fragment into a D meson with light-cone momentum fraction z and intrinsic transverse momentum k ⊥D . The heavy quark mass m Q is taken into account in the amplitudes of both the partonic processes and the resulting elementary cross sections are: where τ 1,2 and ρ are dimensionless quantities defined in terms of the partonic Mandelstam variablesŝ,t andû as: The denominator of A N , Eq. (2), is analogously given by In Eqs. (3) and (9) we consider intrinsic transverse motions in the distributions of initial light quarks, in the elementary process and in the heavy quark fragmentation function, i.e. we consider a fully non planar configuration for the partonic scattering. This has two main consequences: on one side, taking into account three intrinsic transverse momenta makes both the kinematics and the dynamics highly non trivial; on the other side it generates a large number of contributions, other than the Sivers effect, which originate from all possible combinations of k ⊥ dependent distribution and fragmentation functions, each weighted by a phase factor (given by some combination of sines and cosines of the azimuthal angles of the parton and final meson momenta). This topic deserves a full treatment on its own, which will soon be presented in Refs. [17,18]. At present, we only point out that we have explicitely verified that all contributions to the p ↑ p → DX single spin asymmetry from k ⊥ dependent pdf's and ff's, aside from those of the Sivers functions, are multiplied by phase factors which make the integrals over the transverse momenta either negligibly small or identically zero. Therefore, they can be safely neglected here.
Eq. (3) shows how A N depends on the unknown Sivers distribution function; as all other functions contributing to A N are reasonably well known (including the fragmentation function D D/Q [19]) a measurement of A N should bring direct information on ∆ N f g/p ↑ and, to some extent, ∆ N f q/p ↑ .
II. NUMERICAL ESTIMATES
So far, all analyses and fits of the single spin asymmetry data were based on the assumption that the gluon Sivers function ∆ N f g/p ↑ is zero. RHIC data on A N in p ↑ p → DX will enable us to test the validity of this assumption. In fact, as the gg → cc elementary scattering largely dominates the process up to x F ≃ 0.6 (see Fig. 1), any sizeable single spin asymmetry measured in p ↑ p → DX at moderate x F 's would be the direct consequence of a non zero contribution of ∆ N f g/p ↑ . For x F ∼ > 0.6 the competing qq → cc term becomes approximately the same size as gg → cc (see Fig. 1); consequently the quark and gluon Sivers functions could contribute to A N in approximately equal measure making the data analysis more involved, as we shall discuss below.
Since we have no information about the gluon Sivers function from other experiments, we are unable to give predictions for the size of the A N one can expect to measure at RHIC. Instead, we show what asymmetry one can find in two opposite extreme scenarios: the first being the case in which the gluon Sivers function is set to zero, ∆ N f g/p ↑ (x a , k ⊥a ) = 0, and the quark Sivers function ∆ N f q/p ↑ (x a , k ⊥a ) is taken to be at its maximum allowed value at any x a ; the second given by the opposite situation, where ∆ N f q/p ↑ = 0 and ∆ N f g/p ↑ is maximized in x a .
Concerning the k ⊥ dependence of the unpolarized pdf's and the Sivers functions, we adopt, both for quarks and gluons, a most natural and simple factorized Gaussian parameterization where M = k 2 ⊥ and φ k ⊥ is the k ⊥ azimuthal angle. The extra factor 2k ⊥ M/(k 2 ⊥ + M 2 ) in the Sivers function is chosen in such a way that, while ensuring the correct small k ⊥ behaviour, it equals 1 at k ⊥ = M , being always smaller at other values. The azimuthal cos(φ k ⊥ ) dependence is the only one allowed by Lorentz invariance, via the mixed vector product P · (p ×k ⊥ ) where, with our frame choice, P = (0, 1, 0) is the proton polarization vector,p = (0, 0, 1) is the unit vector along the polarized proton motion andk ⊥ = (cos(φ k ⊥ ), sin(φ k ⊥ ), 0). The Sivers functions (11) for both quarks and gluons must respect the positivity bound which means that Eq. (12) can be satisfied for any x a and k ⊥a values by taking For the fragmentation functionD D/Q (z, k ⊥D ) we adopt a similar model, in which we assume factorization of z and k ⊥D dependencesD where D D/Q (z) is the usual fragmentation function available in the literature (see for instance Ref. [19]) and g(k ⊥D ) is a gaussian function of |k ⊥D | 2 analogous to that in Eq. (10), normalized so that, for a fragmenting quark of momentum p c , In Fig. 1(a) we show the unpolarized cross section for the process p p → DX at √ s = 200 GeV as a function of both the heavy meson energy E D and its transverse momentum p T , at fixed pseudo-rapidity η = 3.8 (notice that x F ≃ E D /(100 GeV)). In Fig. 1(b) the same total cross section is presented as a function of x F at fixed p T = 1.5 GeV/c. The x and Q 2 -dependent parton distribution functions f q/p (x, Q 2 ) are taken from MRST01 [20], while the k ⊥ dependence is fixed by Eq. (10) with k 2 ⊥ = 0.8 GeV/c [15]; similarly, the fragmentation functions D D/Q (z, Q 2 ) are from Ref. [19], with the k ⊥ dependence fixed by k 2 ⊥ = 0.8 GeV/c. We have explicitely checked that our numerical results have very little dependence on the k 2 ⊥ value of the fragmentation functions. Finally, we have taken as QCD scale Q 2 = m 2 Q . The dashed and dotted lines correspond to the qq → cc and gg → cc contributions respectively, whereas the solid line gives the full unpolarized cross section. These plots clearly show the striking dominance of the gg → cc channel over most of the E D and x F ranges covered by RHIC kinematics. Fig. 2 shows our estimates for the maximum value of the single spin asymmetry in p ↑ p → DX. The dashed line shows |A N | when the quark Sivers function is set to its maximum, i.e. ∆ N f q/p ↑ (x) = 2f q/p (x), while setting the gluon Sivers function to zero. Clearly, the quark contribution to A N is very small over most of the kinematic region, at both fixed pseudo-rapidity and varying E D , Fig. 2(a), and fixed p T and varying x F , Fig. 2(b). The dotted line corresponds to the SSA one finds in the opposite situation, when ∆ N f g/p ↑ (x) = 2f g/p (x) and ∆ N f q/p ↑ = 0: in this case the asymmetry presents a sizeable maximum in the central E D and positive x F energy region (in our configuration positive x F means D mesons produced along the polarized proton direction, i.e. the positive Z-axis). This particular shape is given by the azimuthal dependence of the numerator of A N , see Eqs. (3) and (11). When the energy E D is small, p T is also very small (for instance, for E D ≤ 23 GeV, p T ≤ 1 GeV/c) and the partonic cross sections dσ/dt depend only very weakly on φ k ⊥a . Therefore, when we integrate over φ k ⊥a the partonic cross sections multiplied by the factor cos(φ k ⊥a ) from the Sivers function, we obtain negligible values. The transverse momentum p T of the detected D meson grows with increasing E D and the partonic cross sections become more and more sensitively dependent on φ k ⊥a : then A N grows and a peak develops in correspondence of φ k ⊥a ≃ 0. Similarly, one can understand the behaviour of |A N | max in the negative and positive x F regions, Fig. 2(b). Only at very large E D and x F the qq → cc contribution becomes important, and a rigorous analysis in that region will only be possible when data from independent sources will provide enough information to be able to separate the two contributions.
By looking at Fig. 2 it is natural to conclude that any sizeable single transverse spin asymmetry measured by STAR or PHENIX experiments at RHIC in the region E D ≤ 60 GeV or −0.2 ≤ x F ≤ 0.6, would give direct information on the size and importance of the gluon Sivers function.
III. COMMENTS AND CONCLUSIONS
We have shown that the observation of the transverse single spin asymmetry A N for D mesons generated in p ↑ p scattering offers a great chance to study the Sivers distribution functions. This channel allows a direct, uncontaminated access to this function since the underlying elementary processes guarantee the absence of any polarization in the final partonic state; consequently, contributions from Collins-like terms cannot be present to influence the measurement. Moreover, the large dominance of the gg → cc process at low and intermediate x F offers a unique opportunity to measure the gluon Sivers distribution function ∆ N f g/p ↑ .
Once more, intrinsic parton motions play a crucial role and have to be properly taken into account. Adopting a simple model to parameterize the k ⊥ dependence we have given some estimates of the unpolarized cross section for D meson production, together with some upper estimate of the SSA in the two opposite scenarios in which either ∆ N f g/p ↑ is maximal and ∆ N f q/p ↑ = 0 or ∆ N f g/p ↑ = 0 and ∆ N f q/p ↑ is maximal. Our results hold for D = D + , D − , D 0 , D 0 . Both the cross section and A N could soon be measured at RHIC.
It clearly turns out that any sizeable contributions to the p ↑ p → DX single spin asymmetry at low to intermediate E D 's or x F 's would be a most direct indication of a non zero gluon Sivers function. | 2019-04-14T02:36:52.520Z | 2004-07-08T00:00:00.000 | {
"year": 2004,
"sha1": "5e031bd856726a000497d71b36fdbf7e3b68bee1",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/hep-ph/0407100",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "132c8292cbfb8df72452f137587eb2b63a18ff11",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
4900094 | pes2o/s2orc | v3-fos-license | Aortitis as a Rare Cause of Aortic Aneurysm and Valve Regurgitation: Is Repair Precluded?
Aortitis is an infrequent cause of aortic root dilatation and aortic valve regurgitation. Valve-sparing procedures have been proposed, but there is not clear evidence of which is the treatment of choice. We report the case of a 38-year-old pregnant lady with a diagnosis of idiopathic aortitis associated with aortic root aneurysm and severe aortic valve regurgitation.
Introduction
Aortitis is an infrequent cause of aortic root dilatation and aortic valve regurgitation, having been described several infectious and noninfectious etiologies for this yet not wellknown entity [1]. e approach to the "valvular problem" is controversial, having been reported increased rates of paravalvular leaks with prosthetic replacement [2]. Valve-sparing procedures have also been proposed [3], but there is not clear evidence of which is the treatment of choice.
Case Presentation
We report the case of a 38-year-old pregnant lady with a diagnosis of idiopathic aortitis associated with aortic root aneurysm and severe aortic valve regurgitation. e patient had a history of cardiac murmur, having had an echocardiogram performed two years before the episode showing a mild mitral regurgitation with a normal aortic valve and aorta. She was asymptomatic until the 4th month of pregnancy when she began with tachycardia and exertional dyspnea which rapidly evolved to rest dyspnea; she was admitted to the hospital, and an echocardiographic study was performed, showing a very severe aortic regurgitation and a 5.5 cm aneurysm of the root and ascending aorta; the left ventricle was markedly dilated with a borderline ventricular fraction. No computed tomography scan was done to avoid potential damage to the fetus.
She was transferred to our unit for urgent surgery for a planned valve-preserving surgery.
She was operated on under general anesthesia and median sternotomy. e aortic arch and the right atrium were cannulated in a standard fashion, and cardiopulmonary bypass (CPB) was established under normothermia. CPB ows were kept high maintaining a mean arterial pressure above 70 mmHg, and the use of vasoconstrictors was avoided [4,5]. e aorta was clamped and opened showing a marked thickening of its wall. e root was dissected, and the coronary ostial buttons trimmed; the tissues around the root had heavy adherences indicating an in ammatory process. Geometric height of the lea ets was measured with a ruler; it was 15 for the noncoronary lea et, 14 for the right, and 14 for the left. A valve-sparing remodeling procedure was done using a dacron straight graft of 24 mm. e technique of the remodeling procedure followed the recommendations of the Homburg group [6]. e post procedure intraoperative echocardiogram showed a residual mild regurgitation of the valve, which was considered acceptable. e jet was central, in its origin and direction, and the e ective height of the repaired cusps was 8 mm.
e fetal status was assessed before surgery and after the completion of the procedure by transabdominal echography, showing an alive fetus. e postoperative course of the patient was uneventful. She was discharged on metoprolol and iron. A transthoracic echocardiogram was performed before discharge showing a mild regurgitation of the aortic valve with a marked unloading of the left ventricle.
She continued her pregnancy, and a healthy girl was born with a vaginal delivery.
She was investigated thoroughly, and all known causes of aortitis or genetic disorders were discarded. No other signs or symptoms of the disease were found. She was diagnosed with idiopathic aortitis at that moment. Anyhow, she was started with immunosuppressant treatment with corticoids.
She underwent clinical and serial echocardiographic follow-up. e 6-month echo showed a progression of the regurgitation which was moderate at that time, and in the 1year echo, it became severe. As she continued to be asymptomatic, a narrower follow-up was decided, but 3 months later, the left ventricle had dilated markedly, and redo surgery was indicated.
Redo surgery was performed through a resternotomy with central cannulation. e Dacron graft was opened through a transverse incision, the aortic valve was measured and resected, and a 23 mm mechanical prosthesis was inserted in a supra-annular position with pledgeted sutures. e lea ets were markedly retracted at that time, with a geometric height of less than 4 mm. e Dacron graft was then sutured, and the patient was weaned from CPB and transferred to the Intensive Care Unit for an uneventful recovery.
Two years later, a new-onset supra-aortic vessel narrowing has been detected, and a diagnosis of Takayasu disease has been made.
Discussion
Aortitis is a rare cause of aortic aneurysms and aortic valve regurgitation.
Aortitis of the ascending aorta is a rare condition that can result in aneurysm and valve dysfunction. Although valve-preserving surgery has been successfully accomplished by others [3], in our opinion, these patients might probably be best served by valve and root replacement.
In our patient, the aorta wall was markedly thickened and the geometric height of the lea ets was low which should have led us to replace the valve instead of performing a valvepreserving surgery as formerly showed by the evolution of the patient. Geometric height is a major prognostic factor in valve repair with heights under 16 mm in tricuspid valves or 19 mm in bicuspid valves considered as restricted, being advisable to replace the valve in those instances [7]. e mechanism because of which the valve was regurgitant was not only the dilation of the aorta but also, and probably most importantly, the retraction of the tissue of the lea ets which was responsible for the recurrence of the aortic insu ciency.
Considering the pregnancy of the patient, we tried to avoid a prosthetic valve substitute and preserved the valve despite its low geometric height. She could continue her pregnancy until term without any damage to the newborn. Anyway, given the evolution of the case, probably she would have been better served with a biological root replacement.
Follow-up of repaired patients is mandatory, with 6month and yearly echocardiographic assessment highly advisable in our opinion, to discard recurrence of the regurgitation, which happened in our patient, who despite being asymptomatic had progressive dilatation of the left ventricle.
Conclusions
Aortic valve repair is probably not the best option for patients with aortitis.
Close follow-up of these patients is mandatory. | 2018-04-27T03:28:06.298Z | 2018-02-13T00:00:00.000 | {
"year": 2018,
"sha1": "81bb4a68f2413939b6e0bd3e8eb14fcde69db329",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/cric/2018/5757081.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "ea584bba51b3d5044936eab3f4df8876764c0fd1",
"s2fieldsofstudy": [
"Medicine",
"Engineering"
],
"extfieldsofstudy": [
"Medicine"
]
} |
56680111 | pes2o/s2orc | v3-fos-license | Implementation of The LDA Algorithm for Online Validation Based on Face Recognition
This paper report work in implementation of computer vision application in face recognition to the on-line validation for distance learning. Face recognition is chosen among many other alternatives of validation because its robustness. The problem with basic validation such as password cannot validate the student in distance learning. This cannot be accepted especially is distance examination. Face recognition algorithm used in this research is Linear Discriminant Analysis (LDA). By using this algorithm, the system capable of recognize the authorized persons about 93% and reject the unauthorized persons 100%.
Introduction
In some of last decades, the use of biometric security system was widely discussed in the world of computer vision and pattern recognition research. Face recognition is one of the biometric technology has now been applied for many applications in the fields of security, such as Access security system, Authentication system, and as a tool in tracking criminals [1]. High security system is one of the reasons the biometric continue to be developed, where the workings of the biometrics cannot be imitated and transferred to the second party. Some government agencies now select system security by using face recognition biometric.
Distance educations require authentication systems in order to find out the validity of their participants. Distance education itself is a system of education where teaching and learning among teachers and learners are not limited by space and time [2]. Some contents of e-learning and distance learning, such as [3][4], should be limited to the only member of the distance learning group. Authentication using face recognition can be a solution to overcome the anxiety of the Government in implementing a system of distance education.
Face recognition system offers some algorithms in the process of face identification. Linear projection method is one of the most commonly used algorithm in the application of face recognition system [5]. The process of this method is by projecting the image pixels into a matrix. Then, matrix is converted to a single-dimensional as vector data to obtain the eigen value and covariance value. This method is applied on the two Principal Component Analysis algorithms (PCA) and Linear Discriminant Analysis (LDA). The purpose of PCA and LDA methods solely to reduce the dimensions of high dimensional space into a low-dimensional space.
In this research, the education is based on distance online system. The proposed security system used the student ID number as user and face recognition as password to access the system. The algorithm used in this research is the LDA. Face recognition is a complex problem in the verification process and identification of a face through the digital images or video frames [6]. The process of face recognition in a digital image or video frame is not an easy job. There are some challenges in the process of face identification on an image as follows [7]: intensity, pose, face rotation, and occlusion.
Face recognition and Algorithms
To overcome difficulties at the identification process of face image, it should take the right algorithm. In general, there are two techniques that are proposed, namely the approach of local features and holistic approach [8]. But lately, the development of new technique in the form of hybrid approach has already been done, i.e. a merger between engineering approach of holistic and local features. Figure 1 shows the data reduction process of the PCA.
Linear Discriminant Analysis
Like PCA, LDA is also an algorithm which aims to reduce dimensions. LDA finds an optimal projection so that it can project the input data on space with smaller dimensions where all patterns can be separately maximized [11] [12]. Separation of the LDA will maximize the dissemination of data-input between different classes and also minimizing the spread of input in the same class. The difference between the class represented by the matrix of the Sb and the difference in class is represented by the matrix Sw [13]. Weight matrix Transpose PCA is the input for the LDA, this shows clearly that most of the algorithm LDA must run view steps through the PCA algorithm. The following algorithm LDA [10]: • Vector data input from transpose weight PCA become input to the LDA: • Calculates the spread between classes (Sb) and grade (S):
The proposed Research Method
This research used a holistic approach with LDA algorithm. The system is tested using 10 dataset classes. Each class consists of 8 image training. Following are the steps in the process of face detection: • Enrollment At this point students enrolled in Distance Education website. This image will be stored in MySQL database as the training image. Figure 3 shows one of the training picture stored in Mysql.
• Training Image
This stage uses LDA algorithm to train images stored in database after previous process. Figure 6 shows the results of LDA training.
Results and Discussion
This system was built using two different programming languages, on the side of the client/participant using the PHP programming language and on the server side using Java. System testing is performed with 10 participants, and 8 pictures for each of them. Each participant performs login tests to the system as much as 10 times to find out how it performs. Table 1 shows the results of the testing participants who have already registered in the system. The results show that the system can work well when a face is unregistered. Average percentage indicates the level of accuracy as much as 93%. Error system shows the percentage of 7% at some login tests. e. User 5 10 9 1 90% 10% 6 User 6 10 9 1 90% 10% 7 User 7 10 10 0 100% 0% 8 User 8 10 10 0 100% 0% 9 User 9 10 9 1 90% 10% 10 User 10 10 9 1 90% 10% Average 93% 7% In addition to conducting trials on the participants who have registered in the database, testing is also done on a user who is not listed in the database. Table 2 shows the percentage of 100%, where the face who tries to login and not listed on the database cannot access the system. User 4 10 0 10 100% 0% 5 User 5 10 0 10 100% 0% 6 User 6 10 0 10 100% 0% 7 User 7 10 0 10 100% 0% 8 User 8 10 0 10 100% 0% 9 User 9 10 0 10 100% 0% 10 User 10 10 0 10 100% 0% Average 100% 0%
Conclusion
From the results of this research it can be concluded that the online system could run on face authentication system using LDA algorithm. Some failure in the system are results of this holistic approach is still caused by several factors such as how big a difference the intensity of light received by each of the pictures, poses, and other additional attributes. | 2019-01-23T00:39:39.116Z | 2017-01-01T00:00:00.000 | {
"year": 2017,
"sha1": "58e89c92b195123cdd0dc175d61572c7f51c7d58",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1742-6596/801/1/012047",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "6eb54f9c250d7d2c8b59ad75debaf670b37f01be",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Physics",
"Computer Science"
]
} |
204009112 | pes2o/s2orc | v3-fos-license | Causality and deceit: Do androids watch action movies?
We seek causes through science, religion, and in everyday life. We get excited when a big rock causes a big splash, and we get scared when it tumbles without a cause. But our causal cognition is usually biased. The 'why' is influenced by the 'who'. It is influenced by the 'self', and by 'others'. We share rituals, we watch action movies, and we influence each other to believe in the same causes. Human mind is packed with subjectivity because shared cognitive biases bring us together. But they also make us vulnerable. An artificial mind is deemed to be more objective than the human mind. After many years of science-fiction fantasies about even-minded androids, they are now sold as personal or expert assistants, as brand advocates, as policy or candidate supporters, as network influencers. Artificial agents have been stunningly successful in disseminating artificial causal beliefs among humans. As malicious artificial agents continue to manipulate human cognitive biases, and deceive human communities into ostensive but expansive causal illusions, the hope for defending us has been vested into developing benevolent artificial agents, tasked with preventing and mitigating cognitive distortions inflicted upon us by their malicious cousins. Can the distortions of human causal cognition be corrected on a more solid foundation of artificial causal cognition? In the present paper, we study a simple model of causal cognition, viewed as a quest for causal models. We show that, under very mild and hard to avoid assumptions, there are always self-confirming causal models, which perpetrate self-deception, and seem to preclude a royal road to objectivity.
1 Introduction: Causal cognition and its vulnerabilities 1
.1 Causal cognition in life and science
Causal cognition drives the interactions of organisms and organizations with their environments at many levels. At the level of perception, spatial and temporal correlations are often interpreted as causations, and directly used for predictions [27]. At the level of science, testing causal hypotheses is an important part of the practice of experiment design [16], although the general concept of causation seldom explicitly occurs in scientific theories [41]. It does occur, of course, quite explicitly in metaphysics and in theory of science [7], and most recently in AI 1 [36,46].
Causal cognition and launching effects
Causal theories may or may not permeate science (depending on how you look at it) but they certainly permeate life. Why did the ball bounce? Why did my program crash? Why did the boat explode? Who caused the accident? Why did the chicken cross the street? We need to know. We seek to debug programs, to prevent accidents, and to understand chicken behaviors. We often refuse to move on without an explanation. Lions move on, mice run away, but humans want to understand, to predict, to control.
In some cases, the path to a causal hypothesis and the method to test it are straightforward. When a program crashes, the programmer follows its execution path. The cause of the crash must be on it. The causal hypothesis is tested by eliminating the cause and verifying that the effect has been eliminated as well. If a ball bounces like in Newton's cradle, in Fig. 1 on the left, we see that the ball was launched by another ball. If a boat explodes like in Fig. 1 on the right, we see that the explosion was caused by another boat crashing. Action movies are chains of causes and effects, packed between a problem and a solution, which are usually presented as a big cause and a big effect.
In other cases, establishing a causal hypothesis may be hard. It is obvious that the collision caused the explosion; but who caused the collision? It is obvious that the bouncing ball extends the trajectory of the hitting ball; but how is the force transferred from the hitting ball to the bouncing ball through all the balls in-between, that don't budge at all? Such questions drive our quest for causes, and our fascination with illusions. They drive us into science, and into magic. In a physics lab, a physics teacher would explain the physical 1 AI is usually interpreted as the acronym of Artificial Intelligence. But the discipline given that name in 1956 by John McCarthy evolved in the meantime in many directions, some of which are hard to relate with any of the usual meanings of the word "intelligence". We are thus forced to either keep updating the concept of "intelligence" to match the artificial versions, or to unlink the term "AI" from its etymology, and to use it as a word rather than as an acronym, like it happened with the words gif, captcha, gulag, or snafu. The latter choice seems preferable, at least in the present context. law behind Newton's cradle. But in a magic lab, a magic teacher would present a magic cradle, looking exactly the same like Newton's cradle, but behaving differently. One time the hitting ball hits, and the bouncing ball bounces. Another time the hitting ball hits, but the bouncing ball does not bounce. But later, unexpectedly, the bouncing ball bounces all on its own, without the hitting ball hitting. Still later, the bouncing ball bounces again, and the hitting ball hits after that. Magic! Gradually you understand that the magic cradle is controlled by the magic teacher: he makes the balls bounce or stick at his will.
Michotte studied such "launching effects" in his seminal book [27]. The action movie industry is built upon a wide range of techniques for producing such effects. While the example in Fig. 1 on the right requires pyrotechnic engineering, the example in Fig. 2 on the right is close to Michotte's lab setups. Observing subtle details, we can usually tell apart a real effect Figure 2: The cause of an effect may appear obvious, but the appearance may deceive from an illusion. But illusions are often more exciting, or less costly, and we accept them. We enjoy movies, magic, superstition. We abhor randomness and complexity, and seek the simplest causal explanations. We like to be deceived, and we deceive ourselves.
Causal cognition as a security problem
Cognitive bias and belief perseverance are familiar and well-studied properties of human cognition, promoted in evolution because they strengthen the group cohesion [21]. As human interactions increasingly spread through media, human beliefs are hijacked, their biases are manipulated, and social cohesion is amplified and abused at ever larger scales. This is also well-known and well-studied in psychology [8,Part VI] and in social sciences [49]. However, on the information technology side, cognitive bias manipulations on the web market and in cyberspace have been so profitable in practice, that there was no time for theory.
We live in times of advertising and publicity campaigns. Brands and crowds are built and steered using the same tools, and from the same materials: human cognition and sensitivities. The AI tools, recently added to the social engineering toolkit, appear to be among the most effective ones, and among the least understood. The practice is far ahead of theory. The research reported in this paper is a part of a broader effort [9,31,33,34,47] towards developing models, methods, and tools for this area, where human and computational behaviors are not just inseparable, but they determine each other; and they do not just determine each other, but they seek to control each other.
The specific motivating question that we pursue is: Is artificial causal cognition suitable as a tool to defend human causal cognition from the rapid strategic advances in cognitive bias manipulation, influencing, and deceit?
There are, of course, many aspects of this question, and many approaches to it. The combined backgrounds of this paper's authors provide equipment only for scaling the south-west ridge: where the west cliff of computation meets the south cliff of communication. Neither of us is equipped to place or interpret our model and results in the context of the extant research in psychology, or social sciences, where they should be empirically validated. We are keen to present it to a mixed audience hoping that there are gates between the silos. Rather than attempt to shed light on the problem from a direction that is unfamiliar to us, we are hoping to shed light on the problem from a direction that is unfamiliar to the reader, while doing our honest best to avoid throwing any avoidable shadows. If the reader sheds some light from their direction, this multi-dimensional problem may emerge in its multi-disciplinary space.
Background: Causality theories and models
Causality is studied in different communities from different standpoints. The following high level overview is mainly intended to put us at a starting position. The reader is welcome to fast-forward at any point.
A very brief history of causality
Causal relations impose order on the events in the world: an event a causes an event b, which causes an event c, whereas the event d is unrelated to a or b, but may contribute to c. An event may have multiple causes, and multiple effects, and they may all be distant from one another in time and in space. We impose order on the world by thinking in terms of causes and effects.
But connecting causes and effects also causes difficulties. In Fig. 1, the force on one end of Newton's cradle causes the effect on the other end without affecting anything in-between; whereas the boats collide and are subsequently engulfed in the explosion, but the explosion is not caused by the collision, but staged. Tellingly, such decouplings are called special effects. Our eye distrusts the physical effect on the left, and accepts the special effect on the right. Recognizing such remote causations and adjacent non-causations requires learning about the unobservable causes of observable effects that fill the world, and about the unobservable effects of observable causes, and even about the unobservable causes of unobservable effects. Imposing the causal order on the world forces us to think about the unobservable, whether they are black holes or spirits of our ancestors.
While acting upon some causes to prevent some effects is a matter of individual adaptations and survival for humans and androids, plants and animals, understanding the general process of causation has been a matter of civilization and cognition. While the efforts in the latter direction surely go back to the dawn of mankind, the early accounts go back to pre-Socratic philosophy, and have been traditionally subsumed under Metaphysics, which was the (post-Socratic) title of Aristotle's manuscript that came after (i.e. it was the "meta-" to) his Physics [1]. We mention just three paradigmatic steps towards the concept of causation: i) Parmenides: "Nothing comes from nothing." ii) Heraclitus: "Everything has a cause, nothing is its own cause." iii) Aristotle: "Everything comes from an Uncaused Cause." Step (i) thus introduces the principle of causation; step (ii) precludes causal cycles, and thus introduces the problem of infinite regression through causes of causes; and step (iii) resolves this problem by the argument 2 that in Christian theology came to be interpreted as the cosmological proof of existence and uniqueness of God. Just like the quest for causes of particular phenomena leads to magic and superstition, the quest for a general Uncaused Cause leads to monotheistic religion. This logical pattern persists in modern cosmology, where the Uncaused Cause arises as the initial gravitational singularity, known as the Big Bang. Although it is now described mathematically, it still spawns untestable theories. The singularity can be avoided using mathematical devices, inflating and smoothening the Uncaused Cause into a field; but preventing its untestable implications requires logical devices.
While modern views of a global cause of the world may still appear somewhat antique, the modern theories of local causations evolved with science. The seed of the modern idea that causes can be separated and made testable through intervention [18,45,48] was sown by Galileo: "That and no other is to be called cause, at the presence of which the effect always follows, and at whose removal the effect disappears" (cited in [7]). As science progressed, philosophers echoed the same idea in their own words, e.g., "It is necessary to our using the word cause that we should believe not only that the antecedent always has been followed by the consequent, but that as long as the present constitution of things endures, it always will be so" [28].
But David Hume remained unconvinced: "When I see, for instance, a billiard-ball moving in a straight line towards another; even suppose motion in the second ball should by accident be suggested to me, as the result of their contact or impulse; may I not conceive, that a hundred different events might as well follow from the cause? [. . . ] The mind can never possibly find the effect in the supposed cause, by the most accurate scrutiny and examination. For the effect is different from the cause, and can never be discovered in it." [22, 4.10] Hume's objections famously shook Immanuel Kant from his metaphysical "dogmatic slumber", and they led him on the path of "transcendental deduction" of effects from causes, based on synthetic a priori judgements, that are to be found mostly in mathematics [24]. Scientists, however never managed to prove any of their causal hypotheses by pure reason, but took over the world by disproving their hypotheses in lab experiments, whereupon they would always proceed to formulate better causal hypotheses, which they would then try to disprove again, and so on. The imperative of falsifiable hypotheses, and the idea of progress by disproving them empirically, suggested the end of metaphysics, and of causality, and it prompted Bertrand Russell to announce in 1912: "We shall have to give up hope of finding causal laws such as Mill contemplated, as any causal sequence which we have observed may at any moment be falsified without a falsification of any laws of the kind that the actual sciences aim at establishing. [. . . ] All philosophers of every school imagine that causation is one of the fundamental axioms or postulates of science, yet, oddly enough, in advanced science such as [general relativity], the word 'cause' never occurs. [. . . ] The law of causality, I believe, like much that passes muster among philosophers, is a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm." [41]
A very high-level view of causal models
In spite of the logical difficulties, the concept of causation persisted. General relativity resolved the problem of action at distance, that hampered Newtonian theory of gravitation, by replacing causal interactions by fields; but the causal cones continued to mushroom through the spacetime of special relativity, even along the closed timelike curves. In quantum mechanics, causality was one of the main threads in the discussions between Einstein and Bohr [5]. It was also the stepping stone into Bohm's quantum dialectical materialism [2,3]. Last but not least, causality remains central in the modern axiomatizations of quantum mechanics [10,11], albeit in a weak form. Nevertheless, even the weakest concept of causality implies that the observables must be objective, in the sense that the outcomes of measurements should not depend subjective choices, which precludes covert signaling.
The most influential modern theory of causation arose from Judea Pearl's critique of subjective probability, which led him to Bayesian networks [35,37]. As a mathematical structure, a Bayesian net can be thought of as an extension of a Markov chain, where each state is assigned a random variable, and each transition represents a stochastic dependency. Extending Markov chains in this way, of course, induces a completely different interpretation, which is stable only if there are no cycles of state transitions. Hence the causal ordering.
Originally construed as an AI model, Bayesian networks turned out to be a useful analysis tool in many sciences, from psychology [19] to genetics and oncology [42]. Broad overviews of the theory and of the immense range of its applications can be found in [25,36,46]. Very readable popular accounts are [38,39], and [44] is written from the standpoint of psychology. The causal models imposing the acyclicity requirement on the dependencies between random variables opened opened an alley towards an algorithmic approach to causal reasoning, based on dependency discovery: given some input random variables and some output random variables, specify a network of unobservable random variables whereby the given inputs can cause the given outputs. A good introduction into such discovery algorithms is [46]. The applications have had a tremendous impact.
A skeptic could, or course, argue that inserting in-between a cause and an effect a causal diagram is just as justified as inserting Kant's synthetic a priori judgements. A control theorist could also argue that the canonical language for specifying dependencies between random variables was provided by stochastic calculus a long time ago. They were not presented as elegantly as in bayesian nets, but they were calculated and graphed. On the other hand, the dependencies modeled in control theory are not required to be acyclic. This is, of course, essential, since the acyclicity requirement precludes feedback. Is feedback acausal? A simple physical system like centrifugal governor 3 obviously obeys the same laws of causation like its feedback-free cousins: an increase in angular momentum causes the valve lever to rise; a wider opening lets more steam out and decreases the pressure; a lower pressure causes a decrease in angular momentum. The only quirk is thus that some physical effects thus cause physical effects on their causes, in continuous time. Whether that is causation or not depends on the modeling framework.
Be it as it may, our very short story about the concepts and the models of causality ends here. The upshot is that there are effective languages for explaining causation. The availability and effectiveness of such languages is the central assumption on which the rest of the paper is based. Such languages, effective enough to explain everything, have been provided by metaphysics, and by transcendental deduction, and by bayesian networks, and by stochastic processes. There are other such languages, and there will be more. The rest of our story proceeds the same for any of them. Henceforth, we zoom out, hide the "innards" of causal models, and study how they are composed and decomposed. Towards this goal, causal processes are presented as string diagrams, like in Fig. 3. String diagrams consist of just two graphic components: • strings -representing types, and • boxes -representing causal processes.
Formally, types and processes are the basic building blocks of a causal modeling language: e.g., they may correspond to random variables and stochastic processes. Informally, a type can be thought of as a collection of events. In a causal process, enclosed in a box, the events of the input type, corresponding to the string that enters the box at the bottom, cause the events of the output type, corresponding to the string that exits the box at the top. The time just flows upwards in our string diagrams. We call the types consumed and produced by a causal process inputs and outputs (and not causes and effects) to avoid the ambiguities arising from the fact that the events produced by one process as effects may be consumed by another process as causes. The input and the output type may be the same. The diagram in Fig. 3 on the left is hoped to convey an idea what is what. Presenting causal processes as boxes allows us to abstract away the details when we need a high-level view, and also to refine the view as needed, by opening some boxes and displaying more details. This is similar to the mechanism of virtual functions in programming: we specify the input and the output types, but postpone the implementation details. A refined view of a causal process in an open box may be composed of other boxes connected by strings. An example 4 is in Fig. 3 on the right. The types of the strings coming in and out show that grades and reference letters, as events produced as effects of a causal process course, are causally determined by students' work and class activities, as well as by the hardness of the course itself. All these causal factors are viewed of events of suitable types. When we open the course box, we see that this causal process is composed from two simpler causal processes: one is exam and homework, the other reference request. Each of them inputs two causal factors at the bottom; exam and homework outputs two effects, whereas reference request outputs one. Each grade, as an event of type Grade Report, is caused both by the course hardness and by the student work ; whereas the reference letters are caused (influenced, determined. . . ) by the class activity and by the course record, which is again an effect of the student work and of the course hardness in the process of exam and homework. The causal dependencies between the random variables corresponding to the types Grade Report and Reference Letter, and their dependencies on the random variables corresponding to Course Hardness, Student Work, and Class Activity, are thus displayed in the string diagram inside the course box. For a still more detailed view of causal relations, we could zoom in further, and open the boxes exam and homework and reference request, to display the dependencies through some other random variables, influenced by some other causal processes.
The causes corresponding to the input types are assumed to be independent on each other. More precisely, the random variables, corresponding to the types Course Hardness, Student Work, and Class Activity, are assumed to be statistically independent. On the other hand, the causal dependencies of random variables corresponding to Grade Report, Record, and Reference Letter are displayed in the diagram. E.g., the content of a Reference Letter is not directly caused by Course Hardness, but it indirectly depends on it, since the student performance in the Record depends on it, and the Record is taken into account in the Reference Letter.
The abstract view of the causal process box match on the left could be refined in a similar way. The causes are boxers' actions, the effects are boxers' states. The images display a particular cause, and a particular effect of these types. The direct cause of the particular effect that is displayed on top is a particular blow, that is only visible inside the box. The cause at the bottom is the boxer's decision to deliver that particular blow. The causal process transforming this causal decision into its effect can be refined all the way to distinguishing good boxing from bad boxing, and to analyzing the causes of winning or losing.
Spelling out a process corresponding to the box match on the right in Fig. 2 might be even more amusing. While the strings outside the box would be the same, the causal dependencies inside the box would be different, as boxers' states are not caused by their blows, but feigned; and the blows are not caused by boxers' decisions, but by the movie director's requests.
A category of causal processes
String diagrams described in the previous section provide a graphic interface for any of causal modeling languages, irrespective of their details, displaying only how causal processes are composed. It is often convenient to arrange compositional structures into categories [26]. The categorical structures capturing causations turn out to yield to string diagrams [13,12,23]. For the most popular models, as mentioned above, the strings correspond to random variables, the boxes to parametrized stochastic processes. The strings are thus the objects, the boxes the morphisms of a monoidal category. When no further constraints are imposed, this category can be viewed as the coslice, taken from the one-point space of the category of free algebras for a suitable stochastic monad. For a categorical account of full stochastic calculus, the monad can be taken over measurable spaces [15,17]. For a simplified account the essential structures, the convexity monad over sets may suffice [23]. For our purpose of capturing causal explanations and taking them into account as causal factors, we must deviate from the extant frameworks [13,15,17,23] 5 in two ways. One is minor, and well tried out: the randomness is captured by subprobability measures, like in [29], to account for nontermination. The other is that causal processes must be taken up to stochastic indistinguishability, for the reasons explained in [46,Ch. 4].
Notation. In the rest of the paper, we fix an abstract universe C of causal processes, which happens to be a strict monoidal category [ The compositional structure is briefly described in the Appendix.
4 Modeling: Causal models as causal factors
Modeling causal models
The experimenter has been recognized as a causal factor in experiments since the onset of science. Already Galileo's concept of causality, mentioned in Sec. 2, requires that the experimenter manipulates a cause in order to establish how the effect depends on it. In modern theories of causation, this idea is refined to the concept of intervention [20]. In quantum mechanics, the effects of measurements depend not only on the causal freedom of the measured particles, but also on experimenters' own causal freedom [4,14]. As mentioned in Sec. 1, similar questions arise even in magic: if the magician manipulates the effects, what causes magician's manipulations?
But if the magician and the experimenter are causal processes themselves, then they also occur, as boxes among boxes, in the universe C of causal processes. And if the experimenter's causal hypotheses are caused by some causal factors themselves, then the universe C contains, among other types, a distinguished type of causal models Ω, where the experimenter outputs his hypotheses. For instance, Ω can be the type of our string diagram models, like in Fig. 3. Or Ω can be the type of (suitably restricted) stochastic processes; or of Bayesian nets, including the running example [25, Fig. 3.3], on which Fig. 3 is based. 6 There are many different languages for describing different causations, and thus many different ways to think of the type Ω. Psychologist's descriptions of causal cognition differ from physicist's view of the causal process of experimentation. What is their common denominator? What distinguishes Ω from other types?
In the Sec. 4.3, we attempt to answer this question by specifying a structure, using three simple axioms, which should be carried by the type Ω in all frameworks for causal modeling. The other way around, we take this structure to mean that a framework where it occurs describes causal modeling; that the blind men talking about different parts of an elephant are talking about the same animal. This is where the assumption stated at the end of Sec. 2 is used.
But stating and using the axioms requires two basic syntactic features, which are explained next.
Parametrizing and steering models and processes
Suppose that we want to explore how learning environments influence the causal process course from Fig. 3 In the first case, each causal process course(school) consumes the causal inputs in the form course(school)(hardness, work, activity); whereas in the second case, all inputs are consumed together, in the form course(school, hardness, work, activity). The difference between the parameter school and the general causal factors hardness, work, activity is that the parameter is deterministic, whereas the general causal factors influence processes with some uncertainty. This means that entering the same parameter school always produces the same causal process, whereas the same causal factors hardness, work, activity may have different effects in different samples.
The convenience of internalizing the indices and writing indexed families as parametric processes is that steering processes can be internalized as reparametrizing. E.g. given a function ς : Teachers −→ Schools, mapping each teacher the the school where they work, induces the reindexing of families Deterministic functional dependencies are characterized in string diagrams in Appendix D.
Axioms of causal cognition
Every framework for causal modeling must satisfy the followin axioms: I : Every causal model models a unique causal process (over the same parameters). II : Every causal process has a model (not necessarily unique). III : Models are preserved under steering.
Axioms formally
We view a causal model as a family P (y) ∈ Ω, indexed by y ∈ Y , i.e. as a parametrized model Y P − − → Ω. The notation introduced in Sec. 3.2 collects all Y -parametrized causal models in the set of processes C(Y, Ω). The parameter type Y is arbitrary. On the other hand, all Y -parametrized causal processes Y ⊗ A p − → B, where the events of type A cause events of type B, are collected in the set C (Y ⊗ A, B).
Axiom I postulates that every parametrized causal model Y P − → Ω induces a unique causal process Y ⊗ A P − − → B with the same parameters Y . A causal modeling framework is thus given by a family of the prediction maps indexed over all types Y, A and B.
Axiom II says that the prediction maps − are surjective: for every causal process Y ⊗A there is a causal model Y P − → Ω that predicts its behavior, in the sense that the process P is indistinguishable from p, which we write P ≈ p. The indistinguishability relation ≈ is explained in the next section.
Axiom III says that for any function ς : X −→ Y and any causal model Y P − → Ω reparametrizing P along ς models steering the modeled process P along it, in the sense that (2)
Axioms informally
Axiom I is a soundness requirement: it says that every causal model models some causal process. If causal processes are viewed as observations, the axiom thus says that for any causal model, we will recognize the modeled process if we observe it.
Axiom II is a completeness requirement: it says that for every causal process that we may observe, there is a causal model that models it. Can we really model everything that we observe? Yes, but our models are valid, i.e. their predictions are consistent with the behavior of the modeled processes only in so far as our current observations go. Given a process p, we can always find a model P whose predictions P summarize our observations of p, so that P and p are for us indistinguishable, i.e. P ≈ p. The less we observe, the less we distinguish, the easier we model.
Indistinguishability relations, be it statistical, computational, or observational, are central in experimental design, in theory of computation, and in modern theories of causation [46,Ch. 4]. The problem of distinguishing between observed processes has been tackled since the early days of statistics by significance testing, and since the early days of computation by various semantical and testing equivalences. Axiom II says that, up to an indistinguishability relation, any observed causal process has a model. It does not say anything about the hardness of modeling. This problem is tackled in research towards cause discovery algorithms [46].
Axiom III is a coherence requirement: it says that steering a process Y ⊗ A p − → B along a deterministic function ς : X −→ Y does not change the causation, in the sense of (2) or
Universal testing
Since any causal modeling language can thus be viewed as a type Ω, living in the process universe C, the family of all causal models ω ∈ Ω, trivially indexed over itself, can be represented by the identity function Ω Figure 4: There is an explanation for every causation mutandis, for an arbitrary process p and a model P assured by axioms I and II, axiom II thus implies p ≈ Id • (P ⊗ A), as displayed in Fig. 4.
Any causal universe C, as soon as it satisfies axioms I-III, thus contains, for any pair A, B, a universal testing process Ω ⊗ A Id − − → B, which inputs causal models and tests their predictions, in the sense that X ⊗A Universal testing is thus a causal process where the predictions of causal models are derived as their effects. It can thus be construed as a very high level view of scientific practice; perhaps also as an aspect of cognition.
Partial modeling
If a causal process has multiple causal factors, then they can be modeled separately be treating some of them as model parameters. E.g. a process in the form Y ⊗ X ⊗ A r − → B can be viewed as a Y -parametrized process with causal factors of type X ⊗ A, or as a Y ⊗ X-parametrized process with causal factors of type A. The two different instances of (1), both surjections by axiom II, would lead to models Y A Ω X Ξ universal testing Id X⊗A Figure 5: Causes of type X are treated as parameters In particular, taking r to be the universal process Ω ⊗ X ⊗ A Id − − → B, and interpreting it as an Ω ⊗ X-parametrized process, leads to an Ω ⊗ X-parametrized model Ξ, such that Id X⊗A ≈ Id A • (Ξ ⊗ A), as displayed in Fig. 5.
Slicing models
Using universal testing processes and partial modeling, causal processes can be modeled incrementally, factor by factor, like in Fig. 6. testing testing partial steered model Figure 6: Separating independent causes allows incremental modeling 5 Construction: Self-confirming causal models In everyday life, we often defend our views by steering ourselves to the standpoints where they seem valid. This assures belief perseverance, but also leads to cognitive bias and selfdeception. In science, experimenters obviously influence the outcomes of their experiments by choosing what to measure. In quantum mechanics, this influence propagates to processes themselves, as the choice of the measurement basis determines the choice of the states to which the measured process may collapse [5,14]. Causal models are thus always among the causal factors of the modeled quantum processes 7 . Formally, a quantum process is thus always in the form Ω ⊗ A q − → B, parametrized by its models in Ω.
Many processes outside the realm of quantum are in this form as well, as causal cognition curls into itself by influencing causation. The placebo and the nocebo effects are ubiquitous: patient's belief in the effectiveness of a medication contributes to its effectiveness, whereas negative beliefs often cause negative effects. In Shakespeare's tragedy, Macbeth is driven to murder the King by the prophecy that he would murder the King. In the early days of Facebook, attracting new members required convincing them that many of their friends were already members. This initially had to be a lie, but many believed it, joined, and it ceased to be a lie.
Other causal models impact their own validity negatively. If the customers of a restaurant come to believe that it is too busy on Fridays, it may end up empty on Fridays: the belief will invalidate itself. If the rumor that it was empty spreads, the customers will swarm back, and the model will invalidate itself again. Such network effects are observed not only in the famous El Farol bar in Santa Fe, and in Kolkatta paise restaurants, but also in financial and stock markets, and in urban agglomerations. Causal models often influence the modeled causal structures, and sometimes impact their own validity. -What are the conditions and limitations of this phenomenon? be used to construct models that would invalidate themselves.
In any case, it turns out that a causal universe C contains a self-confirming causal model for every causal process parametrized over models, as soon as it supports axioms I-III.
The crucial insight is that the partial model Ξ : → Ω, or in the indexed form Ξ(ω, ω), which for every ω ∈ Ω predicts the effect of steering the Ω-parametrized process ω to ω. This "self-model" Ξ • ∆ is displayed in Fig. 8 on the left, where we precompose the given process q with it. Axiom II now gives a model G of the resulting composite process. Fig. 9 shows that Γ = Ξ • ∆ • G provides the claimed causal model of the given process Ω ⊗ A q − → B, parametrized over the possible models. The first step follows from the definition of G in Fig. 8, by substituting G also as the input. The second step folds two copies of G into one. The third step follows Figure 9: Fig. 5. The upshot is that the causal process Id • (Γ ⊗ A) on the right, which is by Sec. 4.4 just the process Γ modeled by Γ, is indistinguishable from the instance q • (Γ ⊗ A) on the left, obtained by steering the process q, which is parametrized over its models, to the model Γ. The causal model Γ thus steers the given causal process q to an instance which confirms Γ as its model.
Summary: Towards artificial causal cognition?
The concept of causality made a remarkable historic voyage, from ancient physics and metaphysics [1] to modern physics [2,10,11], through incisive critiques in philosophy [22,41], through enduring attention in psychology [27,18,44], all the way to its current promise in machine learning and in AI [25,36,37,39,40,46]. How much does the new outlook change the scene of causal cognition? If subjectivity is implied by cognitive self-confirmation, then the results presented in this paper suggest that all causal cognition, human or artificial, is inexorably subjective, as soon as models cannot be eliminated as causal factors. In particular, we have shown how causal models and causal processes can be steered to support each other, and adapted to absorb any new evidence. This is a concerning conclusion, if scientific theories are required to be falsifiable. The fact that causal cognition, as modeled here, permits steering and self-validation under the assumptions as mild as axioms I-III, can also be interpreted as demonstrating that causal cognition cannot be understood just in terms of logic and structure. Such a conclusion would have significant repercussions on the AI research.
Computer science has been recognized as a natural science almost from the outset.This is by now largely accepted in principle, although often misunderstood in the popular narrative, and dismissed as irrelevant by many working computer scientists and engineers. With the advent of the web, and the spread of personal devices, computation propagated through the vast area of economic, political, and social processes, and came to play a role in many forms of human cognition. Changing the angle only slightly, the same process can be viewed as evolution of artificial cognition within a natural process of network computation. Whether the androids of the future will be attracted to causal illusions of action movies remains to be seen; but the humanities of the present cannot avoid studying human cognition together with artificial cognition. Like identical twins, or like parallel universes, or like Lao Tse and the butterfly, the two can be reliably distinguished only by themselves, observing each other from within, but difficult to tell apart for the external observers. The human gave rise to the artificial, and then the artificial gave rise to a new human, and then maybe there was a new artificial, and then they lost count.
Psychology of artificial mind? Herbert Simon, one of the originators AI, anticipated the creation of a whole new realm of "The Sciences of the Artificial" [43]. With the AI technologies permeating market and politics, a psychology of the artificial seems to have become not just a real possibility, but also an urgent task.
the cause a ∈ A . In summary, we have Meaning of the parallel composition and product types. Since the strings in a string diagram correspond to types, drawing parallel strings leads to product types, like A ⊗ U, which is the name of the type corresponding to the strings A and U running in parallel. The events of this type are the pairs a, u ∈ A ⊗ U, where a ∈ A and u ∈ U. The parallel composite A ⊗ U f ⊗t − − → B ⊗ V can thus be defined as the causal process transforming independent pairs of causes into independent pairs of effects, without any interferences between the components:
B Units
Vectors, scalars, covectors. There are processes where events occur with no visible causes; and there are processes where events have no visible effects. Such processes correspond, respectively, to string diagrams c and e in Fig. 11. There are even processes with no observable causes or effects, like the one represented by the diamond s in the middle of Fig. 11. When there are no strings at the bottom, or at the top of a box in a string diagram, we usually contract the bottom side, or the top side, into a point, and the box becomes a triangle. When there are no strings either at the bottom or at the top, then contracting both of them results in a diamond. A diamond with no inputs and outputs may still contain a lot of information. E.g., if causal processes are timed, then reading the time without observing anything else would be a process with no observable causes or effects. If processes are viewed as linear operators, then those that do not consume any input vectors and do not produce any output vectors are scalars. The processes that do not consume anything but produce output are just vectors, since they boil down to their output. The processes that do not produce vectors, but only consume them, are just covectors, or linear functionals.
Invisible string: The unit type I. Since every process must have an input type and an output type for formal reasons, in order to fit into a category of processes, a process that does not have any actual causes is presented in the form I e − → A, where I is the unit type, satisfying for every type A. The unit type is thus degenerate, in the sense that any number of its copies can be added to any type, without changing its elements. It is easy to see that it is unique, as the units in algebra tend to be. A process that does not output any effects is then in the form A c − → I. The unit type is the unit with respect to the type product and to the parallel composition of processes, just like 0 is the unit with respect to the addition of numbers. It can be thought of as the type of a single event that never interferes with any other events.
Since it is introduced only to make sure that everything has a type, but otherwise has no visible causes or effects, the unit type is usually not drawn in string diagrams, but thought of as an "invisible string". In Fig. 11, the invisible string is coming in below e, and out above c, and on both sides of s.
holds for every process A f − → B. This property is clearly analogous to (5), but has a different meaning: the causal process id A inputs causes and outputs effects, both of type A; but it does not modify anything: it just outputs every cause a ∈ A as its own effect. Since the box corresponding to id A thus just passes to the output string whatever comes at the input string, and there is nothing else in it, we do not draw such a box. In string diagrams, the boxes corresponding to the identity processes id A can thus be thought of as "invisible boxes". In a sense, the strings themselves play the role of the identities.
Because of (5), any number of invisible strings can be added to any string diagram without changing its meaning. Because of (6), any number of invisible boxes can be added on any string, visible or invisible, without changing its meaning. This justifies eliding the units not only from string diagrams, but also from algebraic expressions, and writing things like With this notation, the algebra of tensor products, studied in serious math, boils down to a single law, from which everything else follows:
C The middle-two-interchange law
The main reason why string diagrams are convenient is that their geometry captures the middle-two-interchange law: (f ; g) ⊗ (t; s) = (f ⊗ t); (g ⊗ s) Note that the both sides of this equation correspond to the same string diagram, displayed in Fig. 10. The left-hand side of (7) is obtained by first reading the sequential compositions vertically, and then composing them in parallel horizontally; whereas the right-hand side is obtained by first reading the parallel compositions horizontally, and then composing them in sequence vertically. Sliding boxes along stings provides an easy way to derive further equations from the middle-two-interchange low of (7), such as In string diagrams, this is just
D Functions
Causations are usually uncertain to some extent: a cause induces an effect with a certain probability. Those causations that are certain, so that the input always produces the output, and the same input always produces the same output, are called functions. If causations are presented as stochastic processes, then functions are the subfamily deterministic processes.
An intrinsic characterization of functions in a monoidal framewrok is provided by the following definitions.
A data type is a type A ∈ C, equipped with data services, i.e. the operations respectively called deleting and copying, which satisfy the following equations: We define data as the elements preserved by data services. Such elements are precisely those that can be manipulated using variables [30,32].
Remark. If we dualize data services, i.e. reverse the arrows in (8), we get a binary operation and a constant. Transferring the equations from deleting and copying makes the binary operation associative and commutative, and it makes the constant into the unit. The dual of data services is thus the structure of a commutative monoid. The structure of data services itself is thus a commutative comonoid.
Functions are causations that map data to data. A causation A f − → B is a function if it is total and single-valued, which respectively corresponds to the two equations in Fig. 12.
They also make f into a comonoid homomorphism from the data service comonoid on A to the data service comonoid on B. | 2019-10-10T06:24:18.000Z | 2019-10-10T00:00:00.000 | {
"year": 2019,
"sha1": "c52394f3f427abead965341b93491b2130edf290",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "c52394f3f427abead965341b93491b2130edf290",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Computer Science",
"Mathematics"
]
} |
249217187 | pes2o/s2orc | v3-fos-license | Bromodomain and Extra-Terminal Protein Inhibitors: Biologic Insights and Therapeutic Potential in Pediatric Brain Tumors
Pediatric brain tumors have surpassed leukemia as the leading cause of cancer-related death in children. Several landmark studies from the last two decades have shown that many pediatric brain tumors are driven by epigenetic dysregulation within specific developmental contexts. One of the major determinants of epigenetic control is the histone code, which is orchestrated by a number of enzymes categorized as writers, erasers, and readers. Bromodomain and extra-terminal (BET) proteins are reader proteins that bind to acetylated lysines in histone tails and play a crucial role in regulating gene transcription. BET inhibitors have shown efficacy in a wide range of cancers, and a number have progressed to clinical phase testing. Here, we review the evidence for BET inhibitors in pediatric brain tumor experimental models, as well as their translational potential.
Introduction
Pediatric brain tumors are the most common solid tumor in children, and are the leading cause of cancer-related death [1]. Several aggressive pediatric brain tumors, such as medulloblastoma (MB), high-grade glioma (HGG), ependymoma (EPN) and atypical teratoid/rhabdoid tumors (ATRTs), are defined by epigenetic dysregulation that has been shown to reflect disordered developmental processes that occur in susceptible cell types and along specific spatio-temporal patterns [2,3]. In the context of normal development, epigenetic mechanisms allow for transcriptional control independent of DNA sequence, and are crucial components of cell differentiation and specialization [4][5][6]. The most widely studied mechanisms include DNA methylation and histone modifications, both of which are commonly co-opted in pediatric brain tumor pathogenesis.
Histones are the fundamental building blocks of chromatin and assemble in octamers, around which DNA is wrapped to form the nucleosome [7]. Each histone has an amino acid tail that extends from the nucleosome and is enriched in positively charged residues, such as lysine and arginine, that are subject to post-translational covalent modifications that have dramatic impacts on chromatin accessibility [8,9]. At least 16 of these modifications have been described to date and include lysine acetylation, lysine and arginine methylation, serine and threonine phosphorylation, and lysine ubiquitination or sumoylation [10,11]. Specific enzymes, known as "writers" and "erasers", are responsible for catalyzing the transfer of these marks, while "reader" proteins are involved in recognition of the marks to recruit other protein complexes involved in transcriptional control.
Histone lysine acetylation is one of the most well-studied histone post-translational modification (PTM) and decreases the positive charge of histones leading to a relaxed chromatin conformation associated with greater levels of transcription [12,13]. Histone acetylation writers (histone acetyltransferases, or HATs), erasers (histone deacetylases, or Human bulk RNA expression from the GTex database, clustered by brain tissue type [34] (B) Single-cell RNA sequencing atlas of the developing mouse brain showing ubiquitous BRD4 expression throughout cell types of the developing central nervous system [35,36].
BET Inhibitors
Small molecule inhibitors of BET bromodomains were first discovered in 2010 by two groups, based on a thienotriazolodiazepine (JQ1) and benzotriazolodiazepine (I-BET151) scaffold, respectively [37,38]. They were shown to potently inhibit bromodomain binding to acetylated lysine residues, resulting in displacement of BRD4 from nuclear chromatin. Both compounds selectively inhibit BET bromodomains over other bromodomain classes, although neither discriminates between BD1/BD2 within each BET protein nor between BRD2/BRD3/BRD4/BRDT. These compounds have demonstrated impressive pre-clinical Human bulk RNA expression from the GTex database, clustered by brain tissue type [34] (B) Single-cell RNA sequencing atlas of the developing mouse brain showing ubiquitous BRD4 expression throughout cell types of the developing central nervous system [35,36].
BET Inhibitors
Small molecule inhibitors of BET bromodomains were first discovered in 2010 by two groups, based on a thienotriazolodiazepine (JQ1) and benzotriazolodiazepine (I-BET151) scaffold, respectively [37,38]. They were shown to potently inhibit bromodomain binding to acetylated lysine residues, resulting in displacement of BRD4 from nuclear chromatin. Both compounds selectively inhibit BET bromodomains over other bromodomain classes, although neither discriminates between BD1/BD2 within each BET protein nor between BRD2/BRD3/BRD4/BRDT. These compounds have demonstrated impres- sive pre-clinical efficacy in a variety of cancer models, including BRD4-NUT fusions NUT-midline carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies. lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
AZDZ5153
Pan-BET I/II AML; Unique bivalent binding mode carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
BI-894999
Pan-BET I Advanced solid tumors, DLBCL, or NMC carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
BMS-986158
Pan-BET I/II Advanced solid tumors and hematologic malignancies; Ongoing pediatric study carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies. Pan-BET I Advanced solid tumors, NHL; Ongoing pediatric study, good CNS penetration efficacy in a variety of cancer models, including BRD4-NUT fusions NUT-midline carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
CPI-0610
Pan-BET I/II Lymphoma, Multiple myeloma, AML, MPNST efficacy in a variety of cancer models, including BRD4-NUT fusions NUT-midline carcinoma (NMC), acute myeloid leukemia (AML), medulloblastoma, breast cancer, and lung cancer [39][40][41]. In the last decade, several "clinical-grade" pan-BET inhibitors with improved pharmacokinetic profiles have been developed (Table 1). From the initial diazepane-based JQ1 and I-BET151, a variety of scaffolds with potent BET inhibitory properties have been employed successfully. Clinical trials with these agents have shown that thrombocytopenia is an important and common dose-limiting toxicity of pan-BET inhibitors [42]. More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
Medulloblastoma
Medulloblastoma is the most common embryonal brain tumor occurring in childhood, and accounts for approximately 6.5% of all pediatric brain tumors. 1 Several landmark studies, published in 2006-2011, used genomic profiling to identify distinct molecular subgroups, which were summarized in a 2012 consensus report designating four subgroups (WNT, SHH, Group 3, Group 4) [48][49][50][51][52]. In 2017, three independent groups also used DNA methylation analysis, which recently has been demonstrated to have exquisite discriminatory power in the diagnosis of CNS tumors, to further subclassify medulloblastoma subtypes [53][54][55]. Each medulloblastoma subgroup has unique underlying biology, which is reflected in differences in patient outcomes and treatment responses. These subgroups have subsequently been incorporated into the 2016 and 2021 updated World Health Organization (WHO) Classification of Tumors of the Central Nervous System [56,57].
Treatment for medulloblastoma is aggressive and includes surgical resection, radiation therapy, and chemotherapy. While these treatments have led to five-year overall survival rates over 80%, they can lead to significant treatment-related toxicities, and patients with higher risk features can have survival rates as low as 20% [58]. One subgroup that is associated with a poor prognosis is Group 3 medulloblastoma, which are often metastatic at diagnosis and tend to be refractory to therapy or have late recurrence. Molecularly, Group 3 tumors often have MYC amplification or high MYC expression, and high MYC expression has been reported to be an independent risk factor for poor outcome in medulloblastoma [59].
Several groups have reported promising findings of BET inhibition in MYCamplified medulloblastoma. The BET inhibitor JQ1 potently decreases cell viability of MYC-amplified medulloblastoma cell lines, causes G1 arrest and apoptosis, and
Medulloblastoma
Medulloblastoma is the most common embryonal brain tumor occurring in childhood, and accounts for approximately 6.5% of all pediatric brain tumors. 1 Several landmark studies, published in 2006-2011, used genomic profiling to identify distinct molecular subgroups, which were summarized in a 2012 consensus report designating four subgroups (WNT, SHH, Group 3, Group 4) [48][49][50][51][52]. In 2017, three independent groups also used DNA methylation analysis, which recently has been demonstrated to have exquisite discriminatory power in the diagnosis of CNS tumors, to further subclassify medulloblastoma subtypes [53][54][55]. Each medulloblastoma subgroup has unique underlying biology, which is reflected in differences in patient outcomes and treatment responses. These subgroups have subsequently been incorporated into the 2016 and 2021 updated World Health Organization (WHO) Classification of Tumors of the Central Nervous System [56,57].
Treatment for medulloblastoma is aggressive and includes surgical resection, radiation therapy, and chemotherapy. While these treatments have led to five-year overall survival rates over 80%, they can lead to significant treatment-related toxicities, and patients with higher risk features can have survival rates as low as 20% [58]. One subgroup that is associated with a poor prognosis is Group 3 medulloblastoma, which are often metastatic at diagnosis and tend to be refractory to therapy or have late recurrence. Molecularly, Group 3 tumors often have MYC amplification or high MYC expression, and high MYC expression has been reported to be an independent risk factor for poor outcome in medulloblastoma [59].
Several groups have reported promising findings of BET inhibition in MYCamplified medulloblastoma. The BET inhibitor JQ1 potently decreases cell viability of MYC-amplified medulloblastoma cell lines, causes G1 arrest and apoptosis, and
Medulloblastoma
Medulloblastoma is the most common embryonal brain tumor occurring in childhood, and accounts for approximately 6.5% of all pediatric brain tumors. 1 Several landmark studies, published in 2006-2011, used genomic profiling to identify distinct molecular subgroups, which were summarized in a 2012 consensus report designating four subgroups (WNT, SHH, Group 3, Group 4) [48][49][50][51][52]. In 2017, three independent groups also used DNA methylation analysis, which recently has been demonstrated to have exquisite discriminatory power in the diagnosis of CNS tumors, to further subclassify medulloblastoma subtypes [53][54][55]. Each medulloblastoma subgroup has unique underlying biology, which is reflected in differences in patient outcomes and treatment responses. These subgroups have subsequently been incorporated into the 2016 and 2021 updated World Health Organization (WHO) Classification of Tumors of the Central Nervous System [56,57].
Treatment for medulloblastoma is aggressive and includes surgical resection, radiation therapy, and chemotherapy. While these treatments have led to five-year overall survival rates over 80%, they can lead to significant treatment-related toxicities, and patients with higher risk features can have survival rates as low as 20% [58]. One subgroup that is associated with a poor prognosis is Group 3 medulloblastoma, which are often metastatic at diagnosis and tend to be refractory to therapy or have late recurrence. Molecularly, Group 3 tumors often have MYC amplification or high MYC expression, and high MYC expression has been reported to be an independent risk factor for poor outcome in medulloblastoma [59].
Several groups have reported promising findings of BET inhibition in MYCamplified medulloblastoma. The BET inhibitor JQ1 potently decreases cell viability of MYC-amplified medulloblastoma cell lines, causes G1 arrest and apoptosis, and More recently, BD2 selective BET inhibitors, such as ABBV-744, have been developed with the goal to narrow anti-neoplastic selectivity and reduce off-target hematologic toxicities [43,44]. Another exciting development has been in the field of targeted protein degradation. In 2015, three BET proteolysis-targeting chimeras (PROTACs) were discovered: dBET1, MZ1, and ARV-825 [45][46][47]. Proteolysis-targeting chimeras (PROTACs) are bifunctional small molecules that include a "warhead" ligand that recruits a protein of interest, linked to an E3 ligase ligand, that induces target ubiquitination and subsequent proteasomal degradation. In some settings these compounds have shown enhanced potency in cancer cell lines compared to their parental inhibitors, paving a new strategy for targeting BET proteins in select malignancies.
Medulloblastoma
Medulloblastoma is the most common embryonal brain tumor occurring in childhood, and accounts for approximately 6.5% of all pediatric brain tumors. Several landmark studies, published in 2006-2011, used genomic profiling to identify distinct molecular subgroups, which were summarized in a 2012 consensus report designating four subgroups (WNT, SHH, Group 3, Group 4) [48][49][50][51][52]. In 2017, three independent groups also used DNA methylation analysis, which recently has been demonstrated to have exquisite discriminatory power in the diagnosis of CNS tumors, to further subclassify medulloblastoma subtypes [53][54][55]. Each medulloblastoma subgroup has unique underlying biology, which is reflected in differences in patient outcomes and treatment responses. These subgroups have subsequently been incorporated into the 2016 and 2021 updated World Health Organization (WHO) Classification of Tumors of the Central Nervous System [56,57].
Treatment for medulloblastoma is aggressive and includes surgical resection, radiation therapy, and chemotherapy. While these treatments have led to five-year overall survival rates over 80%, they can lead to significant treatment-related toxicities, and patients with higher risk features can have survival rates as low as 20% [58]. One subgroup that is associated with a poor prognosis is Group 3 medulloblastoma, which are often metastatic at diagnosis and tend to be refractory to therapy or have late recurrence. Molecularly, Group 3 tumors often have MYC amplification or high MYC expression, and high MYC expression has been reported to be an independent risk factor for poor outcome in medulloblastoma [59]. Several groups have reported promising findings of BET inhibition in MYC-amplified medulloblastoma. The BET inhibitor JQ1 potently decreases cell viability of MYC-amplified medulloblastoma cell lines, causes G1 arrest and apoptosis, and downregulates MYCtargets, as well as MYC transcription itself [60,61]. Furthermore, JQ1 treatment shows in-vivo efficacy with decreased tumor growth and prolonged survival in both flank and cerebellar orthotopic models of Group 3 medulloblastoma (Table 2). JQ1 has also been shown to induce cellular senescence and suppress transcriptional programs associated with poor prognosis in medulloblastoma patients [62].
BET inhibition has also been shown to be a promising strategy in SHH-driven medulloblastoma. Canonical hedgehog (Hh) signaling occurs via binding of Hh ligands to the PTCH1 transmembrane protein, which leads to the cessation of PTCH1 s role in repressing the G protein coupled receptor Smoothened (SMO) [63]. Activation of SMO leads to positive regulation of the GLI zinc-finger transcription factors, leading to the expression of a Hh transcriptional program that regulates proliferation and cell specification [64]. Aberrant Hh signaling has been associated with several cancers, including basal cell carcinoma, medulloblastoma, and pancreatic cancer [65,66]. BRD4 has been shown to be critical in GLI1 and GLI2 transcription via direct promoter occupation using a mouse 3T3 Gli-luciferase reporter cell line [67]. This interaction can be significantly disrupted with JQ1 treatment, providing a biological rationale for BET inhibition in SHH-driven medulloblastoma. Indeed, JQ1 decreases proliferation and viability of SHH-driven medulloblastoma cell lines in-vitro and in-vivo, even when mutations conferring these cell lines to SMO inhibitor resistance are present. The BET inhibitor I-BET151 has also been studied, and was found to significantly attenuate HH activity in the Light2 reporter cell line [68]. Furthermore, I-BET151 decreases growth and viability of a murine derived (Ptch1 +/− ) medulloblastoma cell line in-vitro, decreases GLI1 transcription, and decreases tumor growth in-vivo.
Several synergistic drug combinations with BET inhibitors have also been investigated, with particular focus on cell cycle inhibitors. In 2019, Bandopadhayay et al. used a functional genomic approach including CRISP/Cas9-mediated loss of function and ORF/cDNA rescue screens to identify key regulators of BET inhibitor response and resistance [69]. They found that resistant cells express transcriptional programs associated with neuronal differentiation, while still maintaining proliferative potential. They also demonstrated that CDK4/CDK6 inhibition delays the acquisition of BET inhibitor resistance, and the combination of JQ1 with LEE01 (a CDK4/6 inhibitor) improved survival, compared to monotherapy in flank and orthotopic xenograft models of MYC-driven medulloblastoma. The CDK2 inhibitor Milcilib also synergizes with JQ1 in Group 3 medulloblastoma in-vivo and in-vivo models [70]. From a mechanism standpoint, combination treatment with JQ1 + Milcilib targets MYC expression, as well as MYC stabilization, respectively; the latter being a known effect of CDK2 inhibition through suppression of MYC residue S62 phosphorylation. Through these mechanisms, combination treatment was found to be significantly more effective than either JQ1 or Milcilib monotherapy in downregulating MYC target expression in Group 3 medulloblastoma models.
Diffuse Intrinsic Pontine Glioma
Diffuse intrinsic pontine glioma (DIPG), now referred to as diffuse midline glioma, H3 K27-altered in the 2021 WHO classification, is a lethal brainstem tumor occurring in childhood. Radiation is the only proven therapeutic, despite dozens of chemotherapeutic trials over the last decades, and it only confers a survival advantage of approximately three months [71]. In 2012, a breakthrough was made when several groups reported that over 80% of DIPG tumors harbor a unique lysine-to-methionine mutation (K27M), the H3 histone tail [72,73]. Subsequent work showed that this mutation leads to widespread loss of the repressive H3K27me3 mark through inhibition of the poly-comb repressor 2 complex (PRC2) [74]. Importantly, there are some genomic regions where PRC2 activity is retained and even increased [75]. ChIP sequencing studies have shown that in DIPG, the pathogenic H3K27M mutation colocalizes with acetylated H3K27, RNA polymerase II, and BRD2/BRD4 at sites of active transcription [76]. Treatment with JQ1 impairs the growth and viability of DIPG cell lines, with RNA sequencing demonstrating specific targeted effects at genes with BRD2/BRD4 promoter occupancy. Furthermore, 10 days of JQ1 treatment was found to significantly prolong survival in a DIPG mouse orthotopic (brainstem) patient derived xenograft model. Nagaraja et al. found similar promising findings with JQ1 treatment in a panel of H3K27M mutant DIPG cell lines, while a H3WT glioblastoma cell line (SU-pcGBM2) showed little response [77]. Transcriptomic analysis after JQ1 treatment showed downregulation of genes associated with central nervous system development, such as NTRK3, ASCL1, and MYT1. Lentiviral shRNA knockdown of BRD4 in two DIPG models decreased tumor growth and prolonged survival after mouse orthotopic brainstem injections with lentiviral-modified and control cell lines.
Drug combinations with BET inhibitors have also been investigated in DIPG. Wiese et al. examined JQ1 with the CREB binding protein (CBP) inhibitor ICG-001 [78]. CBP's main function is as an acetyltransferase which regulates H3K27 acetylation, although it also recruits transcription factors and serves a scaffolding role in multi-unit transcriptional complexes [79]. The authors found that JQ1 + ICG-001 treatment synergistically inhibited DIPG proliferation and invasion potential in DIPG cell lines. RNA sequencing showed that JQ1 downregulated a significant proportion of super-enhancer regulated genes, while surprisingly ICG-001 monotherapy led to an upregulation of these genes. Combination therapy was able to reverse this subset of SE-associated genes which were inadvertently increased with ICG-001 treatment. Another study used H3K27M/PDGFB expressing NSCs as a DIPG model, and showed that JQ1 synergized with the EZH2 inhibitor EPZ6438 (tazemetostat) in-vitro and in-vivo [80]. Taylor et al. showed that targeting the NOTCH pathway with the γ-secretase inhibitor MRK003 also synergized with JQ1 in 2/3 DIPG cell lines tested [81].
Recently, high-throughput chromosome conformation capture (Hi-C) has been used to characterize the 3-dimensional chromatin structure of DIPG cells [82]. This technology enables mapping of tumor-specific regulatory networks, as well as enhancer hijacking events. The BET inhibitor BMS986158 and the BET degrader dBET6 were found to significantly perturb the micro-and macro-chromatin interactions of DIPG cells, and these findings were uniformly more pronounced with dBET6 treatment. This study suggests one possible advantage of targeted protein degradation over catalytic inhibition in this context.
Ependymoma
Ependymoma is a brain tumor that occurs in both children and adults and arises from ependymal cells lining the ventricular system and spinal canal. Ependymoma is the third most common brain tumor in childhood, and can arise anywhere along the craniospinal axis; although in children they tend to occur more commonly in a supratentorial (ST) or posterior fossa (PF) location [83]. Diagnosis of ependymoma is made histologically with three distinct histologic grades (Grade I, II, and III), although prognostic utility is debated given inter-observer variability and the advent of molecular stratification [84]. Standard of care involves maximum safe surgical resection with post-operative radiotherapy [85]. The role of chemotherapy is controversial, although it is commonly used in patients under 18 months of age to delay radiation and there are ongoing efforts to understand if there are benefits to adjuvant and/or maintenance chemotherapy regimens for select patients [86].
In a landmark 2015 study, Patjler et al. used DNA methylation and transcriptomic profiling to identify at least nine clinically and biologically relevant distinct subgroups, which have had major implications in the official WHO classification [87]. Supratentorial ependymomas are made up of ST-sub-ependymoma and two subgroups, defined by characteristic oncogenic fusions (YAP1 and RELA). Of note, RELA ependymomas have been renamed in the most recent WHO classification to ZFTA-fusion positive with the recognition that C11orf95/ZFTA fusions can occur with non-RELA partner genes, such [88][89][90]. Posterior fossa ependymomas are comprised of PF-sub-ependymoma and two subgroups, denoted group A (PFA, associated with EZHIP overexpression and poor clinical outcome) and group B (PFB, associated with relatively favorable prognosis). Spinal ependymomas are made up of SP-sub-ependymoma, SP myxopapillary ependymoma, and SP ependymoma. A recent study by Bockmayr et al. demonstrated that SP myxopapillary ependymoma is molecularly comprised of two distinct subgroups, MPE-A and MPE-B; with the former being associated with significantly higher relapse rate and decreased progression-free survival [91].
Work by Mack et al. used H3K27ac ChIP-seq to identify enhancers and superenhancers in a cohort of 42 primary intracranial ependymomas [92]. They found that genetic knockout of super-enhancer regulated genes impaired growth of ependymoma cell lines in-vivo, which provided a rationale for trialing BET inhibition as a therapeutic strategy. They subsequently showed that JQ1 inhibited the growth of two ependymoma cell lines, one PF-A and the other ST-ZFTA. Another group tested the CNS-penetrant BET inhibitor OTX-015 in three ependymoma cell lines (two PFA, one ST (subgroup not specified)) and similarly found it to be effective with sub-micromolar IC50 values in three-day viability assays [93]. They found that in-vivo treatment prolonged survival in 1/2 of the PFA intracranial patient-derived xenograft models tested.
Embryonal Tumor with Multilayer Rosettes (ETMR)
ETMR is a rare, aggressive CNS embryonal tumor occurring primarily in children younger than three years of age [94]. It was first recognized as a distinct molecular diagnosis after studies identified a unique recurrent microRNA amplification of C19MC on chr19q13.42 in a subset of primitive neuroectodermal tumors, including embryonal tumors, with abundant neuropil and true rosettes (ETANTR), ependymoblastoma, medulloepithelioma and supratentorial primitive neuroectodermal tumors (sPNET) [95][96][97]. Treatment of ETMR involves maximum safe surgical resection, and given the young age at diagnosis craniospinal irradiation is typically contraindicated (although local radiation is often used) [98]. High-dose chemotherapy protocols using agents with known embryonal activity, such as cyclophosphamide, vincristine, methotrexate, etoposide, cisplatin, carboplatin, and thiotepa, are often used [99]. However, given the rarity of the disease no standardized treatment protocol has been established yet. Despite these intensive treatment regimens prognosis remains dismal, with reported five-year survival rates between 0-30%.
In 2019, Sin-Chan et al. comprehensively profiled over 80 primary ETMR samples using global methylation, SNP, transcriptional, and miRNA profiling [100]. Through this analysis they uncovered a C19MC-LIN28A-MYCN feed forward oncogenic circuit. LIN28A had previously been found to be highly expressed in ETMR, and is a pluripotency factor and RNA-binding protein important to neural development. LIN28A has also been implicated in the pathogenesis of many advanced human malignancies, including ovarian carcinoma, germ cell tumors, and Wilms' tumor [101]. The authors hypothesized that this C19MC-LIN28A-MYCN core regulatory circuit could be targeted by BET protein inhibition, and, indeed, found that treatment with JQ1S (the active isomer of JQ1) led to a reduction in viability of ETMR cell lines, and qRT-PCR and western blot showed downregulation of MYCN and LIN28A.
Atypical Teratoid/Rhabdoid Tumor (ATRT)
ATRT is a rare, highly aggressive brain tumor of early childhood [102]. Biallelic inactivation of SMARCB1, a core member of the SWI/SNF (also known as BAF) chromatin remodeling complex, is the core genetic driver event in the vast majority of cases [103,104]. While ATRT has been shown to have a "quiet genome" with few other identifiable pathogenic alterations, this genetic simplicity belies the clinical aggressiveness of these tumors and the difficulties in finding effective targeted therapeutics [105]. Recent studies have used transcriptomics and DNA-methylation to group ATRT into three distinct subgroups: ATRT-TYR, ATRT-SHH, and ATRT-MYC [106][107][108]. ATRT expresses high levels of c-MYC, and ChIP-sequencing shows that SMARCB1 loss leads to significant enrichment of MYC chromatin occupancy at transcriptional start sites (TSS), compared to normal embryonic stem cells [109]. Genetic knockdown of c-MYC decreases ATRT cell growth in-vivo and prolongs survival of mouse PDX models in-vivo, providing a rationale for BET inhibition in this malignancy. Indeed, JQ1 treatment significantly decreases c-MYC transcription, as well as MYC-driven stemness programs regulated by SOX2, Nanog and OCT4 in ATRT. Furthermore, JQ1 treatment significantly prolongs survival in orthotopic xenograft mouse models.
BET Inhibitors in CNS Malignancies
Despite the development of several clinical grade BET inhibitors, few have demonstratable CNS penetrance. While JQ1 is brain penetrant and has widely been used in the pre-clinical setting, it's poor pharmacokinetic properties (namely short half-life) have precluded translation. One CNS penetrant BET inhibitor that has been well characterized is OTX015/MK-8628/Birabresib, which was shown to be effective in glioblastoma (GBM) pre-clinical models [110]. This led to a Phase IIa trial in patients with recurrent glioblastoma (NCT02296476). Twelve patients were enrolled, and the drug was well tolerated with pharmacokinetic studies demonstrating biologically active levels; however, all patients progressed with a median progression-free survival of two months and the trial was terminated [111]. BMS-986378/CC-90010 is an orally bioavailable, CNS-penetrant BET inhibitor that is currently being evaluated in clinical trials. A phase I trial of CC-90010 in patients with advanced solid tumors and relapsed/refractory non-Hodgkin's lymphoma enrolled a total of 69 patients, of whom 10 had high-grade gliomas (NCT03220347). There was an overall response rate of 2.9% (n = 2), with 8.8% (n = 6) of patients achieving stable disease [112]. One patient with a progressive grade II diffuse astrocytoma had a complete response (CR). There are two ongoing studies of CC-90010 in high-grade glioma, including a phase I trial to evaluate the CNS penetration of CC-90010 in patients with progressive/recurrent astrocytoma, anaplastic astrocytoma or GBM (NCT04047303), as well as a Phase Ib study of CC-90010 in combination with temozolomide and radiation in patients with newly diagnosed GBM (NCT04324840).
BET Inhibitors in Pediatrics
In 2019, the first pediatric trial of BET inhibition opened (NCT03936465) investigating the inhibitors BMS-986158 (Arm 1) and BMS-986378/CC-90010 (Arm 2). The primary aims of the trial are to determine toxicities and recommended phase II doses for these agents, while secondary outcomes include: (a) efficacy, (b) pharmacokinetics, and (c) a host of pharmacodynamic and predictive biomarkers. Patients must be <21 and able to swallow intact pills. Each arm has two cohorts, one being with unselected biology (relapsed/refractory solid tumors or lymphoma for Arm 1, CNS tumors for arm 2) and the other enriched for predictive biology. Eligibility for this second criterion includes: MYCN amplification or high copy number gain, MYC amplification or high copy number gain, translocation involving MYC or MYCN, translocation involving BRD4 or BRD3, BRD4 amplification or high copy number gain, and/or histologic diagnosis of NUT midline carcinoma (NMC).
The state of BET inhibitor development in pediatric oncology was recently summarized at a strategy forum organized by ACCELERATE, in collaboration with the European Medicines Agency (EMA) and the US Food and Drug Administration (FDA) [113]. Given that most adult trials of pan-BET inhibitors have been challenged by similar issues (narrow therapeutic index, due to thrombocytopenia, modest anti-tumor activity as monotherapy), it was agreed that opening additional trials to study pan-BET inhibitors, other than BMS-986158 and BMS-986378/CC-90010, was not a worthwhile strategy. Agents with unique properties, such as BDII-selectivity, dual targets e.g., BET/p300, or improved CNS penetrance were highlighted as areas of particular interest worthy of further investigation.
Conclusions
The discovery and development of BET inhibitors was a landmark finding in cancer epigenetics, and many drugs in this class have advanced to clinical trial stages. Results to date have mostly been mixed, with modest anti-tumor activity and relatively narrow therapeutic index, due to hematologic dose-limiting toxicities. However, there is strong biological and pre-clinical rationale for testing BET inhibitors in pediatric brain tumors, given the common thread of epigenetic dysregulation. The ongoing trial of BMS-986378/CC-90010 in relapsed/refractory pediatric CNS tumors will be highly informative to assess the promise of BET inhibitors in this patient population. Based on experience from experimental models and adult clinical trials, it is likely that monotherapy alone will not be sufficient and, therefore, continued research dedicated to investigating synergistic combinations with BET inhibitors will be critical.
Data Availability Statement:
The data presented in this study ( Figure 1) is openly available at the Gtex portal [34] and the Mouse Brain Atlas [35,36]. | 2022-06-01T15:14:37.737Z | 2022-05-26T00:00:00.000 | {
"year": 2022,
"sha1": "0fc4f8067a352a8fac9911ec126365ebedc3bf22",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8247/15/6/665/pdf?version=1653559458",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3dd14dd97d09444762f694381663b78e332a22c9",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
7602927 | pes2o/s2orc | v3-fos-license | A Neuro-Mechanical Model Explaining the Physiological Role of Fast and Slow Muscle Fibres at Stop and Start of Stepping of an Insect Leg
Stop and start of stepping are two basic actions of the musculo-skeletal system of a leg. Although they are basic phenomena, they require the coordinated activities of the leg muscles. However, little is known of the details of how these activities are generated by the interactions between the local neuronal networks controlling the fast and slow muscle fibres at the individual leg joints. In the present work, we aim at uncovering some of those details using a suitable neuro-mechanical model. It is an extension of the model in the accompanying paper and now includes all three antagonistic muscle pairs of the main joints of an insect leg, together with their dedicated neuronal control, as well as common inhibitory motoneurons and the residual stiffness of the slow muscles. This model enabled us to study putative processes of intra-leg coordination during stop and start of stepping. We also made use of the effects of sensory signals encoding the position and velocity of the leg joints. Where experimental observations are available, the corresponding simulation results are in good agreement with them. Our model makes detailed predictions as to the coordination processes of the individual muscle systems both at stop and start of stepping. In particular, it reveals a possible role of the slow muscle fibres at stop in accelerating the convergence of the leg to its steady-state position. These findings lend our model physiological relevance and can therefore be used to elucidate details of the stop and start of stepping in insects, and perhaps in other animals, too.
Introduction
When legged animals stop or start stepping, a transition between posture and locomotion takes place. Thus stop and start of stepping are essential basic actions of the musculo-skeletal system of a leg. There is experimental evidence [1] that a leg does not stop randomly during its step cycle. Rather it stops or starts stepping in a systematic way depending on its position within the stepping cycle [1]. It is therefore reasonable to assume that both processes require coordinated actions of the leg muscles. This is presumably achieved by the interactions of the local neuronal networks that control the activity of the leg muscles. For a deeper understanding of the underlying neurophysiological mechanisms of locomotion in insects, it is thus quite important to study and analyze its elementary processes such as stop and start of stepping. This may open up the way for tackling more complex processes of walking in various conditions.
To be specific, in the stick insect, 3 pairs of antagonistic muscles play the major part in locomotion: the m. protractor and retractor coxae at the thorax-coxa (ThC) joint, the m. levator and depressor trochanteris at the coxa-trochanter (CTr) joint, and the m. flexor and extensor tibiae at the femur-tibia (FTi) joint. The coordination of their activity within a leg is achieved by means of proprioceptive sensory signals. They report load or position, or position and (angular) velocity to the nervous system. The load signals are generated in the campaniform sensilla (CS) [2], the position signals in specialized hairfields [3][4][5], and the position and (angular) velocity signals in the chordotonal organs [6,7]. Chordotonal organs are present in other insects, as well [8,9]. Their sensory signals are conveyed between the local neuronal networks controlling the activity of the muscle pairs. On the efferent side, one can distinguish between slow and fast muscle fibres constituting each of the above muscles according to their contraction kinetics [10,11], or histochemical properties [12]. [10] and [11] showed that the two types are anatomically separated in the extensor tibiae muscle of the stick insect but more importantly that they also have different physiological function: fast muscle fibres are active during stepping, only, whereas slow muscle fibres are responsible for maintaining the static position (posture) of the stick insect. Since [12], in a recent work, showed the existence of slow and fast fibres in the other, aforementioned muscles, too, it seems reasonable to assume that they have analogous function in those muscles, as well.
The question now arises whether and how the neuro-muscular system just described can bring about the stop and start of stepping of an insect leg. One suitable way to try to answer this question is to use appropriate mathematical models. In the accompanying paper [34], we presented a neuro-mechanical model that included slow and fast muscle fibres and their dedicated controlling neuronal networks. In this paper, we apply an extended version of this model in an attempt to unveil and elucidate the details of the stopping and starting of stepping. We have thus extended the model in [34] to include four important new properties: i) all six muscle types have both slow and fast fibres; ii) the slow muscle fibres possess residual stiffness, and iii) are controlled by the activity of the common inhibitor motoneuron CI1 (for the flexor tibiae muscle CI2 and CI3); iv) the effects of the position and (angular) velocity sensory signals are implemented. As a result, we can suggest neuro-mechanical mechanisms that might exist in insects at stop and start of stepping. More generally, we hope to have helped gain a deeper understanding of elementary mechanisms of locomotion in insects, and perhaps in other animals, too.
Methods
The model comprising all three neuro-muscular systems The model introduced in this paper is an extension of the models in [13] and the accompanying paper [34]. Fig. 1 shows the network with all three neuro-muscular systems. Each of them is now equipped with slow muscles, too, and with motoneurons (MNs) that innervate the slow muscles (in short slow MNs), as well as with the corresponding interneurons (INs). The three systems are coupled via position and load signals [2] represented by the levation angle b (hexagon with b in Fig. 1). If b exceeds, or falls below, a critical value (b cr~3 8 o for the protractor-retractor system and b cr~5 0 o for the extensor-flexor system), it will initiate a new (swing or stance) phase of a stepping cycle. For a more detailed explanation, see [13].
The activation kinetics of a muscle fibre during a contraction initiated by the excitation of its MN determine its type. Thus fast muscle fibres have fast activation kinetics and slow fibres much slower ones compared to those of the fast muscle fibres. The slow kinetics of the slow muscle fibres are therefore characterized by small rate constants, which apply during an incoming action potential. The specific values of the activation rate constants of the fast muscle fibres are listed in Table 1 for each muscle type. These values were chosen in earlier versions of the model [13,14] such as to fit the movements of the femur and the tibia during the swing and the stance phase of the stepping leg as seen in the experiments [15].
The values of the activation rate constants of the corresponding slow muscle fibres were set to be 100 times smaller. The relaxation rate constants (b values) were chosen to be identical in both muscle types (b~0:01 ms 21 for all muscle types). Details of the properties of the neuron and muscle models and the neuro-muscular coupling can be found in [14] and in the accompanying paper [34].
However, the elastic properties of the slow muscle fibres differ substantially from those of the fast ones. All types of the slow fibres are assumed to have a positive residual stiffness, while the fast ones are not. Formally that means that the actual value, k(t), of the stiffness (spring constant) of the slow muscles in the absence of an action potential is now calculated as (cf. eqns 1-2 in the accompanying paper [34]). Here is k res , the actual value of the residual stiffness. This value can be affected by the activity of the CI1 (and CI2-CI3) (see below). The residual stiffness ensures that static positions of the stick insect are maintained over a longer period of time with virtually no driving activity of the motoneurons innervating these fibres [16]. The value of the residual stiffness is controlled by the activity of common inhibitory MNs. The common inhibitory MN CI1 innervates slow fibres of five of the six muscles named above. (The slow flexor tibiae muscle is innervated by CI2 and CI3.) [17] performed experiments on the locust and showed that the residual stiffness of the slow muscles is abolished during locomotion (stepping) by the activity of CI1. He suggested that the main physiological role of CI1 (and of the synchronously active CI2 and CI3 in the m. flexor tibiae [18]) is to ensure fast movements of the limbs, especially during the swing phase (e.g. during protraction in the protractor-retractor muscle system). Similar results were obtained in the crab [19] and in the cockroach [20,21]. We implemented the residual stiffness of the slow muscle fibres in accordance with these findings and hypotheses. Table 2 lists its value for each muscle type. The values in Table 2 were chosen such that the stationary position, i.e. the angles b and c at the coxa-trochanter and the femur-tibia joint, respectively, be in the range of angles measured in the stick insect in its standing resting position (Fig. 2). Thus, b in a small range around 30 o (b&30 o ), and c in the range of 80 o to 90 o (c&84 o ) [22]. The position of the femur in the horizontal plane is determined by the angle a. Its stationary value was set to 90 o , and the values of the residual spring constants of the slow protractor and retractor muscles were calculated accordingly (Table 2). These ''basic'' stationary values can, of course, be modified by making use of the recruitment properties of the model (cf. Results and the accompanying paper [34]).
In the present version of the model, we implemented the function of CI1 implicitly. We did not model the mechanism itself that produces the changes in the residual stiffness of the slow muscle fibres during activity of CI1 but the effect, i.e. the changes, only. Hence, the residual values of the spring constants assume their stationary (nonzero) values during inhibition of the innervating CI1. During stepping, when CI1 is active, the residual values vanish in the swing phase and are small in the stance phase of the stepping. To be more specific, CI1 affects the actual value of the residual stiffness k res of a slow muscle in the following way: In this eqn, k res0 is the value of the residual stiffness k res which it assumes during steady state. Furthermore, t stop is the onset of the stop command, and t start is the onset of the command for re-start. V CN (t), with N~1,4,5, denotes the actual value of the membrane potential of the 'retractor' (C1), 'depressor' (C4) and 'extensor' (C5) CPG neurons' (cf. Fig. 1). V th is a threshold value (set to {25mV in the simulations). It is used to decide whether the CPGs are in phases corresponding to the swing phase of the stepping cycle, in particular, whether the protractor-retractor CPG is in the protraction phase. That is, if the neurons CN, N~1,4,5 are sufficiently hyperpolarized, the leg is in the swing (protraction) phase. The third condition in eqn 2 means that the stepping movement is in the stance phase. Obviously, during stepping: t 6 [½t stop ,t start . The lack of residual stiffness in the swing phase increases the speed of the leg movement, in accordance with the experimental findings [17,19,21]. In the model, we lumped together CI1, which innervates five of the six muscle types, CI2 and CI3, which innervate the flexor tibiae muscle. This simplification is not likely to cause noticeable errors in the simulations, since, as experiments showed, the activities of CI1, CI2 and CI3 are synchronous with good approximation [18,21]. , which convey neuronal signals to the CPG from sense organs of other joints of the same leg, or possibly of other legs. g app1 , g app2 are conductances of the driving currents to C1 and C2, respectively. g d1 -g d4 are conductances of the inhibitory currents to IN1-IN4, respectively. g b is the conductance of the sensory input current from the levator-depressor muscle system. The levator-depressor (LD) neuro-muscular system consists of a central pattern generator: CPG, slow and fast levator and depressor muscles as The various sensory, in particular position (angle) and (angular) velocity, signals were also implicitly modelled. The sense organs that produce such signals do exist for each joint of each leg in insects [6,8]; for a survey, see [7]. In the model, these signals were treated as physical quantities and were not encoded in neuronal signals. Their effects were therefore implemented as abstract logical decisions or operations (see Results below). Nevertheless, we have endeavoured to make these effects and functions physiologically viable by providing a putative neuronal network that, at least qualitatively, is capable of reproducing them (cf. Discussion).
A technical note on the usage of muscle fibre recruitment in the model If the recruitment levels of the slow muscle fibres and the contraction forces in the recruited fibres of a pair of antagonistic muscles are given, they determine a unique stationary position of the leg joint (cf. the accompanying paper [34]). It is however more practical to define the stationary angle (e.g. the horizontal position of the femur by the angle a) and to determine the recruitment levels in the appropriate muscles (e.g. in the protractor and retractor muscles).
Here is a short description of the calculation of the angle a~a 0 , which yields the horizontal resting position of the femur. At the stationary angle a 0 , we have [14,34] a~k R,eff k P,eff~l R (a 0 )(l P (a 0 ){l P,min ) 2 l P (a 0 )(l R (a 0 ){l R,min ) 2 ð3Þ where k eff (omitting here and in the subsequent formulae the subscripts indicating the specific muscle type for the sake of simplicity) is the so-called effective spring constant, which is the value corrected for the actual level of recruitment r a ; l(a) is the muscle length at the angle a, l min denotes the minimal length of the corresponding muscle (at which the muscle is completely relaxed). Now, k eff~ra k 0 =r c , where r c is the control (or reference) recruitment level, and k 0 is computed in the model for the recruitment level r c in the first place (cf. the accompanying paper [34]). In the stationary state, we have k 0~kres , the latter being the residual value of the spring constant k. Hence, a~k R,res k P,res r aR r cR r cP r aP ð4Þ according to [34]. From this eqn, we obtain r aR r aP~a k P,res k R,res r cR r cP~a ð5Þ If av1, we set r aP~rcP and r aR~a r cP . If aw1, then analogously r aP~raR =a~r cR =a. (At a~1 trivially r aP~rcP and r aR~rcR .) When the stick insect is standing in the usual (normal) position (cf. Fig. 2), then no calculation of the recruitment levels is needed for the other two muscle pairs. The stationary values of the two corresponding angles b and c are namely predefined by the normal standing position. That is, we have b&30 o and c&84 o . It is, of course, not difficult to apply the reasoning we have just used for the protractor-retractor system to the other two neuromuscular systems, if required.
Results
In the model, we assumed that sensory signals reflecting the positions (angles) and the (angular) velocities of the joints of the middle leg were used to coordinate the movements of the femur and tibia both at stopping and starting the locomotion (stepping). According to our hypothesis, both processes were triggered by mutually exclusive command signals of central origin. First, we deal with the stop of stepping.
Stop of stepping
We recall the definitions of the angles a, b and c [13,14]: a increases during retraction, i.e. moving the femur backward in the horizontal plane; b increases during levation, i.e. lifting the femur off the ground; and c increases during flexion, i.e. moving the tibia towards the body in the vertical plane.
With these conventions in mind, we assumed that the stopping process must start with a and c both in the decreasing phase, i.e. with protraction and extension of the femur and the tibia, respectively. This constraint ensured that the leg movement never finished with a protraction in the stance phase. Such a situation was never observed in experiments ( [1], and Grabowska, unpublished data). It is however self evident that the last phase before steady state is always a stance phase [1], since ground contact must have been established before, or at, a complete still stand. This means that the central command signal to stop became indicated (slow depr. m. etc.), the corresponding motoneurons: MN(DS) etc., 4 inhibitory interneurons (IN7-IN10) connecting the CPG to the motoneurons, and two additional interneurons (IN11-IN12), which convey neuronal signals to the CPG from peripheral sense organs of the leg (ramp symbol). g app3 , g app4 are conductances of the driving currents to C3 and C4, respectively. g d7 -g d10 are conductances of the inhibitory currents to IN7-IN10, respectively. The extensor-flexor (EF) neuro-muscular system consists of a central pattern generator: CPG, slow and fast extensor and flexor muscles as indicated (slow ext. m. etc.), the corresponding motoneurons: MN(ES) etc., 4 inhibitory interneurons (IN13-IN16) connecting the CPG to the motoneurons, and two additional interneurons (IN17-IN18), which convey neuronal signals to the CPG from sense organs of other joints of the same leg, or possibly of other legs. g app5 , g app6 are conductances of the driving currents to C5 and C6, respectively. g d13 -g d16 are conductances of the inhibitory currents to IN13-IN16, respectively. g b is the conductance of the sensory input current from the levator-depressor muscle system. g MN is the conductance of the common (central) input current to all motoneurons. Empty triangles are excitatory synapses; filled black circles on neurons are inhibitory synapses. The tiny black circles on synaptic paths are branching points. The three neuro-muscular systems have been put in frames with reference to Fig. 7 for the sake of their easier identification in that figure. doi:10.1371/journal.pone.0078246.g001 for i~1,3 (cf. Fig. 1). In these eqns, g di,n~1 :6nS is the value of g di during stepping ('normal' value), and g di,red~1 :0nS is its value during steady state ('reduced' value). c STOP is the boolean variable for the stop command, and '^' is the boolean 'and' operator. The last inequality condition in eqn 6 means that the switch in the value of g di occurs only once while the stop command is on. If we set i~13,15, substitute c for a in eqn 6, we obtain a completely analogous condition for stopping the fast MNs of the extensorflexor neuro-muscular system. Once these conditions were met, the activity of the fast MNs was inhibited via the inhibitory INs. Consequently, the fast muscle fibres in all four muscles: protractor, retractor and extensor, flexor relaxed and did not exert any torque on the thorax-coxa and femur-tibia joints, respectively. As far as the corresponding slow MNs are concerned, their activity was enhanced upon onset of the stop command due to the inhibition of their inhibitory INs: Again, expressing this in form of equations, we have for i~2,4 and i~14,16 in the protractor-retractor and the extensor-flexor neuro-muscular system, respectively. In eqn 8, g di,incr~1 0nS is the enhanced conductance of the (central) inhibitory current to the IN (cf. Fig. 1). The disinhibition of the corresponding slow MNs resulted, of course, in stronger activity of the slow muscles innervated by these MNs. The negative angular velocities above were preconditions for establishing and maintaining permanent ground contact by the levator-depressor neuro-muscular system. However, they had to be complemented by one condition on the angular signal b, which had to reach the value at ground contact (&30 o ) in order that both fast and slow levator and depressor MNs became inhibited. In equation form, these conditions read g di~gdi,n during stepping ð11Þ for i~7,8,9,10, j~1,3, and k~13,15; Boole is a boolean variable, and e is a sufficiently small value (e~1 o ). Thus, once the conditions in eqn 10 were fulfilled, a permanent ground contact was established. The (middle) leg came to a complete rest, having performed a retraction in the stance phase, while approaching the stationary angle a 0 . The activity of slow muscle fibres of both the protractor-retractor and the extensor-flexor system was only stopped when the angles a and c were close enough (+3 o ) to their stationary values (a 0 and c 0~8 4 o ) and the angular velocities da=dt and dc=dt were nearly zero. These conditions, like the previous ones, can also be expressed as equations.
g di~gdi,n during stepping ð13Þ for i~2,4 (cf. Fig. 1). Here, e 1 and e 3 are small values (3 o and 0:5s {1 , respectively). Setting i~14,16 and substituting c for a, we obtain analogous conditions for stopping the activity of the slow MNs of the extensor-flexor neuro-muscular system. Fig. 3 illustrates the result of these interacting processes. The top panel shows the three angular movements a(t) (red), b(t) (black), and c(t) (green), and their phase relations. The arrow indicates the central stop command. As it arrives at the start of a retraction (stance) phase, this phase of the stepping cycle is completed before the stopping process becomes effective at the beginning of the next protraction (swing) phase. Then a permanent ground contact is established (black trace). Finally, the complete steady state is preceded by a retraction (stance) phase. The middle and the bottom panel display intracellular activity of the slow and the fast protractor MNs, respectively. Note that whereas the fast MNs are rhythmically active, the slow ones show tonic activity, especially after the inhibition of the fast MNs. This happens, because we additionally assumed that, in the model, the central stop command also enhanced the activity of the slow protractor, rectractor, extensor and flexor MNs (via inhibition of their inhibitory INs by increasing the corresponding g d values, see eqn 8 above and Fig. 1). There is only weak indirect experimental evidence for the prolonged tonic activity of the slow MNs. Records in [10] show tonic activity of the slow extensor MN (SETi) in the stick insect and in the locust in the absence of the fast MN activity at fixed middle leg, that is a prolonged tonic activity of the slow MNs can be seen. The process of stopping described above in detail is independent of the steady-state angle a 0 to which a converges. Fig. 4 demonstrates this result. The different steady-state values of a were set by making use of the recruitment property of the model (cf. Methods and [34]). The steady-state values of b (b 0~3 0 o ) and c (c 0~8 4 o ) were always the same since the projection of the standing position into the vertical plane is unchanged in all cases (cf. Fig. 2). The cases depicted in Fig. 4 do have experimental correlates. In [1], they correspond well to the steady-state horizontal position of the femur of the front, middle and hind leg, respectively (cf. [1], Fig. 2). Moreover, some unpublished observations (Grabowska, unpublished data) also appear to be in agreement with our simulation results. Fig. 5 demonstrates an important putative physiological role, as mimicked by the model, of the slow muscle fibres and of the slow MNs innervating them during the stopping process. It shows that the convergence of the angle a to its steady-state value is markedly faster when the slow protractor and retractor MNs exhibit prolonged and enhanced tonic firing activity (compare the corresponding angular movements: red, blue in Fig. 5). Although this result is shown here for the protractor-retractor neuromuscular system only, it is also valid for the extensor-flexor neuromuscular system. That is, the slow MNs of the latter system work according to the same time schedule during the stopping process as their counterparts in the protractor-retractor system. However, the situation is different in the levator-depressor neuro-muscular system. We found in the simulations that when the slow MNs of this system were tonically firing, a stable and permanent ground contact in the stance phases could not be produced because of the tonic firing of these very MNs. Hence, we attributed the slow MNs in the levator-depressor system the same rhythmic activity as that of the fast MNs, i.e we made no difference between the activities of the slow and fast MNs.
Finally, we show in Fig. 6 that the behaviour of the system remains basically unaffected by the time of onset of the central stop command. In the simulations, we chose a number of different occurrence times of the central stop command that fell within the same stepping cycle (of &500 ms), i.e. occurred at different phases of that cycle, in each case. Every time, the levator-depressor system behaved exactly the same way, whereas the protractor-retractor system showed qualitatively the same behaviour by performing a retraction (in the stance phase) before reaching its stationary position (the angle a approaching its steady-state value a 0 from below). The extensor-flexor system, however, did not always exhibit exactly the same behaviour. Occasionally, as shown in the bottom panel of Fig. 6 (green trace), it carried out first a flexion followed by a partial extension during which the angle c monotonically converged to its steady-state value. This happened because the central stop command arrived just when the flexion started. Note that the extension during stepping finishes a bit earlier than the corresponding protraction (Fig. 6), like in the stick insect [15]. This is why the protractor-retractor system is still in the protraction phase when the extensor-flexor system has already completed its extension phase.
Start of stepping
Theoretically, there are two possible ways of starting from a standing, stationary position: performing first a swing phase (protraction) or a stance phase (retraction). We implemented both possibilities in the model. Indeed, the stick insect seems to make use of both of these possibilities [1] and (Grabowska, unpublished data). Whether a swing phase or a stance phase would be produced first by a (middle) leg depends on the actual steady-state value of a in the standing position. In the model, we defined a critical angle that separates these two cases. We set its value to be 60 o . In the extensor-flexor system of the stick insect, it was found that the relative frequency of the middle leg starting with a swing or a stance phase monotonically depended on the angle c [23]. Our assumption of a critical a angle seems therefore reasonable, even though the choice of the specific value of 60 o remains somewhat arbitrary.
The simulations show examples for both cases. In Fig. 4, top panel in which a 0~4 0 o , the protractor-retractor system starts with a retraction, accompanied by a flexion in the extensor-flexor system, while the leg has still ground contact (b~30 o ), i.e. the leg is in the stance phase. Only when both retraction and flexion are completed will the levator-depressor system be enabled to lift the leg off the ground. This appears to be in good agreement with findings in [1] and (Grabowska, unpublished data). In the model, the stepping starts at a well-defined phase of the activity of the protractor-retractor CPG (C1 and C2 in Fig. 1): when the ''retractor'' CPG neuron (C1 in Fig. 1) is about to reach its maximal depolarization by exceeding a threshold value V thr~1 5 mV. Note that all three CPGs are active and synchronized, because the load and position sensory signals represented by the angle b are active, i.e. b is below its critical value (38 o for the protractor-retractor system, and 50 o for the extensor-flexor system; cf. Fig. 1 and Methods). Summarizing the starting conditions in this case in form of equations, we have g di~gdi,n if (c START~T RUE)^(V C1 (t)wV thr )^(g di =g di,n ) ð14Þ g dj~gdj,n if (g di~gdi,n )^(g dj =g dj,n ) ð15Þ g dk~gdk,n if (g di~gdi,n )^(g dj~gdj,n )^(g dk =g dk,n ) ð16Þ for i~1,2,3,4, j~13,14,15,16, and k~7, 8,9,10. The boolean variable c START represents the central start command. These eqns also show that after the protractor-retractor system is activated, in a retraction (stance) phase, the extensor-flexor neuro-muscular system is activated next with a flexion, provided that the protractor-retraction system is already active. The levator-depressor neuro-muscular system is finally activated, once both of the other systems are already active. Because of the synchronization between the three CPGs, the leg's next lift-off will only occur when the retraction and the flexion have been completed. Note that in all three systems the slow and fast MNs are activated at the same time.
In the two other panels of Fig. 4 in which a 0 w60 o , the stepping starts with lifting the leg off the ground. This kind of start was also observed in the animal [1] and (Grabowska, unpublished data). In the model, the mechanical movement is triggered when the ''levator'' CPG neuron (C3 in Fig. 1) reaches the depolarization level of V thr~1 5 mV. The corresponding equation reads for i~7, 8,9,10. Then the angle b increases by activation of the levator muscles in the levator-depressor system. The levation induces protraction and extension, respectively, in the two other neuro-muscular systems via the load and position sensory signals represented by the angle b (Fig. 1). Here, too, in all neuro-muscular systems, the slow MNs and muscles are simultaneously activated with the fast ones. Note that the extension starts later than the protraction due to the different thresholds of b in the two systems (cf. Methods and [13]).
Discussion
In this paper, we set out to provide an explanation of how locomotion in the stick insect is stopped and started by making use of a suitable neuro-mechanical model. This model is based on our one in the accompanying paper [34]. We complemented it by making physiologically plausible assumptions with regard to the stopping and starting of stepping. Among these assumption two are of fundamental importance. The first one is on the coordinating roles of sensory signals reflecting position and velocity of the leg's main joints. The existence of such sensory signals has been well established in experiments [7][8][9]. Our assumption, more precisely a set of assumptions, relates to their crucial role in coordinating the activities of the leg's neuromuscular systems during locomotion, standing, and during the transition processes between these states. These sensory signals can thus trigger sub-processes in the same or in a different neuromuscular system. For example, the negative value of the angular velocity da=dt results in the inhibition of the fast MNs in the protractor-retractor neuro-muscular system at stopping; the activity of the levator-depressor can only be blocked at stopping, once the fast MNs of both the protractor-retractor and the extensor-flexor system have been inhibited. A model not incorporating the specific set of assumptions we used here would have to have quite different properties and mechanisms of intra- Figure 7. The physiological viability of the model. Active pathways at stopping. The three large boxes in the centre of the figure are the three neuro-muscular systems as indicated from Fig. 1. Their existence is physiologically well established [15,27]. Their inner structure has been partially confirmed in experiments [10,11,24,[27][28][29][30][31]. The small boxes CC and S1-S6 are CPG-like functional units consisting of mutually inhibitory neurons. Such small networks of mutually inhibitory neurons do exist in the nervous system of athropods [32] and other animals [33]. However, such units with the specific roles assigned to them in the present model have not been identified. Triangle CI represents the common inhibitory motoneuron. Its existence is proven in experiments [17][18][19]21]. N1-N5 are neurons whose existence with the specific function they have in the present model has not yet been established. Pentagons a, b, c are sense organs where signals reflecting the position (angle) are generated; pentagons da, dc are sense organs where signals reflecting the (angular) velocity are generated. Such sense organs do exist in insects (e.g. [6][7][8]). Small empty triangle on neurons or on system boxes are excitatory synapses. Filled black squares symbolize inhibition by CI of the slow muscle fibres. Filled black circles are inhibitory synapses. Very small black circles along paths (neuronal connections) are branching points. Red pathways are the connections that are active at stop of stepping. Grey pathways are the connections that are active at start of stepping (see. Figs. [8][9]. For more explanation as to the workings of this network, see text. doi:10.1371/journal.pone.0078246.g007 leg coordination. The second fundamental assumption was not made explicitly, since we used it in our earlier work [13,14]. Its main claim is that neither stopping nor starting of stepping requires any direct interference with the activity of any of the three CPGs. As it should be clear from the Results, and the eqns describing the actions of the sensory signals, the driving currents (the conductances g appk ) to the CPGs were at no point changed. That means that the CPGs ran autonomously all the time without direct interference. All actions on the MNs, hence on muscles, were conveyed via the inhibitory INs connecting the CPGs to the MNs. Thus all transient signals that would arise by directly changing the activity of the CPGs (i.e. changing the conductances g appk ) could be avoided. Equipped with this property, our model is capable of responding to unexpected changes (e.g. sudden stop or start) much faster than models in which the activity of the CPGs is directly manipulated. Quite recently, experimental evidence has arisen that the CPG is not affected during change in the MN activity. It has been shown by Rosenbaum et al. (unpublished results) that the activity of the protractor-retractor CPG remains unaffected during a switch from forward to backward walking of the stick insect when, however, the protractor and retractor muscles, hence MNs, exchange roles. We also successfully modelled this phenomenon the same way, i.e. without changing the activity of the protractor-retractor CPG in the model [14].
We also assigned the slow muscle fibres residual stiffness and took the effects of the common inhibitory MNs on the function of slow muscle fibres into account [17][18][19]21]. Having incorporated these properties into the new version of the model, we could mimic both stopping and starting of locomotion (stepping). It turned out that both processes, though basic elements of locomotion, required precise coordination of the three main neuro-muscular systems of a leg. The simulations with the model highlighted the details of the coordination between those neuro-muscular systems. The constituent processes are physiologically plausible. They are i) sensory signals representing position (angle) or (angular) velocity at the individual joints as well as load signals, represented by the angle b); ii) neuronal signals driving the MNs of the system; iii) activities of the MNs driving the fast or slow muscles they innervate; and iv) contractions of the slow and fast muscles producing the mechanical movement of the individual leg joints. Gradual muscle recruitment in the model ensures that any stationary position (angle a) of the femur can be attained. The steady state, posture, is Fig. 7: Here, the pathways that are active when the starting process commences with a protraction (swing phase) are highlighted in green. The other pathways are suppressed by having been drawn in grey. For explanation of how the network works in this case, see text. doi:10.1371/journal.pone.0078246.g008 maintained in the model by the residual stiffness of the slow muscle fibres, in agreement with experimental findings [16,17]. This property of the model does not contradict the result in the accompanying paper [34] that steady state is maintained by enhanced tonic firing of the slow MNs, hence co-contractions of the slow muscle fibres in the protractor-retractor system. There, the slow muscle fibres did not possess any residual stiffness, and the co-contraction of the slow muscle fibres was thus required to maintain a stable steady state. In reality, both possibilities are likely to exist and co-exist [10,16].
There are important experimental results that underpin the assumption made when constructing the model. First of all, it is well established that slow and fast muscle fibres have their dedicated MNs driving them, and that the fast muscle fibres do not contribute to maintaining the steady state (posture) in insects [10,11]. It has also been experimentally confirmed that there are position-, velocity-and load-dependent sensory signals in the animals [2,[7][8][9]. The existence of residual stiffness in the slow muscle fibres, and the related function of the common inhibitory MNs (CI1-CI3) to remove it, was demonstrated [10,11,[17][18][19]21]. It was also observed that stop of stepping does not occur randomly during a step cycle, and a leg grinds to a halt when it has ground contact, i.e. in the stance phase [1]. The model presented in this paper incorporates all these properties, and, as far as we are aware of, it is the first one to do so.
The model behaviour is in good agreement with the experimental findings. In the simulations, the time evolution of the angles at the leg joints produced coordinated movement of the whole leg both at stop and start of stepping, similar to the observed ones in the stick insect [1] (Grabowska, unpublished data). In particular, our model is capable of reproducing the preferred steady-state positions of the front, middle and hind leg in the stick insect: the front leg (femur) being near to its anterior extreme position, the hind leg near to its posterior extreme position, and the middle leg somewhere between them. (Compare the three panels in our Fig. 4 with Fig. 2 in [1].) The fast MNs fire rhythmical bursts during locomotion due to their rhythmic inhibition by the associated CPG [24]. The slow MNs, by contrast, produce tonic firing, at least in the case of the extensor muscle [10]. In the model, we applied these firing modes both in the protractor-retractor and the extensor-flexor system. In the levator-depressor system, however, we found that tonic firing of the slow MNs resulted in incomplete ground contact in the stance phases. In the levator-depressor system, we therefore implemented identical rhythmic firing of both the fast and the slow MNs. The experimental evidence, we could find is inconclusive as to the activity of the slow MNs in the levator-depressor system [25,26]. Hence this part of the model remains highly hypothetical. However, we succeeded in identifying a putative physiological role of the slow muscle fibres at stopping. The prolonged and enhanced tonic activity of the slow protractor and retractor MNs driving the corresponding slow muscle fibres substantially accelerated convergence of the angle a to its stationary value (Fig. 5). The same is true for the slow MNs and muscle fibres of the extensor-flexor system. This is a hitherto unidentified function of the slow MNs and muscle fibres in both systems, and it is thus an important prediction of our model. It should be mentioned here that there is some uncertainty in the physiological interpretation of the apparent ''stop'' and ''start'' commands in the model. Currently, there is no experimental way to find out whether such single command signals are produced by the brain, and if so when and in which part of the brain they exactly arise.
As pointed out in the Methods and Results, the sensory signals representing position (angle) and (angular) velocity (e.g. a, da=dt) were not implemented in the model as neuronal signals but only formally, by computing their effects on the neuronal and muscular activity in an abstract way. This is also true for the role of the common inhibitor MNs CI1-CI3. Their effect: removing the residual stiffness of the slow muscle fibres during locomotion and restoring it at stopping was implemented only, as an abstract mathematical function obeying logical conditions.
Physiological viability of the model
Now, we show that our model is, at least in qualitative terms, physiologically viable. We present, in Figs. 7-9, a neuronal network that could be regarded as a physiologically meaningful implementation of our model. The network structure and its elements are not entirely hypothetical. The existence of the three main building blocks (neuro-muscular systems) has reliably been established in experiments [15,27]. Basic properties of the local networks of the neuro-muscular systems could also be gleaned from experimental data [10,11,24,[27][28][29][30][31]. In addition, it is known that pairs of mutually inhibitory (nonspiking) neurons or groups of neurons capable of exhibiting bistability do exist in the nervous system of athropods [32] and other animals [33]. All these facts lend the network in Fig. 7 a sufficient physiological basis.
We did not implement this network as a quantitative model to be used in simulations, since there are too many unknown parameter values, especially synaptic weights, in it, hence the attribute ''qualitative''. Nevertheless, we believe that presenting it does make a contribution to bringing an abstract model and a physiological system closer to each other, and to furthering a better understanding of how the model works.
Here we give an account of how the network works at stop and start of locomotion (stepping), respectively. The stop is initiated by a central stop command: red path labelled with ''stop'' into unit CC. When the stop command arrives CC, it inhibits its alternative, the ''start'', command (green neuron). The stop command enhances the activity of the slow MNs in the protractor-retractor and the extensor-flexor system (see red pathways starting at the red neuron in unit CC and ending in red empty triangles on the boxes representing these two systems). The fast MNs of the protractorretractor system are inhibited, the activation of the fast muscle fibres innervated by them stopped, as soon as da=dtv0 (cf. Results), because the red neuron in S2 is activated and sends an inhibitory signal to the MNs of the fast protractor-retractor system.
It also inhibits the other (magenta) neuron in S2. An analogous cascade of events takes place in the extensor-flexor system (unit S4). Now, if the red neurons in the units S2 and S4 are excited, they send a signal each to their red counterpart in unit S3. In order to enter the excited state, the red neuron in S3 also needs an excitation from the sense organ sensing position (angle b) in the levator-depressor system. This sensory signal is evoked, if b is in a close neighbourhood of its steady-state value (30 o ). Then the red neuron in S3 is activated and inhibits both the fast and slow MNs of the levator-depressor system. By this, a permanent ground contact of the leg is established. Note that the slow MNs in both the protractor-retractor and the extensor-flexor system are still active. Moreover, their activity has been boosted by the stop command. The slow MNs of the protractor-retractor system are inhibited as soon as the red neuron in S5 is activated. For that to happen, it is required that the angle a be close to its steady-state value, and that da=dt&0. If this is the case, the red neuron in S5 will receive an excitatory signal originating in the sense organ that signals the value of a, and the inhibition coming from the sense organ signalling the value of da=dt will vanish. Thus the red neuron in S5 becomes excited and sends an excitatory signal to neuron N4, which has already been receiving another excitatory signal from S2. The latter has, on its own, been insufficient to activate neuron N4. Now, it is activated and it inhibits the slow MNs of the protractor-retractor system. The mechanical movement of the femur grinds to a halt. The events in the extensorflexor system are completely analogous. By inhibiting all MNs in each of the three systems, the leg stops moving. The central stop command inhibits the common inhibitor CI directly, and the excitatory signal to it from unit CC also ceases. This abolishes the inhibition by CI of the slow muscle fibres in all muscles. That is, residual stiffness in the slow muscle fibres becomes active, and it determines the stationary position of the leg. Of course, as soon as the red neurons of the units S2-S6 (and CC) become activated, they inhibit their magenta counterparts in the same unit strongly enough to prevent any signal flow through them.
The start of stepping can occur in the model in two different ways: the leg starting with a swing phase or a stance phase. The steady-state value a 0 of the angle a determines which of the two cases will occur. In any case, the foremost action is that the central start command activates the 'start' neuron (green in Fig. 8 and blue in Fig. 9) in unit CC, hence inhibits the 'stop' (red) one whose excitatory output immediately ceases. Let us consider first the case when a 0 wa 0thr~6 0 o . That is, the stepping will start with a swing phase. For this variant, the green pathways in Fig. 8 will be used. Because of the condition for a 0 above, the position signal for a (cf. Fig. 8) excites the green neuron in unit S1, together with the start signal, and inhibits neuron N1. In turn, the blue neuron in S1 is inhibited, and, as a consequence, all blue pathways will be blocked (Fig. 9). The green neuron in S1 then excites the neurons N2 and N3 enabling the excitatory pathways through them to become active. The pathway through N3 from the CPG conveys an excitatory signal to the magenta neuron in S3 when the CPG of the levator-depressor system starts the levation phase. This excitatory signal and the start signal conveyed by the green neuron in unit S1 activate the magenta neuron in unit S3, which inhibits the red neuron in the same unit. This, in turn, abolishes the inhibition of both the fast and slow MNs of the levatordepressor system. Since the CPG is in the levation phase, the angle b increases, i.e. the leg is lifted off the ground. To transmit this information to the other neuro-muscular systems, an excitatory signal is sent via neuron N2 to the magenta neurons in the units S2, S4, and S5, S6. The magenta neurons in these units become activated, hence their counterparts (the red neurons) in the same unit inhibited. Accordingly, the inhibitory signals to the fast MNs disappear directly, whereas the inhibition to the slow MNs indirectly by deactivation of the neurons N4 and N5, respectively. By virtue of the sensory signals represented by the angle b (cf. Fig. 1), the CPGs of the protractor-retractor system and the extensor-flexor system are synchronized to that of the levator depressor system [13,14,34]. The stepping therefore starts with a swing phase (see, for example, Fig. 3 or Fig. 6).
The alternative to this process is one with an initial stance phase. An example is displayed in the top panel of Fig. 4. This process uses the blue pathways in Fig. 9. In this case, the retractor phase of the CPG of the protractor-retractor system triggers the start of the stepping. Since there is now no a-signal to S1 and to neuron N1, the latter is activated, and it activates, together with the central start command, the blue neuron in unit S1. This neuron, in turn, inhibits the neurons N2 and N3, hence no excitatory signal from the b-sensor and the CPG of the levatordepressor system can reach the magenta neurons in units S2 and S4. (The b-signal to unit S3 is irrelevant in this case, because it is present only at the tarsus reaching the ground.) Instead, excitatory signals from S1 arrive these (magenta) neurons, activate them and inhibit the corresponding red neurons simultaneously. This is also true for units S5 and S6, which control the slow MNs in the protractor-retractor and the extensor-flexor system (cf. direct blue pathways to these units from unit S1). The red neurons will also be inhibited in units S5 and S6. This means that the inhibition of both the fast and the slow MNs will be abolished in both the protractor-retractor and the extensor-flexor neuro-muscular system. Since the levator-depressor system is still inhibited, ground contact is maintained, and because the protractor-retractor CPG is in the retractor phase, a stance phase of the stepping cycle ensues. The magenta neuron in unit S3 is activated by simultaneous excitatory signals from S1 and S4, the latter belonging to the extensor-flexor system. The red neuron in S3 is now inhibited, and the inhibition of the fast and slow MNs of the levator-depressor system is abolished. Thus the next swing phase can commence when the CPG of the levator-depressor system reaches the next levation phase of the stepping cycle. Irrespective of whether the stepping starts with a swing or a stance phase, the start signal immediately activates the common inhibitor MN (CI in Figs. [8][9] by inhibiting the red neuron in CC, which has an inhibitory effect on CI and, at the same time sending a permanent excitatory signal to CI (Figs. 8-9).
Conclusions
Having shown that our model is physiologically viable, i.e. it can, at least qualitatively, be implemented by using a neuronal network with physiologically realistic neurons and synapses, we should like to argue that it is physiologically relevant, too. First, it was constructed by using experimental results and physiologically reasonable assumptions. The signals in it are of neuronal origin. Thus there exists a close correspondence between the model and its biological counterpart at several levels of complexity. This makes the interpretation of the simulation results easier and more plausible. Second, our model, in contrast to earlier ones, allows functional differentiation between static and dynamic aspects of movement control. Third, even though our model has been constructed by using experimental findings from the stick insect, the main result achieved with it, namely showing of how intra-leg coordination is organized during stop and start of locomotion (stepping) may be generalized to elucidate analogous processes in other insect species, too. We even venture to suggest that some details of the model could perhaps be used in constructing insectlike robots. In this sense our model might attain a more general relevance and, maybe, significance than just relating to physiological processes in the stick insect. | 2016-05-12T22:15:10.714Z | 2013-11-22T00:00:00.000 | {
"year": 2013,
"sha1": "39d33c15e9670e24b73265a2b9426f7679850d9d",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0078246&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "39d33c15e9670e24b73265a2b9426f7679850d9d",
"s2fieldsofstudy": [
"Biology",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
33040243 | pes2o/s2orc | v3-fos-license | The Distant Siblings—A Phylogenomic Roadmap Illuminates the Origins of Extant Diversity in Fungal Aromatic Polyketide Biosynthesis
Abstract In recent years, the influx of newly sequenced fungal genomes has enabled sampling of secondary metabolite biosynthesis on an unprecedented scale. However, explanations of extant diversity which take into account both large-scale phylogeny reconstructions and knowledge gained from multiple genome projects are still lacking. We analyzed the evolutionary sources of genetic diversity in aromatic polyketide biosynthesis in over 100 model fungal genomes. By reconciling the history of over 400 nonreducing polyketide synthases (NR-PKSs) with corresponding species history, we demonstrate that extant fungal NR-PKSs are clades of distant siblings, originating from a burst of duplications in early Pezizomycotina and thinned by extensive losses. The capability of higher fungi to biosynthesize the simplest precursor molecule (orsellinic acid) is highlighted as an ancestral trait underlying biosynthesis of aromatic compounds. This base activity was modified during early evolution of filamentous fungi, toward divergent reaction schemes associated with biosynthesis of, for example, aflatoxins and fusarubins (C4–C9 cyclization) or various anthraquinone derivatives (C6–C11 cyclization). The functional plasticity is further shown to have been supplemented by modularization of domain architecture into discrete pieces (conserved splice junctions within product template domain), as well as tight linkage of key accessory enzyme families and divergence in employed transcriptional factors. Although the majority of discord between species and gene history is explained by ancient duplications, this landscape has been altered by more recent duplications, as well as multiple horizontal gene transfers. The 25 detected transfers include previously undescribed events leading to emergence of, for example, fusarubin biosynthesis in Fusarium genus. Both the underlying data and the results of present analysis (including alternative scenarios revealed by sampling multiple reconciliation optima) are maintained as a freely available web-based resource: http://cropnet.pl/metasites/sekmet/nrpks_2014.
Introduction
The last 5 years of fungal genomics have been fruitful. We see mass characterization of fungal genomes, with increasing coverage given to both economically important and evolutionarily divergent lineages of eukaryotic microorganisms (Grigoriev et al. 2012;http://1000http:// .fungalgenomes.org, last accessed May 30, 2015. Availability of next-generation sequencing methods sped up the pace of novel genomic projects, whereas the increased awareness and use of deletion libraries and heterologous expression systems allow characterization of larger sets of genes (e.g., Bergmann et al. 2007;Nielsen et al. 2011;Ahuja et al. 2012). However, our knowledge of evolutionary basis underlying the extant genetic diversity is still incomplete.
In particular, the origins of diversity in fungal polyketide synthesis have been the subject of intense investigations, even prior to the advent of whole-genome sequence analysis. Previously, the landmark paper by Kroken et al. (2003) has postulated that horizontal transfers are not required to explain major parts of observed variability in fungal iterative polyketide synthases. However, since then the increased sampling of genomic data from diverse taxonomic groups has enabled researchers from multiple groups to consider evidence in support of individual HGT (horizontal gene transfer) scenarios. Salient examples include the aflatoxin cluster (Slot and Rokas 2011), the fumonisin cluster (Khaldi and Wolfe 2011), as well as the ancient origin of partially reducing polyketide synthases from transfer originating in bacteria (Schmitt and Lumbsch 2009). Thus, the availability of data and advances in phylogenomics have permitted elucidation and study of individual scenarios contributing to the present-day diversity of fungal secondary metabolism. This "classic" approach to proposing and corroborating evolutionary scenarios typically (Schmitt and Lumbsch 2009;Khaldi and Wolfe 2011;Slot and Rokas 2011;Proctor et al. 2013) proceeds by thorough analysis of multiple individual reconstructions of gene histories in context of multiple reference species. The analysis is then followed by serial likelihood ratio tests aiming to partition the adjacent genes according to their associated phylogenies supporting one or more origin scenarios (e.g., duplication, horizontal transfer, and loss). The discussion of evolutionary events leading to extant diversity is thus limited to supporting the cases separately. Conversely, larger scale statistical modelling does not capture particular, individual events leading to known biosynthetic activities (e.g., CAFE analysis of gene family expansions/contractions in multiple lineages; Bie et al. 2006;Bushley and Turgeon 2010).
Consequently, what is still lacking are "phylogenetic roadmaps"-resources that strive to reconcile species and gene histories explicitly. These, by definition, will not provide a conclusive proof in favor of one or the other scenarios but can show the parts of family evolutionary history that are better explained with or without recourse to horizontal transfer versus duplication. A large scale inference followed by reconciliation can enumerate well-supported scenarios given different assumptions as to the cost of particular evolutionary events (i.e., transfer, duplication, loss). Based on maximum parsimony reconciliations we can identify monophyletic clades clearly dominated by speciation events (where orthologous relationships likely point to a conserved chemical structure of the base polyketide chain), as well as indicate potential transfers and pinpoint alternative donor-recipient scenarios. Fast reconciliation methods are now available (Doyon et al. 2011;Bansal et al. 2012), with scalable algorithms for efficiently sampling the space of optimal solutions at random (Bansal et al. 2013), as well as for finding the most likely scenario in a probabilistic framework (ALE, Amalgamated Likelihood Estimation; Szollosi, Rosikiewicz, et al. 2013).
Our work's goal was to provide such a "proof-of-concept" roadmap for a large set of nonreducing polyketide synthases (NR-PKSs), involved in biosynthesis of diverse, well-characterized aromatic compounds (e.g., Cox 2007;Sanchez et al. 2012). This reconstruction and reconciliation was further supplemented by statistical analysis of genomic context and gene structure. For this, it was necessary to consider tradeoffs between scalability and accuracy of methods-choosing methods which produce best results in reasonable computation time (Anisimova et al. 2013) and provide a well-laid foundation for continuous inclusion of new (or corrected) data.
Materials and Methods
The entire workflow of conducted analysis is summarized on figure 1. Individual stages are described in detail below. All of the sequences and other supplementary material (i.e., trees, alignments, and annotations) referenced in this study can be found in the supplementary material, Supplementary Material online, as well as on the MetaSites database website, maintained by the authors: http://cropnet.pl/metasites/sekmet/ nrpks_2014.
Preparation of Reference Genomes
Genomes of 149 model fungi were gathered from the National Center for Biotechnology Information (NCBI)/ GenBank (Benson et al. 2015), JGI-DOE/MycoCosm (Grigoriev et al. 2012), and Ensembl/Fungi (Flicek et al. 2014) repositories. This reference set has also included the PKS lacking genomes (mainly Saccharomycetes and Schizosaccharomycetes). The genome sequence of Blumeria graminis f. sp. hordei was taken from BluGen powder mildew genome database (http://www.blugen.org; Spanu et al. 2010, last accessed January 4, 2015. The complete list of genomes and data sources can be found in the supplementary table S1, Supplementary Material online. All of the reference genomes were curated to ensure that reading frames and gene structure matched the recorded protein translation for all proteincoding genes.
Annotation of Domain Architecture
All protein-coding genes within the reference genomes were annotated with Hidden Markov Models, (HMMs) from Pfam database version 27 (Finn et al. 2014). Additionally, the mixed kingdom (Foerstner et al. 2008) and the fungi-specific (Delgado et al. 2012) models were created on basis of the alignments supplied with the original papers. These covered the typical NR-PKS domains (ketoacyl synthase, starter and main acyl transferase, thioesterase, methyltransferase, phosphopanthotein binding site, NAD-binding reductase). The HMM searches were conducted with the HMMer 3.0 hmmscan program (Eddy 2009) and were followed by double checking against the web-based version of NCBI Conserved Domain Database (Marchler-Bauer et al. 2015). The product template (PT) domain assignment was based solely on checking against hot dog fold superfamily signature in NCBI/CDD.
Prescreening for NR-PKS Subset within Reference Genomes
The selection of full NR-PKS complements from model fungal genomes was supported by three independent criteria: The clustering of ketoacyl synthase sequences based on all versus all Basic Local Alignment Search Tool (BLAST) comparisons (pairwise similarities between individual sequences), the
GBE
presence of architectural similarities (based on matching domain fingerprints), and inclusion in a monophyletic clade with all reference NR-PKS sequences during the preliminary phylogeny reconstruction step.
Initially, the set of ketoacyl synthases was selected from above-mentioned model genomes (including Caenorhabditis elegans as model outgroup), as well as PDB (Rose et al. 2013) and UniProt/SwissProt (UniProt Consortium 2014) sequences. The PDB (432 sequences) and SwissProt (611 sequences) subsets were included only during the clustering and initial selection of fungal PKS cluster, prior to preliminary phylogenetic tree reconstruction.
In the prescreening, only the sequences with at least one ketoacyl synthase domain occurrence (based on either fungal or mixed kingdom HMM matches, of length exceeding the threshold of 50 amino acids) were retained (3,067 in total). To further distinguish between the fragmentary sequences of incomplete PKSs (whether due to mistakes in gene prediction/ annotation or due to evolutionary events) and the loosely associated unrelated ketoacyl thiolases with a partial KS domain signature, we conducted unsupervised clustering analysis with CLANS (Frickey and Lupas 2004). Clustering used a 10 À5 inclusion threshold for hit P values. Based on membership of functionally and structurally annotated PKSs, the single cluster containing both C. elegans FAS and all reference PKS sequences (bacterial, fungal and protistan, 1,659 sequences) was chosen for further analysis.
To ensure the correct selection of the NR-PKS subset, we then employed an additional step based on the FastTree (Price et al. 2010;multithreaded version). One hundred bootstrap iterations were done on replicates obtained with SEQBOOT (Felsenstein 1989). This extended majority rule consensus tree was based solely on the ketoacyl synthase domain. MAFFT-LINSI v. 7.2 (Katoh and Standley 2013) was used to align the sequences, the resulting alignment was pruned with "trimal" (Capella-Gutié rrez et al. 2009) at 70% occupancy threshold for columns.
The final set of NR-PKS homologs was chosen by extracting the monophyletic fungal clade delineated by lowest common ancestor of all reference (experimentally validated) nonreducing polyketide synthases (node with full support from bootstrap). The 414 representatives from model fungal genomes (discarding PDB and SwissProt sequences, but including the fumonisin HR-PKS outgroup sequence) were retained for gene tree inference based on presence of conserved KS-AT module.
No outliers or additional sequences were indicated by either phylogeny or domain architecture, save for Ustilago maydis PKS1/UM04105 (and its two counterparts from Ustilago hordei and Sporisorium reilianum) which lacks an acyl transferase domain and thus was not included in the final analysis. For a single experimentally characterized NR-PKS dbaI (ANIA_07903; Gerke et al. 2012), the sequence was manually updated to match revised AspGD gene model which translates to a full length PKS protein instead of the truncated version.
Gene Tree Construction
The final NR-PKS alignment was constructed by aligning the core KS-AT module with MAFFT-LINSI (Katoh and Standley 2013). This alignment was curated with T-COFFEE transitive consistency score analysis (Chang et al. 2014; default exhaustive "proba_pair" setting-columns scoring 2 or above were kept in the final alignment, in keeping with the results of the above-mentioned article). The filtered alignment (available as supplementary material S2, Supplementary Material online, in NEXUS format; mappings of sequence identifiers to loci/gene names are available as additional supplementary file S3, Supplementary Material online) was used for final phylogeny reconstruction.
As the protein sequence identity levels can be low and homoplasies are expected in the data set, Bayesian inference (BI) was conducted with PhyloBayes-MPI 4 chains 200,000 each, first 40,000 trees discarded as burn-in, every fifth tree sampled; two chains of best convergence chosen for further analysis; CAT-Poisson parameterization was used due to computational resource constraints). Chain convergence was assessed with pairwise comparison using "bpcomp" tool (Lartillot et al. 2009). The analysis was also carried out in maximum-likelihood (ML) framework with IQTREE 0.9.6 (Minh et al. 2013; 1,000 ultrafast bootstrap replicates; exhaustive nearest neighbor search setting "-nni5"; LG+G4+I+F model selected with IQTREE internal model testing). The Bayesian consensus tree and the underlying sampled trees were selected for reconciliation analysis with ALEml and DTL-RANGER; the resulting ALEml tree is hereafter referred to as "amalgamated." Reconstructed gene tree topologies are available in supplementary material, Supplementary Material online (ML, Bayesian, and amalgamated in supplementary files S4-S6, Supplementary Material online). The amalgamated tree with annotated gene structure and domain architecture is shown in supplementary figure S1, Supplementary Material online.
Throughout the text, we refer to the nodes of the amalgamated gene tree by the "g<number>" notation derived from postorder traversal of tree structure. Where the original BI tree or the species tree is referenced, the analogous "go<number>" or "so<number>" notations are used.
Species Tree Construction
The species tree was constructed based on 23 best scoring single-copy orthologs (see supplementary table S2, Supplementary Material online) with best topological scores (over 95%) as reported in FUNYBASE (Marthey et al. 2008). The topological scores were previously introduced by Aguileta et al. (2008) and measured concordance of ML trees based on individual orthologs with a reference supertree inferred on basis of over 120 single copy orthologs identified in that study. The ortholog protein sequences were aligned with MAFFT-LINSI and resulting alignments concatenated (the concatenated alignment is available in the supplementary file S7, Supplementary Material online, in NEXUS format). The gapped positions were curated using 70% occupancy threshold (-gt 0.7) in "trimal," resulting in 15,314 columns in the final alignment (available as supplementary file S2, Supplementary Material online). BI was conducted with PhyloBayes-MPI 1.5a . The inference consisted of 4 chains of 40,000 iterations, first 5,000 trees from each chain were discarded as burn-in, every fifth tree was sampled. As in the case of gene tree reconstruction, two chains of best convergence were chosen for further analysis, CAT-Poisson parameterization was used due to computational resource constraints.
Additional ML reconstruction was conducted with IQTREE. Individual protein models were chosen for each of the 23 genes based on ProtTest v3 (Darriba et al. 2011) results (corroborated based on IQTREE's own model testing). The reconstruction used ultrafast bootstrap with default stopping criterion and exhaustive NNI (Nearest Neighbor Interchange) search (-nni5 option), resulting in convergence after 100 ultrafast bootstrap iterations. Caenorhabditis elegans represented the outgroup used to root the tree.
To obtain the approximate chronogram, the dating was performed on PhyloBayes-MPI consensus tree. Relaxed, log-normal autocorrelated clock with soft bounds under a birth-death prior was used (as implemented in PhyloBayes 3.3f; Lartillot et al. 2009). The dating constraints were introduced based on Fungi/Animalia split dated at 983 Ma (based on Douzery et al. 2004), as well as number of additional constraints based on subsequent inquiries into fungal phylogeny (Sung et al. 2008;Gueidan et al. 2011;Ohm et al. 2012;O'Donnell et al. 2013
Tree Reconciliation
The resulting chronogram (dated species tree) and the ensemble of Bayesian gene trees were reconciled using the ALE approach, as implemented in ALE v.0.3 (40,000 trees discarded as burnin, every fifth tree sampled from both runs). To minimize numerical errors on the large data set, scaled version of the program was compiled with floating point calculations carried out on 128 bit numbers (as implemented by "boost::multiprecision" library "float128" type). The inferred transfers are summarized in table 1.
Additional support for transfers as well as alternative scenarios was explored by sampling the multiple optimal reconciliations with DTL-RANGER (Bansal et al. 2012). Here, the support for duplication/speciation/transfer events as well as mapping of individual events on the species tree was annotated on basis of 1) frequency of inferred events/ mappings across different transfer cost values and 2) highest transfer cost where transfer event is indicated (only for inferred HGT events), with fixed duplication and loss costs (DTL-RANGER dated version with parameters L = {1, 2, 3}, Á = 4, e{5, . . . , 40}). The rationale here is that the horizontal transfer is best supported where "regrafts" between cotemporaneous parts of species tree are both most evident (appear at highly penalized transfer cost) and most consistent (as supported for a range of decreasing cost thresholds). For the purpose of sampling, the reconciliations at random, the algorithms from Bansal et al. (2013) were used with a random sample size of 1,000 optimal solutions per each cost combination.
Synteny Analysis
To establish the enrichment or depletion of specific gene families/subfamilies in the syntenic context of NR-PKS core genes, we introduced an approach based on multiple applications of the Fisher exact test.
First, all homologs within +20/À20 genomic context were extracted and subjected to exhaustive, all against all BLASTP searches. The BLASTP expectation values were then used for clustering with MCL (Enright et al. 2002). Only (bidirectional) BLAST hits with E value less than 1E-10 and majority coverage (>50%) of the longer sequence were considered. The MCL inflation threshold was set to 1.4. The clustering parameters were chosen as a compromise between stringent clustering (E value) and observed property of high inflation thresholds erroneously breaking up the clusters for highly diverse sequences (Chan et al. 2013). For our analysis, the inflation threshold setting was also supported in the average values of silhouette width (Rousseeuw 1987), a cluster quality measure independent of predefined class labels which consistently presented a slight peak at the 1.4 setting.
The 2Â2 contingency tables were set up to contrast inclusion of NR-PKS gene in a given clade (vs. the rest of the tree), with presence/absence of a candidate homolog group member in the vicinity of the said gene. Thus, the resulting candidate homolog groups numbering ten or more sequences (133 clusters) were iteratively tested for association with all possible subtrees containing five or more leaves. As remarked above, this was done with Fisher exact test, corrected for multiple testing with Bonferroni correction (based on total number of tests for all candidate homolog groups). The local P-value minima were marked and test results corresponding to the most significant, nonoverlapping subtrees reported for each homologous group. The corrected P-value threshold for inclusion in reported results was set to 0.0001. The detected associations are summarized in table 2 and an example demonstrating the detection of one of the strongest associations (b-lactamase accessory enzymes) is included in supplementary figure S4, Supplementary Material online. An analogous approach was employed for the gene structure, where exon positions were mapped to the respective protein domain sequence alignments and counts calculated for consensus positions based on the said alignments. The starter acyltransferase (SAT), ketoacyl synthase (KS), main acyltransferase (AT), and PT domains were analyzed based on the realignment (MAFFT-LINSI) of corresponding sequence fragments detected earlier. Due to highly divergent sequences of C-terminal domains (methyltransferase, thioesterase, NADdependent reductase) and the apparent incompleteness of a fraction of gene models, these were excluded from splicing site analysis.
Visualization
The trees/dendrograms and annotation (gene structure, domain architecture, reconciled events) were visualized using custom Python scripts dependent on BioPython (Cock et al. 2009;Talevich et al. 2012) and ETE2 (Huerta-Cepas et al. 2010). Where final, annotated versions of published genomes had yet to appear in NCBI/GenBank, annotated gene names were derived following the practice of combining existing gene numbers with organism-specific prefix (locus prefix tag) based on codes available through genome sequencing project pages (http://www.ncbi.nlm.nih.gov/bioproject/, last accessed January 4, 2015).
The functional annotation for experimentally verified and described nonreducing polyketide synthases was gathered from available scientific literature manually (for complete list and descriptions, see supplementary table S4, Supplementary Material online). Classification of cyclization activities (C2-C7, C6-C11, C4-C9) originates from nomenclature used by Li et al. (2010). The structural model of PT domain was visualized with PyMol 1.7.2 (www.pymol.org, last accessed May 30, 2015).
Inferred History of NR-PKS Complement Is Strongly Supported
The reconstructed species tree (see fig. 2 for simplified representation showing predicted evolutionary events), based on 23 single copy orthologous genes, is largely in accord with previously published results, in particular the comprehensive phylogenomic analysis by Ebersberger et al. (2012) and alignment-free composition vector-based comparison by Wang et al. (2009). Notably, the set of selected orthologs supports the alliance of Dothideomycetes and Transfer events are marked through their respective numbers from table 1 (filled outlines indicate donors, hollow outlines indicate transfer acceptors). Salient broad taxa within higher fungi are highlighted by colored backgrounds (in clockwise direction: Basidiomycota, Leotiomycetes, Sordariomycetes, Dothideomycetes, Eurotiomycetes). Subscript on branches denotes predicted numbers of, respectively, duplications/deletions/genes for respective branch. For ease of reference, some nodes were collapsed (thickened branches, proportional to the number of species/strains). Duplications within the Agaricomycetes clade are not shown due to space constraints (see fig. 4). Eurotiomycetes postulated by both. The individual splits are well supported by both inference methods (BI and ML), with majority of splits within Bayesian consensus tree supported by both methods (see supplementary material, Supplementary Material online).
The reconstruction of NR-PKS phylogeny (the gene tree) was based on the conserved KS-AT fragment of the megasynthases. Again, majority of splits are well supported by both Bayesian and ML results. Notably, although approximately 15% of splits (60/413) present in the Bayesian consensus tree are challenged during ALE reconciliation, the subdivision into major clades (as depicted on fig. 3) is not modified.
Inferred Events Are Largely Supported by Both Reconciliation Methods
Within the limits of comparability, the predictions obtained with DTL-RANGER and ALE align well (see supplementary tables S5-S7, Supplementary Material online, for details). Expectedly, there is a marked decrease of reconciliation cost on the amalgamated topology (which resolves uncertainties in the original Bayesian consensus in favor of the species tree) with reported minimum costs on average 10% lower, across all investigated parameter combinations (see supplementary table S5, Supplementary Material online, for detailed comparison).
Based on the amalgamated topology (where direct comparison is possible) predicted duplications and speciations are largely the same for DTL-RANGER (across varying cost thresholds). In particular, for the following parameter combinations there is over 99% identity in predictions (costs given in <D, T, L> format for, respectively, duplication, transfer, and loss): <4, 10-12, 1>, <4, 18-20, 2>, and <4, 25-28, 3>. Analogous results were observed for mapping of the events to the corresponding nodes of the species tree for the abovementioned parameter values. Overall, although predictions for speciation and duplication nodes correspond to ALE results in overwhelming majority (peaking at 98-99%), this correspondence is worse for transfer predictions (peaking at 92%, considerably worse for many parameter values). The DTL-RANGER predictions on the original tree (compared over the 352 bipartitions shared between the original Bayesian consensus and the amalgamated topology) also align well, showing that simple reconciliation without rearrangement can capture at least part of the events correctly (see supplementary tables S6 and S7, Supplementary Material online, for details).
Pertinently, both the inferred existence and the tentative topological dating of majority of duplication events (~70%) are unperturbed regardless of the assumed parameters. Thus, our prediction of the ancestral nature of most underlying duplications is largely independent of detected transfers.
Biosynthesis of Orsellinic Acid Is an Ancestral Trait of Higher Fungi
The obtained results support ancient origins of biosynthesis of compounds derived from modification of one or more molecules of orsellinic acid, in all three ancestral clades (a, b, and g). The earliest diverging clade a (see figs. 3 and 4) encompasses both basidiomycete and ascomycete sequences, including ones with experimentally established propensity toward orsellinic acid biosynthesis (pks1-Coprinus cinereus, pks14-Fusarium graminearum). Likewise, clade b presents multiple examples of biosynthesis of derivatives of orsellinic acid as the first intermediate metabolites (6-MOS, tropolone, violaceol, mitorubrinic acid-clades a and b). A final argument is provided by the unequivocal placement of Ustilago/Sporisorium clade within the g clade of NR-PKSs (by both BI and ML inference), as well as the confirmed orsellinic acid biosynthesis (Aspergillus nidulans gene orsA, predicted to be of transferred origin) in the early diverging g.C2-C7a clade. Notably, although grouping of these three clades is resolved differently by original gene trees (b as an outgroup to a/g-notion supported by divergent signatures of C-terminal thioesterase, as well as presence of the methyltransferase domain throughout b), the support for the groups themselves is strong and unchallenged.
Ancient Duplications Underlie a Patchwork of Distant Siblings
As remarked above, the predicted events giving rise to the present-day diversity of aromatic compounds are likely ancient (see fig. 2, for graphical summary). The amalgamated likelihood approach predicts them to be mostly duplications followed by extensive losses (849 predicted loss events according to ALE). By this view, extant nonreducing polyketide synthases constitute a patchwork set of distantly related sibling groups.
In particular, repeated emergence of novel cyclization specificities within the g clade (C6-C11-emodins, asperthecin, monodictyphenone; and C4-C9-sterigmatocystin, aflatoxins, fusarubin) is shown to be monophyletic and ancient. The topological dating of relevant nodes provided by reconciliations places the origin in the common ancestral lineage of four major classes of filamentous fungi (Leotiomycetes, Sordariomycetes, Dothideomycetes, and Eurotiomycetes).
Of the total 87 duplication events inferred by ALE reconciliation, 62 are predicted to have occurred in this common ancestral lineage. Furthermore, three ancestral duplication events are predicted to predate the split between Basidiomycetes and Ascomycetes (the two duplications underlying division into a, b, and g clades of NR-PKSs and a singular duplication within the a clade itself). Four duplications are tentatively dated at Dothideomycetes-Eurotiomycetes split. Slightly higher numbers of more recent duplications are inferred in Dothideomycetes and Eurotiomycetes (mostly in phytopathogenic Pleosporales-2 and geni Aspergillus-5 and Talaromyces-3). This seems related to their rich secondary metabolite repertoires and a relatively dense sampling of the closely related genomes from the respective species. Notably, such patterns cannot be seen for other extensive secondary metabolite producers such as the saprobic/pathogenic Metarhizium species. In Metarhizium, the high number of NR-PKS genes seems to stem from less extensive selection and not late duplications (only a singular duplication has affected the Hypocreales lineage-resulting in the doubling of aurofusarin ancestor gene).
By the same token, the extensive thinning of NR-PKS repertoire (874 losses) started after the divergence of filamentous fungi (estimated losses for Sordariomycetes-Leotiomycetes common ancestor-23, for Dothideomycetes-Eurotiomycetes common ancestor-4) and the losses continue following subsequent divergences. As indicated in the previous paragraph, the selective process did not affect all resulting species equally, with some lineages retaining more of the original diversity.
Transfers Are a Source of Additional Diversity
A ranked list of the top detected transfers with their ALEbased mappings are summarized in Table 1 (alternative scenarios found by DTL-RANGER can be examined on the results website). Briefly, 14 of the 25 transfers (56%) are detected consistently even prior to amalgamation (i.e., the bipartition is present in both amalgamated and original gene trees, the transfer is predicted on both topologies by DTL-RANGER across multiple cost thresholds). The detailed results, obtained from sampling of multiple optimal reconciliations at random, are available through the gene tree visualization FIG. 3.-The divergence of major NR-PKS clades, which enabled diversification of biosynthesized aromatic compounds, was facilitated by close linkage with key accessory enzymes and transcriptional factors. The simplified phylogeny (based on amalgamated gene tree) shows selected groups of syntenic homologs as colored shapes (also indicated by numbers in box shape, which reference descriptions in table 2).
Most predicted transfers are fairly recent in comparison to duplications ( fig. 2), in keeping with majority of the transfers being lost over time. The key predicted transfer events are predominantly in clade g (21 transfers) and include several cases of transfers involving well-characterized core biosynthetic genes: HGT of the norsolorinic acid NR-PKS into Podospora anserina genome ( fig. 5A), as well as transfers of pigment-related NR-PKSs (fusarubins- fig. 5B, acquisition of a putative pksP homolog by Aspergillus terreus from Metarhizium, fig. 5C). Very few transfers are predicted in clade b and no transfers are indicated (postamalgamation) for the early-diverging clade a.
The former example of sterigmatocystin biosynthetic cluster in P. anserina constitutes perhaps the best documented individual case of horizontal transfer involving a highly toxic and large secondary metabolite biosynthetic cluster (Slot and FIG. 4.-The phylogeny of ancestral clade a, which predates basidiomycete-ascomycete split, contains conserved NR-PKSs biosynthesizing orsellinic acid (pks14, pks1). Support noted below branches in form BI/ML/C (BI, support from BI; ML, support from ML ultrafast bootstrap; C, whether bipartition is present in Bayesian consensus tree; * denotes full support, % denotes support below 1% level). Domain architecture and gene structure are visualized on the right side (colors: red, SAT; blue, KS; green, AT; purple, PT; orange, ACP; pink, TE; brown, R). Color of the tree edges and letters denote predicted events (D, duplication; TD, transfer donated; TA, transfer accepted). Tree is not drawn to scale.
Rokas 2011). In our analysis, both the support for the event itself and its mappings are consistent. Interestingly, the prediction indicates transfer donor in Dothideomycetes, within R. rufulum lineage. Notably, this origin as unfragmented cluster in Dothideomycetes has been postulated in a recent paper (Bradshaw et al. 2013).
In some cases, the ALE has resulted in a significantly different scenario from simply sampling the optimal reconciliations based on the original tree topology. For a singular Tuber melanosporum NR-PKS (GSTUM_00003432001), its placement in the original BI tree (at 74% BPP, Bayesian Posterior Probability) has introduced an apparent HGT from early Pezizomycotina to Basidiomycetes. This scenario is rejected by ALE where an alternative grouping is incorporated in the final tree instead (bipartition g31 supported at 26% BPP; see fig. 4). This rearrangement shows that consideration of alternative rather than majority bipartitions is required for validating transfer events, even at most restrictive cost settings.
As the sampling of taxa is not even across the species tree, one can expect that increasing coverage of undersampled species will reveal more lineage-specific duplications and transfer events. This is well demonstrated by Eurotiales clade. For this reason, we refrain from making quantitative calls about the singular numbers of events. Still, some observations stand out. In particular, the branch leading to Glarea lozoyensis (Helotiales) appears to be a rather frequent donor (four donated transfers). This is partially explained by both the length of the corresponding branch of the species tree (by the dated chronogram the split between Glarea and B. graminis lines occurred~250 Ma) as well as the other represented species being plant pathogenic rather than saprobic throughout their lifecycle. Also, a high number of transfers (8) are predicted within the g.C2-C7a subclade (represented by orsA and zea1) which presents a sparse distribution of homologs from distantly related species. Although artifactual origins are possible (genes strongly differ in structure), HGT was previously (Xu et al. 2014) invoked as an explanation of the observed diversity of macrocyclic lactones. Additionally, the highly reducing polyketide synthases (indicative of the biosynthesis of lactone part not derived from orsellinic acid molecule) are strongly enriched for the entire g.C2-C7a clade (as well as part of the b clade containing the asperfuranone biosynthetic NR-PKS).
Speciation, Duplication, and Transfer Differentially Affected the Genomic Context
We investigated the overall differences in conservation of genomic context (+20/À20 genes) at each bipartition (in the amalgamated tree) by looking at the average number of conserved homologs shared between neighborhoods of any two members of the respective descendant clades (left and right subtrees of inner nodes in the tree). This estimate was then contrasted with event labels ascribed to bipartition by ALE. Duplications were found to disrupt the synteny most (11% of shared homolog groups on the average are retained), with significantly (Wilcoxon ranked sum test) higher retention for speciation (40%; P < 10 À5 ) and slightly higher for transfers (21%, not significant). The number of the putative counterparts decays with the depth of the node (Pearson correlation between the ratio of shared homologs vs. depth of the node in the tree r = À0.22, P = 5 Â 10 À6 ), confirming that context is also lost over multiple events (in part due to most of the deeper nodes being duplications, while many recent ones are speciations). Notably, we think that these estimates err on the strong side as counterparts were identified on basis of membership in the same CHG (gene family or subfamily) rather than precise orthology/paralogy relationships.
Although for some of the predicted transfer cases the genomic context is partially preserved (e.g., sterigmatocystin in Podospora-17% genes have a counterpart in Rhytidhysteron), there are five cases (g361, g511, g664, g731, g791), including the strongly supported transfer of the pksP into A. terreus (g664), where genomic neighborhood is not shared between predicted donor and acceptor branches at all.
Conserved Splice Junctions and Overrepresented Accessory Enzymes Are Associated with Emergence of Different Cyclizations
The gene groups significantly overrepresented in association with particular monophyletic clades are summarized in table 2 (graphic summary of selected associations is shown on fig. 3). Out of 132 homolog groups, 40 display significant (P < 0.0001) enrichment for at least one monophyletic clade in the tree (possibly indicating the point of origin for acquiring the accessory enzyme). Strikingly, no enrichment of any particular gene family was noted in association with the early diverging clade a (which contains sequences from both Basidiomycetes and Ascomycetes). This trend was not mirrored in exon junction positions which are conserved relative to consensus sequences of KS and AT domains encoded by a clade members.
In terms of splicing site conservation, all three clades (a/b/g) demonstrate specific patterns (see supplementary table S8, Supplementary Material online). In group a, the member genes typically present higher overall number of exons due to multiple splice junctions embedded within the above-mentioned KS and AT domains. In group b, the core NR-PKS domains are typically encoded on a single large exon, with a clear exception of the clade (g182) associated with biosynthesis of meroterpenoid products (terretonin, austinol). The meroterpenoid NR-PKSs show their own characteristic pattern of splice junctions within SAT, KS, and AT domains.
The g clade of NR-PKSs encompasses synthases capable of a number of divergent modes of cyclization (C2-C7, C6-C11, C4-C9; Li et al. 2010). This plasticity of function is perhaps facilitated by the observed tendency toward exon gain within both the SAT and the PT domains. Notably, by relating to available protein structure of the PT domain (norsolorinic acid synthase from Aspergillus parasiticus, C4-C9 cyclization type) one can see that the highest degree of fragmentation results in four individually encoded pieces, each of which contributes to the gating of the cyclization chamber (see fig. 6the model of PksA aflatoxin NR-PKS with delineated structural elements corresponding to coding sequence parts separated by splice junctions).
The subfunctionalization toward different cyclization schemes and/or end products is also shown to be related to recruitment of specific accessory enzymes. The manganesedependent b-lactamases are unambiguously associated with g.C6-C11 subclade (g552, corrected P = 8.6e-43), with the notable exception of endocrocin sister clade (g439; where b-lactamase is not always present in the vicinity of NR-PKS). Taken as an exemplary association, the endocrocin example demonstrates that strong linkage follows recruitment of the accessory enzyme and thus provides an upper constraint on the timing of the original involvement in the biosynthetic pathway.
As with most of the predicted duplications, the topological dating places this tight linkage with b-lactamases (Li et al. 2011), prior to the divergence of major classes of filamentous fungi. As in other clades, the syntenic relationships persist through multiple predicted horizontal transfer events, involving among others: The acquisition of capacity to synthesize alternariol/isocoumarins by A. nidulans/oryzae/flavus (pkgA homologs, predicted to originate from Sordariomycetes, HGT supported solely by ALE) and spread of asperthecin biosynthetic PKS aptA homologs to Arthroderma sp. as well as Metarhizium sp. (with the original donor in Aspergillus fischerii/fumigatus clade-a likely ortholog of the extant neosartoricin NR-PKS nscA involved in prenylated xanthone biosynthesis).
The enrichment analysis suggests that diversification of activity was also associated with recruitment of conserved but divergent groups of transcription factors (of the zincbinding finger variety). Of the five transcription factor groups, only one (ch43) is enriched for two separate subclades of g. The only other case, where a homolog group is strongly enriched in more than one part of the tree, concerns the highly reducing polyketide synthases mentioned in the previous subsection.
As a sidenote, there are multiple cases where accessory enzymes strongly associated with a monophyletic clade are nevertheless present (in singular numbers) around NR-PKS genes from different clades (see table 2, e.g.: Several aflatoxin biosynthetic cluster genes, aurofusarin biosynthesis dehydratase aurZ). Depending on the respective gene family histories, the presence of such "outliers" raises the salient questions of frequency and modes of exchange of accessory enzymes between different clusters, particularly if biosynthetic activities of individual clusters are linked by common end product further down the line (e.g., Andersen et al. 2013).
The Evolution of Pigment Biosynthesis Presents Both Conservation and Diversification
Perhaps, the most striking examples of both conservation and variation concern the evolution of melanin biosynthesis (see fig. 7) and splitting of the biosynthesis of alternative pigments and toxins from this conserved branch. First, the common ancestor of g.C6-C11, g.C2-C7b/c, and g.C4-C9 ( fig. 3) acquired an accessory enoyl-CoA reductase (of the short chain dehydrogenase superfamily). This likely enabled biosynthesis of reduced intermediate compounds (such as versicolorin during the biosynthesis of aflatoxins, tetra/ trihydrohydroxynaphthalene during biosynthesis of melanins, monodictyphenone during the biosynthesis of xanthones). Next, the linkage with copper-dependent laccase enzymes enabled oxidative polymerization (e.g., in biosynthesis of melanins, aurofusarin) in the common ancestor g.C2-C7b/c and g.C4-C9.
Movement between different components of the (twospeed) fungal genome is evidenced in the placement of melanin biosynthetic NR-PKSs (including the genes involved in biosynthesis of melanin through naphtopyrone intermediate) in the vicinity of confirmed housekeeping genes (see table 2). Conservation is further underscored by inheritance of the tetrahydroxynaphtalene synthase (core NR-PKS) as an ortholog in the majority of filamentous fungi ( fig. 7-most of Leotiomycetes, Sordariomycetes, and Dothideomycetes). In select Sordariomycetes (Hypocreales) as well as most Eurotiomycetes, pigment biosynthesis is instead carried out through modified naphtopyrone routes (through YWA1 heptaketide intermediate product-e.g., Chiang et al. 2011). Notably both the speciation patterns and the conserved gene structure of the core melanin biosynthetic NR-PKS (the tetrahydroxynaphthalene synthase) suggest possible application as a source of additional barcode markers across the relevant taxa.
The basic building blocks (accessory enzymes) partaking in biosynthesis of melanins have been further retained for biosynthesis of other compounds (and are present in the extant clusters). Furthermore, possibly due to the increased toxicity of C4-C9 cyclization products (such as sterigmatocystin and aflatoxins, javanicin, and fusarubins) that subclade is associated FIG. 7.-The phylogeny of tetrahydroxynaphthalene synthases (the core enzyme of melanin biosynthesis) mirrors speciation explicitly in the majority of Dothideomycetes, Leotiomycetes, and Sordariomycetes. Visual conventions (branch support, exon, domains) analogous to figure 4. Gene models for MBM_00260 and M7I_3853 were truncated from 5 0 and 3 0 sides, respectively. with significantly overrepresented retention of accessory methyltransferases in the close neighborhood of the core megasynthase.
The analysis of top-scoring, consistently predicted transfer events suggests that biosynthesis of aromatic pigments, following the divergence of genes involved in melanin biosynthesis through the DHN route, could have been influenced by key transfer events (e.g., origin of Pgl1/Fsr1 orthologs in Fusarium as a horizontal transfer from early Pleosporales, reintroduction of a putative pigment biosynthesis gene into A. terreus as a transfer from Metarhizium). As an example of finer resolution, due to the increased taxon sampling of our analysis (relative to earlier phylogenomic analysis by Brown et al. [2012]), the core genes of bikaverin and aurofusarin biosynthesis are now shown to be nonorthologous in origin. These NR-PKSs are now revealed to belong to sister clades which diverged early during the evolution of filamentous Ascomycota and (in case of aurofusarin NR-PKS) underwent an additional Hypocreales' specific duplication. The aurofusarin biosynthetic gene is however shown to be orthologous to experimentally characterized Trichoderma PKS4 (pigment synthesis PKS; Atanasova et al. 2013), following the said duplications ( fig. 8).
Last point of note concerns the replacement of thioesterase domain with NAD-binding reductive domain leading to aldehyde end products of cyclization. This event has likely occurred twice during the evolution of NR-PKSs (respectively, in g.C4-C9 clade leading to fusarubin subclade- fig. 5B-and in the part of the b clade associated with biosynthesis of azaphilones and tropolone-derived compounds). Based on the gene structure and similarity of protein sequences of affected domains, a likely possibility is the recombination (or fusion) between a donor gene from b clade and an acceptor gene from g clade (or vice versa). Regardless of whether the domain has been acquired from an outside or inside source, the implied scenarios show the need for considering mosaic ancestry of multidomain proteins, where different regions of the protein/gene could have conflicting origins.
Toward Phylogenomic Roadmaps of Secondary Metabolite Biosynthesis in Fungi
The reconstruction of evolutionary history of fungal NR-PKSs faced a number of obstacles. Foremost among these was the quality of gene tree inference for a large and diversified gene (sub)family across multiple species. First, the data are strongly saturated worsening the risks associated with long branch attraction and incorrect resolution of deeper nodes in the tree. Second, due to being focused on a conserved fragment of limited length (the KS-AT core modules and intervening region) attempts to put conservative filters on data, commonly accepted in species-focused phylogenomics, would likely introduce stochastic bias (particularly) in the resolution of rapid successions of events. Nevertheless, we have taken several precautions against the common artifacts of phylogeny reconstruction (use of protein sequences, filtering protein alignments based on transitive consistency score, use of different methods for phylogeny reconstruction including models able to take into account the heterogeneity of rates across informative sites).
In its present iteration, the reconstruction uncertainty is also alleviated by conscious inclusion of species trees in phylogeny reconstruction process (so in the absence of prevailing evidence, neutral scenario implied by species tree topology is considered by ALE). The contemporary developments in probabilistic models for inferring both species and gene phylogeny simultaneously ) and for inferring the gene phylogeny based on simultaneous reconciliation with a putative species tree (Szollosi, Rosikiewicz, et al. 2013) formalize this approach in a probabilistic framework. In our example, the application of ALE has also given an added benefit of explicitly modelling the two-step transfers through intermediary donors from extinct lineages .
The results demonstrate the utility of maximally parsimonious reconciliations and phylogenetic inference in uncovering the sources of extant diversity in a highly diverged gene family (duplication, transfer, and selective loss). As a proofof-concept our analysis demonstrates that in considering the evolution of secondary metabolism in microbial Eukaryotes, the scientists have to take into account transfers (at varying degrees) as a viable origin hypothesis for present metabolite diversity. This is even though, in this case, majority of molecular innovation seems ancient in origin and (throughout the history of multiple extant fungal lineages) has been subjected mostly to selective losses. The use of probabilistic methods (ALE) allowed us to address the parameter choice (costs of different events) in an unsupervised way-with additional support lent to early diversification of biosynthetic mechanisms and a limited, but not negligible, number of predicted transfer events.
In the future, the increased sampling of previously uncharacterized taxa of higher fungi will likely increase the number and improve the resolution of individual events. The biases in taxonomic coverage resulting from uneven sampling of divergent taxa are liable to influence what is perceived as an incorrect outlier (errant bipartition due to inclusion of a gene from a rogue taxa) rather than accumulated evidence from multiple, more distantly related species (Szollosi et al. 2015). The information about spread and retention of biosynthetic potentials in different lineages may thus reveal stronger and/or different trends among donor and receiver groups, as well as contrast these across ecological niches occupied by related organisms. The resolution of transfers, in particular can be challenged, should novel organisms from related clades (the increasing public coverage of previously unsampled lineages, such as Chaetothyriales, Lecanoromycetes or Xylonomycetes) imply different distributions of genes. This is one of the reasons why roadmap nature of the resource should be emphasized-one taking into account the "landmarks" (species) available at the time of its creation.
Likewise, the continuous development of novel computational approaches taking advantage of high-performance computing infrastructure results in more accessible methods increasingly able to account for some of the unavoidable biases in the data. For example, both rate heterogeneity and misleading signals, arising due to saturation, are better handled by improved Bayesian and ML methods, as evidenced by a number of studies (e.g., Boussau and Gouy 2006;Lartillot et al. 2009;Husník et al. 2011).
Last but not least aspect of the presented analysis lies in providing, for the first time, a phylogeny-based annotation of NR-PKS core genes involved in fungal secondary metabolism. With increasing number of inquiries into the phylogenetic basis of eukaryotic secondary metabolism based on functional experiments (knockout, deletion mutant libraries) and growing genomic coverage, similar initiatives should complement large tools and databases (Khaldi et al. 2010;Blin et al. 2013;Wang et al. 2014). In particular, a phylogeny-centric resource should facilitate research focused on chemotaxonomy, as well as further dissection of the evolutionary fates of specific compounds and clusters (e.g., adaptation of pigment molecules for protection from both abiotic and biotic stresses, emergence and spread of high toxigenic potential).
Conclusions
We present, for the first time, a phylogeny-based analysis elucidating the origins of diversity in one of the largest and most important groups of core genes involved in fungal secondary metabolism-the nonreducing polyketide synthases. The results support ancient duplications of a limited number of NR-PKSs into subsets catalyzing divergent chemical reactions. The highly diverse representation of these genes in extant fungal genomes is revealed to be a result of subsequent, large-scale selective losses, moderated by a number of more recent duplications and horizontal transfers (associated with, e.g., emergence of fusarubin biosynthesis in fusaria). The complementary analysis of phylogeny-associated traits (genomic neighborhood, splice junctions) shows gain of tightly linked accessory genes and modularization of key PKS domains (PT domain), as crucial traits accompanying the functional diversification. | 2018-04-03T02:17:43.429Z | 2015-11-01T00:00:00.000 | {
"year": 2015,
"sha1": "46e732eaeede836ebdce493e6c0ef17a77df3115",
"oa_license": "CCBY",
"oa_url": "https://academic.oup.com/gbe/article-pdf/7/11/3132/20592899/evv204.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "46e732eaeede836ebdce493e6c0ef17a77df3115",
"s2fieldsofstudy": [
"Biology",
"Chemistry",
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
21707486 | pes2o/s2orc | v3-fos-license | On the concentration phenomenon of $L^2$-subcritical constrained minimizers for a class of Kirchhoff equations with potentials
In this paper, we study the existence and the concentration behavior of minimizers for $i_V(c)=\inf\limits_{u\in S_c}I_V(u)$, here $S_c=\{u\in H^1(\R^N)|~\int_{\R^N}V(x)|u|^2<+\infty,~|u|_2=c>0\}$ and $$I_V(u)=\frac{1}{2}\ds\int_{\R^N}(a|\nabla u|^2+V(x)|u|^2)+\frac{b}{4}\left(\ds\int_{\R^N}|\nabla u|^2\right)^2-\frac{1}{p}\ds\int_{\R^N}|u|^{p},$$ where $N=1,2,3$ and $a,b>0$ are constants. By the Gagliardo-Nirenberg inequality, we get the sharp existence of global constraint minimizers for $2<p<2^*$ when $V(x)\geq0$, $V(x)\in L^{\infty}_{loc}(\R^N)$ and $\lim\limits_{|x|\rightarrow+\infty}V(x)=+\infty$. For the case $p\in(2,\frac{2N+8}{N})\backslash\{4\}$, we prove the global constraint minimizers $u_c$ behave like $$ u_{c}(x)\approx \frac{c}{|Q_{p}|_2}\left(\frac{m_{c}}{c}\right)^{\frac{N}{2}}Q_p\left(\frac{m_{c}}{c}x-z_c\right).$$ for some $z_c\in\R^N$ when $c$ is large, where $Q_p$ is up to translations, the unique positive solution of $-\frac{N(p-2)}{4}\Delta Q_p+\frac{2N-p(N-2)}{4}Q_p=|Q_p|^{p-2}Q_p$ in $\R^N$ and $m_c=(\frac{\sqrt{a^2D_1^2-4bD_2i_0(c)}+aD_1}{2bD_2})^{\frac12}$, $D_1=\frac{Np-2N-4}{2N(p-2)}$ and $D_2=\frac{2N+8-Np}{4N(p-2)}$.
Introduction and main result
In this paper, we study the existence and the concentration phenomenon of normalized solutions to the following Kirchhoff equation (1.1) where N ≤ 3, a, b > 0 are constants and 2 < p < 2 * := 2N N −2 , 2 * = 6 if N = 3 and 2 * = +∞ if N = 1, 2. The potential V : R N → R is a suitable function.
In the past years, equation (1.1), which is a nonlocal one as the appearance of the term R N |∇u| 2 , has attracted a lot of attention. (1.1) is no longer a pointwise identity, which causes some mathematical difficulties and makes the study particularly interesting. see [1,2,3,5,6,15,20,22] and the references therein. The first line to study (1.1) is to consider the case where ρ is a fixed and assigned parameter. see e.g. [12,14,17,18,19,25]. In such direction, the critical point theory is used to look for nontrivial solutions, however, nothing can be given a priori on the L 2 -norm of the solutions. Recently, since the physicists are often interested in "normalized solutions", solutions with prescribed L 2 -norm are considered. A solution with |u| 2 = c corresponds to a critical point of the following C 1 -functional where S c = {u ∈ H 1 (R N )| |u| 2 = c} and When b = 0, equation (1.1) does not depend on the nonlocal term any more, i.e. it becomes the following typical Schrödinger equation (for simplicity, we may assume that a ≡ 1) −∆u + (V (x) − ρ)u = |u| p−2 u, x ∈ R N , ρ ∈ R, N ≥ 1. (1.5) In the literature, there are some papers studying the existence and the concentration behavior of normalized solutions to (1.5), see e.g. [7,13,21,23,24]. Similarly, a solution with |u| 2 = c can be obtained by looking for a minimizer of e V (c) = inf [7,13,24]). Moreover, if 2 < p < 2N +4 N , then e 0 (c) < 0 and e 0 (c) has a minimizer for all c; if 2N +4 N ≤ p < 2 * , then e 0 (c) has no minimizers for all c. In [21], Maeda shows that up to translations, the minimizer of e 0 (c) is unique (see also [9] [15]) and gets a scaling property of the minimizer of e 0 (c) for 2 < p < 2N +4 N , i.e. suppose that u c is a minimizer of e 0 (c), then up to translations, When V (x) ≡ 0, Maeda in [21] proves that e V (c) has a minimizer for all c > 0 if 2 < p < 2N +4 N and V (x) satisfies that 0 = inf ) for some positive constant C, by using the essential scaling property (1.6) and a special L 2 -preserving scaling, [21] shows that there exists y c ∈ R N such that the minimizer of e V (c), denoted Guo et. al. in [8][10] [11] proved that e V (c) has a minimizer if and only if 0 < c < c 0 for some c 0 > 0. Moreover, the minimizer of e V (c) concentrates at the flattest minimum of V (x) as c ր c 0 .
When b > 0, due to the effect of the term ( R N |∇u| 2 ) 2 appeared in the energy functional, p = 2N +4 N is now L 2 -subcritical for (1.3) and (1.4). The existence of normalized solutions to equation (1.1) may be different from the Schrödinger one. When V (x) ≡ 0, by the L 2 -preserving scaling and the well-known Gagliardo-Nirenberg inequality with the best constant [26,16]: when either p ∈ [2, 2N N −2 ) with N ≥ 3 or p ∈ [2, +∞) with N = 1, 2, where the equality holds for u = Q p and Q p is, up to translations, the unique positive least energy solution of it is showed in [27,28] that p = 2N +8 N is the L 2 -critical exponent for the minimization problem (1.4) in the sense that for each Moreover, we collect some known results in the following proposition concerning the minimizers of i 0 (c).
such that i 0 (c) = 0 for all 0 < c ≤ c * and i 0 (c) < 0 for all c > c * .
(ii) i 0 (c) has a minimizer if and only if c ∈ T , where (1.9) Moreover, i 0 (c) has no minimizer for all c > 0. If V (x) ≡ 0, then the case becomes totally different from the autonomous one since the L 2 -preserving scaling argument fails here. As far as we know, there is no paper on this respect. In this paper, we study the existence and the concentration behavior of minimizers for i V (c) when V (x) satisfies the following condition: We call (u c , ρ c ) ∈ S c × R a couple of solution to the equation (1.1) if (u c , ρ c ) satisfies (1.1) and |u c | 2 = c.
Our first result is as follows. [4]). To prove Theorem 1.2 (1), it is enough to show the boundedness of each minimizing sequence for i V (c), which can be obtained by using the inequality (1.7). By using the least energy solution Q p as a test function and a suitable L 2 -preserving scaling, we succeeded in obtaining Theorem 1.2 (2).
Our main result concerns the concentration phenomena of minimizers u c of i V (c) as c → +∞ for 2 < p < 2N +8 N . Since lim To do so, we need to study the property of minimizers for i 0 (c). The theorem below summarizes some properties of i 0 (c). (1) Up to translations, . That is to say, i 0 (c) has a unique minimizer for each c ∈ T . Moreover, Then our main result is as follows.
For any sequence {c n } ⊂ (0, +∞) with c n → +∞ as n → +∞, let (u cn , ρ cn ) ∈ S cn × R be the couple of solution to (1.1) obtained in Theorem 1.2. Then there exists a subsequence of {c n } (still denoted by {c n }) and a sequence {z n } ⊂ R N such that as n → +∞.
Remark 1.5. Although the equation (1.1) is "nonlocal", the concentration behavior of the solution to (1.1) given in Theorem 1.4 when the L 2 -norm of the solution is "large", is similar to that of the solution to (1.5), which is "local" (see Lemma 4.6 in [21]). However, our argument is different.
We give the main idea of the proof of Theorem 1.4. As the problem we deal with is nonlocal, the approach of [21] considering mass concentration of Schrödinger equation with the L 2 -subcritical exponent cannot be applied directly for two reasons. First, since the terms in I 0 (u) or in I V (u) scale differently in space, one cannot hope to get rid of the nonlocal term by scaling in space. Then we cannot follow the argument in [21] to show that the minimizer of i 0 (c) possesses a property like (1.6), which makes it difficult to study the convergence of i V (c) as c → +∞. To overcome this difficulty, the key point is to prove the uniqueness of the minimizer v c of i 0 (c) and to give the accurate expression of v c (see Theorem 1.3). This is not easy and it needs much more analysis. Based on this property, we show that Indeed, to show (1.10), it requires that mc n cn converges to either 0 or +∞, which is the reason why p = 4 is assumed in Theorem 1.4 (see Theorem 1.3 (3)). Second, due to the effect of the nonlocal term, property (1.6) no longer holds for i 0 (c) and the minimizer of i 0 (c) (see Theorem 1.3 (1) above for details), which makes that the special L 2 -scaling using in [21] to get the mass concentration cannot be used here. We overcome this difficulty by using the estimate (1.11) to obtain the following optimal energy estimates for each minimizer u cn of i V (c n ): as n → +∞, i.e. I 0 (uc n ) i 0 (cn) → 1, which gives us a cue that u cn might behave like the minimizer of i 0 (c n ). It is necessary to point out that the nonlocal term plays an important role in the proof of (1.12) and (1.13). Finally, for any c ∈ T , we set w n (x) := c cn u cnmc mc n c cn (x), by (1.12) (1.13) we see that {w n } is a bounded minimizing sequence of i 0 (c) and then w n converges strongly to the minimizer of i 0 (c) in H 1 (R N ). So the theorem is proved.
Throughout this paper, we use standard notations. For simplicity, we write Ω h to mean the Lebesgue integral of h(x) over a domain Ω ⊂ R N . L p := L p (R N ) (1 ≤ p ≤ +∞) is the usual Lebesgue space with the standard norm | · | p . We use " → " and " ⇀ " to denote the strong and weak convergence in the related function space respectively. C will denote a positive constant unless specified. We use " := " to denote definitions and B r (x) := {y ∈ R N | |x − y| < r}. We denote a subsequence of a sequence {u n } as {u n } to simplify the notation unless specified.
The paper is organized as follows. In § 2, we prove Theorem 1.3. In § 3, we prove Theorem 1.2 and Theorem 1.4.
Proof of Theorem 1.3
In this section, we will present an accurate description of i 0 (c) = inf u∈ Sc For simplicity, in what follows we denote For any u ∈ H 1 (R N ) and any t > 0, in what follows we denote By (1.7)(1.8) and the corresponding Pohozaev identity we see that It is proved in [9] that Q p is decreasing away from the origin and as |x| → +∞.
The following lemma, which will be useful in the main proof, can be easily obtained, so we omit the proof.
Proof. By the Lagrange multiplier theorem, there exists µ c ∈ R such that (v c , µ c ) satisfies the following equation By Lemma 2.1 in [17], we see that v c satisfies the following Pohozaev identity and (2.7) Since D 2 > 0 and i 0 (c) ≤ 0, we have Therefore, by Lemma 2.1 and (2.4)-(2.8) we see that v c ∈ S c is a positive solution of the following equation By the fact that w is positive together with the uniqueness of positive solutions (up to translations) to the equation (2.9), we conclude that By direct calculation, we see that Proof.
The following proposition follows from Theorem 1.1 in [27] and Lemma 2.2.
Based on Corollary 2.3, we could get the exact value of c * defined as in Proposition 1.1 when 2N +4 N < p < 2N +8 N , which is not obtained in [27].
(1) It is proved in Lemma 2.5 of [27] that c → i 0 (c) c 2 is strictly decreasing on (c * , +∞). For readers' convenience, we give a detailed proof of this fact here.
So the function c → i 0 (c) c 2 is strictly decreasing on (c * , +∞). To prove the lemma, since i 0 (c) c 2 < 0 for all c > c * , we see that either i 0 (c) By (1)
Proof of Theorems 1.2 and 1.4
In this section, we first consider the minimization problem (1.3) when V (x) satisfies condition (V ). We need the following compactness result, see e.g. [4].
Proof of Theorem 1.2 So we see from Proposition 1.
For any c ∈ T , let {u n } ⊂ S c be a minimizing sequence for i V (c), i.e. I V (u n ) → i V (c) as n → +∞. By Lemma 3.1, to prove that i V (c) has a minimizer it is enough to prove that {u n } is uniformly bounded in H.
Indeed, if {u n } is uniformly bounded in H, then up to a subsequence, there exists u c ∈ H such that u n ⇀ u c in H. Lemma 3.1 shows that u n → u c in L q (R N ), 2 ≤ q < 2 * , which implies that |u c | 2 = c, i.e. u c ∈ S c . By the weak lower semicontinuity of the norm in H, we have i V (c) ≤ I V (u c ) ≤ lim inf n→+∞ I V (u n ) = i V (c), i.e. u c is a minimizer for i V (c). So i V (c) has at least one minimizer for each c ∈ T .
Let us next prove the boundedness of {u n }. If 2 < p < 2N +8 N , for any c > 0, since I V (u n ) → i V (c), by (1.7) we have for n large enough 1) which and 0 < N (p−2) 4 < 2 imply that {u n } is uniformly bounded in H.
Then for n large enough, we see that which implies that {u n } is uniformly bounded in H.
Since i V (c) has a minimizer u c , by the Lagrange multiplier theorem there exists ρ c ∈ R such that (u c , ρ c ) ∈ S c × R is a couple of solution to the equation (1.1).
(2) Let ϕ ∈ C ∞ 0 (R N ) be a radial cut-off function such that 0 ≤ ϕ ≤ 1, ϕ ≡ 1 on B 1 (0), ϕ ≡ 0 on R N \B 2 (0) and |∇ϕ| ≤ 2. For where A t > 0 is chosen to satisfy that u t ∈ S c . In fact, By direct calculations and the Dominated Convergence Theorem we have is bounded and has compact support, we see So by (2.3) and the exponential decay of Q p and |∇Q p |, we see that if p = 2N +8 N , then So the theorem is proved.
Next we give some preliminary lemmas to prove Theorem 1.4. Proof.
(1) For any c > 0 and any sequence {c n } ⊂ (0, +∞) satisfying that c n → c as n → +∞, let u c ∈ S c be a minimizer of i V (c), then cn c u c ∈ S cn and i V (c n ) ≤ On the other hand, let {u cn } ⊂ S cn be a sequence of minimizers for i V (c n ). Since (2) Note that i V (c) ≥ i 0 (c) for any c > 0 when 2 < p < 2N +8 N . For c = 1, suppose that u 1 is a minimizer of i V (1), then for any c → 0 + , cu 1 ∈ S c and By Proposition 1.1 we have to discuss the following two cases. Case 1: 2 < p < 2N +4 N . By Proposition 1.1, if 2 < p < 2N +4 N , then i 0 (c) < 0 and i 0 (c) has a minimizer v c ∈ S c for all c > 0. Hence by (1.7) we see that , which and 0 < N (p−2) 4 < 1 imply that lim c→0 + |∇v c | 2 = 0. Then lim c→0 + i 0 (c) = 0. So combining (3.4) we have lim N , then i 0 (c) = 0 for c > 0 small. Then we conclude from (3.4 Since c → +∞, we may assume that c > c * , then i 0 (c) < 0 and i 0 (c) has a unique minimizer for all 2 < p < 2N +8 N . Let ϕ ∈ C ∞ 0 (R N ) be a radial cut-off function as given in the proof of Theorem 1.2 (2). For any x 0 ∈ R N and R > 0, Set is the unique minimizer of i 0 (c) given in Theorem 1.3 and A R,c > 0 is chosen to satisfy that u R,c ∈ S c . In fact, 0 < A R,c ≤ 1 and lim R→+∞ A R,c = 1.
It is standard to show that u R,c → Q c in H 1 (R N ) as R → +∞ for each c > c * . Then by continuity, we see that for any c > c * , If lim c→+∞ mc c = 0, then → 0 as c → +∞, where we have used the fact that Q p (x) is continuous. Therefore for any ε > 0, by (3.5)-(3.7) there exists R ε > 0 large enough such that We conclude from Lemma 2.6 (1) that i V (c) i 0 (c) → 1 as c → +∞.
(2) Since → +∞. By the Gagliardo-Nirenberg inequality (1.7), we see that In order to prove Theorem 1.4, we need the following two crucial lemmas.
as c → +∞, where m c is given in Theorem 1.3.
(3.8)
Then we see that .
Proof.
As c → +∞, we may assume that c > c * . Then by Corollary 2.3 (1), Lemmas 3.4 and 2.7, we see that Proof. For any sequence {c n } ⊂ (0, +∞) with c n → +∞ as n → +∞, by Theorem 1.2, there exists (u cn , ρ cn ) ∈ S cn × R a couple of solution to the equation (1.1) with I V (u cn ) = i V (c n ). Then u cn is a minimizer of i V (c n ). Without loss of generality, we may assume that c n > c * for all n. By Lemma 3.4 we see that | 2018-05-03T00:02:28.000Z | 2018-05-03T00:00:00.000 | {
"year": 2019,
"sha1": "411f232822d021a54de5558c4ae53ab2c6be7a30",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1805.01059",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "411f232822d021a54de5558c4ae53ab2c6be7a30",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics",
"Physics"
]
} |
259683852 | pes2o/s2orc | v3-fos-license | A pilot case series for concurrent validation of inertial measurement units to motion capture in individuals who use unilateral lower-limb prostheses
Introduction Inertial measurement units (IMUs) may be viable options to collect gait data in clinics. This study compared IMU to motion capture data in individuals who use unilateral lower-limb prostheses. Methods Participants walked with lower-body IMUs and reflective markers in a motion analysis space. Sagittal plane hip, knee, and ankle waveforms were extracted for the entire gait cycle. Discrete points of peak flexion, peak extension, and range of motion were extracted from the waveforms. Stance times were also extracted to assess the IMU software’s accuracy at detecting gait events. IMU and motion capture-derived data were compared using absolute differences and root mean square error (RMSE). Results Five individuals (n = 3 transtibial; n = 2 transfemoral) participated. IMU prosthetic limb data was similar to motion capture (RMSE: waveform ≤4.65°; discrete point ≤9.04°; stance ≤0.03s). However, one transfemoral participant had larger differences at the microprocessor knee joint (RMSE: waveform ≤15.64°; discrete ≤29.21°) from IMU magnetometer interference. Intact limbs tended to have minimal differences between IMU and motion capture data (RMSE: waveform ≤6.33°; discrete ≤9.87°; stance ≤0.04s). Conclusion Findings from this pilot study suggest IMUs have the potential to collect data similar to motion capture systems in sagittal plane kinematics and stance time.
Introduction
Motion capture equipment is considered the gold standard for collecting gait data, but is impractical to use in clinical practice for several reasons: high costs, lack of portability, and the need for specialized personnel. 1 Clinicians, such as prosthetists and physical therapists, evaluate prosthetic alignment and gait deviations using visual observation. Therefore, the patient's quality of rehabilitation can be subject to factors such as clinician experience, time allotted for the appointment, and patient fatigue. Existing low-cost and portable wearable sensors called inertial measurement units (IMUs) could provide an objective method for practical real-time clinical measurements of gait, such as stance time or sagittal plane asymmetry between intact and prosthetic limbs, that could inform rehabilitation recommendations in individuals who use lower-limb prostheses. 2 For instance, targeting specific muscle groups for strengthening, gait trainings, or prosthetic adjustments. However, IMUs must first be compared to motion capture equipment to determine their validity in prosthetic limbs.
Numerous studies have validated IMUs against motion capture systems in healthy individuals, 3 with a clinical threshold of 5 degrees of error over gait cycle waveforms 4,5 and up to 15°in peak flexion-extension sagittal plane data. 6,7,8,9,10,11,12,13 Several studies have also validated IMUs against motion capture systems in individuals with various lower-limb pathologies, such as stroke, 14 Parkinson's Disease, 15 and knee osteoarthritis. 16 However, there is still a need to assess the ability of IMUs to adequately capture motion given additional considerations specific to this population. For instance, the influence of prosthetic technology (e.g. microprocessor knees) has not been determined. It is possible that prosthetic componentry could interfere with IMU componentry, such as magnetometers, which are known to have interference issues when near metal that produce heading errors. Determining the ability of IMUs to collect prosthetic limb data with or without interference could help inform the integration of IMUs into clinical practice.
Further, as wearable sensors become more feasible to use in research and clinic, there is a need to determine how IMU data compares to previous studies that used motion capture data. In individuals with unilateral lower-limb loss, recent studies have used IMUs to measure symmetry and repeatability, 17,18 coronal plane kinematics, 19 pose estimation, 20 sensory feedback, 21 spatiotemporal parameters (e.g. stance time), 22 walking speed in daily life, 23 and compare IMU algorithms to detect gait events. 24,25,26,27 However, no literature in individuals who use lower-limb prostheses has directly compared IMU-derived data to motion capture, aside from a 2014 case study in a transfemoral user that found IMU-derived knee and ankle flexion data was within 4 degrees of motion capture data. 28 Lower-limb kinematic (e.g. joint angle) and stance time data from motion capture have been associated with adverse clinical events, such as falls, in individuals who use lowerlimb prostheses. 29 Improving symmetry between the prosthetic and intact limb in these parameters is typically considered a primary goal in rehabilitation, as more symmetry is viewed as a reduction in gait deviations. However, these gait parameters rely on motion capture equipment to calculate, 30 which require a large unobstructed space, specialized expertise to run, analyze, and maintain, and is cost prohibitive. These barriers make motion capture systems infeasible for use in clinics. Commercially available IMUs have the ability to detect gait events, yet no research has directly compared these data to motion capture. The ability to quantify lower-limb sagittal plane kinematics and stance time (e.g. asymmetry between intact and prosthetic limbs) of prosthesis users in clinical practice could aid prosthetists and physical therapists with decisions regarding prosthetic alignment, targeted rehabilitation exercises, and potentially insurance justification.
Therefore, IMU-derived kinematic and temporal data were compared to motion capture in individuals who use unilateral lower-limb prostheses. Based on a recent metaanalysis in healthy controls, 3 we hypothesized IMU data on the prosthetic limb, compared to motion capture, would show: 1) full gait cycle waveform root mean square errors at the hip, knee, and ankle ≤5.0°, 2) peak parameters at the hip, knee, and ankle ≤15.0°, and 3) stance durations ≤0.02 s (s). Findings could highlight considerations of using IMUs with lower-limb prosthetic devices and comparing IMU to motion capture-derived data in this population.
Participants
This study was approved under the North Texas Regional Institutional Review Board (#2020-048). Individuals who used unilateral transtibial or transfemoral prostheses were recruited from prosthetic clinics in the Dallas-Fort Worth metroplex to complete a single visit at the University of North Texas Health Science Center. Individuals were included if they: were between the ages of 18 and 95 years, walking with a prosthesis for at least 1 year, and able to walk independently for at least 5 minutes or 100 yards. Individuals were excluded if they had: pain, open wounds or discomfort on their lower-limbs or trunk on the day of testing, limb loss or deficiency on other limbs, major musculoskeletal injury or surgery in the last year besides amputation or revision, or other comorbidities that would make standing, turning, or walking unsafe. All participants provided written informed consent to voluntarily participate in this study, and written permission to use data and images.
Equipment
Seven IMU sensors (iSen, STT Systems, San Sebastian, Spain) were used in this study to collect and store lowerbody IMU data within iSen software (v2022.0). The iSen IMU system was selected due to its low cost option in comparison to other more prominent IMU systems, yet still provided similar specifications in components and accuracy. Each IMU (46 g) had a tri-axial accelerometer (±16°g), gyroscope (±2000°/s), and magnetometer (±13G) with company-reported pitch/roll accuracy within 0.5°and heading accuracy within 2.0°. Calibration prior to each walking trial was performed according to iSen instructions, (i.e. standing straight, facing forward, neutral posture). A 14-camera motion capture system (Cortex, Motion Analysis Corp, Santa Rosa, CA) was used to collect and store data from 32 reflective markers using a modified full-body Helen-Hayes marker set. Calibration prior to each participant's data collection was performed, according to Cortex instructions, with a wand, participant static, and participant dynamic collection. Potential sources of error from using IMU and motion capture equipment generally include movement artifacts between equipment and clothing or skin, as well as environmental factors, such as magnetic interference. All equipment was placed by the same individual (first author, MGF) and all data collections were conducted in the same room. All data was collected at a sampling rate of 100 Hz.
Data collection
Demographic data were collected and managed using Research Electronic Data Capture (REDCap) software hosted at the University of North Texas Health Science Center. 31,32,33 Participants also completed the following clinical outcome measures: Socket Comfort Score (SCS), Prosthesis Limb Users Survey of Mobility 12-Item Short Form (PLUS-M), and Amputee Mobility Predictor with Prosthesis (AMPPRO). After demographic and clinical outcome measure data was collected, researchers placed IMUs and reflective markers on participants as shown in Figure 1. Participants wore fitted clothing with their shirt tucked into their shorts or pants.
IMUs were secured to each participant as recommended by STT Systems for the lower-body model: one on the sacrum, one on each thigh, one on each shank, and one on each foot. IMUs were secured to body segments with elastic straps provided by STT Systems, excluding foot IMUs, which were secured directly to the dorsal aspect of the shoe with velcro. Additionally, due to the conical shape of transfemoral prosthetic sockets, the thigh IMU was secured directly to the transfemoral prosthesis using self-adherent wrap (Coban 1580 Series, 3M™, Saint Paul, MN) to prevent slippage. IMUs and reflective markers on the prosthetic limb were matched to the placement of the intact limb. In a similar manner to the intact limb, the prosthetic knee was bent to visualize a central pivot point for transtibial participants, then compared to the intact limb marker placement for accuracy. The ankle joint marker and knee joint marker for transfemoral participants for the prosthetic limb was placed at the same height as the intact limb and aligned medially. Participants walked across 6 m of level ground at their self-selected habitual walking speed for five trials, with 30 s rest periods between each trial. Figure 2 depicts the process used for data analysis. IMU data was processed to calculate joint angles and gait event detection within proprietary iSen software provided by STT Systems and exported to MATLAB for analysis. Motion capture data was processed using a custom MATLAB code based on Cortex definitions to derive joint angle data, with means and standard deviations in Supplemental Table 1. IMU and motion capture data were time-aligned and vertically aligned. Vertical alignment was necessary to make the necessary comparisons across waveforms for this manuscript, as the motion capture system and IMU system would be calibrated to different starting values inherent to differences in how their systems perform calibration. For instance, what an IMU defines as neutral during a walking trial may actually contain some flexion or extension, based on how the participant was standing when the examiner hit the "calibrate" button before the start of the walking trial. Alignments of IMU and motion capture data were manually confirmed by a qualified study personnel (MGF) through visual inspection. If alignments were not deemed appropriate, that trial was not used for analysis, and a different trial was selected to provide a total of three trials. Three of the five trials were chosen to allow for averaging across multiple steps, and is in line with a previous study that found between one and ten trials are needed depending on the motion capture system used and outcome measures collected. 34 Three of the participant's five walking trials were used for analysis. Typically, the first three trials were selected due to the last two trials showing evidence of IMU slippage distally on the thigh segment, determined by visual inspection of the participant walking and noted during the study visit and IMU drift during visual inspection of the data. IMU slippage was only noted on the prosthetic thigh of participants who used a transfemoral prosthesis. Once three walking trials were selected, the three middle steps of each limb from each trial was extracted to minimize effects from the participant accelerating or decelerating, and to minimize IMU drift or distal slippage that tended to occur towards the end of the walking trial.
Data analysis
Each step, for both IMU and motion capture-derived data, was defined as occurring between two initial contact events. For IMU-derived data, steps were extracted based on the automatic detection of gait events identified within iSen IMU software. Additionally, toe-off events were needed to calculate single and double limb support times, which were also automatically detected within iSen IMU software. For motion capture-derived data, steps were extracted based on the maximum distance between the sacrum marker and heel marker for each limb, and toe-off events to calculate single and double limb support times were extracted using the minimum distance between the toe marker for each limb and the sacrum marker. 35 All gait events were confirmed by a qualified study personnel (MFG) and cut steps were deemed appropriate by inspecting the placement of gait events for hip, knee, and ankle waveforms. Any steps that were not appropriate (e.g. large deviations due to IMU slippage) were not used, and a different walking trial was included to ensure three middle steps for each of the three walking trials were analyzed for each limb of each participant.
Lower-limb sagittal plane kinematic and stance time parameters were extracted from each cut step. Lower-limb sagittal plane waveforms at the hip, knee, and ankle, as well as peak flexion and extension values were extracted across the entire gait cycle. Range of motion was calculated from peak flexion and extension values. Stance time parameters included single and double limb support times for both prosthetic and intact limbs. These parameters were extracted from the three middle steps of three walking trials of each limb, then averaged. IMU and motion capture-derived data were then compared by calculating absolute differences in seconds and degrees. Additionally, root mean square error (RMSE) across the entire waveform and at discrete points of peak flexion and extension at the hip, knee, and ankle during the entire gait cycle were also calculated. RMSE waveform values were a single RMSE averaged across the entire waveform, while RMSE discrete point values were taken at single points. Throughout this manuscript, error and RMSE are defined as the difference between the IMU and motion capture data. Peak parameters and stance durations were selected to assess accuracy (e.g. magnitude of difference between peaks for IMUs and motion capture through RMSE) and precision (e.g. consistency of each peak through standard deviation). Peak parameters and stance durations were also selected to provide the ability to compare to previous studies. 3,30 Further, stance duration was also chosen to examine the ability of IMUs to accurately detect gait events (e.g. foot strike).
Results
Five individuals who use unilateral lower-limb prostheses (n = 3 transtibial users; n = 2 transfemoral users) participated in this study. Demographic characteristics are reported in Table 1. Absolute differences and group (transtibial and transfemoral) RMSE of discrete points between IMU and motion capture-derived data are reported for each participant in Table 2. Average RMSE values between IMU and motion capture-derived waveform data are reported for each participant in Table 3. Representative joint angle waveforms from the middle step are depicted with RMSE values across the entire gait cycle for transtibial users in Figure 3 and transfemoral users in Figure 4. Following our hypothesis, each results section reports all waveform RMSEs, discrete point RMSEs, and then stance time RMSEs for each participant. RMSEs are further discussed based on their context within our hypothesis based on previous healthy control data (5°for waveform RMSEs, 15°for discrete point Figure 2. The flow of IMU (dark blue) and motion capture (mint green) data processing. The left hip sagittal plane joint angles from a walking trial of a representative participant was used as an example. Data was exported to MATLAB (Panel A), time-aligned (Panel B), and vertically aligned (Panel C). Then, the 3 middle steps were extracted using initial contact gait events (Panel D). All data was processed using a custom MATLAB code and confirmed by visual inspection. Figure 3). Participant 1 showed the highest intact RMSE at the knee joint, while Participants 2 and 3 showed the highest intact RMSE at the ankle joint (Table 3, Figure 3). These were driven by differences in peak knee flexion values for Participant 1, peak dorsiflexion values for Participant 2, and peak plantarflexion values for Participant 3 ( Table 2, Table 2).
Transfemoral participants
Prosthetic limbs. For both transfemoral participants, hip, knee, and ankle waveform RMSEs were all ≤3.84° ( Table 3), with the exception of Participant 5's prosthetic knee joint with higher error values of RMSE ≤15.64°, driven by differences in peak knee extension values (Table 2, Figure 4). These higher prosthetic knee joint RMSEs in Participant 5 were caused by magnetometer interference, detailed in the discussion. Double limb support times had RMSEs of ≤0.03s and prosthetic limb support times had RMSEs of ≤0.03s (Table 2).
Intact limbs. Intact limbs had higher error values than prosthetic limbs at the ankle joint (RMSEs ≤3.45°) ( Table 3, Figure 4). Higher RMSE in ankle range of motion were driven by peak ankle plantarflexion values (Table 2, Figure 4). Intact limb support times had RMSEs of ≤0.02s (Table 2).
Discussion
To our knowledge, this was the first study to directly compare gait data between IMUs and motion capture systems in unilateral lower-limb prosthesis users. The iSen IMU system provided similar data compared to other IMU systems used in prior studies, such as Xsens, 6,8,11 Re-haGait, 12 Sparkfun Electronics, 13 and Synertial. 7 IMUs provided similar prosthetic limb data compared to motion capture (RMSE: waveform ≤4.65°; discrete point ≤9.04°; stance ≤0.03s) as previous comparisons in healthy controls (RMSE: waveform ≤5.0°; discrete point ≤15.0°; stance ≤0.02s), 3 with the exception of the microprocessor prosthetic knee joint in one transfemoral user due to magnetometer interference. However, RMSE across participants tended to vary widely, so comparisons across participants should be made with caution. Additionally, prosthetic limb RMSE tended to be less than previous control data from healthy control participants, which could be explained by: 1) differences in how the IMUs were secured to each limb (prosthetic with direct velcro and selfadherent wrap; intact with elastic straps), or 2) intact limb skin and muscle motion that does not occur on the prosthesis. Potentially for similar reasons, lower-limb kinematic IMU data was more similar to motion capture in the stance phase than the swing phase. Overall, findings from this case series indicate IMUs could collect lower-limb sagittal plane kinematic and stance time data to eventually inform rehabilitation (e.g. quantify asymmetry over time to inform considerations for adjustment to the prosthesis over time, prosthetic alignment, muscle group strengthening). It is suggested clinicians not use the IMU system to document absolute values of range of motion (i.e. patient had X degrees of knee hyperextension during gait) but rather for: 1) comparison of range of motion values between intact and prosthetic limbs for symmetry (i.e. patient had X% symmetry) or 2) documentation of absolute differences in stance time in seconds. Future studies should include a larger sample size to determine if findings are generalizable to a larger population of unilateral lower-limb prosthesis users. The portability of IMUs could allow researchers to include participants that have been underrepresented in gait literature (e.g. diabetes, older in age, less active) to better reflect the overall population of lower-limb prosthesis users. 30,36 Prosthetic limbs compared to previous control data Prosthetic limbs had similar or less RMSE than healthy control participants in previous studies. A clinical threshold of 5°(°) of error has been used for motion capture systems and recently applied to IMUs. 4,5 Recent meta-analysis 3 found multi-sensor waveform RMSEs in healthy controls at the hip (2.7-6.3°), 6,7,8,9 knee (0.7-4.6°), 6,7,8,10,11 and ankle (4.0-7.8°). [6][7][8]11 Compared to these values, prosthetic limbs in this study showed similar or lower RMSEs in transtibial participants at the hip (2.05-2.74°), knee (2.77-4.65°), and ankle (1.06-2.09°), and transfemoral participants at the hip (2.23-2.70°), and ankle (1.80°-3.32°). However, Participant 5's RMSEs at the microprocessor prosthetic knee tended to be higher than previous control data (15.64°). Discrete points of peak flexion and extension RMSE in control limbs at each joint have ranged 2.7-15°. 12,13 Prosthetic limbs in this case series had similar or lower RMSEs of 1.55-13.04°, with the exception of the transfemoral prosthetic knee joint (≤40.87°) of Participant 5 due to magnetometer interference with the microprocessor knee.
In the same meta-analysis, IMU single limb support time RMSEs were 0.02s. 6,37,38,39 These values aligned with prosthetic limb support times in this case series of all participants (0.01s -0.03s) except Participant 1 (0.04s), potentially due to increased IMU slippage and motion in the anatomical limb discussed later. This is the first paper to compare stance time calculations between IMU and motion capture for this population. While healthy populations show stance time differences within 0.02s between IMUs and motion capture, the slightly larger RMSE value of 0.04s in this study may suggest either the proprietary algorithm of stance detection for this IMU system is not as accurate as the software algorithm used in the healthy population, 3 or that the software algorithm was less able to detect heel contact and toe-off due to the prosthetic foot exhibiting different properties than an intact foot.
Prosthetic limbs compared to intact limbs
Control limb data mentioned above was similar to intact limb data in this case series. Therefore, prosthetic limbs also had similar or less RMSE than intact limbs, which may have been influenced by how the IMUs were secured to each limb. IMUs on the intact limb were typically secured using STT systems' recommended method of elastic straps, which allowed more slippage distally and movement between the IMU and individual's body segment. However, IMUs on the prosthetic limb were typically secured directly to the prosthesis with velcro and then wrapped tightly in selfadherent wrap to prevent slippage. Additionally, anatomical limbs produce skin and muscle motions that do not occur on the prosthesis. These findings are supported by a 2014 case study of a single transfemoral user that suggested the participant's intact limb RMSE was nearly four times higher than prosthetic limb RMSE due to skin and muscle motions. 18 Transtibial participant comparisons Participant 1 had the most conically-shaped limb, so increased IMU slippage down the thigh and anatomical motion could explain why they had higher RMSE values, particularly at the intact knee joint, than the other two transtibial participants. Of the three transtibial participants, Participant 2 was the youngest and most active, potentially explaining why they had the least RMSE at all joints. Participant 3 was the only participant that used a K2 level foot, which is classified by Medicare as typical for the limited community ambulator. 40 This difference in foot componentry may explain why Participant 3 had higher prosthetic limb RMSE at the knee. They may have employed proximal compensation strategies to ensure prosthetic foot clearance, or intact limb compensations due to the reduced range of motion available in the prosthetic foot, which were observed during their visit.
Transfemoral participant comparisons
The only study that has previously compared lower-limb sagittal plane kinematics between IMUs and motion capture was a 2014 case study that reported knee and ankle RMSEs in one transfemoral prosthesis user. 28 Compared to knee error in the 2014 case study (RMSEs: prosthetic ≤1.0°; intact ≤4.0°), participants in this case series had higher prosthetic knee error (RMSE ≤15.64°) and intact knee error (RMSE ≤6.33°). Additionally, compared to ankle error in the 2014 case study (RMSE ≤2.0°at the prosthetic and intact ankle), transfemoral participants in this case series had similar prosthetic ankle error (RMSEs ≤3.32°) but higher intact ankle error (RMSEs ≤3.45°). These differences between the 2014 case study and this case series may have been due to differences in IMU systems. Differences may have also been due to prosthetic foot componentry, as the prosthetic knee was the same as Participant 5, and the 2014 case study did not list the participant's prosthetic foot.
Participant 4 had larger differences between IMU and motion capture data during hip range of motion compared to all other participants, potentially due to proximal compensatory strategies, such as the circumduction observed during their visit, from undergoing an amputation before reaching walking age. Perhaps these compensatory motions might lead to the IMU not correctly displaying motion that is "out of plane." Participant 5 had a large RMSE at the prosthetic knee joint due to a technical issue with the IMU magnetometer. The IMU magnetometers were disabled in Participants 4 and 5 to avoid interference with the microprocessor knees. Disabling the magnetometer is available under certain IMU systems such as the one utilized in this study and can be as simple as changing the option in the software. However, this is not true for all systems. We suggest clinicians ensure that the IMU system they purchase allows for magnetometers to be disabled. Upon data analysis, a qualified study personnel (MGF) found the prosthetic knee data still looked as if the magnetometers were enabled, so data files were sent to the IMU company, iSen, for inspection. iSen staff concluded there still appeared to be magnetic interference, but could not explain why, since the display showed the magnetometers were disabled. After troubleshooting, our research group at the University of North Texas Health Science Center found the disabling of the magnetometers only took effect if data collection was initiated with the magnetometers disabled. Participant 4 was able to come in for retesting, while Participant 5 could not be retested due to time constraints. Thus, Participant 4's data was collected with the magnetometers actually disabled, while Participant 5's data was collected while the system incorrectly showed the magnetometers were disabled. Therefore, the magnetometer should be disabled when collecting data from participants who use microprocessor knees. The general purpose of magnetometers is to sense the Earth's magnetic field for orientation in direction. In Participant 5, the prosthetic knee waveform indicates a steady oscillation from standing (knee extended) to an atypical amount of nearly 40 degrees of hip extension, then knee flexion for an extended period of time. This oscillation is not reflective of how this participant ambulated during visual observation or video. Disabling the magnetometer still provided data similar to motion capture for Participant 4, and was reflective of the gait of the participant from visual observation and video.
Limitations and future work
For clinical use, while the user interface was generally intuitive, experience with biomechanics data was still required in order to process and interpret IMU data. Another limitation is that motion capture data was collected using a marker set and modeling technique that makes inherent assumptions regarding body segments that do not typically hold true for prostheses (e.g. body segment lengths remain constant during movement). 41,42 However, differences between IMU and motion capture-derived data were present on both intact and prosthetic limbs, and RMSEs were typically higher in intact limbs compared to prosthetic limbs. This suggests both intact limb and prosthetic limb data was similarly represented, regardless of the motion capture marker set and modeling techniques used in this case series.
While steps are commonly selected in biomechanical analysis to represent steady-state walking, it should be noted this would tend to reflect the least problematic data (e.g., potential IMU slippage during the end of the walking trial), potentially impacting clinical implications. For clinical implementation, it is suggested that IMUs are securely instrumented to minimize slippage. This slippage, similar to marker movement, will change the position data and ultimately the joint angles. IMU slippage, however, will have larger impacts to the output joint angle data than motion capture marker movement. Further, IMUs were only compatible with microprocessor knees when magnetometers were disabled. While the process for disabling the magnetometers was simple using our IMU system, this could limit clinical utility of IMUs for insurance justification and rehabilitation in MPK users. Findings should also be generalized with caution, as this study included a small sample size with data from one IMU company, particularly as iSen software is proprietary.
This study included only five participants from a convenience sample that could complete this study in a limited time frame. Future studies should include more participants to determine if our findings are generalizable to a larger sample of unilateral and bilateral lower-limb prosthesis users. Future work could also determine if IMU RMSEs are influenced by: the method in which they are secured to body segments (e.g. elastic strap compared to direct velcro), motion capture marker modeling techniques, IMU settings, or IMU company. These future directions could help inform a recommended clinical data collection protocol.
Conclusions
Findings from this case series suggest IMUs are capable of providing lower-limb kinematic and stance time data comparable to motion capture systems. Prosthetic limbs tended to have less error than intact limbs or previous control limb data, potentially due to increased movement of the IMUs on anatomical limbs. We suspect for similar reasons, IMU-derived lower-limb kinematic data tended to be more similar to motion capture-derived data in stance than swing. However, error varied across participants, suggesting comparisons within individuals may be more accurate. Therefore, we suggest clinicians not use the IMU system to document discrete peaks of range of motion, but rather for asymmetry or comparison across multiple timepoints within participants. Future studies should include larger sample sizes and other IMU systems to assess generalizability of findings in this case series. Clinicians and researchers could eventually use IMUs to collect gait data that better reflects real-world conditions in prosthesis users to help inform rehabilitation. | 2023-07-12T06:58:07.378Z | 2023-01-01T00:00:00.000 | {
"year": 2023,
"sha1": "4fc178168fb1c5ae3002ed30137cfaff18071581",
"oa_license": "CCBYNC",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/20556683231182322",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "ebf101e6cf3fadc6103648b0ab483883101fb7d6",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Medicine"
]
} |
8174682 | pes2o/s2orc | v3-fos-license | Glutamate-Bound NMDARs Arising from In Vivo-like Network Activity Extend Spatio-temporal Integration in a L5 Cortical Pyramidal Cell Model
In vivo, cortical pyramidal cells are bombarded by asynchronous synaptic input arising from ongoing network activity. However, little is known about how such ‘background’ synaptic input interacts with nonlinear dendritic mechanisms. We have modified an existing model of a layer 5 (L5) pyramidal cell to explore how dendritic integration in the apical dendritic tuft could be altered by the levels of network activity observed in vivo. Here we show that asynchronous background excitatory input increases neuronal gain and extends both temporal and spatial integration of stimulus-evoked synaptic input onto the dendritic tuft. Addition of fast and slow inhibitory synaptic conductances, with properties similar to those from dendritic targeting interneurons, that provided a ‘balanced’ background configuration, partially counteracted these effects, suggesting that inhibition can tune spatio-temporal integration in the tuft. Excitatory background input lowered the threshold for NMDA receptor-mediated dendritic spikes, extended their duration and increased the probability of additional regenerative events occurring in neighbouring branches. These effects were also observed in a passive model where all the non-synaptic voltage-gated conductances were removed. Our results show that glutamate-bound NMDA receptors arising from ongoing network activity can provide a powerful spatially distributed nonlinear dendritic conductance. This may enable L5 pyramidal cells to change their integrative properties as a function of local network activity, potentially allowing both clustered and spatially distributed synaptic inputs to be integrated over extended timescales.
Introduction
Pyramidal cells are the principal excitatory neurons in the cerebral cortex and those in layer 5 (L5) form its primary output [1][2][3]. Tufted L5 pyramidal cells integrate synaptic input from local circuits together with long-range inputs from other cortical regions and thalamic nuclei [4,5]. A substantial fraction of the long range thalamic and cortico-cortical inputs form synapses onto the highly branched apical tuft in L1 and L2 [6,7], which is electrically remote from the soma [8]. The tuft therefore receives many different types of signals including information on different sensory modalities, motor control and emotional state [5,9,10]. These inputs are likely to span a wide range of temporal scales, ranging from precisely timed millisecond bursts to more sustained ratecoded signals. This raises the question as to how an individual tufted L5 pyramidal cell combines and transforms such temporally and spatially diverse signals.
In vitro studies show that voltage-dependent synaptic NMDA receptors (NMDARs) can sustain local regenerative responses called NMDAR spikes in the fine dendrites of pyramidal cells [29][30][31][32][33]. Such dendritic thresholding units could substantially increase the computational power of the neuron by enabling the dendritic tree to act like a feed-forward neural network [33,34]. Dendritic recordings and two-photon uncaging experiments in acute slices indicate that activation of NMDAR spikes requires input from a substantial number of clustered synapses [29,31,35,36] and that activation of multiple branches are required to trigger a Ca 2+ spike in the apical dendrite [29,37], which couples the electrically remote tuft to the axon initial segment, where action potentials (APs) are generated [38]. On the other hand, in vivo experiments show that synaptic input evoked by sensory stimuli are dispersed over the dendritic tree of pyramidal cells in visual [39] and barrel cortices [40] as well as spontaneous synaptic input in L5 neurons of motor cortex [41], casting doubt on whether such highly correlated spatio-temporal patterns of synaptic input occur naturally in cortex, although they have been reported in hippocampus [42]. Moreover, NMDAR spikes are also sensitive to inhibition [43,44] and are therefore likely to be less robust in vivo, where inhibition is stronger [16,17], than under in vitro conditions where inhibition is often reduced or absent. Nevertheless, NMDAR spikes have been reported in L4 spiny stellate neurons of barrel cortex following whisker stimulation [45] and in L2/3 pyramidal cells in somatosensory cortex during hindlimb stimulation [46]. Moreover, dendritic patch-clamp recordings from L2/3 pyramidal cells in primary visual cortex suggest that NMDAR spikes contribute to orientation selectivity [47]. Large and widespread Ca 2+ transients have been recorded in the tufts of L5 pyramidal cells when sensory input occurs during motor activity, but evidence suggests that Ca 2+ spikes rather than NMDAR spikes were responsible [10]. Indeed, inhibition of dendritic targeting somatostatin expressing (SOM) interneurons, which have been shown to transiently reduce inhibitory tone in the dendrites of L2/3 cells during active behaviour [48], could underlie the increase in gain of the tuft region of L5 pyramidal cells. These findings show that NMDAR spikes occur in a range of different cortical neurons in vivo, but it is unclear to what extent nonlinear NMDAR conductances contribute to synaptic integration in L5 pyramidal cells and what spatio-temporal patterns these cells respond to in vivo.
We have investigated the potential impact of in vivo-like background network activity on synaptic integration in L5 cortical pyramidal cells using a morphologically and biophysically detailed model that reproduces a wide range of experimentally measured synaptic, dendritic and somatic behaviours [29]. Our simulations show that asynchronous background synaptic input onto the apical dendrites at rates observed in vivo [48][49][50] can profoundly alter the integrative properties of L5 pyramidal cells. Our results suggest that, by activating nonlinear NMDAR conductances distributed over the dendritic tuft, background excitatory input enables pyramidal cells to integrate spatio-temporally distributed patterns of synaptic input in L1 and L2/3 and that the level of background inhibition could tune the spatio-temporal integration window.
Results
To investigate how background network activity influences synaptic integration in tufted L5 cortical pyramidal cells, we adapted an existing model consisting of a detailed morphology ( Fig. 1A) and 9 membrane conductances ( [29] Text S1. Supporting Information). Groups of synaptic inputs were generated with the software neuroConstruct [51] to mimic stimulus-evoked L1/2 input (i.e. cortico-cortico or thalamocortical sensory, motor or emotional) and ongoing background network activity onto the apical dendritic tree. Excitatory synaptic inputs were simulated with random Poisson trains of postsynaptic conductances, consisting of a linear AMPA receptor (AMPAR) component and a voltage-dependent NMDAR component, with peak amplitudes and kinetics that matched experimentally measured quantal excitatory postsynaptic current (EPSC) waveforms and the AMPAR/NMDAR ratio in these cells [29,35,[52][53][54]. Single quantal conductances were used because the ,5 synaptic contacts that typically make up a unitary synaptic connection between two cortical pyramidal cells are usually distributed over different dendritic branches [53]. We first compared integration of excitatory synaptic input in the tuft under quiescent conditions and during modest levels of background excitatory synaptic input, to enable comparison to in vitro experimental results. However, to understand the role of background input under more physiological conditions, we then examined the properties of pyramidal cell integration when background excitation and inhibition were balanced, as observed in vivo.
Number of synapses required to trigger NMDAR spikes in distal dendritic branches
We examined how many synapses onto an individual terminal dendritic branch in the apical tuft were required to trigger a regenerative NMDAR event in the absence of background synaptic input. This condition roughly approximates the in vitro conditions used to study synaptic integration with glutamate uncaging, where background activity is reduced due to long range inputs being severed in acute slices and inhibition is blocked by MNI-glutamate [55]. Since the most likely location for NMDAR spike generation is in the high impendence terminal branches of the tuft [29,35,56], we focused on these dendritic compartments (Fig. 1B). To stimulate a particular branch, synapses were randomly distributed along the branch and near-coincident stimulus-evoked synaptic input was mimicked by driving each synapse with a 200 Hz random train for 5 ms. This resulted in each quantal synaptic input (subsequently referred to as 'synapse') firing once, on average. When few coincident synapses were stimulated, the EPSP in the stimulated distal branch was small and increased approximately linearly from the resting potential (258 mV; similar to 257 mV measured in slices [36]) as the number of synapses increased. However, above 15 synapses, the duration of the EPSP increased dramatically and exhibited a flat top, hallmarks of a local regenerative NMDAR spike (Fig. 1C). We identified the presence of NMDAR spikes in a dendritic branch using a 230 mV voltage threshold criterion (Fig. S1). By stochastically varying the spatial and temporal patterns of synaptic input onto a dendritic branch, and measuring the occurrence of NMDAR spikes over 10 trials, it was possible to build an inputoutput (I-O) relationship for each individual branch. That is, the relationship between the number of stimulated synapses and the probability of generating an NMDAR spike (P(NMDAR spike)). Fig. 1D shows the fits of a sigmoid function (Materials and Methods) to the I-O relationships for all 28 terminal branches in the model (mean length = 72652 mm). Between 10 and 25 temporally coincident synapses were required to trigger an NMDAR spike with a probability of 0.5 in the terminal branches. On average 18 synapses were required and this increased to 30 coincident synaptic inputs to reliably trigger an NMDAR spike on
Author Summary
In the brains of awake animals, networks are active even when there is no input from the outside world. Neurons embedded within cortical networks experience this intrinsic ongoing firing as 'background' synaptic input. While the effect of this background input on the integration properties of neurons has been studied in the cell body region, little is known about how asynchronous background activity affects integration in distal dendrites, which contain nonlinear mechanisms that boost and dampen synaptic input. Our simulations, using a model of a cortical L5 pyramidal cell, show that the nonlinear NMDA receptor conductance activated by distributed background activity could increase the gain of the dendrite, enabling synaptic inputs to be integrated more effectively over the dendritic tree and over longer time intervals than previously thought possible. This mechanism could potentially enable the integrative properties of individual neurons to change as a function of the activity of the network in which they are embedded. Our work suggests that background network activity could play a key role routing and transforming information as it flows through the cortex. every trial. These values are in good agreement with experimental estimates from dendritic recordings [31] and glutamate uncaging experiments on L2/3 [35] and L5 neurons [36] in acute slices and previous modelling studies [29,44]. Given that the average number of spines per terminal branch is approximately 140, assuming a spine density of 2 mm 21 [57], these simulations suggest that nearly coincident activation of a substantial fraction (13-20%) of the synapses on a terminal branch is required to activate a dendritic spike reliably, under quiescent conditions.
Background excitatory synaptic input reduces the threshold for NMDAR spikes
We next examined how asynchronous background excitatory synaptic input affected integration in the tuft. We did not include background inhibition in these initial simulations, in order to study the effects of background excitatory inputs in isolation. We estimated the rate of background excitatory synaptic input in vivo from both anatomical and functional measurements. Tufted L5 pyramidal cells are innervated by ,15,000 excitatory synaptic contacts (assuming 1 per spine), of which 3500 occur on the apical trunk and dendritic tuft in layers 1-3 [57,58]. Since many of these contacts arise cortically from other L5 and L2/3 pyramidal cells, which typically fire at ,1 Hz in vivo [49,50], a L5 pyramidal cell is expected to experience a background quantal synaptic input rate of ,1750 Hz on its apical dendritic tree, assuming a release probability of ,0.5 per synaptic contact [59]. To mimic a modest level of background excitatory synaptic input, we randomly distributed 900-1500 quantal synaptic inputs over the apical tuft (including the Ca 2+ hot zone; Blue region, Fig. 1A) and drove each synapse with a different 0.85 Hz Poisson train, resulting in a background synaptic input rate of 765-1250 Hz. These modest levels of uncorrelated background excitatory input rarely triggered spontaneous NMDAR spikes (,1% of trials, which were excluded from the analysis whenever possible), but did depolarize the dendritic membrane potential by 2-5 mV, within the range of experimental recordings [36,47] (Fig. 1E, F) and generated voltage noise at the soma with a standard deviation of 0.86-1.26 mV.
We next examined how the branch I-O relationship was affected by background excitatory synaptic input. These simulations revealed that many fewer coincident synaptic inputs were required to trigger an NMDAR spike than for quiescent control conditions (Fig. 1G). On average, the number of synapses required to trigger an NMDAR spike with 50% probability fell from 18 to 6 in the presence of 1500 background synapses (Fig. 1H). A lower threshold for NMDAR spike generation and an increased in slope (gain) of the branch I-O relationship was also observed with background input when the time window for synaptic input was increased from 5 to 15 ms and the input frequency scaled down to maintain ,1 quantal conductance per synapse (Fig. 1I). These simulations show that NMDAR spikes can be triggered reliably in terminal branches by much smaller numbers of nearly-coincident synaptic inputs (i.e. activation of 4-7% of the total synapses) in the presence of modest levels of background excitatory synaptic input, than under quiescent conditions.
Background excitatory input decreases the number of stimulated branches required to trigger action potentials To investigate whether background excitatory synaptic input also affects the spike output of the pyramidal cell model, we determined the number of stimulated branches required to trigger somatic APs. To do this we randomly selected a group of N branches within our set of 28 terminal branches in the tuft and simultaneously stimulated each branch with 30 near-synchronous synaptic inputs (5 ms window) to ensure a high probability of triggering an NMDAR spike in each branch under all conditions (Fig. 1H). We then systematically altered the number of stimulated branches and determined whether the neuron fired APs or not. This rather artificial synaptic input configuration was chosen to quantify the impact of excitatory background input when multiple branches were stimulated. Fig. 2A shows an example where stimulation of 4 apical branches produced only a subthreshold somatic potential during quiescent control conditions. Systematically increasing the number of stimulated branches, while averaging across many combinations of branches to minimize branch specific effects, revealed that in the absence of background input 11 near-synchronous stimulated NMDAR spikes were required to trigger APs with 50% probability, and 15 were required to trigger an AP reliably (Fig. 2B).
By contrast, in the presence of background excitatory synaptic input, the relationship between the probability of triggering APs (P(AP)) and the number of stimulated dendritic branches exhibited a lower activation threshold and was much steeper ( Fig. 2A, B). Indeed, the number of stimulated branches required to trigger an AP with 50% reliability was reduced from 11 to ,4 for 1500 background excitatory synapses. The enhanced efficacy of NMDAR spikes was not dependent on the strong activation of individual branches, since reducing the number of synapses per stimulated branch from 30 to 15 still produced a leftward shift in the I-O relationship (Fig. 2B, grey). Thus, even the modest level of background excitatory input used here substantially increased the efficacy of NMDAR spikes in triggering APs. These results suggest that, in the absence of inhibition, modest levels of background excitatory synaptic input arising from ongoing network activity could increase the probability of NMDAR spikes occurring in the dendritc tuft and increase AP generation.
Background excitatory input extends the temporal integration window in pyramidal cells
To examine how the L5 pyramidal cell model integrated inputs on longer time scales than the nearly coincident synaptic input examined so far, we desynchronized the stimulation of different dendritic branches located over the apical tuft. To do this we stimulated each branch with a near-synchronous (5 ms) burst of synaptic input, but activated different branches at random times within a time window that ranged from 5 ms to 200 ms and examined how desynchronization of branch activation affected the probability of triggering APs. In the absence of background excitatory input, 4 stimulated branches produced only a subthreshold response (Fig. 3A1). The average number of stimulated branches required to trigger an AP with 50% reliability gradually increased from 11 to 21 as stimulation of the branches was desynchronized from 5 ms to 200 ms (Fig. 3B). This result suggests that under quiescent conditions, L1 synaptic input would need to be both synchronous and strong enough to trigger NMDAR spikes in a substantial fraction of the terminal branches in order to produce an AP.
By contrast, background excitatory synaptic input substantially reduced the number of stimulated branches required to trigger APs (e.g. 4 branches, Fig. 3A2) and asynchronous branch activation within a 100 ms window became just as effective at triggering APs as coincident activation of the dendritic branches (Fig. 3B). The number of stimulated branches required to trigger an AP only slightly increased when branches were stimulated over a 200 ms window (Fig. 3B, C). The longer the time window over which the NMDAR spikes occurred, the larger the absolute reduction in the number of stimulated branches required to trigger APs with background input, compared to control conditions (Fig. 3C). These results suggest that, in the presence of modest levels of background excitatory input, clusters of synchronous synaptic inputs could be integrated over timescales that are considerably longer than both the membrane time constant of L5 pyramidal cells (,10-20 ms [60]) and the NMDAR-mediated synaptic integration window in individual terminal branches of L2/3 pyramidal cells recorded in acute slices using glutamate uncaging (,10 ms integration window [35]), where background network activity and inhibition were largely absent.
Background excitatory synaptic input enables integration of spatially and temporally distributed excitatory synaptic input To investigate how background excitatory synaptic activity affects the integration of spatially distributed input, we examined the I-O relationship of the model when stimulus-evoked synaptic input was randomly distributed across the entire apical tuft (Fig. 4A), rather than being clustered on selected branches. In the absence of background network activity, near synchronous activation (5 ms duration random burst at 200 Hz) of 100 excitatory quantal synapses typically generated subthreshold responses (Fig. 4B1), while 160 synapses triggered an AP with a 50% probability (Fig. 4C). By contrast, 100 near-synchronous synapses triggered numerous NMDAR spikes and somatic APs in the presence of 1500 background excitatory synapses (Fig. 4B2). Indeed, only 60-70 nearly synchronously activated synapses were required to reach P(AP) = 0.5 in the presence of background excitatory input (Fig. 4C). When the temporal coincidence of the stimulus-evoked input was relaxed from 5 ms to 50 ms (and the input frequency scaled down to ensure the same total amount of excitation per synapse), the number of spatially distributed inputs required to reach P(AP) = 0.5 rose to .200 in the absence of background input (Fig. 4C). However, 60-70 spatially and temporally distributed synapses were still sufficient to reach P(AP) = 0.5 in the presence of background excitatory input, only marginally more than for coincident input (Fig. 4C). Temporally dispersing the synaptic input further to 100 ms (Fig. 4B1,B2) or 200 ms required a greater number of synaptic inputs, but the number of synapses required in the presence of background excitatory input remained approximately 3-fold less than in its absence (Fig. 4C,D). These simulations show that modest levels of background excitatory synaptic input can extend both spatial and temporal integration in a L5 pyramidal cell model, when compared to quiescent conditions.
Synaptic integration during balanced background excitatory and inhibitory synaptic input
So far we have examined how synaptic integration in the apical tuft of a L5 pyramidal cell model is altered by background excitatory synaptic input. However, under physiological conditions ongoing cortical network activity consists of both excitatory and inhibitory conductances in an approximately balanced configuration [15,16,18]. Although our understanding of cortical inhibition is far from complete, our knowledge of the properties of inhibitory synaptic inputs onto pyramidal cells has expanded recently, as a result of a number of in vitro and in vivo studies [48,[61][62][63][64][65][66]. These studies show that Somatostatin (SOM) expressing interneurons (which include Martinotti cells) innervate the dendritic tuft of L5 pyramidal cells with GABA A receptor mediated synaptic input [63], while neurogliaform (NGF) cells, which have dense axonal plexi, form numerous mixed GABA A and GABA B receptor mediated synapses [62]. These dendrite targeting inhibitory interneurons [62,63,67] have been shown to powerfully control dendritic gain [68] and neuronal firing during behaviour [48]. Moreover, distributed inhibition has been shown to be particularly effective in shunting excitation in branched structures such as the dendritic tuft [69].
To examine how synaptic integration in L5 pyramidal cells might operate under more physiological conditions, when network activity delivers balanced excitatory and inhibitory background synaptic input, we added 60 SOM-like GABA A receptor mediated synapses and 13 neurogliaform (NGF) cell-like mixed GABA A / GABA B synapses onto the apical dendritc tree. These inputs were driven at firing rates reported for passive touch whisker experiments in awake animals and are likely to reflect the upper end of the rates observed during active touch, which fall to about half this level on average [48]. The GABA A receptor component had a fast rise and a decay time of 10 ms [66], while the GABA B receptor-mediated K + conductance had a slow rise and decay time of 50 and 80 ms, respectively, with peak values of 0.5 nS and 0.06 nS and an activation delay of 10 ms [70] (Text S1 Supporting information). This produced somatic IPSPs with properties comparable to those measured experimentally [61,62] (C) Probability of triggering action potentials (P(AP)) versus number of stimulated synapses, spatially distributed over the apical tuft. Synapses were driven with random trains with a window of 5 ms, 50 ms, 100 ms and 200 ms, and the frequency was scaled to maintain ,1 quantal conductance per synapse, in the absence (red-yellow traces) and presence of background excitatory input (blue traces). (D) Number of distributed synapses required to trigger an AP with 50% probability (P(AP) = 0.5), for different levels of temporal dispersion in the stimulus evoked synapses (input window), in the absence (red) and presence of background excitatory input (blue), and during AMPAR-only background synaptic inputs which depolarized the dendrites to comparable levels (green). doi:10.1371/journal.pcbi.1003590.g004 (Fig. S2). Although the inhibitory synapses were fewer in number than the excitatory synapses, their higher firing rate and slower kinetics effectively counterbalanced the AMPAR component of the 1500 excitatory synapses, producing a time averaged GABAR/ AMPAR conductance ratio of 1.5, comparable to that estimated from synaptic currents [16,71]. We refer to this experimentally constrained configuration of asynchronous background excitation and inhibition as ''balanced background synaptic input''.
When the balanced background synaptic input included SOMtype and NGF (SOM+NGF) mediated inhibition, the NMDAR spike threshold was reduced compared to quiescent conditions, but not as strongly as during background excitation alone (Fig. 5A). The fact that the NMDAR spike threshold was similar for SOM+NGF and for a SOM-only case where the number of SOMinputs was doubled (26SOM), to maintain the same total GABA A receptor mediated conductance, suggests that the slow lowamplitude GABA B component and the GABA A component are equally effective at inhibiting NMDAR spikes (Fig. 5A). Nearsynchronous activation of 6 branches triggered an AP burst (P(AP) = 0.5) during SOM+NGF mediated inhibition (Fig. 5B1). However, SOM+NGF mediated inhibition was more effective at counteracting the background excitation-mediated increase in AP probability in response to multiple asynchronously activated branches than the 26SOM input alone (Fig. 5C1, C2).
For spatially distributed synaptic input, the number of nearly synchronous inputs required to trigger APs was only marginally higher for 26SOM-only balanced background inhibition than background excitation alone (Fig. 5D1, D2). However, for combined SOM+NGF inhibition, the number of synapses required to trigger an AP burst increased, indicating that the presence of the slow inhibitory component was effective at lowering dendritic gain for spatially distributed synaptic inputs (Fig. 5D1, D2). The fact that NMDAR spikes were evident in the voltage recordings from the terminal branches indicates that nonlinear dendritic integration underlies the AP output during distributed input in the presence of SOM+NGF inhibition (Fig. 5B2). Balanced background inhibition mediated by SOM+NGF was also effective at counteracting asynchronous inputs, increasing the number of spatially distributed inputs requited to trigger an AP burst with P(AP) = 0.5, from 100 with background excitation alone to 170 for a temporal window of 100 ms (Fig. 5D1,D2). Nevertheless, this was still far lower than under quiescent condition when 273 synapses were required. These simulations show that GABA A and GABA B receptormediated inhibitory conductances can counteract the effects of background excitatory input on synaptic integration in the dendritic tuft by raising NMDAR spike threshold and lowering the dendritic gain. Our simulations predict that during periods of lighter dendritic inhibition, for example when both sensory and motor systems are engaged [10,48], the presence of balanced background synaptic input increases dendritic gain and extends the spatio-temporal integration properties of L5 pyramidal cells.
In the subsequent sections we investigate the mechanisms underlying network activity-dependent spatio-temporal integration in L5 pyramidal cells.
Mechanisms underlying the changes in spatio-temporal integration during background excitatory input
Several dendritic mechanisms could be involved in the network activity-dependent changes in synaptic integration we observe, but two are likely to be particularly important. Background excitatory inputs activate glutamatergic synapses that: 1) depolarize the dendrite (Fig. 1F) and 2) generate glutamate-bound but mostly silent NMDARs over the dendritic tree. Depolarization [33], glutamate spillover [72] and glutamate-bound NMDARs from paired-pulse synaptic activation [73,74] have been shown to facilitate the generation of local NMDAR spikes in basal dendrites. In contrast to these relatively local effects, background network activity confers a spatially distributed NMDAR conductance that could provide an additional nonlinear resource for regenerative activity across the dendritic tree.
Since the threshold of NMDAR spikes is voltage-dependent [33], we first examined the effect of the depolarization induced by excitatory background activity on NMDAR spike properties. To do this we applied spatially distributed current injections (Fig. S3A1) that depolarized the apical tree to a level comparable to that during background excitatory synaptic input (Fig. S3A3, inset). Current injections accounted for a substantial part of the change in the singlebranch I-O relationship observed with background input (Fig. S3A2). However, when spatially distributed AMPAR-only synapses were used to depolarize the apical tree to a comparable level of depolarization obtained with 1500 AMPAR/NMDAR synapses, they were less effective at altering the branch I-O relationship, due to the shunt introduced by these conductances (Fig. S3A2). Similarly, AMPAR-only background activity that generated comparable dendritic depolarization only accounted for a fraction of the changes observed in the neuronal I-O relationship in response to clustered (Fig. S3A3) and spatially distributed inputs (Fig. 4C). Moreover, current injections that produced depolarisations comparable to that obtained during background synaptic input at the Ca 2+ -spike initiation zone and at the soma recovered only a small part of the leftward shift in the neuronal I-O relationship (Fig. S3B, C). These results show that, while depolarization is important in determining NMDAR spike threshold [33], when depolarization is mediated by AMPAR-only background synaptic conductance it does not fully account for the lowering of NMDAR spike threshold or the increased efficacy of apical input in triggering APs during background input. This suggests that the spatially distributed NMDAR conductance arising from background network activity is also playing a key role in synaptic integration.
NMDAR component of background synaptic input extends the duration of NMDAR spikes
Comparison of the properties of NMDAR spikes in apical branches during different levels of background excitatory synaptic input revealed that the duration of NMDAR spikes systematically increased with the amount of background activity ( Fig. 6A1-4). The mean duration increased from 45 ms in the absence of background input to 79 ms in the presence of 1500 background synapses ( Fig. 6B1-2). Examination of the distribution of spike duration revealed an increased dispersion and some very long events (.150 ms; Fig. 6C1-2). Depolarization with current injections into the apical tuft accounted for part, but not all, of the increase in NMDAR spike duration and the contribution of depolarization became smaller at higher background rates (Fig. 6D). Replacing the current injections with AMPAR synaptic conductances accounted for less than half of the shift in the decay time distribution with 1500 background excitatory synapses ( Fig. 6D-E). These results are consistent with the idea that both depolarization and the presence of glutamate-bound NMDARs over the dendritic tree contribute to the lengthening of NMDAR spikes during background input.
Propagation and multiplication of dendritic regenerative events during background excitatory synaptic input
The finding that spatially distributed NMDAR conductance arising from background excitatory synaptic input contributes to the lengthening of dendritic spikes, suggests that glutamate-bound NMDARs in the vicinity of stimulated branches are recruited by depolarization. To investigate the extent of this effect, we monitored the voltage in each terminal branch, while triggering an NMDAR spike in only one branch at a time, with and without the background input ( Fig. 7A; Movie S1). In the absence of background input, triggering an NMDAR spike in branch 11 with 30 nearly synchronous synaptic inputs produced a large voltage depolarization (Fig. 7B, lower red trace) that spread into the neighbouring branch (10) producing a similar voltage profile (Fig. 7B, black trace). However, the voltage in more distant branches (12 and 13) was markedly attenuated. When the same branch was triggered in the presence of background excitatory synaptic input (upper red trace) both the duration of the NMDAR spikes in branches 10 and 11 were extended and an additional regenerative event was observed in branch 12 and branch 13 (blue traces), suggesting that glutamate-bound NMDARs had become unblocked in neighbouring branches. Since it is difficult to distinguish between regenerative events and passive voltage propagation in the presence of background input (Fig. 7C), we used a more stringent voltage threshold criterion to identify regenerative events in neighbouring branches (peak background 2peak control $13 mV). The average number of additional activated branches ranged from 0.46 with 900 background excitatory inputs, to 2.19 with 1500 background excitatory inputs (Fig. 7D). These results show that additional regenerative events can be generated in neighbouring branches during background synaptic activity (Movie S1). This nonlinear dendritic amplification potentially explains why fewer stimulated branches are required to trigger APs in the presence of background excitatory synaptic input (Fig. 4A).
Lowered threshold, extended duration and spread of regenerative NMDAR events during background excitatory synaptic input in a passive L5 pyramidal cell model L5 pyramidal cells contain several nonlinear voltage-dependent dendritic conductances that could interact with synaptic input in complex ways. We therefore examined whether any of the basic changes in synaptic integration we observed in the presence of background excitatory input occurred in a passive L5 model, where all the non-synaptic nonlinear conductances were replaced by a uniform passive leak conductance. Although there were some differences between the resting potential of this passive model compared to the active model (the passive model is more hyperpolarized) asynchronous background synaptic input lowered NMDAR spike threshold and elongated the duration of NMDAR spikes (Fig. 8A, B). These results confirm that synaptic AMPAR and NMDAR conductances are sufficient to change NMDAR spike threshold and duration during asynchronous excitatory background synaptic input.
We also examined whether propagation of regenerative events into neighbouring dendritic branches that we observed during background excitatory input could be generated by synaptic AMPARs and NMDARs in the passive model. Fig. 8C shows an example where we stimulated a branch and examined the response in the other distal branches of the tuft. Regenerative events in neighbouring branches were less evident, nevertheless the voltage depolarization in neighbouring dendritic branches during background input was larger than the linear sum of the depolarizations observed with the stimulus alone and the background input alone (Fig. 8C, D). This suggests that, in the presence of background excitatory input, the spread of voltage depolarization from the stimulated branch has a nonlinear regenerative component. Indeed, the duration of the voltage depolarisations in neighbouring branches also increased and some continued to build up well after the peak depolarization in the stimulated branch (Fig. 8E). As for the full active model, regenerative events were observed in neighbouring branches when a single branch was stimulated during background input (Fig. 8F).
These results suggest that synaptic AMPARs and NMDARs activated by background excitatory input can mediate active propagation into neighbouring branches, under certain conditions. However, the presence of dendritic Na + and Ca 2+ conductances in real cells [75] and in the original model are likely to facilitate this process. The spread of NMDAR-mediated regenerative events across branches and the increased decay time are likely to be interlinked: spread of voltage into neighbouring branches recruits NMDAR conductance activated by background input, potentially triggering a regenerative event. Recruitment of NMDAR conductances will help sustain the depolarization, lengthening the NMDAR spike duration, which in turn propagates more effectively through the dendritic tree. Simulations show that during background activity, a few synchronous distributed inputs can trigger an 'avalanche' of activity in multiple branches that feeds and sustains the depolarization for more than one hundred milliseconds. These events, through the recruitment of Ca 2+ currents in the trunk, can then initiate somatic APs (See Movie S2). Propagation of voltage and activation of spatially distributed NMDAR conductance in neighbouring branches can therefore account for the increased efficacy of NMDAR spikes and the extended spatio-temporal integration we observe in the L5 pyramidal cell model, during background excitatory synaptic input.
Effect of inhibition on NMDAR spike initiation, duration and regeneration in the tuft
Having established that depolarization and spatially distributed NMDAR conductances are principally responsible for extending spatio-temporal integration during background excitatory synaptic input, we next examined how inhibition counteracts these effects. Comparison of individual NMDAR spikes during background excitatory synaptic input and during balanced background synaptic input, showed that the presence of SOM+NGF inhibition had little effect on their amplitude but did shorten their duration [44] (Fig. 9A). Inspection of the NMDAR spike decay time distribution showed that inhibition truncated the longer NMDAR spikes, making the distribution less skewed (Fig. 9B). Moreover, SOM+NGF inhibition was more effective than 26SOM at truncating NMDAR spikes, explaining why GABA B receptor inhibition was effective in modulating temporal integration (Fig. 5). However, NMDAR spike duration was still substantially longer in the presence of SOM+NGF based balanced background synaptic input than under quiescent conditions (Kolmogorov-Smirnov test, P,0.05; Fig. 9B). We next examined the effect of background excitation and inhibition on the spread of NMDAR spikes into neighbouring branches. The average fractional increase in the number of branches exhibiting regenerative potentials was reduced from 2.19 to 1.2 when inhibition was added to background excitation (Fig. 9C). Interestingly, SOM+NGF and 26SOM inhibition were equally effective, consistent with their comparable effects on NMDAR spike initiation (Fig. 5A). These simulations show that dendritic inhibition counteracts the effects of background excitatory input by increasing the threshold for NMDAR spikes and reducing spike duration and spread in neighbouring branches (see Movie S1 and S2).
Discussion
Our simulations suggest that the background synaptic input arising from the ongoing cortical network activity observed in vivo, can extend the spatial and temporal properties of dendritic integration in L5 pyramidal cells. Spatio-temporal integration depends on background network activity, because it introduces spatially distributed synaptic NMDAR conductance over the dendritic tree and depolarizes the dendrites. Increasing the background excitatory synaptic input to levels expected in vivo has three main effects: 1) the number of coincident synaptic inputs required to trigger an NMDAR spike in a branch is reduced; 2) the duration of NMDAR spikes is increased, and 3) NMDARmediated regenerative events spread and trigger additional regenerative events in neighbouring dendritic branches. These mechanisms markedly reduce the number of spatially distributed synapses or stimulated terminal dendritic branches required to trigger APs. Inclusion of fast (GABA A ) and slow (GABA B ) receptor-mediated inhibitory conductances in the background synaptic input in an approximately balanced excitatory/inhibitory configuration only partially counteracted these effects. Our results show that in vivo-like balanced background input could enable L5 pyramidal cells to integrate spatially distributed and/or temporally dispersed synaptic inputs onto the tuft more effectively. Our findings suggest that the activity state of the cortical network can dynamically control the integrative properties of L5 pyramidal cells by adjusting the density of glutamate-bound NMDARs on the dendritic tree. This prediction has important implications for cortical processing.
Mechanisms underlying dendritic integration in active networks
Our simulations show that spatially distributed synaptic NMDAR conductances arising from ongoing network activity extend spatial and temporal integration in the dendritic tuft of L5 pyramidal cells. This network activity-dependent nonlinear dendritic mechanism is distinct from the well-documented effects of noise and linear synaptic conductances on neuronal I-O relationships [20,[23][24][25]27]. While AMPAR-mediated conductances do contribute to integration, they summate poorly during asynchronous input due to their rapid decays (t = 2 ms), producing a mild dendritic depolarization, voltage noise and reduced membrane resistance [19]. By contrast, the slow kinetics of NMDARs (t = 70 ms) [54,76] summate effectively, allowing background network activity to set the density of NMDAR conductance on the dendritic tree. The dynamic nature of the spatially distributed NMDAR conductance contrasts with the properties of other spatially distributed voltage-dependent dendritic conductances (e.g. Na + , Ca 2+ and K + channels), which are typically modulated slowly, through plasticity [77][78][79] and homeostatic mechanisms, although the effective number of channels available could change rapidly with voltage due to inactivation.
Several properties of the GluN2A/B containing NMDARs present in L5 pyramidal cells are important for dendritic integration in the presence of background synaptic activity. Their strong block by Mg 2+ below 260 mV [80,81] prevents their involvement in signalling except when the dendritic branch on which they are located is depolarized. This, together with their high affinity for glutamate, means that GluN2A/B-containing NMDARs can remain glutamate bound and 'silent', yet primed and ready to contribute current if the voltage of the dendritic branch depolarizes [73,74]. The steep voltage dependence of the NMDAR conductance introduces a highly nonlinear threshold, Figure 8. Background excitatory input lowers threshold, extends duration and regeneration of NMDAR spikes in a passive model. Fig. 7) in the passive model for a single trial, during nearly synchoronous activation of branch 11 (red traces) in the presence (upper trace; blue) and absence (lower trace; black) of background activity from 1500 excitatory inputs. Linear sum of the depolarization during control and during background input only, shown in green. (D) Peak depolarization induced by an NMDAR spike in branch 11 versus terminal branch number in absence (black) and presence (blue) of background activity. Peak depolarization during background activity alone shown in grey and peak value of linear sum of control during stimulation of branch 11 and background-only shown in green. (E) Voltage traces from branch 11 (red) and branch 7 (blue) in the absence (dashed lines) and presence of 1500 background excitatory inputs (solid lines). Grey and green trace as in (D). (F) Number of additional branches activated during background input when a single branch is stimulated (empty bar). This was estimated by stimulating the terminal branches in turn for 10 trials each and applying the 13 mV criterion, which identified regenerative peaks, although the number of branches exhibiting regenerative events is probably overestimated due to passive spread of voltage. Solid bar shows original model for comparison. doi:10.1371/journal.pcbi.1003590.g008 enabling regenerative depolarization, while the slow kinetics of the conductance extends the time course of local EPSPs beyond the local effective membrane time constant of the fine dendritic branches [31,56]. Our simulations show that these properties of NMDARs enable them to extend spatial and temporal integration in L5 pyramidal cells during in vivo-like network activity. Moreover, our control simulations indicate this effect is robust across a range of NMDAR kinetics: the slower the decay of the synaptic NMDAR conductance the longer the absolute window for temporal integration in the absence of other counteracting conductances (Fig. S4). However, faster NMDAR kinetics would reduce the impact of background synaptic input unless there was a proportionally higher background input rate or a larger NMDAR/AMPAR peak amplitude ratio to maintain the same time-averaged NMDAR conductance. Another key property of NMDARs that sets them apart from nonlinear dendritic Na + conductances [82] is the fact that they remain active during bursts of backpropagating APs and the sustained depolarisations during up-states [83].
(A) Probability of triggering an NMDAR spike (P(NMDAR spike)) in terminal branches versus number of nearly synchronous stimulated synapses in a passive model (solid lines), where all non-synaptic active dendritic conductances have been removed, in the absence (black) and presence of 1500 background excitatory synapses (blue). Original model shown for comparison (dashed lines). (B) NMDAR spike decay time distribution (N = 100, triggered with 30 synapses each) lines as for (A). Excitatory decay times distribution during control and during background activity. Inset, mean voltage (6SD) in randomly selected terminal branches (N = 100) with and without background activity in passive (filled symbols) and original model (open symbols). (C) Membrane voltage traces from all 28 terminal apical branches (as in
By increasing the gain of dendritic integration, background excitation could trigger an unconstrained positive feedback loop, integrating over longer timescales and resulting in prolonged global activation of the dendritic tuft. In practice, cortical microcircuits operate in a balanced configuration with inhibition closely tracking excitation [16,18,48,71,84,85]. Indeed, strong inhibition targeted to the tuft [43,48] may be required to dampen excessive NMDAR excitation during highly active network states and prevent dendritic 'chain reactions' from reaching critical levels. Consistent with this, recent experimental evidence shows that apical inhibition has a stronger effect on the I-O relationship of CA1 pyramidal neurons than somatic inhibition [68]. The apical dendrites of pyramidal cells receive inhibition from SOM expressing interneurons, which include Martinotti cells [48]. SOM-expressing interneurons respond to synaptic inputs from neighbouring pyramidal cells and deliver fast GABA A receptor mediated inhibition to the tuft region [63]. Interestingly, in vivo recordings show that SOM cell firing is suppressed for about 1 s during active whisker touch behaviour, suggesting that the tuft region is disinhibited during this behaviour [48]. Consistent with this, the G exc /G inh ratio is higher during in vivo up-states [13], and engaging the motor and sensory circuits together during whisker touch behaviour results in global tuft activation [10]. Our simulations, which reproduced the excitation/inhibition balance and the firing rates of SOM and NGF cells measured in awake animals [48], show that inhibition reduces NMDAR spike threshold. However, our simulations of asynchronous background inhibition cannot rule out the possibility that fast feed-forward inhibition could veto NMDAR spike generation more effectively if it is present in L1 and L2 and targeted to the same branches as the excitation. Our simulations suggest that slow GABA B -mediated inhibition arising from NGF cells [61,62,65] is particularly effective at truncating NMDAR spikes and reducing dendritic gain. This, together with the fact that the firing rate of NGF cells is elevated during active touch [48], suggests that NGF mediated inhibition may be particularly well placed to control spatiotemporal integration in the tuft (Fig. 5). Such co-variation of background excitatory and inhibitory conductances [85] in recurrent cortical networks [43] is likely to extend the dynamic range over which pyramidal cells can operate by adjusting their integrative properties to match the excitatory drive. These findings extend previous work showing that inhibitory conductances can terminate NMDAR spikes [44] by showing how background inhibition arising from specific interneuron types could counteract the effects of background excitation and shape spatio-temporal integration in L5 pyramidal cells. Our results suggest that during active whisking the lowered level of inhibition [48], could promote NMDAR-mediated distributed signalling, while strong dendritic inhibition could keep it in check at other times, thereby preventing runaway excitation.
Potential interactions between background synaptic input, NMDAR spikes and voltage-gated dendritic conductances
The dendrites of pyramidal cells contain many different voltagegated conductances [86]. While it was not possible to explore all possible interactions between background input and the 9 conductances in our model, the simulations we carried out do point to some general principles of interaction and highlight specific conductances that could play a key role in shaping the properties of spatiotemporal integration. At the most basic level, some excitatory dendritic conductances in the model such as the Na + and Ca 2+ conductances clearly aid background induced regenerative activity (Fig. S5). Indeed, we found that Ca 2+ spikes occur in the tuft during strong tuft stimuli, and help to electrically couple the tuft to the soma, as found experimentally [37]. By contrast, increasing I h reduced the NMDAR spike threshold in the presence of background input, but had negligible effect on NMDAR spike duration (Fig. S5). Indeed, more complex effects are seen when conductances have activation and/or inactivation properties with significant voltage dependences around 255 mV, because background synaptic input depolarizes the dendrite by several millivolts compared to the quiescent condition. For example, transient A-type K + conductances inactivate strongly over this voltage range making their impact weaker than expected (Fig. S6). By contrast, the sustained slow K + conductance (delayed rectifier) strongly activates during NMDAR spikes and truncates their duration (Fig. S6). Increasing the K + conductances to match levels recently reported experimentally [36], increased the threshold of NMDAR spikes, limited their spread to neighbouring branches and was particularly effective at reducing their decay time (Fig. S7). However, the time course of the resulting dendritic spikes generated under these conditions was markedly different from NMDAR spikes recorded experimentally in L5 pyramidal cells [29,36] and another recent study estimated much lower densities of K + conductances in the tuft region [87] comparable to that in the original L5 model [29]. Nevertheless, slow K + conductances appear well placed to modulate spatio-temporal integration in L5 pyramidal cells. SK type channels in spines, which are activated by local Ca 2+ influx though NMDARs [88], may also truncate NMDAR spikes (Fig. S8). Thus the level of expression of the various dendritic conductances and their precise voltage and Ca 2+ dependencies are likely to tune the spatiotemporal integration. Irrespective of the natural configuration of the dendritic conductances present in vivo, our results using a passive L5 model, show that the spatially distributed NMDAR conductance arising from ongoing network activity will extend spatio-temporal integration unless other dendritic conductances are specifically configured to counteract this basic property. Indeed, spatially distributed glutamate-bound NMDARs arising from ongoing network activity extend the toolbox of dendritic conductances available to pyramidal cells, potentially enabling them to perform a wider range of behaviours.
Enhanced spatial and temporal integration during background network activity
The fine dendrites of pyramidal cells have traditionally been thought to act as coincidence detectors on fast time scales [89], due to their fast local membrane time constant [31,90]. However, when depolarized, slowly decaying synaptic NMDAR conductances set the dendritic EPSP waveform, extending the integration window to tens of milliseconds [35]. The voltage dependent properties of NMDARs have also been shown to extend the time window for integration, underlying the supra-linear response of the second of a pair of synaptic inputs onto the basal dendrites of L5 pyramidal cells [32]. Moreover, temporal integration can be extended further when pairs of bursts are used [73,74]. Our simulations extend these findings by showing that a spatially distributed NMDAR conductance arising from in vivo-like background synaptic input can extend both the spatial and temporal integration in L5 pyramidal cells by providing both depolarization [27] and a spatially distributed NMDAR conductance that sustains the NMDAR spike plateau depolarization. Interestingly, these mechanisms act in the opposite direction to passive membrane properties, which reduce the temporal integration window during elevated synaptic input, due to increased membrane shunting by synaptic conductances.
Our simulations show that background synaptic input can lower the number of coincident synaptic inputs required to trigger an NMDAR spike in a distal branch. This suggests that NMDAR spikes are much more likely to occur in vivo during network activity than in vitro when the network is largely quiescent. Another important consequence of spatially distributed NMDAR conductance arising from background network activity is that it enables active propagation of NMDAR-mediated regenerative events into neighbouring branches, resulting in a multi-branch regenerative plateau potential. This prediction is consistent with recent in vivo recordings of NMDAR spikes in L2/3 pyramidal cells, which reported that 83% of events occurred in multiple branches [46]. Spatially distributed NMDAR conductance is therefore well suited for integrating excitatory input that is distributed across the branches of the dendritic tree [39,40]. Our simulations show that under in vivo-like levels of background input our L5 model could integrate spatially and temporally dispersed synaptic input. This suggests that spatial clustering of synaptic inputs required for triggering NMDAR spikes under quiescent in vitro conditions [35,36,64] may not be necessary in vivo. Hence, while spatial clustering of synaptic input onto the dendrites of pyramidal cells has been found [42,91], our results suggest it is not a prerequisite for activating the tuft of L5 pyramidal cells. Indeed, our finding that network activity extends spatio-temporal integration may explain how NMDAR spikes contribute to orientation selective tuning in L2/3 pyramidal cells in mouse V1 [47], when the synaptic inputs onto these cells are spatially distributed [39].
Widespread NMDAR spikes have been reported in the dendrites of L4 spiny stellate neurons of rat barrel cortex, during whisker deflection [45]. The multiple hot spots of Ca 2+ influx observed are consistent with a small number of distributed thalamo-cortical inputs being amplified by the dendritic NMDAR conductance arising from local network activity, but alternative scenarios are also possible. In L2/3 pyramidal cells in somatosensory cortex, both spontaneous and hind paw-evoked NMDAR spikes are localized to ,30 mm regions of the dendritic tree and usually encompass multiple branches. These events are effective at triggering APs but do not appear to involve a dendritic Ca 2+ spike [46]. More widespread [Ca 2+ ] signals have been observed in the tuft of L5 pyramidal cells in awake animals performing active whisker touch, when vibrissal sensory input is combined with L1 input arising from the primary motor cortex [10]. Similar tuft responses could be reproduced in vitro when apical trunk Ca 2+ spikes were paired with dendritic tuft depolarization or synaptic input. This suggests that in vivo tuft plateau potentials and Ca 2+ influx are caused by apical dendritic Ca 2+ spikes spreading into the tuft when it is depolarized by L1 input. Although, Ca 2+ conductances are clearly important [10], network activity-dependent distributed NMDAR conductance could also contribute to this widespread Ca 2+ influx. Indeed, recent in vivo recordings from CA1 pyramidal cells show that both NMDARs and Ca 2+ channels are required to generate the slow, widespread regenerative dendritic events that underlie complex spike bursts [92]. These results suggest that the densities of both glutamate-bound NMDARs and Ca 2+ channels can set the spatial spread of regenerative dendritic events in pyramidal cells. Our results highlight the role of spatially distributed NMDAR conductances in pyramidal cells under in vivo conditions, by showing how they could enable these cells to integrate both spatially clustered and spatially distributed inputs over longer timescales than previously thought.
Implications for cortical processing
The finding that synaptic input and nonlinear membrane conductances can produce local regenerative dendritic events has lead to the idea that individual dendritic branches can operate as independent thresholding units [33,73,93] enabling a pyramidal cell to act as a powerful pattern separator [34]. In this scenario, individual branches could act as feature detectors [77] and neuronal firing could report the coincident occurrence of these features. However, our results showing that NMDAR spikes can spread into neighbouring dendritic branches suggest that they may not operate independently and are more effective at triggering APs than anticipated, consistent with recent experimental findings [46]. This enhancement in spatio-temporal integration during network activity is likely to come at the cost of loss of power to discriminate different spatial patterns of synaptic input. That said, our results do not exclude the possibility that L5 pyramidal cells can still act as coincidence detectors and perform pattern discrimination, since our simulations show that altering the level of dendritic inhibition or the density of voltage-gated channels on the dendrite can tune the spatio-temporal properties of the tuft. Coincidence detection and pattern discrimination could be enhanced by raising the threshold for dendritic spikes, truncating their duration and decoupling dendritic branches. This could be achieved by expressing high levels of K + channels [36,94], which have been shown to compartmentalize dendritic responses in CA1 pyramidal cells and through experience-dependent adjustment of synaptic weights onto specific branches through NMDARdependent Ca 2+ influx [95]. Moreover, recent work suggests that GIRK channels can also modulate NMDAR spike duration in a branch specific manner [78]. Our results suggest that these slowly modulated dendritic mechanisms are complemented, in vivo, by faster dynamic modulation conferred by the background synaptic excitation and inhibition arising from ongoing network activity. Thus, both the past and the present activity state of the network determines which spatio-temporal patterns of synaptic input a L5 pyramidal cell can respond to.
Network activity-dependent spatio-temporal integration could have important implications for cortico-cortical and thalamocortical signalling. While global network activity is maintained at a relatively constant level due to tight inhibitory control [16,18,71,85], the activity of sub-networks of excitatory neurons is likely to vary widely, due to common receptive field properties and preferential local synaptic connectivity [96]. If the elevated activity in L2/3 pyramidal cell ensembles observed during sensory input [96] increases synaptic input onto the apical dendrites of subsets of L5 pyramidal cells [4,97], the resulting increase in NMDAR conductance will enable those L5 pyramidal cells to integrate both spatio-temporally clustered and dispersed signals. This property suggests that the tuft could combine different modalities, such as motor, sensory and emotional state, which span a wide range of timescales. This effect is predicted to be particularly strong when the apical tuft is disinhibited, as has been observed when motor and sensory systems are engaged together [48]. Although speculative, network-dependent enhancement of spatio-temporal integration could also enable L5 pyramidal cells to communicate selectively across sensory and motor cortical areas, in a context dependent manner. Such reconfigurable functional network connectivity has been reviewed in the context of neuronal gain modulation [98] and could be involved in high-level cortical function including attention [76,99,100] and learning [9]. Our results strengthen these concepts by providing a new mechanism that links network activity to the spatio-temporal properties of dendritic integration.
Materials and Methods
For full methodological details see Supplemental Experimental Procedures.
The L5 pyramidal cell model
The version of the L5 pyramidal cell model [29] available from ModelDB (Accession number: 124043) was used as a starting point for our simulations. Synaptic input was added and all simulations were generated in neuroConstruct (neuroConstruct.org) [51] and run in the NEURON 6.2 simulation environment [101]. The default parameters of the model are listed in Text S1 and any changes are noted in the figure legends. Simulations were carried out with a fixed integration time step of 0.025 ms and voltages were recorded from the soma and the proximal segment of terminal dendritic branches.
Synaptic input
The AMPAR conductance (gAMPA) component of each quantal glutamatergic synapse was modelled with an instantaneous rise and a single exponential decay: Where the maximal peak conductance (g max AMPA) was 0.5 nS and the decay time constant (t decay ) was 2 ms [56].
The NMDAR conductance (gNMDA) component of each quantal glutamatergic synapse was modelled with an exponential rise and decay, and a voltage-dependence function to mimic Mg 2+ block: g max NMDA is the maximal peak unblocked conductance (1 nS). Time constants for the rise (t rise ) and decay (t decay ) were 3 ms and 70 ms, respectively [76,97]. The NMDAR model also included Ca 2+ -dependent inactivation (CDI) of the NMDAR current. This was implemented by multiplying the gNMDA by e 2h , where the change in h at each time step (h9) is as follows: and t h = 1000 ms.
Quantal GABAergic synapses were modelled with an exponential rise and decay: The fast GABA A receptor component had a maximal peak conductance (g max GABA) of 0.5 nS, a t rise = 0.3 ms and a t decay = 10 ms [66]. A is the peak normalization factor as defined above and the Cl 2 reverse potential was 275 mV. Synapse arising from neurogliaform (NGF) cells, also had a slow GABA B receptor component with a t rise = 50 ms and a t decay = 80 ms [70] and an onset delay of 10 ms to account for the activation of the K + channels. The maximal peak conductance (g max GABA = 0.06 nS) was estimated from [61] who reported that the GABA B component accounts for ,1/5 of the peak IPSP for NGF cell to L5 pyramidal cell connections. The conductance was obtained by scaling 0.1 nS by the ratio between the driving forces of the GABA A and GABA B components using a K + reverse potential of 287 mV.
Two groups of synaptic input were defined. Those that mimicked a stimulus-evoked event and those that mimicked asynchronous background network activity. For simulations where specific dendritic branches were stimulated, one of the 28 terminal branches was randomly selected and the stimulus evoked quantal synaptic inputs randomly placed along it. For spatially distributed stimulus-evoked inputs, quantal synaptic inputs were randomly distributed over the entire apical tuft (Blue region, Fig. 1A) Each stimulus-evoked input was driven by an independent Poisson train with a frequency that was inversely related to duration of the stimulus, thereby maintaining an average of a single quantal conductance per synapse, independent of stimulus duration. For background excitatory synaptic input, 900-1500 glutamatergic synapses were placed randomly over the apical dendritic region (starting from ,500 mm from soma). Each of these synapses was driven independently with a Poisson train with a mean rate of 0.85 Hz. Background inhibitory synaptic input was implemented either by 130 GABA A receptor mediated synapses, each driven with a 3 Hz random train (26SOM configuration) or by a mix of 65 pure GABA A receptor-mediated synapses firing random trains at 3 Hz and 13 GABA A /GABA B receptor-mediated connections (10 synapses each) firing random trains at 14 Hz (SOM+NGF configuration). In both cases this gave a time averaged GABA A receptor conductance of 1.5 times the background AMPAR conductance, comparable to that measured experimentally [16,71].
Analysis
Voltage traces from the simulations were analysed in IGOR Pro using NeuroMatic software (http://www.neuromatic. thinkrandom.com/). NMDAR spikes were identified with a threshold crossing criteria of 230 mV ( Fig. S1; unless stated otherwise). Dendritic branch and neuronal I-O functions were fit using a sigmoid function with the form: where n is the rate, P 0 is the offset, P max is the maximum probability of activation and x 50 is the number of synapses or stimulated branches at which P(x) reaches half maximum. Errors are presented 6 standard deviation. Distributions were compared with the Kolmogorov-Smirnov test and considered significant at the p,0.05 level. Figure S4 Effect of changing amplitude and time course of the NMDAR synaptic conductance on NMDAR spikes. (A) Average probability of triggering an NMDAR spike (P(NMDA spike)) in a terminal branch versus number of nearly synchronous stimulus-evoked synapses for peak synaptic NMDAR conductances of 1.0 nS (blue) and 0.5 nS (red) with (filled marker, solid lines) and without (empty markers, dashed lines) background activity from 1500 excitatory synapses. Because of the higher AMPA/ NMDA ratio, NMDAR spikes were identified by eye from their decay time (.15 ms) rather than peak depolarization for this simulation. (B) P(NMDA spike) versus number of nearly synchronous stimulus-evoked synapses for quantal NMDAR conductance amplitudes of 1.0 nS (blue) to 1.5 nS (red) with (filled marker, solid lines) and without (empty markers, dashed lines) background activity of 900 excitatory synapses (reduced because increased gNMDA triggered NMDAR spikes with background activity from 1500 excitatory synapses). (C) P(NMDA spike) versus number of nearly synchronous stimulus-evoked synapses for NMDAR decay time constants of 70 ms (original, blue) and 35 ms (red) with (filled marker, solid lines) and without (empty markers, dashed lines) background activity from 1500 excitatory synapses. Note: to compensate for the faster decay kinetics the background input rate was doubled (1.7 Hz) in order to deliver the same time averaged NMDAR conductance and depolarization. (D) NMDAR spike decay time distributions (N = 100 across randomly selected branches) during control (dashed lines) and during background excitatory activity of 1500 synapses (solid lines), in the original model (70 ms; blue) and in the model with faster NMDAR kinetics (35 ms; red). With a decay constant of 35 ms NMDAR spikes are significantly shorter, both during control and during background excitatory activity. However, the increase over control produced by background was significant (Kolmogorov-Smirnov test, P,0.05). (TIF) Figure S5 Effect of changing dendritic Na + , Ca 2+ and I h conductances on changes in NMDAR spike threshold and duration during background excitatory activity. (A1) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulus-evoked synapses in the original model (black) and in a model where Na + channels were removed from the apical tree (red) measured in the absence (empty markers, dashed lines) and presence of background activity from 1500 excitatory synapses (filled markers, solid lines). (A2) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (A1). (B1) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulus-evoked synapses in the original model (black) and in a model where L-type Ca 2+ channels (Ca-L) were removed from the apical tree including Ca 2+ initiation zone, during control (black) and background excitatory activity (B2) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (B1). (C1) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulus-evoked synapses in the original model (black) and in a model where Ih current was exponentially increased from 0 to 40 pS/mm 2 along the apical tuft (red) measured in the absence (empty markers, dashed lines) and presence of 1500 background excitatory inputs (filled markers, solid lines). (C2) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (C1). (TIF) Figure S6 Effect of increasing dendritic K + conductances on changes in NMDAR spike threshold and duration during background excitatory activity. (A1) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulus-evoked synapses in the original model (black) and in a model where K + A-type (K A ) channels density was doubled (60 pS/mm 2 ) in the apical tree (red), measured in the absence (empty markers, dashed lines) and presence of 1500 background excitatory input (filled markers, solid lines). (A2) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (A1). (B1) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulus-evoked synapses in the original model (black) and in a model where K + delayed rectifier (K dr ) channels density was increased ten-fold (10 pS/mm 2 ) in the apical tree (red), during control (empty markers, dashed lines) and background excitatory activity (filled markers, solid lines). (B2) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (B1). (TIF) Figure S7 Effect of increasing K + conductances on the dendritic tuft to match levels reported by Harnett et al. 2013. (A) Average probability of triggering an NMDAR spike (P(NMDA spike)) versus number of nearly synchronous stimulusevoked synapses in the original model (black) and in a model where apical K A density was increased to 77 pS/mm 2 and apical K dr density to 23 pS/mm 2 to match estimates from [36] (blue), during control (empty markers, dashed lines) and background excitatory activity (filled markers, solid lines). (B) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (A). The decay time was computed as the time when the voltage decayed to 30% of the peak depolarization (instead of 37%) to account for the fluctuations present when the K + conductance was increased (see G). (C) Average profile of NMDA spikes (N = 100) triggered in the original model (grey) and in a model with K + conductances equal to 1 and 2 times the levels reported by [36] (blue and red respectively). Twice as much K + conductance was required to completely counteract the background NMDAR component, matching the average NMDA spike recorded during background excitation with 6000 AMPAR-only synapse, which matched the depolarization during mixed AMPAR+NMDAR excitatory background (green). Note: the time course of this dendritic spike is markedly different from NMDA spikes recorded experimentally in L5 pyramidal cells [29] (D) Average number (N = 28) of additional branches activated during background input 6 SE when a single branch is nearly synchronously stimulated in the modified model (blue) compared to the original model (grey). (E) Example voltage traces from branch 12 (red) and branch 9 (blue) in the absence (dashed lines) and presence of 1500 background excitatory inputs (solid lines). (F) Peak depolarization induced by an NMDAR spike in branch 12 versus terminal branch number in absence (empty marks, dotted line) and presence (filled marks, solid line) of background activity. (G) Local voltage profile of an NMDA spike (dark blue, upper panel) triggered in the modified model by the near synchronous activation of 30 glutamatergic synapses. Current generated by each of the 30 NMDAR synapses (light blue, middle panel) compared with the current densities of the three major ions (Na + in green, K + in red and, Ca 2+ in yellow). Note: another recent study [87] estimated a much lower density of K dr (3 pS/mm 2 ) in the apical dendrites of L5 pyramidal cells. (TIF) Figure S8 Effect of Ca 2+ -dependent local feedback mechanisms on NMDAR spikes and spatio-temporal integration. (A) Individual quantal synaptic NMDAR conductance with Ca 2+ -dependent inactivation (CDI) present in the original model (black trace) and when CDI was removed (red line), both recorded at 260 mV. (B) Ca 2+ current through the NMDAR quantal conductance with (black trace) and without (red line) CDI at 260 mV, together with scaling factor for NMDA conductance (lower plot). (C) Branch voltage during a single NMDAR spike with (dashed traces) and without (solid traces) CDI of NMDARs, in the absence (red) and presence of background activity from 1500 excitatory synapses (blue). The effect of CDI became evident during the longer plateau potential, because Ca 2+ accumulation lead to shortening of NMDAR conductances. (D) NMDAR spike decay time distribution (N = 100 trials across randomly selected branches) for conditions in (C). During background excitatory input the average NMDAR spike duration increased 4.8 fold with background excitation without CDI, compared to 1.4 in the original model. (E) Average NMDAR spike decay time in the absence (control, open markers) and presence of background excitatory input from 1500 excitatory synapses (solid marker), with different levels of negative feedback implemented with an SK-like Ca 2+dependent K + conductance at each synapse (SK syn ). SK syn at a density of 40 pS/synapse reproduced the average decay time in the original model (orange symbol). (F) Cumulative distributions of NMDAR spike decay times with different levels of SK syn (light blue 0 pS, dark blue 40 pS, red 120 pS) in the absence (control, dashed lines) and presence of background excitation (solid markers). All levels tested show a significant increase during background excitation compared to control (Kolmogorov-Smirnov test: P,0.05). (G) Average probability of action potential (P(AP)) (N = 20 trials) with different levels of negative feedback (light blue 0 pS, dark blue 40 pS, red 120 pS) in the absence (control, open markers, dashed lines) and presence of background excitatory input (filled symbols, solid lines), for different numbers of distributed synapses stimulated with random trains at 10 Hz for 100 ms.
(TIF)
Movie S1 Movie 1 shows the membrane voltage in the Layer 5 Pyramidal cell model during nearly synchronous stimulus evoked synaptic stimulation with 30 synapses on a terminal dendritic branch (each synapse firing at 200 Hz from 100-105 ms). The voltage response to the synaptic input is shown under 3 conditions: 1) Control with no background activity. 2) With background activity from 1500 excitatory synapses firing randomly at 0.85 Hz. 3) With balanced background synaptic input including inhibition from SOM-like (GABA A ) and NGF-like (GABA A and GABA B ) mediated inhibition. While the spread of depolarisation in the control case is limited to the stimulated branch and its near neighbours, the excitatory background input causes regenerative depolarisation in branches further away. The spread of depolarisation is also widespread for balanced background synaptic input, but the presence of inhibition shortens the duration compared with background excitation alone. (AVI) Movie S2 Movie 2 shows the membrane voltage in the Layer 5 Pyramidal cell model during nearly synchronous stimulus evoked synaptic stimulation of 100 synapses spatially distributed over the apical tuft (10 Hz from 100-200 ms) under 3 conditions: 1) Control with no background activity. 2) With background activity from 1500 excitatory synapses firing randomly at 0.85 Hz3) With balanced background synaptic input including inhibition from SOM-like (GABA A ) and NGF-like (GABA A and GABA B ) mediated inhibition. Only in the case of stimulation in the presence of excitatory input did the spatially distributed synaptic input trigger somatic action potentials which back propagates into the apical tree. (AVI) Text S1 Supporting information. Supplemental methods, results, table and references. (DOCX) | 2016-01-22T01:30:34.548Z | 2014-04-01T00:00:00.000 | {
"year": 2014,
"sha1": "0dcf679e2132b76e9e8deb2a750623539e5d89b2",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/ploscompbiol/article/file?id=10.1371/journal.pcbi.1003590&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "0dcf679e2132b76e9e8deb2a750623539e5d89b2",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
16988737 | pes2o/s2orc | v3-fos-license | Microwave-Assisted Synthesis of Highly-Crumpled, Few-Layered Graphene and Nitrogen-Doped Graphene for Use as High-Performance Electrodes in Capacitive Deionization
Capacitive deionization (CDI) is a promising procedure for removing various charged ionic species from brackish water. The performance of graphene-based material in capacitive deionization is lower than the expectation of the industry, so highly-crumpled, few-layered graphene (HCG) and highly-crumpled nitrogen-doped graphene (HCNDG) with high surface area have been introduced as promising candidates for CDI electrodes. Thus, HCG and HCNDG were prepared by exfoliation of graphite in the presence of liquid-phase, microwave-assisted methods. An industrially-scalable, cost-effective, and simple approach was employed to synthesize HCG and HCNDG, resulting in few-layered graphene and nitrogen-doped graphene with large specific surface area. Then, HCG and HCNDG were utilized for manufacturing a new class of carbon nanostructure-based electrodes for use in large-scale CDI equipment. The electrosorption results indicated that both the HCG and HCNDG have fairly large specific surface areas, indicating their huge potential for capacitive deionization applications.
Scientific RepoRts | 5:17503 | DOI: 10.1038/srep17503 also is one of the largest volume applications for membrane technology, and most new installations for seawater desalination will use (or at least consider) RO technology.
However, despite improvements in different approaches and the easy accessibility of seawater, most of the desalination methods have not achieved a great deal of success in terms of water treatment, implying the need for more effective and economical methods.
Numerous innovative approaches have been applied to decrease the cost of seawater desalination 5 . One new promising method is electrosorption with porous electrodes that was introduced as a suitable route for eliminating different ions from aqueous media as well as suspensions 6 . Interestingly, electrosorption can operate at low direct voltages and powers, which distinguishes this method from other suggested methods. Also, electrosorption generates no secondary pollution, which is common in other conventional techniques such as thermal approaches, RO, and distillation columns 7 .
The recently-improved electrosorption can remove charged ionic species from seawater as a cost-effective and energy-efficient method [8][9][10] . The fundamental theory of electrosorption comprises making the ions move toward oppositely polarized electrodes via applying an electric field 7 . By placing the electrodes with appropriate surface area and high electrical conductivity into the direct electrostatic field, they easily can generate an intense electrical double layer for absorbing charged ions from solutions 7,9,11 . More beneficially, as the electric field is eliminated, the ions absorbed on the surface of electrodes are effortlessly emancipated 11 . Its characteristics of low energy consumption, being an environmentally-friendly process without harsh chemicals, easy regeneration of electrodes, and the cleaning process make it attractive for use in the field of water treatment. To enhance the electrosorptive procedure, highly-porous materials, such as carbon nanotubes, carbon fibers, carbon aerogels, and graphene, are highly-desired materials for use in electrosorptive electrodes or capacitors 11,12 . High conductivity, appropriate pore size distribution, and specific surface area are the essential factors for realization of high-performance water-treatment procedures with the use of carbon nanostructures 5,6,12 .
Recently, graphene, due to its many promising properties, has been introduced as a unique material and has received considerable attention in the scientific community [13][14][15][16][17][18] . In particular, large specific area, outstanding tensile modulus, and high electrical conductivity at room temperature make graphene an ideal material for electrosorption and for the fabrication of electrosorption electrodes 19 . The suitability of graphene for NaCl electrosorption from aqueous suspensions recently was suggested in several studies [9][10][11][20][21][22][23][24] . Also, the specific electrosorptive capacity of graphene is about 23.18 μ mol/g, which is quite significant 11 . Developing an economical and industrially-scalable method for synthesizing graphene, however, is critical for its large-scale applications.
It is well known that electrosorption capacity improves as the surface area of carbon nanostructures increases. Actually, the large electrosorptive capacity of graphene definitely is highly dependent on the geometry, level of charge separation at the interface, and surface properties of the electrode materials 8 . Therefore, to manufacture high-performance electrodes for electrosorptive applications, materials with high surface areas, such as crumpling graphene, in particular few-layered graphene, can be more suitable. Crumpling the 2D graphene and nitrogen-doped graphene (NDG) with chemical treatment can be an approach to achieve materials with high surface area and pore structures. Thus, highly-crumpled, few-layered, nitrogen-doped graphene and pure graphene with large surface area seem to have these properties. Also, NDG is promising material due to the desirable pseudo-capacitance obtained from the quick Faradic redox reactions among the nitrogen groups and the ions in the electrolytes 20,25 . Furthermore, the presence of nitrogen in the graphene frameworks could increase the electrical conductivity 26 . Nitrogen doping can also increase the wettability of the interface between the electrolyte and electrodes 27 . Xue et al. 20 showed the specific capacitance of NDG to be much higher than that of pristine graphite. Three reason for having higher specific capacitance, i.e., increased specific surface area 11 , increased pseudo-capacitive effect of electrodes by nitrogen doping and improving the wettability of the interface between the electrolyte and the electrode 27 , were introduced for NDG. Thus, the combination of highly-crumpled graphene and nitrogen doping can be effective novelties for increasing the specific capacitance of electrodes.
To this end, a promising and efficient procedure was developed to fabricate highly-crumpled, few-layered graphene (HCG) and nitrogen-doped graphene (HCNDG) nanosheets with large surface area. Also, the method of synthesizing HCG and HCNDG seems to be novel, simple, cost-effective, and industrially scalable. HCG and HCNDG, due to their unique specific surface areas, provide significant improvement in terms of electrosorptive performance.
Material and Methods
Synthesizing HCG and HCNDG. Figure S1 shows the schematic procedure of exfoliation of graphite flakes to HCG and HCNDG with cyanamide. The mechanism for the exfoliation of the graphite to HCG and HCNDG includes the generation of semi-stable diazonium ions, which then initiate a radical reaction with the flakes. Then, treated graphite with higher dispersibility in dimethylformamide (DMF) opens a new gateway for expanding graphite to graphene flakes by placing between layers. Pristine graphite and deionized water were poured into a vessel. The reaction vessel was sonicated after adding cyanamide and isoamyl nitrite, and the contents were poured into a Teflon vessel for performing the reaction under microwave irradiation. Then, the mixture was mixed vigorously, and, after cooling to room temperature, it was filtered and washed. Thus, a semi-stable diazonium ions were produced, resulting in a radical reaction with the graphite flakes. The high-dispersibility product, due to functional group, was sonicated for 1 hr in DMF. The resulting black ink-like dispersion was left to sit for 24 hr to separate the large, unstable graphite aggregates. The dispersed cyanamide-treated graphene was collected using low-speed centrifugation. The centrifuged supernatant was dried. The resulting HCG, comprised of graphene flakes, was stored for further treatment. Then, to prepare HCNDG, the resulting HCG materials were placed in a furnace under a nitrogen atmosphere and heated to 900 °C to change the functionality, including the amine group, into nitrogen-containing species, such as C 2 N 2 + , C 3 N 2 + , and C 3 N 3
Preparation of HCG and HCNDG-Based Electrodes.
To manufacture the HCG and HCNDG-based electrodes, the prepared HCG or HCNDG powder and polytetrafluoroethylene (PTFE) as binder were used in the weight ratio of 92% and 8%, respectively 11 . To achieve good adhesion, the pristine mixture of HCG, HCNDG, and PTFE powders was ground for 3 hr. Ethanol (around10 mL) was added dropwise to the mixture to increase the moisture of the samples, and, then, the mixture was pressed on a graphite sheet. Each HCG-based and HCNDG-based electrode was 70 mm wide, 140 mm long, and 0.5 mm thick. The HCG-based and HCNDG-based electrodes were placed into a capacitive deionization instrument for further investigation, as shown schematically in Fig. 1.
Batch Mode Electrosorptive Experiment. Batch adsorption experiments were performed in a recycling system to investigate the electrosorptive capacity of HCG-based and HCNDG-based electrodes. Figure 1 shows that the electrosorption unit cell consist of a power supply, a peristaltic pump, and a conductivity meter (DDS-307), which were used to prepare aforementioned recycling system. A peristaltic pump was utilized to feed the solution continuously into the cycle. The volumetric flow rate and temperature of the solution were selected based on previous studies 9 for optimal performance, which were 25 ml/min and 20 °C, respectively. In detail, the CDI unit consists of two symmetrical rectangular pieces of silicon with a space defined by a gasket between them. The gasket size was 4 mm, and the distance between the electrodes was 3 mm. The silicon plates have fluidic inlet and outlet ports with the same diameter of 4 mm. Also, a direct voltage of 1.0 to 2.0 V was used, and the solution's temperature was kept at 25 °C. In each experiment, the conductivity's variation in the NaCl solution was monitored continuously and measured at the outlet using an ion conductivity meter. The resulting solution had an initial conductivity of about 51 μ S/cm. An ion conductivity meter was used to measure the change in the concentration of the NaCl solutions at the outlet of the unit cell. To obtain the concentration of NaCl in different solutions during the experiments, a relationship between conductivity and concentration was introduced at the above-mentioned conditions. Figure S2 illustrates the correlation that results from the relationship of the solution's conductivity (μ s/cm) and its concentration (mg/L). The linear equation obtained from the experimental results is shown by Eq. 1:
= .
+ . , ( ) C 0 5109 K 6 0449 1 where C and K represent the solution's concentration(mg/L) and its conductivity (μ S/cm), respectively. Here, the regression coefficient (R 2 ) of Eq. (1) is 0.99. The correlation of Li et al. 9 is reproduced in Figure S2 for comparison. This figure shows that the present data and the correlation given by Eq. (12) are in good agreement with the correlation suggested in 9 . Also, the electrosorptive capacity (amount of NaCl adsorbed per unit mass of the HCG-based and HCNDG-based electrodes) can be calculated using 9 : where q e is the electrosorptive capacity (μ mol/g), C 0 and C e are the initial and final concentration of NaCl (mg/L), respectively, V is the volume of the solution in the cycle (50 mL), and M is the mass of the HCG-based or HCNDG-based electrodes. Characterization instruments and methods are described in detail in the Supplementary Information.
Result and Discussion
First, highly-crumpled graphene (HCG) and nitrogen-doped graphene (HCNDG) were synthesized with a low-cost and efficient approach. Then, both HCG-based and HCNDG-based electrodes with relatively high specific surface areas were fabricated, and their performance in capacitive deionization were investigated. The results showed that the specific surface area in the crumpled morphology has a key role in determining the electrosorption performance of both of the HCG-based and HCNDG-based electrodes. Figure 2a presents the Fourier transform infrared spectroscopy (FTIR) spectra of HCG and HCNDG. While there is no cue of functional groups in pristine graphite, it can be seen that both of the HCG-based and HCNDG-based samples illustrate significant peaks in their spectra.
Characterization.
Scientific RepoRts | 5:17503 | DOI: 10.1038/srep17503 The FT-IR spectrum of HCG shows two new peaks at 1151 and 2268 cm −1 , which were attributed to the C-N and C≡ N stretching vibrations on the HCG generated by the attachment of the -C≡ N chain of cyanamide during the diazonium treatment. Also, the peak at 3436 cm −1 can be assigned to N-H stretching vibration.
For HCNDG, the peak at 1151 cm −1 , corresponding to the stretching vibration of C-N, shifted to 1193 cm −1 . More importantly, the peak at 1550 cm −1 was associated with the stretching vibrations of C = N, which provided evidence of the doping of nitrogen atoms into the HCG framework. In addition, two other peaks, one at 2851 cm −1 and one at 2921 cm −1 , are consistent with the sp 3 C− H and sp 2 C− H, respectively, which resulted from hydrogenation at the edges and the size reduction. It can be seen that the peaks at 3436 and 2268 cm −1 , which were attributed to the C≡ N and N-H stretching vibrations, respectively, were removed in the FT-IR spectrum of the HCNDG after heating at 900 °C. However, the new peak at 1550 cm −1 that was associated with the stretching vibrations of C = N was a suitable substitute for above-mentioned groups.
Raman characterization is a strong measurement for analyzing structure, sp 2 and sp 3 hybridized carbon atoms in carbon-based nanomaterials, functionalization, and exfoliation by following alterations in hole and electron doping [30][31][32] . The Raman spectra of the pristine graphite, HCG, and HCNDG are presented in Fig. 2, panel (b). While the pristine graphite is weak in terms of D intensity, the fairly strong D, G and 2D bands in the HCG and HCNDG samples can be seen at 1351, 1578, and 2699 cm -1 , respectively. The ratio of the intensities of the D-band to those of the G-band (I D /I G ) was considered to be the amount of disordered carbon (sp 3 -hybridized carbon) relative to graphitic carbon (sp 2 -hybridized carbon). In edge-functionalization and/or nitrogen doping studies of graphene, the higher intensity ratio of I D /I G indicates the higher disruption of aromatic π -π electrons, implying partial damage of the graphitic carbon produced by expansion and/or doping 33 . The I D /I G ratio of HCG was relatively higher than that of pristine graphite, which confirmed the successful functionalization via a diazonium reaction under microwave irradiation. Also, the Raman results of HCNDG showed a higher I D /I G ratio than the pristine graphite as well as HCG, implying that the N-doping procedure was performed. Also, the G and 2D peaks in the spectra of HCG and HCNDG retained their intensities after the diazoniation, which confirmed that the quality of the graphene layers was preserved.
Obviously, Raman spectroscopy can distinguish a single layer from a few layers by focusing on the shape, size, and intensity of the 2D bands 34,35 . According to the results of Ferrari et al. 34 , as the layer of graphene increases, the 2D band becomes much broader and up-shifted. Accordingly, a noticeable change in sizes, shapes, and intensities of the 2D peaks of HCG and HCNDG are observed clearly as compared with the pristine graphite. It can be seen that the 2D band of the pristine graphite includes a coupled peak, i.e., D1 and D2 peaks, which produced a broad peak 34,35 . However, single, sharp 2D peaks were shown in the Raman spectra of HCG and HCNDG. This change in the 2D bands observed in the treated samples potentially verified the presence of single layered-or few layered-graphene and nitrogen-doped graphene.
The nature and amount of nitrogen-containing groups in pristine graphite, HCG, and HCNDG were studied by X-ray photoelectron spectroscopy (XPS), which are illustrated in Fig. 3a. The survey spectra of HCG and HCNDG have C1s at around 286 eV, N 1s at 399 eV, and O1s at 530 eV. Based on the results, all samples present a very small amount of oxygen, and pristine graphite doesn't illustrate any nitrogen molecule in its structure. Upon exfoliation and functionalization, the intensity of the N 1s peak increased considerably. The cyanamide functionalities (especially the -C≡ N chain) may explain the higher content of nitrogen in the HCG sample. More interestingly, the amount of nitrogen in HCNDG presents a similar result with HCG after thermal treatment in a nitrogen atmosphere. The insignificant decrease in the N component that was obtained after thermal treatment in the nitrogen atmosphere was in agreement with the FTIR, Raman, C 1s, and N 1s results. Figure 3, panels (b) and (c), present the XPS C1s spectra of HCG and HCNDG, respectively.
Both HCG and HCNDG mainly had peaks around 284.6 eV, which corresponded to the C-C/C = C network. Although the intensity of the oxygen element are insignificant, the minor O component in HCG and in HCNDG presents two peaks in the form of the C = O and O-C = O groups at around 286.8 and 288.5 eV, respectively. Also, the C-N peak centered at 285.6 eV was observed in both the HCG and HCNDG spectra.
The N 1s spectrum of HCG (Fig. 3d) composed of two peaks at about 399.2 and 401.8 eV, which would be assigned the bonding configurations of amine functional groups and C≡ N or that was detected by FTIR discussed above. However, Fig. 3e (HCNDG spectrum) shows an asymmetric N1s spectrum, which can be decomposed into four peaks, i.e., pyridinic N (~ 398.5 eV), pyrrolic N (~ 399.8 eV), graphitic N (~401 eV), and N-oxides of pyridinic N (~402.1 eV). Interestingly, the functional groups, including the nitrogen groups in HCG, change into more thermally-stable graphitic N after annealing, as illustrated in Fig. 3e. The distributions of the content of N species and their interpretation are presented in Table S1. Figure 4a-d show the SEM images of highly-crumpled, few-layered graphene and exfoliated HCG, which were achieved by using a diazonium reaction under microwave irradiation along with physical cracking by a sonicator, demonstrating a noticeable reduction in thickness. The proposed method is able to expand the graphite layers swiftly by initiating a semi-stable diazonium ion and radical reaction with layers. The SEM images of HCG ( Fig. 4a-d) showed homogenous, crumpled structures. Such a worm-like surface, with crumpled and curved sheets, was due to strict functionalization and nitrogen impurity. The HCNDG (Fig. 4e-h) also showed the distinct, crumpled structures with 2D geometry, as did HCG. The crumpled structures of HCNDG were attributed to nitrogen doping 8 . More evidence of this is presented by transmission electron microscopy (TEM) and atomic force microscopy (AFM) in the following section. Figure 5 shows the TEM images and select area electron diffraction (SAED) patterns of HCG and HCNDG. Figure 5a,b show the TEM images of HCG, in which the crumpled structure can be distinguished. This figure shows a large number of few-layered graphene nanosheets with wrinkled morphology and folded edges. Figure 5c-e show the TEM images of the HCNDG sample, which was comprised of a few-layered graphene with wrinkled morphology and folded edges. A set of highly-crumpled, individual graphene flakes with suitable transparency and without observable graphite crystalline structure in Fig. 5a-e confirmed that these wrinkles resulted from the crumpling of graphene rather than stacking, which is in agreement with the sharp 2D bands in the Raman spectra. As further evidence, the crystalline structures of HCG and HCNDG were verified by selected area electron diffraction (SAED), as shown in Fig. 5f,g, respectively. Selected area electron diffraction (SAED) of both samples illustrated a ring-like diffraction pattern with dispersed bright spots (that were stronger in HCNDG). Such an amorphous structure was attributed partially to the existence of functional groups with abundant defective edges for HCG and nitrogen doping defects for HCNDG, which was in agreement with the high intensity of the D band in the Raman spectra of both samples 36,37 . A ring-like diffraction pattern demonstrated the loss of long-range ordering in the sheets 36 .
AFM was used for the further morphological characterization of the thin-layered HCNDG and for the investigation of the thicknesses of the HCNDG in the final product. AFM samples were prepared by sonicating HCNDG sheets in ethanol for two minutes in a bath sonicator without any additives. Figure 6 shows typical AFM images in which one highly-crumpled sheet was investigated, and, interestingly, the sheets were single-layer sheets, which is in agreement with the sharp 2D bands of the Raman results. As shown in Fig. 6, the HCNDG sheets had thicknesses of about 1 nm, which can be considered to be the thickness of one layer, including the wrinkled morphology.
To investigate the effects of exfoliation and functionalization on the specific surface area of HCG and HCNDG, N 2 adsorption-desorption isotherms were measured by a surface area analyzer (Quantachrome Autosorb-1 analyzer at 77 K). The N 2 adsorption-desorption isotherms of the HCG and HCNDG are illustrated in Figure S3 and Table S2. The comparison between the BET results of this study and previous studies 8,25 of graphene materials showed a significant increase in the specific surface area. Also, it can be seen that the N 2 adsorption amount of the HCNDG was greater than that of HCG. The specific surface area of HCNDG after annealing at the high temperature of 900 °C was 1689 m 2 g −1 , but that of HCG was just 1568 m 2 g −1 . This lower specific surface area of HCG suggested that the introduction of functional groups between the 2D graphene sheets can fill the porous area. By looking at the dependence of electrosorptive capacity on the specific surface area, the HCG and HCNDG with appropriate surface area can be the suitable candidates for fabricating capacitive deionization electrodes. Note that the highly-crumpled nature of the samples can be an advantageous property for high-capacity ion storage. Electrosorption of NaCl. The influences of the characteristics of HCG and HCNDG electrodes with respect to specific surface area and the surface functional groups on electrosorption capacities are investigated in this section. According to previous studies 9, 11 , the acid treatment phase of carbon nanostructures with increasing specific surface area plays a key role in the preparation of graphene with high electrosorption capacity. Also, when the pristine graphite is exfoliated adequately, it decreases the possibility of severe aggregation, resulting in the high specific surface area. Here, a cost-effective and potentially industrially-scalable method for synthesizing HCG and HCNDG is introduced for preparing samples with significant specific surface area, which plays a key role in the electrosorption capacity of electrodes.
The electrosorption capacity is the amount of ions adsorbed on one gram of the electrode. It commonly is investigated as a function of concentration or conductivity of solutions over the charging procedure. One of the most important parameters that influences the electrosorptive performance of electrodes is the bias potential 11 . As expected, higher voltage means higher electrosorptive capacity. Figure 7a,b illustrate the conductivity transients in solutions and the electrosorptive capacity at various applied voltages for HCG and HCNDG, respectively. Figure 7 illustrates the conductivity transients in solutions during batch-mode experiments at six different applied voltages, i.e., 1.0, 1.2, 1.4, 1.6, 1.8, and 2.0 V. It can be seen that all of the charge processes were performed for 40 min. Obviously, the electrosorptive capacity of both electrodes increased with the directed voltage, and the electrosorptive capacity of HCNDG electrodes was insignificantly larger than that of the HCG electrodes. When the voltage increased from 1.2 to 2.0 V, the electrosorptive capacity of both electrodes increased. It can be seen that at a low voltage of 1.2 V, the decrease of conductivity with time is not obvious. When the applied voltage was greater than 1.2 V, the conductivity obviously decreased with time, indicating that the higher voltage leads to a higher electrosorptive capacity because of the stronger Coulombic interaction 23 . Although the quantity, 2.0 V, is greater than the Nernst potential for the breakdown of water, it should be noted that the intrinsic resistance of the electrodes permits an overvoltage, and no electrolysis took place. In graphene-based electrodes, similar results have been reported by 7,9,11,23,24 . As mentioned above, both electrodes show approximately similar trends with excellent capacities in electrosorption of NaCl, although the electrosorptive performance of HGNDG was more than that of the HCG electrodes, which can be attributed to the higher specific surface area. Also, after doping with nitrogen, the transferred electron number increased, implying greater electrical conductivity and the generation an intense electrical double layer for absorbing charged ions from solutions 37 . A special parameter of a relatively high specific surface area can be the most influence factor in enhancing the electrosorption capacity 21 for graphene-based electrodes. Herein, the HCNDG exhibited higher electrosorption capacity than the HCG-based electrodes in similar experimental conditions. This may be due to the fact that (i) HCNDG itself has a higher specific surface area 11 , (ii) the specific capacitance of electrode material is an important factor for its electrosorption performance 20 , and due to the increased pseudo-capacitive effect of nitrogen doping, HCNDG possesses excellent capacitive performance compared to that of HCG-based materials, and (iii) nitrogen doping can improve the wettability of the interface between the electrolyte and the electrode 27 .
As pioneers in this field with graphene-based electrodes, Xu et al. 20 concluded that the electrochemically-active nitrogen atoms (pyridinic and pyrrolic nitrogen) could simply control local electronic structures, which is beneficial for the improvement of the bonding between the nitrogen atoms and Na + ions in the solution. During this phenomenon, numerous Na + ions are accommodated on the electrode's surface, including pyridinic and pyrrolic nitrogen, resulting in the possible oxidation/reduction reaction between H 2 O and pyridinic or pyrrolic nitrogen as follows 20,27 : Thus, the wettability of the interface of HCNDG sheets is enhanced, which can be an appropriate parameter in enlarging the available surface area and improving the charge storage of an electrical double layer capacitor 38,39 . Therefore, a greater number of Na + ions can be easily accessible to adsorb by HCNDG sheets to form a more effective electrical double layer and accommodated in the pores of the graphene sheets. Moreover, improving the electrical conductivity of the graphene sheets by nitrogen doping may also play a key role in enhancing the electrosorption capacity 26 . Also, as another effective parameter, an increase in the voltage from 1.0 to 2.0 V leads to a sharp increase in the electrosorptive capacity from 3.0547 to 29.4978 μ mol/g for HCG and from 4.2124 to 33.5152 μ mol/g for HCNDG, which are about 10-fold and 8-fold increases, respectively.
As a result and consistent with previous results 7,9,11,23,24 , the applied voltage of 2.0 V did not generate any visible gas bubbles, implying lack of in situ water electrolysis. Based on previous results reported by Li et al. 11 accompanied by the above-mentioned reason, the applied voltage of 2.0 V can be the optimum working voltage for HCG-based and HCNDG-based electrodes in terms of electrosorptive capacity and energy consumption. The conductivity at different applied voltages obviously decreases with time. Due to electrolysis, voltages greater than 2.0 V are prohibited in aqueous media.
Several electrosorption experiments at 2.0 V and different initial concentrations were conducted to investigate the electrosorptive isotherms of the HCG and HCNDG electrodes. Initial NaCl concentrations of 25, 50, 100, 200, 300, 400, and 500 mg/L were used for both electrodes. Figure 8a,b illustrate the electrosorptive capacities of HCG and HCNDG electrodes, respectively, at 2.0 V. It can be seen that the amount of NaCl adsorbed tends toward zero when the solution tends to be diluted. Figure 8a,b also compare the experimental data with the Langmuir adsorption isotherm model In Eq. (3), C, q, q m , and K L are the equilibrium concentration (mg/L), extent of NaCl adsorbed on electrodes (in μ g/g of HCG or HCNDG), the maximum amount of adsorption capacity corresponding to the monolayer coverage (μ mol/g), and the Langmuir constant (which is associated with the energy of adsorption), respectively. Also, in Eq. (4), K F and 1/n are the Freundlich constant and the tendency of the adsorbate to be adsorbed, respectively. Table 1 lists the Langmuir and Freundlich parameters and regression coefficients (r 2 ). According to the results reported in Fig. 8b and Table 1, the Langmuir isotherm model provided better fits to the experimental data for both the HCG and HCNDG electrodes. It also was observed that the monolayer adsorption was the dominant mechanism in both HCG and HCNDG electrodes during the electrosorption process. As mentioned above, q m is the maximum adsorption based on the Langmuir isotherm model. Thus, for HCG and HCNDG electrodes, the equilibrium electrosorptive capacities are 119.05 and 113.64 (μ mol/g), respectively, at the applied voltage of 2.0 V and flow rate of 25 mL/min.
A comparison between the electrosorptive capacity and specific surface area of the HCG-based and HCNDG-based electrodes at different applied voltages and the flow rate of 25 mL/min are depicted in Fig. 8c,d. Both HCNDG and HCG present promising electrosorptive capacities, indicating that the specific surface area plays the most important role in the electrosorptive capacity. Also, consistent with the electrosorption results, the HCNDG provided larger electrosorptive capacity than the HCG. In addition, the gap obtained in the electrosorption results between HCG and HCNDG was attributed to the higher specific surface area and transferred electron number in the presence of N-doping. The transferred electron number means lower charge transfer resistance, implying higher electrosorption capacity 14 . Moreover, previous results 9 concluded that the conductivity of carbon nanostructures also plays a key role in the electrosorptive process. When a nitrogen atom is doped into graphene, it usually has three common bonding configurations within the carbon lattice, including graphitic, pyridinic, and pyrrolic N. The pyridinic N and pyrrolic N bonds contribute one and two electrons to the π system, respectively, and graphitic N refers to N atoms that are substituted for the C atoms in the hexagonal ring. Clearly, the spin density and charge distribution of carbon atoms can be changed by the adjacent nitrogen dopants 40 , which can induce the "activation site" on the surface of the graphene. Such activated sites are able to participate in the electrosorption. Thus, by looking at the high effective surface area and charge distribution, both HCG and HCNDG have significant potential as candidates in capacitive deionization applications. Table S3 show a comparison of the electrosorption capacities of various graphene-based electrodes for better understanding. As a comparison with other materials, since active carbon (AC) is the most cost-efficient material as compared with other carbon nanostructures and has a high specific surface area, it is the commonly-used material for the CDI electrodes 24 . Also, AC can be prepared from natural or synthetic sources. Other carbon materials used in CDI research are, for example, ordered mesoporous carbon, carbon aerogels, carbide-derived carbons, carbon nanotubes, and carbon black 41 al. 24 reported that AC/graphene electrodes have an electrosorption capacity of 0.85 mg/g at the applied voltage of 2.0 V and initial NaCl conductivity of 50 μ S/cm. Our results showed the electrosorption capacities of 1.723 and 1.959 mg/g in the presence of HCG-and HCNDG-based electrodes, respectively, at the same experimental conditions. As mentioned above, as the electric field is eliminated, the ions absorbed on the surface of electrodes are effortlessly emancipated. A good regeneration is one of the most important properties for electrode materials. Figure 9 illustrates the electrosorption-desorption cycles of the HCG and HCNDG electrodes, which are conducted by repeating several charging (2.0 V) and discharging (0 V) processes in NaCl solution with an initial conductivity of around 51 μ S/cm. Note that regeneration of HCG and HCNDG electrodes and repeatability of the electrosorption process were realized in the unit cell. The electrosorption capacity decline of electrodes have not been observed in the unit cell after over 3 charge-discharge cycles, which is in agreement with recent studies 9,11,42 .
Conclusions
An industrially-scalable, cost-effective, and simple approach was used to synthesize HCG and HCNDG with large specific surface area. Highly-crumpled graphene and highly-crumpled nitrogen-doped graphene with appropriate specific surface area were used as the electrodes for capacitive deionization that showed considerable potential. The results suggested that the electrosorption performances of both electrodes were very high, and they are well-suited for capacitive deionization applications. The study also suggested that the Langmuir isotherm model would provide better fits of the experimental data for both electrodes, implying that the monolayer adsorption on the electrodes was the key mechanism. Also, in the presence of HCNDG and HCG, respectively, high electrosorption capacities of 29.4978 and 33.5152 µmol/g and high specific surface area of 1689 and 1568 m 2 /g illustrated that highly-crumpled materials can be the best options for a new generation of capacitive deionization devices. | 2018-04-03T01:41:05.350Z | 2015-12-08T00:00:00.000 | {
"year": 2015,
"sha1": "a2cd553259d3eadf51b40b4696b9f0f97476114e",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/srep17503.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a2cd553259d3eadf51b40b4696b9f0f97476114e",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Medicine",
"Materials Science"
]
} |
248834254 | pes2o/s2orc | v3-fos-license | A Rule of Solute Segregation at Grain Boundaries
The control of solute segregation at grain boundaries (GBs) is essential in engineering alloy properties, however the structure-activity relationship of the key parameter-the segregation energies-still remains elusive. Here we propose the electronic and geometric descriptors of GB segregation based on the valence, electronegativity and size of solutes and the non-local coordination number of free surfaces, with which we build a predictive framework to determine the segregation energies across different solutes, matrices, GB structures and segregation sites. This framework uncovers not only the coupling rule of solutes and matrices in GB segregation, but also the origin of solute-segregation determinants. The contribution of solutes essentially stems from their d- and s-state coupling in alloying, whereas that of matrix GB interfaces is determined by matrix free surfaces. Our scheme builds a novel picture for the solute segregation at GBs and provides a useful tool for the design of advanced alloys.
Introduction
Solute segregation at grain boundaries (GBs) plays an essential role in governing the mechanical, functional, and kinetic properties of metallic materials because it controls the chemical composition and local structure of GBs (1)(2)(3)(4). This critical segregation behavior can be quantitatively characterized by the segregation energies, which measure the preference of solute atoms to segregate to GBs or remain inside the bulk. Generally, the segregation energies depend on the electronic and geometric properties of solutes and matrices. However, it still presents a fundamental challenge to rapidly evaluate the GB segregation energies and to understand the underlying mechanism of solute segregation.
While many studies of segregation energies focus on the numerical calculations of individual systems using density functional theory (DFT) and molecular dynamics (MD) methods (5)(6)(7)(8)(9)(10)(11)(12), some frameworks have been proposed for predicting the segregation energies. Early empirical thermodynamic models attempted to determine the segregation energies by using macroscopic properties, such as the bulk modulus and shear modulus in the McLean model (13) and the solution enthalpy and solute strain energy in the Miedema model (14,15). However, these models rarely explain the electronic origin of GB segregation and are rather limited in predicting the segregation energies (16,17). Some machine learning methods can be used to predict the segregation energies accurately (18,19), but generally cannot provide a clear physical picture of segregation behaviors. Other effective descriptors were also proposed to determine the segregation energies, such as the parameters based on the d-band properties (20)(21)(22), excess volumes (23) and coordination numbers (18). Essentially these descriptors characterize the structures of matrices, however, none of them can reflect the electronic effect of solutes. In addition, some descriptors like the d-band properties and excess volumes, derived from DFT calculations, are time-consuming to obtain and even with some numerical uncertainties. Moreover, they are hardly applicable to complicated GBs such as those in polycrystals.
Here, we find that the valence and electronegativity of solutes and matrices and the size of solutes combined together with the generalized coordination number of segregation sites control the segregation energies. With these easily accessible parameters, we build a predictive analytic framework for evaluating the segregation energies across the different solutes, matrices, GBs and segregation sites. This framework reveals that the solute effect of GB segregation originates from the solute d-and sp-state hybridization with neighboring atoms in alloying and the matrix interface geometric effect from the matrix free surfaces. These results provide a deep understanding of the GB-segregation mechanism and a physical guidance for designing the alloy structures with targeted properties.
Results
The effect of different solutes in a given matrix. First, we focus on the role of different transitionmetal (TM) solutes in GB segregation for a given matrix. Generally, the shift of a solute atom from the matrix to GB involves the bond breaking and forming and is limited by the volume of segregation sites. We thus attempt to describe the segregation behaviors from the alloying effect between solute and matrix (the interatomic bonding effect of alloying atoms) and from the size of solutes. The tight-binding (TB) and Friedel's models (24,25) have revealed in TMs that the d-band width Wd (defined as the second moment of the projected d-band density of states (26)) is dominant in the formation of the cohesive force and the s-band contribution dictates the mole volume and compressibility (27). It is known that Wd depends strongly on the valence-electron number (Sv) and the interatomic bond strength does on modified Pauling electronegativity (χ) (28). We thus adopt , which can describe the surface bond breaking reasonably (29,30), to study the segregation energies (Eseg). ψ, which is a rough piecewise linear function of Wd and cohesive energy (see Fig. S1), provides a rough description of the segregation energies in the W matrix (the structures are shown in Fig. S2). According to the TB models, the alloying and size effects of solutes contain the d-d state and d-sp state hybridization of the neighboring atoms as well as the charge transfer (ΔQ) between the solute and matrix atoms. It is essential to quantify these effects on the d-band width projected on the solute atoms W d sol = f (Vdd, Vd-sp, ΔQ) in order to determine Eseg (where V is the interatomic matrix element). The d-d orbital bonding interaction can be reflected by the interatomic matrix element (Vdd) as follow: where di,j represent the internuclear distance between i and j atoms. m is the mass of an electron. rd is spatial extent of the d-orbitals. ηdd is a dimensionless proportionality constant that is formally related to the type of d-d orbital bonding interaction. According to the TB models, the interaction between two atoms contains the bonding and antibonding states, corresponding to the attraction (V) and overlap (Pauli repulsion, S) for the electrons. S is proportional to V and the bond energy can be expressed as ∆E=2S|V|-2|V|=2αV 2 -2|V| as demonstrated in Ref. [(33)], in which the repulsive interaction (V 2 ) is approximately the square of the attractive interaction (V) and thus decays as 1/d 12 . The interatomic distance d depends on the radius of solute atoms (R) and the alloying and size effects of solute segregation are proportional to ]. Therefore, we introduce a Lennard-Jones (L-J) potential like formula to determine W d sol and the solute segregation energies as follows, Here, R is the radius of an atom X which is defined as half the length of a X-X single bond in a crystal or molecule. σ with a certain range of 1.18~1.21 has a minor effect in determining the segregation energies and thus is taken as the constant 1.20 for all considered matrices. All the parameters in Dseg can be obtained by table looking up (Table S1). We first study the solute segregation at the ideal W GBs, by adopting 8 symmetric tilt GB structures (Σ3(111), Σ3(112), Σ5(210), Σ5(310), Σ7(213), Σ9(114), Σ11(323) and Σ13(510)) with five different substitutional sites along each GB structure. The symmetric GBs are formed by the two identical free surfaces rotating at a certain angle around the rotation axis, and the ideal GBs experience no sliding or migration when the solute atoms segregate. The ideal GBs and substitutional sites are illustrated in Fig. 1 and Fig. S3. The results show that Eseg exhibits a brokenline relationship with Dseg for the considered TM solutes with reasonable accuracy regardless of the GB structures and segregation sites ( Fig.1 and Table S2). The turning point of the broken-line scaling is around Dseg = -5 (Cr and V) and Dseg = -10 (Hg and Au), dividing the solutes into the three groups.
These results essentially originate from the different d-and s-state properties of TM atoms in alloying. The TB models show that W d sol dominates the trend of the cohesive force of IIIB-VIII group TMs and the s-bands dictate that of the IB and IIB group TMs (27). The segregation behavior of the solutes in groups 1 and 2 follows the trend of the cohesive energy of IIIB-VIII group TMs, reflecting the dominant role of the d-states in alloying IIIB-VIII group solutes with W. On the other hand, the behavior of the solutes in group 3 is opposite to that in group 2, corresponding to the dominant role of the s-states in alloying IB-IIB group solutes with W.
The size effect is secondary compared to the electronic bonding of solutes, in stark contrast with the viewpoints that proposed the solute-size effect to play a major role in GB segregation (5,6). A L-J potential like formula(34) of D seg implies that the size of solutes exhibits mainly a twobody effect in GB segregation with negligible many-body effects. As ψ sol reflects W d sol and s-band filling and the size of solutes does significantly the s-band depth (which is the energy of the s-band bottom) (27), the descriptor D seg uncovers the coupling nature between the d-and s-states of solutes in determining the segregation energies.
For the three group solutes in the W matrix, the corresponding slopes and intercepts of the linear scaling are labeled with ki and bi (i=1-3). Notably, k1, k2 and k3 are linearly dependent on each other for all the GBs and segregation sites (see Fig. 2a and b), and so are b1, b2 and b3 (see Fig. 2c and d). Interestingly, for each solute group, the slope is also a linear function of the intercept, such as k2 and b2 (Fig. 2e). These results indicate a sole physical origin for all slopes and intercepts (that will be further discussed below).
We now consider the segregation energies in W with GB relaxation during solute segregation. For Σ5(210), Σ7(213), Σ9(114), Σ11(323) and Σ13(510) GBs, they experience GB sliding or migration and lose their mirror symmetry after inserting the solute atoms (5,35,36) (see Fig. S5). Nevertheless, Dseg is still able to determine the trends of the segregation energies for various GBs and segregation sites (see Fig. S6 and Table S3), further indicating that Dseg reflects the core elements of the alloying and size effects of TM solutes for GB segregation.
In the case of other metal matrices, such as Nb, V, Ir, Rh, Os, Tc (the structures are illustrated in Fig. S4), Al (37), Ni (38), Zr (39) and Mo (17), we find that Dseg always exhibits a broken-linear relationship with Eseg (Fig. 3, Figs. S7 and S8 and Tables S4-S6). These broken-line relationships lead to the d 5 state at the turning point of groups 1 and 2 and the d 10 state at that of groups 2 and 3. For the TM matrices, the solute itself exhibits this feature. For the main-group matrices such as Al, an Al atom has three valence electrons and forms the d 5 state with Ti (3d 2 ) and the d 10 state with Co (3d 7 ) at the turning points. Importantly, this feature is crucial for the prediction of segregation energies.
Furthermore, Dseg is also applicable to the solute segregation of twist GBs and even polycrystal GBs (that are the irregular grain boundaries between multiple grains with different orientations in polycrystals illustrated in Fig. S9) (19). Dseg exhibits a broken-line relationship with Eseg in Mo Σ5(100) twist GB with the mean absolute error (MAE) of 0.13 eV at site 1 and 2 ( Fig. 3d and Fig. S8d). More importantly, it can determine the trend of segregation energies for a given segregation site in polycrystals of 15 matrices such as Ti, Zr, Ta, Co, Fe, Ni, Mo, W, Pd, Pt, Au, Ag, Cu, Mg and Al, with the accuracy of 0.05 eV (see Figs. S10-S12 and Table S7), comparable to that of the smooth overlap of atomic positions (SOAP) based machine learning work (19). The deviation of some data relative to the fitting lines is likely due to the fact that all data at polycrystal GBs are obtained with MD calculations (which generally bring out larger errors compared with those obtained by DFT calculations). Interestingly, some main-group elemental solutes also follow the Dseg-determined relations of TM solutes, indicating that the valence origin of Dseg is useful to understand the behavior of all kind solutes (see Fig. S8b).
Overall, all the above findings demonstrate that the descriptor Dseg quantifies effectively the alloying and size effects of solutes at the GB segregation across different kind GBs and matrices.
The geometric effects of matrix GBs. We now attempt to understand the geometric effects of GB structures in segregation behavior, which corresponds to the various GB interfaces and the environment of segregation sites. These effects are influenced by many factors such as GB interface sliding and migration as well as the local deformation. We start with the simple ideal symmetric tilt GBs to discuss the geometric effects of segregation energies. Since the symmetric tilt GBs are formed by two identical grains (or free surfaces), we use the generalized coordination numbers (CN ̅̅̅̅ ) of the unrelaxed free surfaces (before GB formation) without any solute and vacancy to characterize GB geometric properties. The expression of CN ̅̅̅̅ is as follows: Here, i represents the ith neighbor of a given surface atom. n is the number of neighboring atoms. The unrelaxed free surfaces do not experience any change of bond length and bond angle, and thus the identification of neighbors for a given surface site does not need a cutoff radii, which avoids the numerical uncertainty. CNmax is the max coordination number in the bulk. m denotes the order of approximation. The ordinary coordination number CN corresponds to the zeroth order approximation CN ̅̅̅̅ (0) , reflecting the geometric environment of a surface atom with its nearest neighbors. We use the first-order approximation CN ̅̅̅̅ (1) , that is, the effect of second-neighboring atoms is included. Notably, this approximation was first proposed to determine the adsorption of small molecules on TM surfaces and nanoparticles (40,41). If one adopts the second-or higherorder approximation of CN ̅̅̅̅ (m) (m ≥ 2), the effects of farther neighboring atoms are involved (CN ̅̅̅̅ (1) and CN ̅̅̅̅ (2) are shown in Table S8). Fig. 4(a-c) shows that for a given solute atom segregated in the ideal W matrix, Eseg is a brokenline function of the CN ̅̅̅̅ (1) of the segregation sites on the unrelaxed free surfaces for each GB.
Moreover, most of the segregation sites comply with the same CN ̅̅̅̅ (1) -determined relation. In contrast, the ordinary CN (namely CN ̅̅̅̅ (0) ) fails to determine the segregation energies (Fig. S13). CN ̅̅̅̅ (2) is almost the same as CN ̅̅̅̅ (1) in determining Eseg (see Fig. S14), indicating that the geometric effect of a segregation site is mainly from its nearest and second-nearest neighbors with a negligible contribution from farther neighbors. The turning point of the broken-line scaling is around CN ̅̅̅̅ (1) = 4, which is roughly the boundary between site 1 and site 2. The reverse folding of site 1 likely stems from the fact that the structure of site 1 at GBs is close to that of the bulk and is thus different from that of surfaces. Besides the ideal W matrix, the CN ̅̅̅̅ (1) of the unrelaxed free surfaces is also able to describe the effect of different segregation sites for the GBs with multiple metastable structures, such as reflection and isosceles of {112} twin boundaries (44) in the W matrix (see Fig. S15). Although these two GB structures exhibit two different sets of segregation energies for a given solute (such as Ir, Pd or Pt), each set can be determined accurately by the same CN ̅̅̅̅ (1) (see Fig. S15). The different interface structures are reflected by the slopes and intercepts of the CN ̅̅̅̅ (1) -determined scaling. These findings also hold for the other five GBs (with both symmetric and asymmetric structures) in the W matrix, such as Σ5(210), Σ7(213), Σ9(114), Σ11(323) and Σ13(510) GBs (Fig. S16). Notably, the segregation site 1 of the relaxed GBs generates two linear relations, mainly caused by the different structural deformations of various GB structures: one is for the ones with GB sliding and the other is for the ones without GB sliding. These two lines are also predictive, if both are included, the MAEs of the predicted segregation energies are decreased from ~0.25 eV to ~0.18 eV. The prediction error of CN ̅̅̅̅ (1) is around 0.2 eV that is within that of the DFT semi-local functional calculations (42,43). Clearly, the accuracy of our scheme is usually better than 0.2 eV (which is the upper limit of the average error in our study) and the prediction error is smaller than 10% of the range of the segregation energies. Considering that the range of the studied segregation energies is ~2.5 eV for the ideal GBs and ~6 eV for the relaxed GBs (Fig. S16), the 0.2 eV error is sufficient for the determination and prediction of the segregation energies. In particular, our scheme is also effective for the segregation with low segregation energy: the 0.03 eV error for the Mo solute segregating in W matrix GBs with the range of segregation energies of ~0.30 eV (Fig. S13e) and the 0.01 eV error for the Ir, Pd or Pt solutes segregating in the twin boundaries of W matrix with the segregation-energy range of ~1.5 eV (Fig. S15). These results imply that the geometric contribution of matrix GBs to the segregation energies is non-local compared with the effect of solutes and mainly determined by the unrelaxed matrix free surfaces regardless of the GB deformation. The non-local CN ̅̅̅̅ (1) can be used to quantify the geometric effect of the segregation energies for both the symmetric and asymmetric GBs formed by the low-index and high-index surfaces.
Our findings are also comparable to the literature results. Hu et al. considered the geometric effect of W and Ta matrix GBs with the linear combination of the bimodality property of the dbands and the strength of sp-d hybridization in the GB structures (that are obtained by DFT calculations) (20). CN ̅̅̅̅ (1) is found to linearly depend on the linear combination of their descriptors (see Fig. S17), supporting the effectiveness of CN ̅̅̅̅ (1) in evaluating the GB segregation. Huber et al. use the excess volume and ordinary CN to determine the geometric effect of segregation sites and the excess volumes are calculated from the relaxed GB structures with DFT calculations (18,23). In comparison, CN ̅̅̅̅ (1) originates from the unrelaxed free surfaces and is independent of DFT calculations, avoiding the numerical uncertainty.
The effect of a given solute in different matrices.
To exclude the effect of lattice structures, we study the segregation energies of a given solute in 8 body-centered cubic (bcc) (Cr, Mn, Fe, V, W, Nb, Mo and Ta) and 7 face-centered cubic (fcc) (Ag, Au, Cu, Rh, Ir, Pd and Pt) GBs. The corresponding segregation energies in different matrices are the broken-line function of the ψ m of matrices at one segregation site for one lattice structure (see Fig. S18). This reflects the fact that the solute atoms only have a negligible effect on the d-band width of matrices due to their low concentration and local effects, which is different from the effect of matrices on the d-band width projected on solute atoms W d sol . The d-band width of matrices thus dominates the types (electronic effects) of matrices on GB segregation. Therefore, the segregation energies of different solutes and matrices at each segregation site can be determined by the linear combination of the Dseg of solutes and the ψ m of matrices ( Fig. S19 and Table S9). The slopes and intercepts of these broken-line relationships (as well as those of the scaling between Eseg versus Dseg) can be characterized as the linear function of CN ̅̅̅̅ (1) of segregation sites (Fig. 4d and Fig. S20), which reveals a coupling mechanism between the solutes and matrices.
A general scheme to determine the segregation energies. Therefore, we can propose an entire expression to determine the segregation energies, by combining the Dseg of solutes, the ψ m of matrices, and the CN ̅̅̅̅ , fcc (Ag, Au, Cu, Pd, Pt, Ni and Al) and hexagonal close-packed (hcp) (Os, Tc, Ti, Zr, Co and Mg) matrices. These include the major binary alloy systems of interest in the field. The studied 24 GBs contain the ideal symmetric tilt and twist GBs, the asymmetric GBs with sliding and migration and even complicated polycrystal GBs. The MAE of prediction is 0.10 eV in 8 W GBs and Ir and Os GBs, 0.05 eV for 2 Nb GBs, 0.06 eV for V GBs, 0.09 eV for Rh and Tc GBs, 0.11 eV for 12 Al GBs, 0.16 eV for Ni GBs, 0.14 eV for Mo GBs and 0.14 eV for Zr GBs and 0.05 eV for polycrystals of 15 matrices. These results demonstrate that our scheme is universal and predictive in determining the segregation energies across different TM solutes, matrices, GB structures and segregation sites.
Our scheme can also deduce or rationalize the experimental segregation tendency. The experimental results in W show that there is almost no segregation for Ti or Ta solutes at GBs, whereas Ag solute is prone to segregate to all sites around GBs (45). These observations likely stem from the fact that the Ti/Ta and Ag solutes sit close to the turning points of our scheme, the top and bottom respectively. More importantly, combining our scheme with the White-Coghlan model (46) (which exhibits a better description of the GB enrichment with temperatures by a distribution of segregation energies, compared to the Langmuir-McLean isotherm relationship by a single persolute value(18)), one can also predict the GB solute concentration (cGB) curves for non-interacting solutes in the W matrix, which is determined as: where N represents the total number of segregation sites (fixed as 5 here), E seg X,i is the segregation energy of solute X segregated to the position i at GBs and T is temperature. cbulk is the solute concentration in the matrix, which is fixed as 2 at.% here (18,20). Fig. 5 and Fig. S21 show that the predicted GB solute concentrated curves based on our framework are close to the DFT-calculated ones across a wide temperature range regardless of solutes and GB structures, further verifying the validity of our framework. Compared with the late TM solutes, the early TM solutes have the solute concentrate depending more strongly on temperature in the W matrix, suggesting temperature as an effective way to engineer the early TM solutes at GBs.
Discussion
Overall, our framework provides an effective tool to analyze the solute segregation at GBs with the easily accessible intrinsic properties. The geometric effect of matrices in GB segregation can be determined by only using our proposed generalized coordination number but not the previously suggested ordinary coordination numbers ( Fig. 4 and Fig. S13). Our electronic descriptor ψ m can quantify the effect of matrix types in the segregation energies effectively. The previous studies have shown that the size of solutes cannot determine the segregation energy alone (5,6), nor can the electronegativity and valence-electron number of solutes. Accurate prediction of the segregation energy can only be achieved through the combination of these three parameters into a descriptor Dseg (that accounts for the alloying and size effect of solutes), as we found. Our scheme identifies the geometric and electronic descriptors for matrices and a single descriptor for solutes and uncovers the coupling rule between the different properties of matrices and solutes with simple and general equations. The proposed descriptor Dseg essentially implies the local effects of solute atoms and the similar matrix deformation with different solute atoms, whereas CN ̅̅̅̅ (1) reflects the non-local effects of segregation sites at GBs and is well suited to single-crystal matrices such as the symmetric and asymmetric tilt GB structures. For polycrystal matrices, one may introduce more complicated geometric features such as bond length and angle to characterize the geometric structures of GBs, which will thus be a natural extension of this study.
In summary, we have identified the main factors that control the segregation energies of solutes at GBs: the valence and electronegativity of solutes and matrices, the size of solutes and the nonlocal coordination number of free matrix surfaces. This enables us to build a predictive analytic framework to quantitatively determine the segregation energies, which holds across the different solutes, matrices, GBs and segregation sites. The determinant of solutes, initially inspired by the TB models, reflects the hybridization of d-and sp-states of solutes and neighboring atoms in alloying, whereas the geometric determinant of GB matrices, indicating a non-local geometric effect, does the free-surface-determined nature of GB interfaces. Our framework uncovers the electronic origin for the behavior of alloying elements in GB segregation, the connection between free surfaces and their resulting GB interfaces, and the coupling mechanism of solutes and matrices, all of which propose a simple and clear physical picture for the solute segregation at GBs. All these findings are thus expected to serve as quantitative guidelines for the future alloy design, particularly considering that all involved parameters are predictable.
Method
Our calculations are performed using Vienna ab-initio simulation package (VASP) code (47). The interactions between ions and electrons are described by the projector augmented wave potential (PAW) method (48). The Perdew-Burke-Ernzerhof (PBE)(49) exchange-correlation functional is adopted for all the studied systems unless otherwise stated. The cutoff energy of 500 eV is used in the calculations and the k-point mesh is given in Table S10. The convergence tests of the cutoff energy, k-point mesh and functionals are summarized in Fig. S22. The cell volume and atomic position are relaxed for each GB supercell without solutes and then kept fixed when the solutes are inserted. The negligible difference between the segregation energies of the relaxed and fixed cell can be found in Table S11. For the optimization, all the atoms are relaxed until the force on each of them is less than 0.02 eV/atom in our calculations.
The segregation energy Eseg is defined as: E seg = E GB -E Bulk (6) where EGB is the total energy of the structure with the solute atoms at GBs and EBulk is the total energy of the structure with the solute atom inside the bulk. | 2022-05-18T06:47:03.560Z | 2022-05-17T00:00:00.000 | {
"year": 2022,
"sha1": "81f3742243a9b66f5040be41e51c6a760fdf9963",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "81f3742243a9b66f5040be41e51c6a760fdf9963",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
222279298 | pes2o/s2orc | v3-fos-license | Short-term Impact of Item-based Loyalty Program on Customer Purchase Behaviors
Studies based on the analysis of a new design of loyalty program, item-based loyalty programs (IBLPs), indicate that customers are more interested in item-based reward points than in traditional price discounts. However, we are still unaware of customer responses to the different point settings on IBLP items. This study investigates an analysis with Tobit II to explore IBLPs’ short-term (4 months) impact on customers’ purchase behaviors using data from two newly opened Japanese supermarket chains that have implemented this new IBLP program from the beginning. The results showed that different types of customers are differently affected by IBLPs, and that heavy customers are more inclined to purchase more items with more spending money than others. The results also indicated that customers’ purchase behaviors are affected by IBLPs’ different point levels. Moreover, to an IBLP with different points, the responses from different types of customers are different. The findings of this study have important guiding significance in IBLP design and marketing management.
Introduction
To attract and retain more customers, as part of customer relationship management strategy, various industries use loyalty programs (e.g., issue points) to make more profits. The total points issued by companies in Japan were at least worth nearly 85 billion yen in 2014, and this number may increase to 1.1 trillion yen in 2022. The loyalty programs are used by industries such as airlines and grocery stores to increase their marketing promotions by identifying customers' performances [1]. By involving the loyalty programs, companies collect more data to enhance their targeted promotions and can further gain more repeated business [2]. Correspondingly, customers can earn reward points by their purchase rates that can help their closer to the redemption thresholds [2][3][4].
Some of the researches done on loyalty program suggest that it can successfully foster customer's loyalty [2,3,5,6]. However, as a loyalty program is widely used in the current marketplace, it can also become a redundant resource in businesses [4]. According to some studies, the design of loyalty program is directly related to its effectiveness [1,4,5,7], which means that new designs for loyalty program are important and worth studying.
For instance, a new design of loyalty program "item-based loyalty program (IBLP)" is proposed by Zhang et al. [7]. The difference between an IBLP and a conventional loyalty program is that the price discounts for each item are replaced by reward points in the former, which means that customers can earn extra points when purchasing specific items. Zhang et al. 's findings indicate that the customers showed more interest toward the item-based reward points than toward traditional price discount. However, in the study, they only explored the difference between non-members and members. As some researchers suggest that comparing the behavior of members with non-members cannot conclusively establish the causal relationship [3], it is important to further indicate the different responses to the IBLP from other types of member classification. Therefore, in this study, we focus on a new IBLP-based design that assigns different points to different products, and try to explore this IBLP's short-term (4 months) impact on customers' purchase behaviors using Tobit II. The data we used were from two newly opened Japanese supermarket chains (belonging to the same company named Hankyu Oasis) that have this new IBLP program in practice since the beginning. To identify the effects of the new IBLP on customers' purchase behaviors, the research findings and conclusion of this study have important guiding significance in IBLP design and marketing management. This paper is an extended version of our previous conference papers [8]. The remainder of this paper is organized as follows: an overview of related studies is provided in Sect. 2 with related empirical hypotheses. In Sect. 3, the target customer, related data set, and model are given. The analysis model and results that we used are provided in Sects. 4 and 5. Finally, we summarize the research results and provide our perspective regarding promising future research in Sect. 6.
Overview of Related Studies and Hypothesis Development
Many studies have been conducted to investigate the effects of IBLPs on customers' purchase behaviors. In this study, based on our case study of a supermarket in Japan (Hankyu Oasis), we define an IBLP as a design in which member customers can earn different points as rewards for purchase of different items. According to the prior researches and our purpose mentioned above, in this study, we try to solve three questions by testing the corresponding hypotheses.
The Effect of IBLP on Different Types of Customers
Findings from Zhang et al.'s [7] IBLP study, which examined the differences in responses between non-members and members, indicate that IBLPs may lead to an increase in non-members' total spending, but, at the same time, may also lead to a decrease in members' spending [7]. However, in this study, they did not consider the different types of members, which also lead to different responses [3]. Therefore, for the IBLP data we have, our first hypothesis will be: H1 Different types of customers respond differently to Oasis's IBLP.
According to the prior studies, customers can be classified into different types based on their basic demographic information [5], their attitude or behaviors to the firm, and so on [3,4,9,10], which may better predict customer responses [4]. Therefore, drawing on the previous study [3], in this study, we use the spending level (heavy, moderate, light) to classify customers. Thus, to prove that the difference exists between the levels we used, as the first step, sub-hypothesis H1a will be: H1a Oasis's IBLP affects light, moderate, and heavy customers' purchase behaviors differently.
Moreover, according to Lal et al.'s study [3], in this study, we use customers' purchase incidence and weekly spending to represent their purchase behavior. As heavy and light customers may be less concerned with the items' additional reward points, sub-hypothesis H1b for customers' purchase behaviors will be: H1b Oasis's IBLP affects moderate customers' purchase behavior more than heavy and light customers' purchase behavior.
The Effect of Different Points on Customers' Purchase Behaviors
A few studies focus on the effect of IBLPs' different point levels on customers' purchase behaviors. Kumar et al. [11] pay attention to the effect of different values of coupons on their coupon elasticity. Their finding indicates that for the same brand, a higher value of coupon has a higher effect on the coupon elasticity. Therefore, as coupon values are similar to IBLP points, hypothesis H2 will be: H2 Different points in Oasis's IBLP affect customers' purchase behavior differently.
Similar to H1, to prove that the difference exists between the groups classified by different points in customer purchase behaviors, sub-hypothesis H2a will be: H2a Different points in Oasis's IBLP have different effects on customers' purchase behavior.
Moreover, according to the results of a coupon elasticity study [11], the higher point group may have a greater influence on customers' behaviors. Thus, sub-hypothesis H2b will be: H2b Customers' purchase behavior is more affected by higher point items than lower point items.
How Customers with Different Characteristics Respond Differently to an IBLP with Different Points
Lal et al. [3] suggest that grocery retailing is one of the least profitable industries and the competition for a shopper is fierce. To improve its competitiveness, a firm needs to increase customers' switching cost by constantly reminding customers about special benefits they can earn only in their store [5,7]. However, given that most firms are offering loyalty programs now, it is very difficult for firms to create differences to attract customers by operating a conventional loyalty program. Zhang et al. [7] find that the IBLP they designed successfully reduced customers' responsiveness to competitors' promotions, which means this new design does differentiate their program from other competitors. Nevertheless, to further improve the IBLP effects, many topics still need to be explored. Therefore, based on Liu [3] and Zhang et al.'s [7] results, this study analyzes the details of an IBLP. In our third hypothesis, we examine how different points in IBLPs affect different types of customers differently. The result of this hypothesis may provide us an opportunity to improve the effect of Oasis's IBLP by making the IBLP design more precise.
H3 Different types of customers respond differently to IBLPs with different points.
If H1 and H2 are supported, we can draw a conclusion that different types of customers respond differently to Oasis's IBLP. Thus, we need to examine which kind of customers are more affected by which kind of points in the IBLP. Similar to what we do in H1b and H2b, we bring up the following sub-hypothesis: H3a Higher point items affect moderate customers' purchase behavior more than they do heavy and light customers' purchase behavior.
Data Set from Oasis's IBLP
As mentioned above, to avoid the learning effect, the data set are selected from two newly opened Japanese supermarket chains (Oasis) that have been using the new IBLP program since the beginning. In their normal design, the member customers can earn one reward point for their each 200 yen spending in the store. Based on this basic design, an advance IBLP design called point plus is used to allow the member customers gain more points (ten levels from 5 to 100 points) according to their purchase items' monthly changing classification.
As shown in Fig. 1, based on the whole data we have, a store selection period is used to estimate the open date of the stores; based on their open day, the data collection period is selected for analysis. In detail, as the prior studies indicate that customers' behavioral loyalty is related more to the short-term period, we used the data from a 4-month period (2016.1.1 to 2016.4.30) and classified the data collection period into two parts: initialization period to classify customers and analysis period to estimate the model.
Target Customers
In this study, to rule out the self-selection effect, we use three criteria to identify our target customers: date of membership, visit frequency, and last visit date. In other words, our target customers are the customers who have the data of purchase behaviors more than one at the beginning to the end of our research period. Therefore, the customers who joined the membership early, have more than one time of store visit history, and visited the store at least in the last month (April) are selected for analysis.
Based on our classification of customers and spending level mentioned in Sect. 2, 1245 target customers are picked for analysis, including 416 of light, 415 of moderate, and 414 of heavy spending category. Figure 2 shows the proportion of different Fig. 1 The time periods we set customer types and their corresponding spending levels in January 2016. Of the total spending, 62% has been done by the heavy types, which is two times than that by the moderate, and six times than that by the light customers. The details of basic statistics of customers with different types are shown in Table 1.
IBLP Items
Based on Oasis's IBLP design, ten levels of points including 5, 10, 15, 20, 25, 30, 35, 40, 50, and 100 point are used to bind the items with different price, which indicate that the items with higher price usually have a higher level of point setting. In addition, the 5-and 10-point levels have the most item categories, and the levels of 25 and 100 only have two.
Tobit II Model
Prior studies have used many methods such as a hierarchical linear model (HLM) that do not require independent observations and can accommodate individual heterogeneity to model customers' reaction to loyalty program [3]. However, using an HLM to model purchase frequency and transaction size separately can lead to [12] that does not require the normality assumption and is usually used to measure customers' purchase behavior [6,10]. Therefore, to test the effect of different IBLP items on customers' purchase behavior (purchases decision and weekly spending), the Tobit II model is involved in our analysis. The Tobit II model has two steps as follows:
Profit Model (Purchase Incidence)
To predict IBLP items' purchase incidence, we assume that customer i purchases the IBLP items from Oasis's store in week t ( Z it ), when the utility of doing so ( Z * it ) is positive. Therefore, Equation 1 means that when customer i purchases in the store in week t, the value of Z it will be 1, else it will be 0. Z * it represents the utility of customer i's purchase in week t, which can be calculated as: In Eq. 2, 0 represents the baseline IBLP items' purchase incidence. Freq it represents customer i's shop visit frequency in week t, log(t) represents the time trends.
it is the random error. To ensure the distribution closer to normal, the variables of independent and dependent are all log-transformed [7,10].
OLS Regression Model (Weekly Spending)
Therefore, based on the Step 1, when the customer i has made a purchase in week t ( Z it = 1 ), the logarithm of weekly spending of customer i in week t (Japanese yen) y it will be: In Eq. 3, 0 represents the baseline weekly spending, and a set of point-related variables (a total of ten, e.g., Point 5it ) is used to represent the spending on corresponding items (e.g., 5-point). log(t) represents the time trends. it is the random error.
These two models are used to perform analysis on all types of customers. (1)
Results and Hypothesis Testing
In this study, the Tobit II model is used to test the effect of IBLP point levels on the purchase behaviors with different customer types. As shown in Tables 2 and 3, for IBLP items' purchase incidence and weekly spending, the results indicate that all the tested variables in our model are significant. Moreover, customers did not have a significant upward trend in their IBLP items-based purchase behaviors.
Therefore, based on the analysis results, the hypotheses are tested as follows:
Test of Hypotheses 1
To test H1, two steps are used to test H1a and H1b. At first, based on the basic statistics that show that the heavy customers have spent more money on the IBLP items, the analysis of variance analysis and Z-test are used to test H1a. As shown in Tables 4 and 5, the results of variance analysis and Z-test indicate that a significant difference exists among the three types of customers' purchase incidence and weekly spending on IBLP items (p < 0.05). H1a is supported. In contrast, based on the results of the Tobit II model as shown in Table 2, a higher value in heavy customers is observed in the parameter of purchase probability baseline 0 , which means that the heavy customers have a higher probability to purchase more IBLP items. Moreover, as shown in Table 3, the heavy customers also have the largest value in the parameter of purchase probability baseline 0 . It means that the heavy customers spent more money on the IBLP items initially. However, except the beginning, the heavy customers not always the highest purchasers, therefore, H1b is partly supported. The heavy customers' purchase behavior is most affected by Oasis's IBLP initially.
Test of Hypotheses 2
Similarly, to test H2, two steps are used to test H2a and H2b.
For H2a, the analysis of variance analysis and Z-test are used. According to the large sample distribution theory [12], only the variables that have enough samples can use the method of Z-test. Therefore, in this study, we select the points 5, 10, and 20 for analysis as they have more than 100 samples. As shown in Tables 6 and 7, the results of variance analysis among total spending are significant (p < 0.05), but the result of their Z-test is not significant (p > 0.05), which indicates that to some extent, the customers do have different reactions to different point levels. H2a is partly supported.
In contrast, for H2b, the results for all customers' purchase behaviors are given in Tables 8 and 9. All of the variables are significant, and for the different point levels, the highest is the level of 25, which means that a higher point does not have a higher effect on customers. H2b is not supported.
Test of Hypotheses 3
To test H3a, as the coefficients of different points for heavy, moderate, and light customers are all significant, we try to explore H3 based on the results shown in Tables 2 and 3. Based on the results, the top three coefficients for heavy customers are 25-, 5-, and 30-point items, which are 0.670, 0.570, and 0.532. As for moderate customers, the coefficients of 25-, 100-, and 5-point items, 0.780, 0.625, and 0.591, are higher than other points. Finally, 5-, 30-, and 100-point items' coefficients for light customers are 0.586, 0.576, and 0.547. Among all these points, the 25-point Furthermore, moderate customers' weekly spending is more affected by 100-, 40-, 25-, and 5-point items, and light customers spend more money on 50-, 30-, 20-, 15-, and 10-point items than other customers. These results suggest that higher point items may have more effect on the moderate customers. Nevertheless, the purchase behaviors of light customers are more affected by the high-level point items. Moreover, the heavy customers are less sensitive to high-point items than others. Thus, H3a is partly supported.
This result indicates that heavy customers are also concerned about how to accumulate more points and maximize their benefits faster by purchasing items in IBLPs. However, heavy customers are less sensitive to the difference among points in IBLPs. In contrast, although moderate and light customers spend less money on IBLP items than heavy customers, they are more affected by certain points in IBLPs. It suggests that all customers pay attention to IBLPs, and, moderate and light customers choose more efficient ways to get more points.
Conclusion
This study focuses on the IBLP-based design that assigns different points to different products, based on the data from two newly opened Japanese supermarket chains (Oasis) that have used this new IBLP program since the beginning. The Tobit II model is used to explore an IBLP's short-term (4 months) impact on customers' purchase behaviors. In detail, the effects of IBLPs' different point levels (5-100, total ten levels) on different types (light, moderate, heavy) of customers' purchase incidence and weekly spending are analyzed.
The analysis results showed that all different types of customers are differently affected by the IBLP. In contrast to prior studies [2,3], our results indicated that the heavy customers are more inclined to purchase more IBLP items with more spending money than others. Moreover, the results also indicated that customers' purchase behaviors are affected by the IBLP's different point levels. The behaviors of customers are affected by the items with a higher point level, but the highest effect belongs to the 25-point level. At the end, we also find that for the IBLP with different points, the responses from different types of customers are different. For example, a higher point item may have more effect on the moderate customers. This finding can help managers to improve the effect of IBLPs by arranging more targeted items to different types of customers.
This study only considers the effects of IBLPs in the short term. As the stores usually use multiple programs to promote sales, it would be useful to involve other programs' influences in future studies. | 2020-08-13T10:09:50.943Z | 2020-08-10T00:00:00.000 | {
"year": 2020,
"sha1": "848b390726a2bc9a635c6c867d613aff7f62c848",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s12626-020-00062-5.pdf",
"oa_status": "HYBRID",
"pdf_src": "SpringerNature",
"pdf_hash": "78c005ea0665d7f87a6074edb207f207fd0f9982",
"s2fieldsofstudy": [
"Business",
"Economics"
],
"extfieldsofstudy": [
"Computer Science",
"Business"
]
} |
19628147 | pes2o/s2orc | v3-fos-license | Comparison of efficacy between TACE combined with apatinib and TACE alone in the treatment of intermediate and advanced hepatocellular carcinoma: A single-center randomized controlled trial
ABSTRACT Objective: This study was designed to compare the clinical efficacy and safety of transcatheter arterial chemoembolization (TACE) combined with apatinib and TACE alone in the treatment of intermediate and advanced hepatocellular carcinoma (HCC). Methods: From March 2015 to August 2015, a total of 44 patients with moderate and advanced HCC, who were admitted in the Navy General Hospital of China, were included into this study. These patients were randomly divided into 2 groups: group A and group B. Patients in group A underwent TACE alone, while patients in group B underwent the combined treatment of TACE with apatinib. Differences in preoperative general data between these 2 groups were not statistically significant (P > 0.05). All patients were followed up for 12–18 months. Changes in α-fetal protein (AFP) at 3 months after treatment and the objective response rate (ORR) at 3, 6, 9 and 12 months after treatment were compared between these 2 groups. Furthermore, progression-free survival (PFS) and the incidence of adverse reactions were also compared between these 2 groups. Results: AFP levels in groups A and B significantly decreased after 3 months of treatment, compared with the levels before treatment, and the differences were statistically significant (P < 0.05). However, at 3 months after treatment, the difference between these 2 groups was not statistically significant (P > 0.05). ORR at 3, 6, 9 and 12 months after treatment was 36.36%, 27.27%, 13.64% and 9.09%, respectively, in group A; and 60%, 50%, 45% and 35%, respectively, in group B. At 3 and 6 months after treatment, the differences between these 2 groups were not statistically significant (P > 0.05); while at 9 and 12 months after treatment, the differences between these 2 groups were statistically significant (P < 0.05). The median PFS was 6.0 months in group A and 12.5 months in group B, and the difference was statistically significant (P < 0.05). The incidences of complications were related to oral apatinib, such as hypertension, hand-foot syndrome and proteinuria, were higher in group B than in group A, and the differences were statistically significant (P < 0.05). These symptoms all alleviated after symptomatic treatments. Conclusions: For intermediate and advanced HCC, the long-term curative effect of TACE combined with apatinib is better than that of TACE alone. The former can obviously prolong the PFS of patients and has a confirmed safety.
Introduction
Primary hepatocellular carcinoma (HCC) is one of the most common malignant tumors in the world, and is the second cause of cancer-related deaths. 1 According to the Barcelona Clinic Liver Cancer (BCLC) clinical staging system, the standard treatment of HCCs in BCLC stages B and C are transcatheter arterial chemoembolization (TACE) and oral sorafenib. 2 Sorafenib is the first molecular targeted drug approved for the treatment of liver cancer. However, its therapeutic effect is limited; and no other alternative therapy can work once drug resistance occurs. Vascular endothelial growth factor receptor (VEGFR)-2 is closely associated to the occurrence of liver cancer. Apatinib mesylate is a novel VEGFR-2 inhibitor that has the highest selectivity, it has 10 times binding affinity of VEGFR-2 tyrosine kinase than that of sorafenib. This drug can block the migration and proliferation of vascular endothelial cells, reduce tumor microvessel density, and inhibit tumor growth. 3 In addition, apatinib can also inhibit multiple ATP binding sites, reverse multidrug resistance and improve the efficacy of traditional chemotherapy drugs. 4,5 Studies have revealed that the number of vascular endothelial growth factor (VEGF)positive cells in residual liver cancer tissues significantly increased after TACE operation, and VEGF-A level significantly increased in patients with rapid progression after TACE, promoting neovascularization. 6,7 Therefore, due to the characteristics of the selective inhibition on VEGFR-2, apatinib may inhibit tumor neovascularization after TACE, and thereby play an anticancer role. When concurrently used, it may increase anti-tumor treatment efficacy.
At present, the phase II clinical trial of apatinib for patients with BCLC stage B or C of HCCs has been completed. The trail reveals that apatinib is effective in treating moderate and advanced HCCs. The objective response rate (ORR, complete remission [CR]C partial remission [PR]) was 5.00%, and the disease control rate (DCR, CRCPRC stable disease [SD]) was 43.80% in 121 HCCs as evaluated at the end of 3 treatment cycles. Serum AFP level was significantly decreased after the treatment. The therapeutic effects of apatinib for liver cancer has been recognized to a certain extent. 8 The phase III clinical trial is being performed. In this single-center randomized study, efficacy and adverse reactions between TACE combined with apatinib and TACE alone in treatment of moderate and advanced HCCs are compared with provide more clinical bases for the combined treatment of TACE and apatinib for moderate and advanced HCC. At present, similar studies have not been reported.
Inclusion and exclusion criteria
Inclusion criteria: (1) patients with HCC diagnosed by biopsy pathology; (2) patients who were in stage B or C according to the BCLC staging system, and could not tolerate surgical resection or refused surgical resection; (3) patients with Child-Pugh A or B liver function; (4) patients without heart, lung oror kidney dysfunction; (5) patients with an Eastern Cooperative Group Performance Status (ECOG) score of 2 points; (4) patients who did not receive other treatments before the operation.
Exclusion criteria: (1) patients with severe dysfunction of the heart, lung, kidney and other important organs; (2) patients with chemotherapy contraindications: white blood cell count <3£ 10 9 / L, platelet count <40£ 10 9 /L, or hemoglobin <60 g/L; (3) patients with severe coagulation disorders (prothrombin time >18 seconds or hemorrhagic tendency); (4) patients with hepatic artery-portal vein fistula or tumor thrombi in the middle trunk or in the left and right trunks of the hepatic vein; (5) patients with large amounts of ascites or refractory ascites; (6) patients with metastasis in other distant organs; (7) patients who discontinued treatment due to severe adverse reactions of TACE (such as Child-Pugh score C, liver abscesses, hepato-renal syndrome,etc.) or stopped taking apatinib for more than one month due to various reasons.
General clinical data
From March 2015 to August 2015, a total of 58 patients with moderate and advanced HCCs, who were admitted in the Navy General Hospital of China, were enrolled into this study. All patients were confirmed to have HCCs by biopsy and were in line with the diagnostic criteria of primary liver cancer. Among these patients, 14 patients were excluded based on exclusion criteria, including 6 patients with distant metastasis, 2 patients with portal vein trunk tumor thrombosis, one patient with refractory ascites, one patient with renal dysfunction and 4 patients who have received other strategies before. All patients were fully informed of the study and provided a signed informed consent that stated their willingness to participate in the randomized controlled trial. According to the random number table method, the 44 included HCC patients were randomly divided into 2 groups: group A and group B (n D 22, each group). General clinical data such as age, gender, the etiology of liver cancer (hepatitis B, hepatitis C, alcoholic liver and others), the primary tumor diameter and Child-Pugh grade are shown in Table 1. Differences in preoperative general data between these 2 groups were not statistically significant (P > 0.05, Table 1).
Therapeutic methods
Patients in group A underwent TACE alone, while patients in group B underwent the combined treatment of TACE and oral apatinib.
TACE method The Seldinger method was adopted: Conventional femoral artery puncture and intubation were performed, and angiography of proper hepatic artery, superior mesenteric artery, arteria phrenica, left gastric artery and adrenal gland was performed to determine the source of the blood supply of the tumors. General situations of the portal vein were explored by indirect portal vein angiography through the superior mesenteric artery or splenic artery. Next, conventional superselective catheterization of the hepatic artery and other feeding arteries were performed, in which the superselective catheters were inserted to each feeding artery of the tumor lesions using a microcatheter. Subsequently, 30 mg of epirubicin powder, 50 mg of oxaliplatin powder, 10-20 ml of lipiodol and 10-20 ml of iodine contrast agent were prepared into the emulsifier and was slowly infused via the transcatheter; and the feeding arteries were embolized using Embosphere particles (300-500 mm; Merit Medical, USA). The endpoint of the embolization was the stagnation of blood flow in the feeding arteries. Routine antiemetic, stomach protecting treatments and other symptomatic treatments, as well as liver protection therapy, were performed after the operation. Abdominal CT, blood biochemistry, and blood routine index were rechecked at 45 d after the operation. TACE would be performed again when the following criteria were met: (1) lipiodol was not found in the lesions, and enhanced CT revealed residual or recurrent lesions; (2) patients had Child-Pugh A or B liver function; (3) patients did not have contraindications of TACE treatment such as serious hepatic and renal damage, uncorrectable coagulation dysfunction, or dyscrasia. When lesions were completely necrotized and AFP was reduced to normal levels, TACE treatment could be stopped. Patients in group A were observed until the lesions progressed, while patients in group B continued to take apatinib orally. Patients who encountered disease progression withdrew from this study and turned to other treatments.
Administration method of apatinib Apatinib was orally taken for the first time 4 d after TACE, and the initial dose was 500 mg/day. The dose was adjusted after 1-2 weeks, according to the tolerance of patients to the drug. The original dose was maintained when the patient had good tolerance or mild adverse reactions. If obvious adverse reactions appeared, the dose was reduced to 250 mg/day; and adverse reactions were observed for a period of time. If patients could still not tolerate the treatment or serious adverse reactions appeared, the discontinuation of the administration of the drug was considered; and the situation of the patient was observed for 1-2 weeks. When adverse reactions degraded or disappeared, the dose was recovered gradually to the original dose. Apatinib was stopped 4 d before the next course of TACE, and was resumed 4 d after TACE. If stopped taking the drug for more than one month, the patient was considered for exclusion from the study.
Evaluation method of therapeutic effect
Enhanced CT or MRI, as well as the examination of serum AFP value, were performed at 3, 6, 9 and 12 months after treatment. The curative effects were evaluated according to the modified response evaluation criteria in solid tumor (mRECIST), which has 4 levels: CR, all enhanced imaging of the target lesions in the arterial phase disappeared; PR, the total reduction of the diameter of the target lesions (enhanced arterial phase) was 30%; SD, the diameter of the target lesion did not reduce to that in PR and did not increase to that in disease progression (PD); PD, the diameter of the target lesion (enhanced imaging in the arterial phase) increased by at least 20% compared with the baseline value, or new lesions appeared. 9 ORR D (CRCPR) / total number of cases £ 100%. Secondary endpoints of the study also included progression-free survival (PFS), which was defined as the interval from the beginning of intervention treatment to the time of PD or death. The starting point was the beginning of the first treatment, and deadline was the time of PD or the last follow-up. Other treatments such as radiotherapy or radiofrequency ablation were performed based on the patient's condition when the patient's disease progressed.
Statistics analysis
Statistical analysis was conducted using statistical software SPSS 21.0. Measurement data were expressed as mean § standard deviation (x § SD). Intra-group comparison was conducted using paired t-test. Intergroup comparison was conducted using independent samples t-test. Count data were expressed as percentage and counts, and were compared using Chi-square test. Data with non-normal distribution were analyzed using Wilcoxon rank-sum test. The statistics of median PFS was conducted using the Kaplan-Meier method, and the survival rate curve was drawn. Intergroup difference in survival was analyzed using the log-rank test. Inspection level was set as a D 0.05.
Results
In group A, 22 patients were treated with 95 times of TACE, while in group B, 22 patients were treated with 75 times of TACE. There were no serious complications in these 2 groups of patients after TACE. In group B, 2 patients withdrew from the group. Among them, one patient developedmassive upper gastrointestinal tract hemorrhage on the second day after oral administration of apatinib and one course of TACE. The treatment was stopped and the patient withdrew from the group. The other patient developed side effects such as severe hypertension, oral and tongue ulcers. This patient decided to stop the administration of the drug for more than a month, and thereby withdrew from the group. Practically, the curative effects in a total of 42 patients were evaluated (22 patients in group A and 20 patients in group B). There were no significant difference in the basic situations of these 2 groups such as age, gender, Child-Pugh classification, ECOG score, etiology, BCLC staging, primary tumor diameter, AFP level and TACE number between these 2 groups of patients (Table 1). All patients were followed-up for 12-18 months. All enrolled patient must be followed up for at least 12 months. The cutoff date for follow-ups was August 15, 2016, namely the date when the last enrolled patient had been followed up for 12 months. The median duration of follow-up for all the patients was 9.65 months.
Comparison of the clinical curative effect between these 2 groups
At 3 months after treatment, AFP levels in groups A and B both significantly decreased. These levels decreased to 67.20 (3.5-265.8) in group A and 43.80 (2.5-209.6) in group B. Furthermore, compared with levels before treatment, the differences were statistically significant (Z: -2.289 and -2.953, respectively; P < 0.05, both). There was no significant difference between the 2 groups (Z D -0.126, P D 0.9 > 0.05). Moreover, AFP levels decreased to normal levels in 5/19 patients in group A, who previously had elevated AFP; while AFP levels decreased to normal levels in 6/16 patients in group B, who previously had elevated AFP. As shown in Table 2, the ORR at 3 and 6 months after treatment were 36.36% (8/22) and 27.27% (6/22), respectively, in group A; while ORR was 60% (12/20) and 50% (10/20), respectively, in group B. Even though there were differences between these 2 groups, these differences were not statistically significant (both, P > 0.05). The ORR at 9 and 12 months after treatment were 13.64% (3/22) and 9.09% (2/22), respectively, in group A; while ORR was 45% (9/20) and 35% (7/20), respectively, in group B. Differences between these 2 groups were statistically significant (P < 0.05, for both). The oral use of apatinib after TACE was advantageous in controlling and reducing the lesions, thereby improving ORR (Fig. 1).
Comparison of PFS between the 2 groups
All patients were followed-up for 12-18 months. The median PFS was 6.0 months in group A and 12.5 months in group B, and the difference between these 2 groups was statistically significant (X 2 D 6.576, P D 0.01 < 0.05; Fig. 2).
Comparison of adverse reactions between these 2 groups
Different extents of postembolization syndrome (PSE) appeared in groups A and B after TACE, including fever, abdominal pain, nausea, vomiting, etc. The incidences in these 2 groups are listed in Table 3. Differences between these 2 groups were not statistically significant. There was no significant difference in the incidence of bone marrow suppression induced by chemotherapy drugs (P > 0.05).
In group B, apatinib-related adverse reactions included the following: hypertension (16 patients, 80%), hand-foot syndrome (11 patients, 55%), proteinuria (9 patients, 45%), diarrhea (4 patients, 20.0%), and oral ulcers (2 patients, 10%). Among those patients, grade 3 or 4 adverse effects occurred in 3 patients. Severe hand-foot syndrome(grade 3) occurred in one patient, severe diarrhea(grade 3) occurred in one patient and massive upper gastrointestinal hemorrhage(grade 4) occurred in one patient. The first patient with severe hand-foot syndrome stopped the treatment of 2 weeks. After symptomatic treatment, hand-foot skin reaction degraded, and the treatment was resumed. The latter 2 patients with severe hand-foot syndrome were excluded due to severe adverse effects. In the rest of patients, their syndromes were alleviated after drug reduction and symptomatic treatments. The incidences of complications related to oral apatinib such as hypertension, hand-foot syndrome and proteinuria were higher in group B than in group A, and the differences were statistically significant (P < 0.05, Table 3). No severe complications such as hypertensive crisis, heart failure, or liver/kidney failure were observed in these 2 groups.
Discussion
China is one of the countries that have a high incidence of HCC in the world. When seeing a doctor after onset, most patients are already in the late stage of the disease. Hence, they cannot tolerate radical surgery. TACE can effectively prolong the survival time of patients with HCC, and improve the 2-years survival rate of patients. Therefore, TACE has become the firstchoice non-surgical treatment of HCC. 10 The occurrence and metastasis of HCC depend on neovascularization. However, chemoembolization usually cannot completely destroy these tumors, and the local hypoxia environment induced by embolization also stimulates neovascularization in tumors. 11 At present, it has been confirmed that there are a variety of angiogenic factors in patients with HCC. Among these factors, vascular endothelial growth factor (VEGF) is the strongest angiogenic factors in the body. 12 After TACE, the expression of VEGF in remnant peripheral tumor tissues becomes higher, more invasive and metastatically active. 13 These factors increase the invasion and metastasis ability of tumors, and become the basis of disease progression or the production of new lesions. Therefore, inhibiting high expression levels of VEGF in tumor cells induced by TACE may become an important link to improve the long-term efficacy of TACE. Sorafenib is currently the only oral targeted drug approved to treat advanced HCC. It is a multi-kinase inhibitor that plays an anti-angiogenesis role by exerting on RAF kinase, VEGFR-2, VEGFR-3 and PDGFR-b tyrosine kinase. 14 Llovet et al. reported the clinical trial of SHARP. Results revealed that sorafenib could prolong the PFS of patients with advanced HCC by 3 months. 15 However, its response rate is very low in HCC , 16 and its curative effect has some limitations when concurrently used with TACE. A multi-center study revealed that TACE combined with sorafenib could prolong mTTP by nearly 2 months, but the difference was not statistically significant. 17 A study in China also revealed that TACE combined with sorafenib in the treatment of HCC in BCLC stage B did not show significant advantages, compared with TACE alone.
As a highly selective VEGFR-2 blocker, apatinib can block the migration and proliferation of vascular endothelial cells, decrease tumor microvessel density, and inhibit tumor growth. Its affinity is 10 times of that of sorafenib. 3 In a randomized, open-labeled, multi-center Phase II clinical study, the effects of 2 different doses of apatinib on the mOS were compared. The mOS of patients who received 850 mg/d vs. 750 mg/g of apatinib were 9.71 vs. 9.82 months respectively. The time to progression (TTP) were 4.21 vs. 3.32 months in patients with BCLC stage B and C in the phase II clinical trial. 8 These were longer compared with those in the study of heart and kidney protection (SHARP). 15 Therefore, apatinib combined with TACE can inhibit neovascularization in tumor tissues after TACE, and thereby plays an anti-cancer role. Furthermore, the concurrent administration of apatinib may increase anti-tumor efficacy, providing more choices for patients with advanced HCC This study revealed that for patients with intermediate and advanced HCC, TACE combined with apatinib treatment could effectively prolong PFS compared with TACE alone; and bring a higher long-term ORR. However, the difference in short-term ORR was not statistically significant. The cause may be as follows: (1) the number of cases in the group was too small to reveal a statistically significant difference; (2) VEGF level increased after TACE and was inhibited by apatinib again, and an amount of time was required to achieve efficacy, thereby allowing this drug to primarily improved the long-term efficacy instead of short-term efficacy in patients. In addition, these 2 treatment options could effectively decrease AFP levels and inhibit tumor growth. This need to be further confirmed by large-sample multi-center randomized controlled studies.
One patient developed severe diarrhea in the TACE combined with apatinib group, who stopped the treatment of more than one month and withdrew from the study. The symptoms gradually relieved after stopping treatment and symptomatic treatment. The remaining patients could all tolerate the adverse reactions, and all symptoms alleviated after symptomatic treatments. There was no significant difference in the incidences of PES after embolization in the 2 groups. The adverse reaction rate was higher in the TACE combined with apatinib group than in the TACE alone group, which mainly manifested as hypertension, hand-foot syndrome, proteinuria and oral ulcer. Among these patients, one patient developed grade 3 hand-foot syndrome, and symptoms improved 2 weeks after stopping treatment and symptomatic treatment. Afterwards, the administration of apatinib was recovered. Overall adverse reactions in patients were acceptable and serious adverse reactions did not occur. Furthermore, its safety was confirmed.
We conducted the first randomized control study to investigate the effects of TACE combined with apatinib in the treatment of HCC. The curative effect of TACE combined with apatinib on intermediate and advanced HCC was better than TACE alone, and the former significantly prolonged the PFS of patients. After symptomatic treatments, most of the adverse reactions did not affect the treatment. Furthermore, its safety was confirmed, which is worth of popularization and application in clinic. However, The shortcomings of our study is small number of enrolled patients and a short time of follow-up. OS is not discussed in this paper. Therefore, perspective randomized controlled study in multi-center need to be conducted to demonstrate the effects and safety of TACE combined apatinib in treatment of HCCs. | 2018-04-03T04:27:25.798Z | 2017-05-26T00:00:00.000 | {
"year": 2017,
"sha1": "143700dead061c9471caf3bca509aeb8898dc9f3",
"oa_license": "CCBY",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.1080/15384047.2017.1323589?needAccess=true",
"oa_status": "HYBRID",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "7d0ae945005086eb2bc61b8884041ce6fe4fd288",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
5088108 | pes2o/s2orc | v3-fos-license | Duality Products of Multivectors and Multiforms, and Extensors
In this paper we study in details the properties of the duality product of multivectors and multiforms (used in the definition of the hyperbolic Clifford algebra of multivefors) and introduce the theory of the k multivector and l multiform variables multivector (or multiform) extensors over V studying their properties with considerable detail.
Introduction
In this article we first briefly recall the exterior algebras of multivectors (elements of V ) and multiforms (elements of V * ) associated with a real vector space V of finite dimension. Next we explore the structure of the duality product < , > of multivectors by multiforms used in the definition of the hyperbolic Clifford algebra Cℓ(V ⊕ V * , < , >) of multivecfors [5]. We detail some important properties of the left and right contracted products among the elements of V and V * . Next, we give a theory of the k multivector and l multiform variables multivector (or multiform) extensors over V (defining the spaces ext l k (V ) and * ext l k (V )) introducing the concept of exterior product of extensors, and of several operators acting on these objects as, e.g., the adjoint operator, the extension operator, and the generalized operator procedure. We study the properties of these operators with considerable detail. Extensors fields (with representatives that are appropriate mappings from an open set U ⊂ M to ext l k (V ) or * ext l k (V )) and the operators acting on them will be shown in forthcoming papers to provide a valuable and simplifying tool (and an improvement over a previous presentation [1,2,3,4] which uses only the multivector and extensor calculus) in the study of some of the fundamental ingredients of the differential geometry 1 of an arbitrary manifold M equipped with an arbitrary connection.
Multivectors and Multiforms
Let V be a vector space over R with finite dimension, i.e., dim V = n with n ∈ N, and let V * be its dual vector space. Recall that Let us consider an integer number k with 0 ≤ k ≤ n. The real vector spaces of k-vectors over V, i.e., the set of skew-symmetric k-contravariant tensors over V , and the real vector spaces of k-forms over V , i.e., the set of skew-symmetric k-covariant tensors over V, will be as usually denoted by k V and k V * , respectively.
Given a vector space V over the real field R, we define V as the exterior direct sum V = n r=0 ⊕ r V = n r=0 r V.
To simplify the notation we sometimes write simply V = R + V + 2 V + · · · + n−1 V + n V.
As it is well known the set of multivectors over V has a natural structure of vector space over R and we have An element of V will be called a multivector over V . If X ∈ V we write: In what follows we shall need also the vector space V * = n r=0 r V * .
An element of V * will be called a multiform over V . As in the case of multivectors we simplify our notation and as in the case of multivectors we simply write: and if Φ ∈ V * we write Of course, V * has a natural structure of real vector space over R. We have, We recall that k V is also called the homogeneous multivector space (of degree k), and to k V * the homogeneous multiform space (of degree k).
Let us take an integer number k with 0 ≤ k ≤ n. The linear mappings such that if X = X 0 + X 1 + · · · + X n and Φ = Φ 0 + Φ 1 + · · · + Φ n , then are called the k-part operator (for multivectors) and the k-part operator (for multiforms), respectively. X k is read as the k-part of X and Φ k is read as the k-part of Φ. It should be evident that for all X ∈ V and Φ ∈ V * : The linear mappings are called the grade involution operator (for multivectors) and the grade involution operator (for multiforms), respectively. The linear mappings are called the reversion operator (for multivectors) and the reversion operator (for multiforms), respectively. Both of V and V * endowed with the exterior product ∧ (of multivectors and multiforms!) are associative algebras, i.e., the exterior algebra of multivectors and the exterior algebra of multiforms, respectively.
Duality Scalar Product
The duality scalar product of a multiform Φ with a multivector X is the scalar Φ, X (i.e., the real number) defined by the following axioms • For all α, β ∈ R : α, β = β, α = αβ.
• For all Φ p ∈ p V * and X p ∈ p V (with 1 ≤ p ≤ n): where {e j , ε j } is any pair of dual bases over V.
Note that for all ω ∈ V * and v ∈ V it holds We present two noticeable properties for the duality scalar product between p-forms and p-vectors • For all ω 1 , . . . , ω p ∈ V * and v 1 , . . . , v p ∈ V : The basic properties for the duality scalar product are the non-degeneracy and the distributive laws on the left and on the right with respect to addition of either multiforms or multivectors, i.e., • Φ, X = 0, for all Φ =⇒ X = 0, Φ, X = 0, for all X =⇒ Φ = 0, and, 4 Duality Contracted Products
Left Contracted Product
The left contracted product of a multiform Φ with a multivector X (or, a multivector X with a multiform Φ) is the multivector Φ, X| (respectively, the multiform X, Φ|) defined by the following axioms: • For all Φ p ∈ p V * and X p ∈ p V with 0 ≤ p ≤ n : • For all Φ p ∈ p V * and X q ∈ q V (or X p ∈ p V and Φ q ∈ q V * ) with 0 ≤ p < q ≤ n : where {e j , ε j } is any pair of dual bases for V and V * .
• For all Φ ∈ V * and X ∈ V : if Φ = Φ 0 + Φ 1 + · · · + Φ n and X = X 0 + X 1 + · · · + X n , then Note that the (q − p)-vector Φ p , X q | and the (q − p)-form X p , Φ q | have frame independent character, i.e., they do not depend on the pair of frames {e j , ε j } chosen for calculating them.
The left contracted product has the following basic properties: • Let us take • For all X ∈ V and Φ, Ψ ∈ V * : • For all Φ ∈ V * and X, Y ∈ V : The left contracted product satisfies the distributive laws on the left and on the right.
• For all Φ, Ψ ∈ V * and X, Y ∈ V : • For all X, Y ∈ V and Φ, Ψ ∈ V * : Proof. We present only the proof of the property given by Eq. (25), the other proofs being somewhat analogous. First note that if Ψ q−p ∈ q−p V * and {e j , ε j } is any pair of dual bases for V and V * , we can write Then, using the definition (21) and the above equation we have and the result is proved.
Right Contracted Product
The right contracted product of a multiform Φ with a multivector X (or, a multivector X with a multiform Φ) is the multiform |Φ, X (respectively, the multivector |X, Φ ) defined by the following axioms: • For Φ p ∈ p V * and X p ∈ p V¨with n ≥ p ≥ 0 : • For Φ p ∈ p V * and X q ∈ q V¨(or X p ∈ p V and Φ q ∈ q V * ) with n ≥ p > q ≥ 0 : where {e j , ε j } is any pair of dual bases for V and V * .
• For all Φ ∈ V * and X ∈ V : if Φ = Φ 0 + Φ 1 + · · · + Φ n and X = X 0 + X 1 + · · · + X n , then The right contracted product satisfies the following basic properties: • For all Φ ∈ V * and X, Y ∈ V : • For all X ∈ V and Φ, Ψ ∈ V * : The right contracted product satisfies also the following distributive laws: • For all Φ, Ψ ∈ V * and X, Y ∈ V : • For all X, Y ∈ V and Φ, Ψ ∈ V * : We present two noticeable properties relating the left and right contracted products.
• For all Φ p ∈ p V * and X q ∈ q V with p ≤ q : • For all X p ∈ p V and Φ q ∈ q V * with p ≤ q : Proof. We present only the proof of the property given by Eq. (42), the proofs of the other properties are analogous. Let Φ p ∈ p V * and and the result is proved.
Extensors
Let ⋄ 1 V, . . . and ⋄ k V be k subspaces of V such that each of them is any sum of homogeneous subspaces of V, and let ⋄ 1 V * , . . . and ⋄ l V * be l subspaces of V * such that each of them is any sum of homogeneous subspaces of V * .
If ⋄ V is any sum of homogeneous subspaces of V, a multilinear mapping is called a k multivector and l multiform variables multivector extensor over V.
If ⋄ V * is any sum of homogeneous subspaces of V * , a multilinear mapping is called a k multivector and l multiform variables multiform extensor over V.
The set of all the k multivector and l multiform variables multivector extensors over V has a natural structure of real vector space, and will be denoted by the highly suggestive notation ext( . When no confusion arises, we use the more simple notation ext l k (V ) for that space.
We obviously have that: The set of all the k multivector and l multiform variables multiform extensors over V has also a natural structure of real vector space, and will be denoted by ext , and when no confusion arises, we use the simple notation * ext l k (V ) for this space. Also, we have, 6 Algebra of Extensors
Exterior Product of Extensors
We define the exterior product of τ ∈ ext l k (V ) and σ ∈ ext Note that on the right side appears an exterior product of multivectors (respectively, an exterior product of multiforms).
The linear mappings are called the ⋄ V -part operator (for multivectors) and ⋄ V * -part operator (for multiforms), respectively.
It should be evident that for all X ∈ V and Φ ∈ V * : Thus, ⋄ V -part operator and ⋄ V * -part operator are the generalizations of k and k . Let τ be a multivector extensor of either one multivector variable or one multiform variable. If is called the duality adjoint of τ.
The basic properties of the adjoint of multivector extensors are: Proof. We present only the proof of the property given by Eq. (57), the other is analogous. First, observe that if X ∈ ⋄ 1 V, then there exists µ integer numbers p 1 , . . . , p µ with 0 ≤ p 1 < · · · < p µ ≤ n such that thus, we can put Now, with the above observation, we have but, by the definition (14) we have that and from Eq. (59), we can write Now, if we see τ s l (X s l ), Φ s l as a scalar product of s l -vectors, then we have and from Eqs. (60), (61), and taking into account Eq. (59) we have and the result is proved.
Let σ be a multiform extensor of either one multivector variable or one multiform variable. Then, if σ ∈ ext( ⋄ for each X ∈ ⋄ 2 V (respectively, for each X ∈ ⋄ 4 V ) is called the duality adjoint of σ. The basic properties for the adjoint of multiform extensors are: The linear mapping ( ) △ will be called the duality adjoint operator.
Extension Procedure for Operators
Let λ be a linear operator on V, i.e., a linear map It can be extended in such a way as to give a linear operator on V, namely λ, which is defined by where {e j , ε j } is any pair of dual bases for V and V * . We emphasize that λ is a well-defined linear operator on V. Note that each k-vector ε j 1 ∧ · · · ∧ ε j k , X λ(e j 1 ) ∧ · · · ∧ λ(e j k ) with 1 ≤ k ≤ n does not depend on the choice of {e j , ε j } , and the linearity follows just from the linearity of the duality scalar product. We call λ the extended of λ (to multivector operator).
The extended of a vector operator λ has the following basic properties: • For all α ∈ R, v ∈ V, and X, Y ∈ V : We observe that the four basic properties as given by Eq.(67), Eq.(68), Eq.(69) and Eq.(70) are completely equivalent to the extension procedure of a vector operator.
Let λ be a linear operator on V * , i.e., a linear map V * ∋ ω −→ λ(ω) ∈ V * . It is possible to extend λ in such a way to get a linear operator on V * , namely the operator λ, defined by such that where {e j , ε j } is any pair of dual bases for V and V * . We emphasize that λ is a well-defined linear operator on V * . We call λ the extended of λ (to multiforms).
The extended of a form operator λ has the following basic properties.
There exists a relationship between the extension procedure of a vector operator and the extension procedure of a form operator.
Let us take a vector operator (or, a form operator) λ. As we can see, the duality adjoint of λ is just a form operator (respectively, a vector operator), and the duality adjoint of λ is just a multiform operator (respectively, a multivector operator). It holds that the duality adjoint of the extended of λ is equal to the extended of the duality adjoint of λ, i.e., Thus, it is possible to use the more simple notation λ △ to mean either (λ) △ or λ △ .
We present some properties for the extended of an invertible vector operator λ.
• For all Φ ∈ V * , and X ∈ V : We present only the proof for the property given by Eq.(78), the other proofs are analogous. Proof. Let us take X ∈ V and Φ, Ψ ∈ V * . A straightforward calculation, using Eq.(57), Eq.(27), Eq.(75) and Eq.(72), yields whence, by the non-degeneracy of duality scalar product the result follows.
We present now some properties for the extended of an invertible form operator λ.
Action of Extended Operators on Extensors
Let λ be an invertible linear operator on V . As we saw above, the extended of λ, denoted λ, maps multivectors over V into multivectors over V. However, it is possible to extend the action of λ in such a way that λ maps multivector extensors over V into multivector extensors over V . We define the linear mapping It that way λ can be thought as a linear multivector extensor operator.
For instance, for τ ∈ ext 0 1 (V ) the above definition yields and for τ ∈ ext We give some of the properties of the action of the extended of a vector operator λ on multivector extensors.
• For all τ ∈ ext l k (V ) and σ ∈ ext s r (V ) : • For all τ ∈ * ext l k (V ) and σ ∈ ext s r (V ) : We present the proof of the property given by Eq.(91), the other proofs are analogous. Proof. Without any loss of generality, we give the proof for the particular whence, the expected result follows.
We present now some of the properties of the action of the extended of a form operator λ on multiform extensors.
Generalization Procedure for Operators
Let γ be a linear operator on V, i.e., a linear map V ∋ v −→ γ(v) ∈ V. It can be generalized in such a way to give a linear operator on V, namely γ ⌣ , which is defined by where {e j , ε j } is any pair of dual bases for V and V * . We note that the multivector γ(e j ) ∧ ε j , X| does not depend on the choice of {e j , ε j } , and that the linearity of the duality contracted product implies the linearity of γ ⌣ . Thus, γ ⌣ is a well-defined linear operator on V.
We call γ ⌣ the generalized of γ (to multivector operator).
The generalized of a vector operator γ has the following basic properties.
Let γ be a linear operator on V * , i.e., a linear map V * ∋ ω −→ γ(ω) ∈ V * . It is possible to generalize γ in such a way as to get a linear operator on V * , namely γ ⌣ , which is defined by where {e j , ε j } is any pair of dual bases over V.
We emphasize that γ ⌣ is a well-defined linear operator on V * , and call it the generalized of γ (to a multiform operator). The generalized of a form operator γ has the following basic properties.
There exists a relationship between the generalization procedure of a vector operator and the generalization procedure of a form operator.
Let γ a vector operator (or, a form operator). As we already know, the duality adjoint of γ is just a form operator (respectively, a vector operator), and the duality adjoint of γ ⌣ is just a multiform operator(respectively, a multivector operator). The duality adjoint of the generalized of γ is equal to the generalized of the duality adjoint of γ, i.e., It follows that is possible to use a more simple notation, namely γ We give some of the main properties of the generalized of a vector operator γ.
• For all Φ ∈ V * , and X ∈ V : We prove the property given by Eq.(109), the other proofs are analogous.
Proof. Let us take X ∈ V and Φ, Ψ ∈ V * . A straightforward calculation, by using Eq.(57), Eq.(27), Eq.(103) and Eq.(106), yields and by the non-degeneracy of duality scalar product, the expected result follows. We present some properties for the generalized of a form operator γ.
Action of Generalized Operators on Extensors
Let γ be a linear operator on V . We can generalize the action of γ ⌣ in such a way γ ⌣ is to map multivector extensors over V into multivector extensors over V . We define the linear mapping for each X 1 , . . . , X k ∈ V and Φ 1 , . . . , Φ l ∈ V * . Thus, we can think of γ ⌣ as a linear multivector extensor operator.
Conclusions
In this paper we studied the properties of the duality product of multivectors and multiforms (used for the definition of the hyperbolic Clifford algebra of multivefors [5] ) introducing a very useful notation for the left and right contracted products of multiforms (elements of V * ) and multivectors (elements of V ). Next, we introduced a theory of the k multivector and l multiform variables multivector (or multiform) extensors over V (defining the spaces ext l k (V ) and * ext l k (V )) defining the exterior product of extensors, and several important operations on them. This algebraic theory will play an important role in a presentation of the differential geometry of a manifold M of arbitrary topology discussed in forthcoming papers. | 2014-10-01T00:00:00.000Z | 2007-03-18T00:00:00.000 | {
"year": 2007,
"sha1": "4ab0f37898f0b5c681017962dfd8bc386c502f13",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "4ab0f37898f0b5c681017962dfd8bc386c502f13",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics",
"Physics"
]
} |
16697018 | pes2o/s2orc | v3-fos-license | Profile of Expression of Genes Encoding Matrix Metallopeptidase 9 (MMP9), Matrix Metallopeptidase 28 (MMP28) and TIMP Metallopeptidase Inhibitor 1 (TIMP1) in Colorectal Cancer: Assessment of the Role in Diagnosis and Prognostication
Background Studies on the pathomechanism of colorectal cancer (CRC) expansion indicate a significant role of metalloproteinases and their inhibitors in the extracellular matrix. The results of the analysis of a profile of transcriptional activity of genes encoding metalloproteinases were the basis of the hypothesis indicating changes in the expression of genes encoding MMP9, MMP28, and TIMP1 as an additional diagnostic and prognostic marker of CRC. Material/Methods The material consisted of samples obtained from resected tumors and healthy tissue samples from 15 CRC patients (aged 46–72 years) at clinical stages (CSs) I and II–IV. Gene expression analysis was done using microarrays. Microarray data analysis was done using the GeneSpring 11.5 platform. The results were validated using the qRT-PCR technique. Results We found high levels of expression of MMP9 at each CS, as well as in the tissues at the early stage of CRC. Additionally, we observed high levels of expression of TIMP1 and low levels of MMP28 genes in CS II–IV. No statistically significant differences based on the stage of CRC were observed. Conclusions MMP9 gene profile may be a complementary diagnostic marker in CRC. The results suggest a crucial role of MMP9 at the early stage of carcinogenesis in the large intestine. The increase in MMP9 and TIMP1 mRNA concentration and the decrease in MMP28 in the large intestinal tissue may be a confirmation of cancer, but it may not indicate the advance of CRC.
Background
As the population is aging, both the rate of cancer incidence and the number of cancer-related deaths are constantly increasing. It is also related to colorectal cancer (CRC), which is the most common malignancy of the digestive tract. A significant therapeutic problem experienced by clinicians who treat colorectal patients is connected with high invasiveness of CRC and its ability to form metastases in distant organs. Absence of CRC symptoms, scant symptoms, or symptoms ignored by patients frequently result in treatment at a late advanced stage of the disease. Therefore, early diagnosis is of crucial importance. Recently, much hope has been placed on the constant development of molecular biology techniques in terms of development of early diagnosis and possible treatment options. The most recent diagnostic and therapeutic trends are focused on the molecular basis of CRC. Significantly changed transcriptional gene activity may be a complementing prognostic and diagnostic marker in CRC, and it may be the basis for establishing new therapeutic goals. Studies on the pathomechanism of cancer expansion demonstrate a significant role of extracellular matrix metalloproteinases (MMPs) in this process [1][2][3]. Their proteolytic activity significantly influences extracellular matrix components (EMCs) through the degradation of its components and regulation of the activity of signal particles [4,5]. Matrix metalloproteinases participate in regulating the ability of cells to differentiate, proliferate, and in the process of apoptosis, adhesion or migration. At some stages, MMPs may also prepare malignant cells for migration, invasion, and dissemination to distant organs. They also participate in malignant transformation itself. However, tissue inhibitors of metalloproteinases (TIMPs) are mainly responsible for controlling the functioning of MMPs, in addition to the other functions they perform. Consequently, their role is to minimize the degradation of EMCs. As a result, MMPs and TIMPs take part in the reconstruction of EMCs, thus influencing transformation of malignant cells and cancer progression.
The aim of the present paper is to analyze a profile of transcriptional activity of genes encoding MMP9, MMP28, and TIMP1, depending on the clinical stage of CRC. The analysis aims at considering them to be potential complementing diagnostic and prognostic markers of CRC.
Material and Methods
The material consisted of samples obtained during the resection of the large intestine. The samples were collected from 15 CRC patients aged 46-72 years. The inclusion criterion was: CRC patients in all stages of the disease undergoing elective classic surgical procedures. The exclusion criteria were: patients reoperated on due to the underlying disease; unclear histological confirmation of CRC; patients with genetic, systemic, or metabolic conditions (with the exclusion of obesity as an isolated condition); and previous radiotherapy or chemotherapy. The study was approved by the Bioethics Commission of the Medical University of Silesia (KNW/0022/KB1/21/I/10). Written informed consent was obtained from all study participants.
The material consisted of CRC and healthy tissue samples. The samples were collected from CRC from the external margin to exclude the presence of necrotic tissues in the specimens. Control samples were obtained from the macroscopically healthy intestine, from the part of the specimen maximally distant from the lesion and at least 5 cm distant from the margin considered to be healthy. The tumor tissue was divided into 2 parts. One part was referred for standard histological assessment and the other for molecular analysis, as in the case of the control. The tissue collection was done immediately after the resection of the intestinal fragment, thus limiting to the minimum the effect on cancer. The tissues were dissected using classic surgical techniques only. No electrical devices or ultrasound were used. Prior to the molecular analysis, the material was stored in RNAlater™ (Qiagen) at the temperature of -80°C to prevent sample degradation.
Twenty-eight samples were obtained (15 CRCs and 13 controls). In accordance with the 7 th edition of the AJCC/UICC staging system of CRC, cancer tissues represented different stages of the disease: 3 Clinical Stage I (CSI), 5 Clinical Stage II (CSII), 3 Clinical Stage III (CS III), and 4 Clinical Stage IV (CSIV). The genetic profile of cancerous tissues was analyzed. The profile was compared to the genetic profile of the control tissues. The samples with CRC were divided into 2 groups: "low stage of cancer" LSC (CSI) and "high stage of cancer" HSC (CSII-CSIV).
Methods of molecular analysis
The molecular analysis was started by extracting total RNA from the obtained fragments of the large intestine. In further stages of the study, RNA was the array for the assessment of intestinal transcriptome, using expression microarrays HG-U133A (Affymetrix ® ) and validation of the array experiment with qRT-PCR, based on mRNA concentration profiles: MMP9, MMP28, TIMP1, and the control of endogenous GAPDH and b-actin.
RNA purification and analysis
Total RNA from the sample of the large intestine was isolated with a total RNA isolation kit (Total RNA Prep Plus, A&A Biotechnology). Next, the extracted RNA was purified and digested with DNase I, using columns of an RNase Minikit (Qiagen) in accordance with the manufacturer's instructions. Qualitative assessment of the obtained RNA extracts was performed using 1% agarose gel electrophoresis, stained with ethidium bromide. Additionally, the degree of total RNA integrity was assessed based on the RNA Integrity Number (RIN) parameter -rRNA ratio (28s/18s).
RNA concentration was assessed spectrophotometrically at a wavelength of 260 nm, using Gene Quant II. The total RNA was the array for transcriptome assessment using expression microarray HG-U133A (Affymetrix ® ) and the number of mRNA copies in µg of the total RNA assessed using qRT-PCR (validation of the array experiment).
Transcriptome assessment with expression microarray HG-U133A (Affymetrix ® ) Large intestine transcriptomes were assessed with mRNA expression microarrays, using HG-U133A? (Affymetrix ® , CA). The isolated total RNA was the array for the synthesis of marked cRNA (biotinylated complementary RNA), the synthesis of which was performed using the 3' IVT Express Kit. The obtained particles were hybridized with HG-U133A microarray. At the next stage, the microarrays were washed and marked by immunofluorescence using the Fluidics Station 450 and the Hybridization Wash and Stain Kit. Next, fluorescence intensity of the transcriptomes was read using the GeneChip Scanner 3000 7G and the Affymetrix ® GeneChip ® Command Console ® Software (AGCC) software. Experiment quality control was done at the subsequent stages of transcriptome assessment, starting with the quality assessment using 1% agarose gel electrophoresis of total RNA after extraction from intestinal samples, and of reverse transcription products (cDNA), transcription (cRNA), and cRNA after fragmentation and immediately before the preparation of the hybridization cocktail.
Validation of the array experiment with qRT-PCR
Validation of the array experiment results was done with qRT-PCR, which allowed precise assessment of diagnostic and prognostic values of the determined changes in mRNA concentration of MMP9, MMP28, and TIMP1. Starters synthesized by Oligo IBB PAN were used for amplification. The qRT-PCR reaction was conducted using the SYBR Green Quantitect RT-PCR Kit and the Opticon™ DNA Engine Sequence Detector. The number of mRNA copies in 1 µg of the total RNA extract was determined based on the standard curve made for commercially available DNA specimens of the b-actin gene. For each test, negative control (without RNA array) and endogenous control (of mRNA of b-actin and GAPDH genes) were done. The specificity of the qRT-PCT reaction was assessed based on electrophoretic separation of amplimers at 6% polyacrylamide gel stained with silver salts with size marker pBR322/HaeIII and Tm value determined based on the amplimer melting curve, for each qRT-PCR product. The qRT-PCR reaction was performed in a reaction mix prepared with the use of a robot integrated with ABI PRISM™ 877 thermal cycler, containing 1× Tth.
The statistical analysis of the results
The statistical analysis of the array experiment results was started with global normalization of all microarrays accepted for the comparative analysis using the RMA-express software. Next, hierarchical clustering with Statistica v. 9.0 (StatSoft, Tulsa, OK) was used to assess the stage of differentiation of the intestinal transcriptomes. During the next stage of the comparative analysis of transcriptome groups, we used GeneSpring GX 11.5 (Agilent Technologies) for the analysis of array experiment results.
Statistical significance was p<0.05. For each parameter analyzed, the most significant elements of the descriptive statistics were determined, including mean, standard deviation, median, extreme values (minimal and maximal), and lower (25%) and upper (75%) quartiles. The hypothesis on the compliance of the distribution of a given parameter with the normal distribution was verified using the Shapiro-Wilk test. In the case of the normal distribution of a given parameter, differences between groups that were statistically significant in reference to particular variables were verified at the first stage using one-way analysis of variance (ANOVA), and then the nature of particular relationships was determined with the post hoc Tukey test. Comparisons between the 2 groups were made using the t test for independent samples. In the case of lack of compliance between the distribution of a given parameter and the normal distribution, comparisons between the 2 groups were made using the Mann-Whitney U test. To increase the probability of obtaining the correct results, the Benjamini-Hochberg correction was used.
Validation of the array experiment results was performed with the qRT-PCR technique, which allowed a precise assessment of the diagnostic and prognostic values of the changes in mRNA of MMP9 and MMP28, as well as TIMP1.
Results
The assessment of the differentiation of transcriptome groups characteristic of the samples from the healthy large intestine and CRC was started with the array experiment. After the initial acceptance of microarrays for the comparative analysis carried out in accordance with the instructions of the microarray producer (Affymetrix ® ) and result normalization with the RMA method, transcriptome clustering was done with the use of the hierarchical clustering method (Figure 1).
The obtained results demonstrate that all transcriptomes are clustered in 3 main groups, differentiated by the stage of CRC: C -control group -normal large intestine samples, LSC -CRC samples in CSI and HSC -CRC samples in CSII, CSIII, and CSIV.
1307
The next stage of the analysis included a comparison of LSC and HSC transcriptomes with the control group. The analysis was started with the characteristics of the dispersion of fluorescence signals of 44 mRNA of metalloproteinases and their inhibitors in the transcriptome groups, which was based on the descriptive statistics (median and interquartile range) (Figure 2). The obtained results indicate that mRNA concentration profile in the transcriptome groups depends on the stage of CRC.
To determine whether the observed differences were statistically significant, concentration profiles of 44 mRNA of metalloproteinases and their tissue inhibitors in healthy intestinal tissues (C) and in LSC and HSC were compared using ANOVA. The obtained results indicated that out of 44 mRNA assessed by the oligonucleotide microarray method (HG-U133A, Affymetrix ® ), 11 mRNA differentiate the compared groups with the assumed p<0.05. mRNA heat maps of metalloproteinases and their inhibitors were simultaneously generated for 11 ID mRNA (GeneSpring GX), demonstrating an mRNA concentration profile differentiating CRC from the control. Compared to the mean mRNA concentration in transcriptomes (yellow), particular metalloproteinases were characterized by varied transcription activity. In some cases, an increase in transcription activity was observed with the increase in the stage of CRC. Metalloproteinase 28 was an exception; the transcription activity decreased with an increase in the stage of CRC.
We also used a test that increased the precision of ANOVAthe post hoc Tukey test -indicating which of 11 mRNA selected in the variance analysis (ANOVA) as the particles differentiating the compared transcriptome groups could be used to differentiate the defined groups of large intestine samples. The analysis indicates that out of 11 selected mRNA, 2 differentiate healthy intestinal tissue samples (C) from LSC, 9 differentiate control samples (C) from HSC, and 5 differentiate LSC from HSC.
Based on these initial analyses (one-way ANOVA combined with post hoc Tukey), 2 genes were extracted that differentiate the following: a macroscopically healthy tissue (C) from a cancerous tissue in both LSC and HSC. The next stages of the comparative analysis included the selection of mRNA differentiating healthy intestine samples and CRC in LSC (C vs. LSC) and healthy intestinal samples and CRC in HSC (C vs. HSC) and CRC in LSC and CRC in HSC. The nonparametric Mann-Whitney U test with the Benjamini-Hochberg correction was used. Validation of the array experiment results was done using the qRT-PCR technique ( Table 1).
The results of the analyses indicate that the profile of the expression of genes encoding MMP9, MMP28, and TIMP1 differentiates the cancerous tissue from the healthy intestine. However, the profile does not demonstrate statistically significant differences depending on the stage of CRC. We did not confirm the result of the one-way ANOVA combined with the post hoc Tukey analysis in relation to mRNA of MMP28.
Discussion
Colorectal cancer, the most common cancer of the digestive tract and generally one of the most frequent malignancies in human pathology, is a disease of complex and multi-factorial etiology where the accumulation of genetic and epigenetic changes in the cell plays a crucial role [6][7][8][9][10][11]. The invasiveness of CRC and the ability of cancerous cells to form metastases in distant organs are one of the most significant therapeutic problems, which result in a number of outcome failures [12]. Matrix metalloproteinases play an active part in that process, degrading EMCs, including basement membranes, which in consequence may result in the infiltration of the parenchyma, adjacent structures, and distant metastases after crossing the blood vessel basement membrane [13].
Currently, CRC treatment is based mainly on radical surgery frequently supported by systemic treatment: chemotherapy and radiotherapy [7,14,15]. The effectiveness of treatment strictly depends on the stage of the disease at diagnosis; faster diagnosis of LSC offers a better prognosis. Together with the increase in the stage of CRC, the probability of successful outcome decreases despite adjunctive therapy. The detection of the disease at its earliest and preferably asymptomatic stage is therefore crucial. Hope is offered by modern techniques that use oligonucleotide microarrays to compare gene expression in cancerous and healthy tissues.
Finding the differences in the expression of genes encoding particular proteins may be helpful in explaining the mechanisms resulting in the occurrence and development of CRC, which in turn is the key to establishing new therapeutic possibilities. Analyzing the problem of solid tumors as a whole, including digestive tract cancers and CRC in particular, more and more reports stress the role of MMPs in CRC formation and progression [3,[16][17][18][19][20][21][22][23][24][25][26][27].
The result of the Mann-Whitney U test with the Benjamini-Hochberg correction used to select mRNA differentiating healthy intestinal samples from CRC in LCS confirmed the result of the analysis of one-way ANOVA combined with the post hoc Tukey analysis only in the case of MMP9. However, the test result comparing healthy intestinal samples with CRC in HSC confirmed the result of ANOVA analysis in MMP9, MMP28, and TIMP1. These results were confirmed using qRT-PCR.
The existing data on the prognostic role of MMPs and their inhibitors as markers of CRC progression are not conclusive.
Our study suggests that MMPs and TIMPs may be a substantial element of controlling CRC stage by indicating the complexity of proliferative and migration mechanisms. It seems that an overexpression of MMP9 and TIMP1 and silencing of MMP28 could become a direct independent target of anti-cancer therapy. Because there is a difference in the expression of MMP9 gene in cancerous tissues compared to healthy tissues, the gene demonstrates a diagnostic value; it may serve as an early marker of cancerous transformation due to the changes in its activity as early as at the cancer initiation process. However, the increase in expression of MMP9 encoding gene, the decrease in transcription activity of MMP28, and the increase in transcription activity of TIMP1 gene may be a confirmation of the existence of cancer. Unfortunately, even though it seems that differences in the expression of these genes exist, as shown by the heat map, when LSC is compared with HSC, these differences do not demonstrate statistical significance. Consequently, they cannot be used in differentiating the stage of CRC. In turn, MMP9 seems to take part in a number of key stages of CRC progression, from carcinogenesis, angiogenesis, and the development of primary tumor, to metastases in distant organs. Using immunochemistry techniques, Zeng et al. [28] demonstrated that MMP9 expression correlates with metastases in human CRC, and it is also related to an increase in TIMP1 activity. Furthermore, an increase in mRNA of MMP9 in CRC tissue compared to the concentration in a healthy intestinal tissue increased with disease progression, being a negative prognostic factor for diseasefree survival and overall survival [29]. TIMP1 levels in serum and TIMP1 expression in cancerous tissues are associated with poor clinical outcomes in many types of cancer [29,30].
Although a study demonstrated an increase in the activity of genes encoding MMP9 and TIMP1 taking part in tissue reconstruction and regulating the transformation and progression of cancerous cells, no close connection was observed in regulation between MMP9 and TIMP-1 in CRC [30]. In our study, MMP9 and TIMP1 in CRC were overexpressed compared to in healthy tissues, which is similar to the results obtained by Walkiewicz et al. [31]. Their study demonstrated high levels of expression of MMP9 and TIMP1 genes, particularly in the tissues from an early stage of CRC. MMP9 expression was also increased in higher stages of CRC, but at a much lower level, which may also suggest a role of these proteins in the process of initiation of tumor progression. | 2018-04-03T04:26:35.457Z | 2017-03-15T00:00:00.000 | {
"year": 2017,
"sha1": "38772a3b9629e6adc5e610bf8deab724d960d771",
"oa_license": "implied-oa",
"oa_url": "https://europepmc.org/articles/pmc5363457?pdf=render",
"oa_status": "GREEN",
"pdf_src": "PubMedCentral",
"pdf_hash": "38772a3b9629e6adc5e610bf8deab724d960d771",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
10772648 | pes2o/s2orc | v3-fos-license | The Use of Thermal Imaging in the Evaluation of the Symmetry of Muscle Activity in Various Types of Exercises (Symmetrical and Asymmetrical)
In order to achieve higher efficiency of training and thus better athletic performance, new research and diagnostic methods are constantly being developed, particularly those that are non-invasive. One such a method is thermography, suitable for quantitative and therefore objective evaluation of variables, such as changes in the temperature of the skin covering working muscles. The aim of this study was to use a thermal imaging infrared camera to evaluate temperature changes of symmetric body surfaces over symmetrically working muscles of male scullers after exercising on a two-oared rowing ergometer and compare these to asymmetrically working muscles of handball players after an endurance training session containing elements of an actual game. In the scullers, the mean temperature of body surfaces was always lower post than pre exercise, with no significant differences in an average temperature drop between the opposite sides, indicating that the work of the muscles involved in the physical exertion on the rowing ergometer was symmetrical. In contrast, in the handball players, skin temperatures in symmetric areas over the asymmetrically working muscles showed statistically significant differences between sides, which was associated with the functional asymmetry of training. This study indicates that thermal imaging may be useful for coaches in the evaluation of technical preparations in sports in which equal involvement of symmetric muscles is a condition of success, e.g. in scullers.
Introduction
The success in modern sports depends on rational training, brought to perfection by coaches and researchers collaborating with athletes in the training process. The optimal use of the abilities of athletes in training and competition requires the evaluation of the effectiveness of stimuli, and constant monitoring of training effects in relation to completed tasks. The decisive factor is data on the athlete's functional potential and his/her ability to use it in an optimal way (Bernstein, 1988). This means that sport performance is the sum of tactical and technical skills, motor and coordination skills, as well as optimization of their development.
A rise in expertise in rowing sports in recent years has resulted in the increasingly detailed studies on rowing technique. Rowing is a sport in which the body moves in a closed cyclical manner. The rowing cycle consists of relatively independent periods with and without resistance. Each period is governed by specific goals for the sculler and the movement of the boat, along with Journal of Human Kinetics -volume 49/2015 http://www.johk.pl spatial and temporal indicators dependent on the individual characteristics of a sculler and the type of boat being used. It is of immense importance for two-oared scullers that the muscles are equally involved in the performed movement, to achieve maximum velocity of the boat and an ideal course in the water. In order to achieve higher training efficiency, and thus better athletic performance, new research and diagnostic methods are constantly being developed, particularly those that are non-invasive. One such method is thermography, which is suited for the quantitative and therefore objective evaluation of changes in the temperature of body surfaces covering active muscles (Akimov and Son'kin, 2011;Lubkowska, 2010, 2012). Non-contact thermography enables diagnosis of the distribution of skin surface temperatures from athletic movement (Fröhlich et al., 2013). Monitoring muscle activity by means of thermovision is possible, as shown by a correlation between performed mechanical work and the metabolic processes that produce heat (Anwajler and Dudek, 2009).
Thermal imaging techniques that measure body surface temperatures may be used to assess the degree of involvement of selected muscle groups during exercise. They can therefore be used as a tool for coaches to evaluate not only the dynamics underlying body surface temperatures from exercise, but also to assess the symmetry of muscle activity. In available literature, however, there are only a few papers that mention the use of thermography for the aforementioned purposes (Awrejcewicz and Zagrodny, 2011;Awrejcewicz 2012;Parkin et al., 2001;Vardasca et al., 2012). Our research indicates the potential usefulness of thermal imaging in the evaluation of the distribution of skin surface temperatures from exercise, especially in athletes whose success depends on symmetrical muscle activity.
The aim of this study was to use a thermal imaging infrared camera to evaluate temperature changes of symmetric body surfaces over the symmetrically working muscles of scullers after exercising on a two-oared rowing ergometer and compare these to the asymmetrically working muscles of handball players after endurance training containing elements of an actual game. Another aim of the study was to show the possibility of using this method to improve monitoring of training progress.
Participants
The study involved two groups of male athletes: scullers and handball players, who represented sport disciplines with a very different nature of physical activity. Training and muscle activity of scullers is symmetric, in contrast to asymmetric training of handball players -set plays are performed with the dominant hand, so functional asymmetry has great significance.
The group of scullers included 18 male athletes from the university rowing club, average age of 20.77 years (SD=1.78), professionally involved in two-oared sculling. The study was performed in the precompetitive period. The average scullers' training experience was 5.44 years (SD=1.77). The number of training sessions varied from 5 to 6 per week. The second study group comprised 16 professional handball players, average age of 22.54 years (SD=2.01), from a local 1st division team. The handball players' average training experience was 8.31 years (SD=2.15). They trained 6 to 8 times per week.
Each athlete provided written consent before participating in the research, according to the Declaration of Helsinki. The study was approved by the Pomeranian Medical University Ethics Committee, (KB-0012/151/12). All participants were healthy and not injured.
Procedures
The study was conducted during a regular training session, after a 30 min warm-up. Indexes of the resting heart rate were monitored by a Polar sport tester.
After recovery of the heart rate to resting values, the scullers were subjected to a maximal exercise on a Concept II two-oared rowing ergometer simulating a distance of 2000 m. This is a basic test used in rowing to reflect the conditions of a rower, and is also used in competitions in hall rowing. The data obtained during these tests provided information on exercise intensity, the heart rate and the pace of rowing. Duration of the exercise test on the Concept II in the group of scullers averaged 6 min (5.32-6.17), with the stroke rate in the range 28-32 (M= 30) strokes per min. During the exercise they © Editorial Committee of Journal of Human Kinetics achieved a heart rate of 190 (185-201) bpm, corresponding to a maximal effort for these rowers.
The handball players were tested during an endurance training session that contained elements of an actual game. Training lasted 90 min, during which the players' heart rates were monitored by means of Polar sport testers, and then individual mean training heart rates (HRt) were determined. The authors also estimated individual percentage ratios of the mean heart rate at training to the maximum heart rate ( , which allowed defining the intensity of effort during the training session, which ranged from 71.96 to 82.35% (M=75.77%; SD= 3.202).
The rooms in which the training sessions took place had similar conditions; the air temperature was 20 o C and air humidity was 55%.
For each participant, two series of thermograms were performed on selected body surfaces in a standing position: the front and rear surfaces of the upper limbs (arm and forearm), the chest, front and rear surfaces of the thighs, and the back. The analyzed areas are presented in Figure 1.
The study used a ThermaCAM SC500 camera (FLIR Systems Inc., Wilsonville, OR, USA). Two measurements were taken, at 25°C and humidity of 60% at a distance of 3 m. The first 20 min before training and the second following training. During the measurements the subjects were dressed in shorts and sport shoes to expose the selected surfaces. The thermography technique allows recording of skin temperatures and speed of temperature change to a resolution of <0.1°C. The following software was used for measurement analysis: Agema Report 5.4.1, Agema Report Viewer 5.4, and Agema Image Viewer 1.02 (AGEMA Infrared Systems, Sweden). The aforementioned series were as follows: 1) in the scullers group: Series 1 -before the training session (adaptation time: 30 min); Series 2 -directly after maximal physical exertion on the Concept II rowing ergometer.
2) in the handball players group: Series 1 -before the training session (adaptation time: 20 min); Series 2 -directly following 90 min of endurance training that contained elements of an actual game.
Duration of exercise and thermal imaging recording were identical for all the participants. The subjects began and finished training one after the other, so the time duration between exercise and testing time was the same for each participant. All participants were right-handed, apart from one handball player who was lefthanded (it was included in the analysis and recorded as the dominant limb).
The selected surface areas were continuously exposed during the exercises. Mean temperature (Tmean) within the selected areas was recorded. Computerized image analysis allowed selection of the measurement area on the thermograms. The tests were performed in accordance with the standards of the European Association of Thermology (Fuijmasa, 1995). Skin emissivity was assumed to be 0.98.
Statistical analysis
Statistical characteristics of the examined variables are presented using arithmetic means and standard deviations. Data were found to be normally distributed (Kolmogorov-Smirnov test). To assess the significance of changes in mean temperatures ΔTmean (before and immediately after the exercise) between the analyzed symmetric body surfaces, we used the Student ttest with statistical significance of p<0.05. Calculations were performed using Statistica 10 software (StatSoft, Inc., Tulsa, OK, USA).
On the basis of the thermograms, changes in Tmean were recorded in the studied areas. Immediately after the exercise mean temperatures were always lower than before in all the subjects. The decreases in temperature in the analyzed body areas between the series of tests are shown in Tables 1 and 2.
In the scullers group (Table 1) there were no statistically significant differences between the mean drops in temperature of the compared symmetric areas over the groups of muscles active during the tests, with differences in the mean temperature drop less than 0.5°C between the opposite sides of the symmetrical areas.
Journal of Human Kinetics -volume 49/2015 http://www.johk.pl Moreover, the Student's t-tests showed that there were no significant differences between temperature drops on the opposite sides in the scullers.
In contrast to this and as expected, in the handball players there were significant differences (Table 2) in the changes of temperature between the opposing symmetric body areas over the muscles involved in training. Differences were statistically significant for the areas of the right and left arm, and the left and right forearm (front and back). The Student's t-tests showed significant differences in temperature drops between these opposite sides.
Discussion
Working muscles and their increased metabolism cause a pronounced increase in core body temperature. The generated heat is received by the flowing blood and is transferred to the surfaces of the body, where it is lost. The most effective way of removing heat is through evaporation of sweat from body surfaces, which is why immediately after maximal exercise the body surface had a lower temperature than before the exercise (Schlader, 2010).
In a better-trained body, adaptive changes result in a smaller internal temperature increase and greater intensity of sweating. Consequently, cooling of the body is more effective and at the same time the surface temperature of the body after physical exertion is lower than in untrained subjects (Aoyagi, 1997;Smorawiński and Grucza, 1994;Torii, 1992). Thus, changes in the surface temperature of the body can be an indicator of the load of the locomotor system, providing information on the efficiency of systems responsible for removing endogenous heat generated during exercise.
This study does not focus on measuring the absolute temperature changes on the skin surface of an athlete participating in different kinds of physical exercise. We did not compare changes in the response to physical exercise, as the intensity of muscle activity was known to be distinctly different. Instead, the aim was to compare such responses between physical activities that involved muscle groups used in a decidedly symmetric or asymmetric manner. Correspondingly, we chose the groups of scullers and handball players due to their specific lateralization. The exercises performed by the scullers were general and required symmetry in the activity of muscles, while in handball players, the involvement and activity of the muscles were distinctly asymmetric.
During exercise, changes in mean temperatures of symmetrical body surfaces directly above the muscles engaged in physical exertion should be similar, as it is where body heat is generated. This excess heat can only be dissipated by increasing blood circulation in the skin, whereby the heat is conveyed more quickly Journal of Human Kinetics -volume 49/2015 http://www.johk.pl to the surface of the skin. At a specific moment, due to increasing loading imposed by the physical activity, the body is no longer able to shed the excess heat. Neither the muscles nor the skin can receive more blood and this is the ground for our assumption that asymmetry in the activity of muscles may result in differences in surface temperatures between the highly and less active groups of muscles; we expected this could be shown by thermal imaging. Our experiment was conducted to identify significant dependence of the measured temperature field on the surface of a selected segment (thigh muscle) with various types of loading. Similar to actual rowing in a boat, while rowing on an ergometer, the cycle of rowing comprised two phases: drive (stroke) and recovery. The drive phase includes a catch, stroke and finish. The recovery phase includes a movement forward and preparation for the next catch. Sculling involves a relatively high number of muscles, requiring intense activity of the upper and lower limbs, trunk, back and abdomen, which is not common in other sports. The symmetry in the activity of individual scullers is important for performance of rowing teams.
In literature, differences between body surfaces temperature up to 0.5°C are deemed negligible (Jones, 1998;Niu, 2001). In the case of any difference greater than 0.5°C, we may speak of asymmetry of temperature distribution, which may be the result of pathological conditions in tissues. In this study, similar to other papers on the changes in the surface temperature in response to exercise or cryogenic temperatures Lubkowska, 2010, 2012;Chudecka et al., 2014), we observed no asymmetry that would exceed 0.5°C. Based on this, we concluded there was no asymmetry in the tested individuals. However, it was expected that local temperature variations would occur due to the differences in the physical activity between the studied groups of scullers and handball players. Accordingly, statistical analysis did show statistically significant differences in the group of handball players, between the left and right arm and forearm (although not exceeding the 0.5°C asymmetry threshold presented in literature). This was caused by the specificity of training in this sport, higher activity of the favored hand, for example during throws at the goal or passes.
In this context, the use of thermal imaging may be very useful for coaches, providing valuable information about the symmetry of muscle activity and thus could be used in training to improve the movement technique and motor coordination. Its advantages also include the non-invasive character, simplicity, and the short time of taking thermal images, as well as the precision of temperature measurements. In sports that use symmetric muscle activity, i.e. in two-oared sculling, it is important to engage both sides of the body in the same way in order to most effectively transform the power of the muscles into the speed of the boat, and at the same time achieve the straightest course. At the highest sporting level, this is one of the most important elements in achieving the best results.
An important aim of diagnostic tests is to identify the indicators that could help define the highest level of preparation, not only in the physical, but also in the technical skills that have a significant effect on the speed of the boat. It seems that symmetry of movement in two-oared sculling is one such an indicator, as it ensures the most economical expenditure of energy. Asymmetric activity of the human body in twooared sculling not only adversely affects the rowing technique, but it also results in lower race times of the boat.
Conclusions
The non-invasive and non-contact method of thermal imaging can be used to assess the activity of muscles of athletes during training, particularly in the context of a comparison of symmetric body areas.
In the group of examined scullers, during a maximal two-oared exercise on a Concept II rowing ergometer, the monitored surface temperature changes of symmetrical areas were similar, indicating that the muscle exertion was similar, especially in the context of results obtained in the group of handball players, with statistically significant differences between the temperatures of symmetrical areas of upper limbs. | 2018-04-03T01:26:14.683Z | 0001-01-01T00:00:00.000 | {
"year": 2015,
"sha1": "997fa07493724983390e2fa287095983bf71f2f7",
"oa_license": "CCBY",
"oa_url": "https://content.sciendo.com/downloadpdf/journals/hukin/49/1/article-p141.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "997fa07493724983390e2fa287095983bf71f2f7",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Medicine"
]
} |
208172000 | pes2o/s2orc | v3-fos-license | Prevalence and antimicrobial susceptibility patterns of Salmonella and Shigella isolates among children aged below five years with diarrhea attending Robe General Hospital and Goba Referral Hospital, South East Ethiopia
Background Diarrheal diseases are responsible for high level of morbidity and mortality, particularly in children below 5 years. Salmonella and Shigella spp. are pathogenic microbes responsible for the major diarrheal associated mortality. The purpose of this study was to determine the prevalence, factors associated with Salmonella and Shigella isolates infections and their antimicrobial susceptibility patterns among diarrheic children aged below 5 years attending BRGH and GRH, Ethiopia. Methods A health institution based cross-sectional study was conducted from April to July 2016. One stool samples was collected from 422 diarrheic children under the ages of five and were cultured on to Hektoen Enteric (HE) and Salmonella-Shigella agar. Isolation identification of the Salmonella and Shigella isolates were conducted using standard bacteriological methods. Antibiotic susceptibility was done by Kirby–Bauer disk diffusion method. The isolates were defined as multidrug resistant if it was resistant to two or more antimicrobial agents. Descriptive statistics were employed and logistic regression models were constructed to determine factors associated with Shigella/Salmonella prevalence. Results The prevalence of Salmonella and Shigella isolates were 6.9 and 4.3%, respectively. Children aged between 1 to 3 years were significantly associated with Salmonella infection [AOR = 19.08, 95% CI (2.68–135.86)]. The odd of prevalence of Salmonella/Shigella isolates was significantly associated with absence of latrine, absence of hand washing after latrine, and in unimmunized children in adjusted odd ratio. Unimproved water sources and hand washing before meal had also higher odd of prevalence although the difference was not significant. All Salmonella and Shigella isolates were resistant to amoxicillin (100%). In addition, all Shigella isolates were completely resistant to chloramphenicol, and tetracycline, and were multidrug resistant. However, all Salmonella and Shigella isolates were susceptible to ciprofloxacin and ceftriaxone. Conclusion There was a relatively low prevalence of Salmonella and Shigella species in the study areas and were significantly associated with lack of personal hygiene and environmental sanitation. There were also higher drug resistance and multidrug resistant pattern. Personal hygiene and environmental sanitation, including access to latrine and supply of safe drinking water are suggested. Checking susceptibilities of Shigella and Salmonella isolates causing diarrhea is also suggested.
(Continued from previous page)
Conclusion: There was a relatively low prevalence of Salmonella and Shigella species in the study areas and were significantly associated with lack of personal hygiene and environmental sanitation. There were also higher drug resistance and multidrug resistant pattern. Personal hygiene and environmental sanitation, including access to latrine and supply of safe drinking water are suggested. Checking susceptibilities of Shigella and Salmonella isolates causing diarrhea is also suggested.
Keywords: Children, Drug susceptibility test, Prevalence, Associated factors, Salmonella, Shigella, Ethiopia Background Diarrhea is a major cause of morbidity and mortality among children < 5 years old in sub-Saharan Africa [1]. It is the second cause of death (16%) after pneumonia in children under 5 years of age worldwide, with developing countries being the most affected [2]. The most common causes of infectious diarrhea include: viruses, bacteria, protozoa and unidentified and mixed infections [3]. Among bacterial pathogens, Salmonella and Shigella are of particular concern as causes of enteric fevers, food poisoning and gastroenteritis [3,4].
Salmonella, with its more than 2500 different serotypes, is a leading cause of food-borne infections worldwide. Salmonella spp. causes self-limiting gastroenteritis and the more severe forms of systemic typhoid fever. Typhoid fever is principally caused by Salmonella enteric serotype Typhi in humans. Salmonella outbreaks are related to unhygienic food preparation, cooking, reheating and storage practices that are contaminated with the pathogen [3,4]. In addition, poor access to good latrine, poor sanitation and hygienic status, hand washing habit before and after meal and / or latrine, absence of proper sewage disposal system were responsible for typhodial type of salmonella infections [5][6][7][8][9]. Shigellosis is caused by Shigella spp. and it is a worldwide problem although more prevalent in developing countries [10][11][12]. Shigellosis is common in areas where living standards are very low and access to safe and adequate drinking water and proper waste disposal systems are often very limited, or even absent [6][7][8][9][13][14][15][16][17]. Shigella spp. are limited to the intestinal tract of humans and cause bacillary dysentery leading to watery or bloody diarrhea.
Studies conducted in Ethiopia also revealed an increasing trend in the prevalence of Salmonella isolates and Shigella spp. [5][6][7][8][9]14]. Antimicrobial resistance of Salmonella and Shigella are emerging global challenges, especially in developing countries where there is an increased misuse of antimicrobial agents in humans and animals [10,11]. In Ethiopia in the past three decades, studies indicated that Salmonella and Shigella have developed varied rate of resistance against the first line antibiotics such as ampicillin, tetracycline, co-trimoxazole, chloramphenicol [5-9, 13, 14, 16-22], second generation fluoroquinolones such as norfloxacin and/or ciprofloxacin [7,9,13,16,19,20,22], the third generation cephalosporins (ceftriaxone) [7-9, 18, 20]. However, there is no published information on the magnitude of Salmonella and Shigella infection and their response to commonly prescribed drugs among diarrheic children aged below 5 years in Bale Robe General Hospital (BRGH) and Goba Referral Hospital (GRH), South Eastern Ethiopia. Understanding the magnitude and drug susceptibility pattern of Salmonella and Shigella are important in designing public health intervention measure in the area. The purpose of this study was to therefore determine the prevalence, associated factors and antimicrobial susceptibility patterns of Salmonella and Shigella isolates among children aged below 5 years with diarrhea in RGH and GRH, South Eastern Ethiopia from April 2016 up to July 2016.
Study area, design and period
Institution based cross-sectional study was conducted at BRGH and GRH from April 2016 up to July 2016 in order to determine the prevalence and antimicrobial susceptibility patterns of Salmonella and Shigella species in diarrheic children under the ages of five. BRGH and GRH are located are 430 Km and 445 Km, respectively away from Addis Ababa in South Eastern Ethiopia. The two hospitals provide health services for two major towns of the zones and to the surrounding districts.
Study population
The source population was all children under the ages of five and who lived either in Robe town, Goba town or in the adjacent rural villages during the study period. The study population was all diarrheic children under the ages of five in the selected hospitals who obtained medical service during the study period. Study population who had taken antibiotics 2 weeks before the study and with incomplete demographic information were excluded from the study.
Sample size determination and sampling techniques
The sample size (n) was determined though a single proportion formula by taking an estimated prevalence of 0.5 (for unknown prevalence) [23]. In addition, 10% of the sample size was added to the normal sample size to minimize errors [24]. A total of 422 study population was included in the study both for socio-demographic survey and stool sample collection. Simple random sampling methods were used to sample study population. Accordingly, 283 and 139 study participants were enrolled from Goba Referral Hospital and Robe General Hospital in the study, respectively.
Data and specimen collection
A structured questionnaire was used to collect data on socio-demographic characteristics (age, sex and residence) of the study population (children), and associated factors (status of drinking water, availability of latrine, hand washing habit before meal and after toilet, immunization status, availability of waste disposal system, presence of domestic animal and milk treatment status). Such data were retrieved from children parents or guardians by experienced physician or nurse (Table 1; Additional file 1). These factors were selected based on literature survey and considering personal hygiene habits and environmental sanitation of the people in the study areas. Children parents or guardians were also informed to bring freshly passed stool and rectal swab of the study subjects in a sterile stool cup using clean applicator stick. Accordingly, one stool sample was collected from 422 children under the age of five. The collected stool was placed immediately in Cary Blair transport (Oxoid, UK) and transported to laboratory for immediate processing. Instructions were given to children parents or guardians on how to collect stool samples from study children.
Culture and identification
Approximately, 1 g of stool specimen was suspended in Selenite-Cystine broth (Oxoid, UK) contained in sterile test tube for overnight. A loopful of the suspension of the specimen was streaked on to two different media namely, Hektoen Enteric (HE) Agar and Salmonella-Shigella (SS) agar, both from Oxoid, UK. These plates were incubated aerobically at 37°C overnight to enhance the recovery of the two pathogens. The isolates were purified by subculturing on nutrient agar plates. A colorless colony with or without black center on SS agar media, and a blue green colony with or without black center on HE agar were presumably isolated as Salmonella-like isolates. A colorless colony on SS and a green, moist and raised colony on HE agar were presumably isolated as Shigella-like isolates [3,25,26]. Colonies exhibiting characteristic reactions of Salmonella and Shigella-like were further characterized by the pattern of biochemical reactions after inoculation on to Triple sugar iron agar, lysine iron agar, Simon's citrate agar, and MIU test (motility test, Indole and Urease production) for final identification using the standard procedures [3,25,26].
Antibiotic susceptibility pattern
The antimicrobial susceptibility testing of Salmonella and Shigella isolates was done by Kirby-Bauer disc diffusion method according to Clinical and Laboratory Standards Institute [27] guidelines. Colonies from pure cultures of Salmonella or Shigella were taken and transferred to a tube containing 5 ml sterile distilled water. It is mixed gently until a homogenous suspension was formed. The suspension was incubated at 37°C until the turbidity became adjusted to a 0.5 McFarland standard and then was uniformly inoculated on to Muller Hinton Agar (Oxoid, UK) by using sterile cotton swab under laminar hood. The inoculated plates were left at room temperature to dry for 3-5 min A sterilized forceps was then used to lightly press the antibiotic discs manually on the surface of a Muller-Hinton plate to make firm attachment. Accordingly, each isolate was subjected to 11 antibiotics discs on Muller Hinton agar (Oxoid, UK): amoxicillin (AML) (2 μg), ceftriaxone (CTR) (30 μg), ciprofloxacin (CIP) (5 μg), chloramphenicol (C) (30 μg), doxycycline (DO) (30 μg), and tetracycline (TE) (10 μg) (all from Oxoid, UK). The plates were then incubated at 37°C for 24 h. The control disc was impregnated with sterile distilled water. Diameters of the zone of inhibition was measured to the nearest millimeter using a metallic caliper, and the isolates were classified as sensitive, intermediate and resistant according to the standardized table supplied by the CLSI [27]. Salmonella/Shigella isolate was defined as multidrug resistant if it was resistant to two or more antimicrobial agents tested [9,18,28].
Data quality control
In order to generate quality and reliable data, all questions were prepared in a clear and precise way and translated into two local languages (Afan Oromo and Amharic). Completeness of the questionnaire was checked whether the necessary information was properly full filled or not. All the instruments used for sample processing were checked for sterility and proper functioning. For quality control, strains of E. coli ATCC 25922, Shigella flexneri ATCC 12021, and Salmonella Typhimurium ATCC 14028 were obtained from Ethiopian Public Health Institute. The sterility of prepared media was checked by incubating one of the prepared media for 24 h at 37°C. Growth media that showed bacterial growth were discarded.
Data analysis
The data was analyzed using IBM SPSS Statistics for Windows version 20 (IBM Corp., Armonk, NY, USA). Prevalence/ isolation rate (the outcome or dependent variable) was expressed as percentage of stool samples of study subjects showing cultured-confirmed Salmonella or Shigella isolates divided by the total number of screened stool samples of study subjects. Sociodemographic factors and associated factors were independent factors from which predictors factors for the Abbreviations: a refers to unimproved water (drank raw without any chemical treatment) or improved (water drank after chemical treatment) outcome variables were identified. Descriptive statistics was used to analyze the rate of isolation of Salmonella or Shigella isolates. Bivariate and multivariable logistic regression analyses were used to compute crude ratio (COR) and adjusted odds ratio (AOR). The independent variables were checked for the presence of little or no multicollinearity using Variance Inflation Factor (VIF) (VIF = 1, not correlated; between 1.0-5.0, moderately correlated and > 5.0 highly correlated). The goodness of fit of the employed model was evaluated using the Hosmer-Lemeshow test. A p-value > 0.05 indicated that the model fit well to run the logistic regression analysis. The logistic regression analysis was done by step-wise manner. First, each independent variable and outcome variable was first evaluated by bivariate logistic regression to calculate Crude odd ratio (COR). The COR of independent variables whose 95% confidence interval (CI) excluding one were fitted to multivariable logistic regression to calculate the adjusted odd ratio (AOR). In AOR, the 95% confidence interval (CI) of odds ratio (OR) excluding one was significantly associated with the corresponding dependent variable.
Ethical consideration
This study was approved by College of Natural and Computational Science of Madda Walabu University. Ethical clearance was obtained from Bale Zone Health Office. Permission to conduct this study was obtained from each hospital administration. Both oral assent and written informed consent were obtained from the parents or guardians of the study subjects before administration of the study. Information of the study subjects and that of the parents or guardians of the study subjects were kept confidential.
Socio-demographic data
A total of 422 stool samples were collected from study participants in BRGH (n = 139) and GRH (n = 283) and the response rate was 100%. Among study subjects, 179 (42.4%) were male and 243 (57.6%) were female. Out of the total study subjects, 176 (41.7%) and 246 (58.3%) lived in urban and rural areas, respectively. The age of the study participants ranged from 3 months to 60 months (mean age 31 months ±4.7 months). Among them, 129 (30.6%) were infants, 86 (20.4%) were between 1 to 3 years and 207 (49%) were between 3 to 5 years. Most of the participants utilized an improved water (81.28%). Availability of latrine, handwashing before meal and after toilet, and child immunization were recorded from 59.24, 90.3, 81.04 and 73.7% of the respondents, respectively ( Table 1).
Prevalence of Salmonella and Shigella isolates
The isolation rate of Salmonella and Shigella isolates are shown in Table 1. Salmonella and Shigella isolates were positive in 29 (6.9%) and in 18 (4.3%) study subjects, respectively. 22 (7.8%) of the Salmonella isolates were from GRH, and 7 (5.0%) were from BRGH. Of the 18 Shigella isolates, 15 (5.3%) were from GRH and 3 (2.2%) were from RGH. A higher isolation rate of Salmonella and Shigella isolates were recorded from study subjects in rural areas (9.3 and 5.3%) than urban areas (3.4 and 2.3%) ( Table 1).
Associated factors for Salmonella and Shigella species infections
Associated factors for the prevalence of Salmonella and Shigella infection were shown in Tables 2 and 3, respectively. Children whose ages were between 1 to 3 years were 19 Table 2). The odd of prevalence of Shigella infection were significantly higher in the absence of latrine [AOR = 4.58; 95% CI (1.1-18.9)], in children whose parents didn't wash their hand after latrine [AOR = 3.31; 95% CI (1.02-10.90)]. Children who drank from unimproved water sources and that was served by parents who didn't wash their hands before meal were also more likely exposed to Shigella infection although the difference was not significantly varied from their counterparts. Unimmunized children had about 3.63 times higher infection risk than those who were immunized [AOR = 3.63; 95% CI (1.04-12.65)] (Table 3).
The results of multiple drug resistant (MDR) patterns of Salmonella and Shigella are presented in Table 5. Most of the Salmonella isolates (n = 23, 79.31%) and Shigella isolates (n = 18,100%) were found to be multiple drug resistant (resistant to two and above antimicrobial drugs). 14 (48.3%) and 9 (31.03%) of Salmonella isolates displayed MDR against two and three drugs, respectively. On the other hand, 2 (11.1%) and 16 (88.9%) Shigella isolates were resistant to three and four drugs, respectively (Table 5).
Discussion
This study was the first report on the prevalence and antimicrobial susceptibility pattern of Salmonella and Shigella isolates in RGH and GRH. The isolation rate of Salmonella isolates (6.9%) in our study (Table 1) was comparable with the study in Jimma, South west Ethiopia (6.2%) [6], but higher than Hawassa town, South Ethiopia (2.5%) [7], Kenya (3.5%) [29] and Botswana (3.0%) [30], and was lower than Bahir Dar, North Ethiopia (7.8%) [31] and Arba Minch, South Ethiopia (12.6%) [9]. Shigella isolates had the isolation rate of 4.3% (Table 1) in conformity with the results from Arba Minch (4.8%) [9] and Nepal (4.6%) [32], but it was lower than that results from Bahir Dar (9.5%) [31], South Ethiopia (8.3%) [33], and Botswana, Gaborone (21%) [30]. This might be due to variations in personal hygiene and environmental sanitation, and food handling practices of the community and due to climatic differences which affect the viability of infective pathogens. In our study, children within the age range of 1-3 years were more susceptible to diarrhea caused by Salmonella Abbreviations: a refers to unimproved water (drank raw without any chemical treatment) or improved (water drank after chemical treatment); AOR Adjusted Odd Ratio, COR Crude odd ratio. Factors of independent variables with bolded COR/AOR were significantly associated the dependent variable (p-value ≤0.05) (12.8%) and Shigella (7%) isolates. This finding was in agreement with several studies [6,9,33,34]. Children at this age are naturally taking contaminated soils, food and water into their mouth and may acquire disease causing microbes including pathogenic Salmonella and Shigella spp. from the environment easily. Children from rural areas were more susceptible to diarrhea due to Salmonella and Shigella infection than children from urban areas, a result in conformity with study of Vargas et al. [35]. The reason might be because less awareness on personal hygiene and environmental sanitation and a reduced provision of health facilitates in rural areas compared to urban areas. Our study revealed that study subjects who used water from unimproved sources, and had no access to latrine had higher odd of prevalence of Salmonella and Shigella isolates than their counterparts although the difference was significant for the former factor (Tables 2 and 3). It was indicated that the consumption of contaminated food and/ or water is responsible for diarrheal diseases caused by Salmonella and Shigella isolates [3,4]. Children whose parents did not wash their hands before serving food to them and after toilet had higher odd of prevalence of Salmonella and Shigella spp. in concordance with several studies [8,9,[15][16][17]. This was a result of fecal-oral transmission of these bacteria from the parents to their children during feeding and/or handling [3,4]. Children whose parents didn't possess waste disposal facility and reared domestic animals had higher odd of prevalence of Salmonella and Shigella isolates although the difference was not significant as compared to their counter parts (Tables 2 and 3). This was in concordance with studies in Arba Minch [9], in Thailand [15] and in Guinea-Bissau [36]. It is revealed that unhygienic living circumstances and close relations between humans and animals may substantially contribute to the occurrence of salmonellosis and shigellosis [11,37].
The antimicrobial resistance of enteric pathogens such as Salmonella and Shigella have been increased all over the world as a result of indiscriminate use of antimicrobial Fig. 1 Antibiotic susceptibility pattern of Shigella isolates against 11 antimicrobial drugs using agar well diffusion method on three Petri Plates. Control discs are indicated as Control at the centre of the plate. They were impregnated with sterile distilled water Table 4 Antimicrobial susceptibility pattern of Salmonella and Shigella isolates agents [18,32,34,35,38]. High level of resistance of Salmonella isolates to amoxicillin (100%) revealed in this study was comparable with studies in different areas of Ethiopia (100%) [6,9,21]. The resistance of Salmonella isolates against chloramphenicol (24.1%) was comparable with study in Addis Ababa (21.7%) [8], but lower than earlier studies in Harar (62.3%) [21] and in Addis Ababa (83.7%) [18]. The current use of this drug is limited in RGH and GRH. The rate of resistance of Salmonella isolates to tetracycline in our study (79.3%) was higher than in Harar (71.4%) [21], in Hawassa town (0%) [7] and in Mozambique 15% [28]. It was revealed that antibiotic resistance of Salmonella isolates from diarrheic children in Southeastern Africa was conferred by tem-like βlactamases for ampicillin, floR genes and CAT activity for chloramphenicol, tetA genes for tetracycline, and dfrA1 genes for trimethoprim-sulfamethoxazole/ co-trimoxazole [28]. The susceptibility of Salmonella isolates in our study to ciprofloxacin and ceftriaxone (100%) was higher and consistent with some studies in Ethiopia and elsewhere [6,8,28], However, ciprofloxacin resistant (6.67%) Salmonella isolates was reported from Nekemte Hospital [20]. It was reported earlier that ciprofloxacin was effective and well tolerated for treatment of typhoid fever in children [39,40]. In addition ceftraxione resistant isolates were reported from Hawassa town (75%) [7] and Nakemete Hospital (3.33%) [20]. Ceftriaxone is safe to use, including in children, is slowly bactericidal against Salmonella serovar Typhi in vitro, and is able to penetrate and kill intracellular bacteria [41].
Our study showed that Shigella isolates were 100% resistant to three antibiotics (amoxicillin, chloramphenicol, and tetracycline). The resistance to amoxicillin in our study (100%) was very high and in conformity with previous studies in Ethiopia [6,7]. The resistance of Shigella isolates to tetracycline in our study (100%) was also higher than earlier studies in Addis Ababa (97.3%) [18], in Bahir Dar (93.8%) [19], in Gonder (89.7%) [17] and in Mozambique 66% [28]. Similarly, the resistance of Shigella isolates to chloramphenicol (100%) was higher than studies in Arba Minch (62.5%) [9], and in Gonder (67.8%) [14]. The resistance of Shigella isolates to many drugs is due to the widespread use of antibiotics in medicine, veterinary medicine, and agriculture [3,42].
Shigella isolates produce R plasmids that code for several resistance genes and can confer multiple antibiotic resistance [3]. Similarly, analyses of antibiotics resistance genes of Shigella isolates causing diarrhea in children under age of 5 years in South eastern Africa revealed the presence of oxa-1-like β-lactamases for ampicillin, dfrA1 genes for trimethoprim-sulfamethoxazole/co-trimoxazole, tetB genes for tetracycline and Chl acetyltransferase (CAT) activity for chloramphenicol [28]. All Shigella isolates in our study exhibited 100% susceptibility to ciprofloxacin and ceftriaxone, a result consistent with studies in Ethiopia [6,19,20] and in Mozambique [28]. A relatively higher rate of resistance of Shigella isolates to ciprofloxacin than our study was reported from Mekelle Hospital (6.7%) [16] and Gonder Hospital (9.2%) [22], and to ceftraxione was reported from Hawassa town (54.5%) [7] and Addis Ababa (4.3%) [8].
The treatment of diarrheal diseases caused by Salmonella and Shigellla species should consider the age restrictions associated with usage of certain antimicrobials in children. Empiric therapy may be started with oral cotrimoxazole or metronidazole, but in severe cases parenteral treatment with ceftriaxone or fluoroquinolones (ciprofloxacin) might be considered [39][40][41]44]. But, the use of fluoroquinolones and tetracyclines in the empirical treatment of diarrhea in small children is constrained by its adverse effects on musculoskeletal events and its adverse effect on the teeth younger than 8 years of age, respectively [45]. Third generation cephalosporins such as ceftraxione is the drug of choice for treatment of severe acute diarrheal caused by pathogens such as Salmonella and Shigella spp. and they have fewer adverse effects on children and effective against fluoroquinolone resistant strains of these bacterial species [45].
Conclusions
The overall prevalence of Salmonella and Shigella isolates was 6.9 and 4.3%, respectively. Children in the ages of 1 to 3 years had relatively higher isolation rate of both Salmonella and Shigella isolates compared to other ages. Drinking from an unimproved water sources, absence of latrine, absence of hand washing habit before meal and after toilet by parents, and unimmunized children resulted in higher odd of prevalence of Salmonella and Shigella infection in children. The Salmonella and Shigella isolates displayed high rate of resistance to commonly used drugs such as ampicillin,tetracycline, chloramphenicol and doxycycline. This indicates that these drugs have a reduced efficacy in the treatment of diarrhea caused by Salmonella and Shigella isolates. But, all Salmonella and Shigella isolates were highly susceptible to ciprofloxacin and ceftriaxone drugs of choice recommended for diarrheal diseases caused by these pathogens in the two hospitals. Multiple antimicrobial resistances were high among Salmonella and Shigella isolates. The limitations of this study were failure of identification of Serogroup of Salmonella and Shigella species as a results of financial constraints and lack of laboratory facilities in the study areas. Periodic awareness on personal hygiene and environmental sanitation, frequent microbiological analyses of food and water are suggested to reduce diarrhea caused by Salmonella and Shigella spp. The choice of the drugs for the treatment of diarrhea caused by Salmonella/Shigella isolates should be supported by an in vitro susceptibility studies of individual drugs.
Additional file 1. Consent form and socio-demographic data and clinical morbidities of children aged below five years with diarrhea attending Robe General Hospital and Goba Referral Hospital, South East Ethiopia. | 2019-11-20T16:42:29.155Z | 2019-11-20T00:00:00.000 | {
"year": 2019,
"sha1": "5c15be467abd5df1921fab7e7b03708880771430",
"oa_license": "CCBY",
"oa_url": "https://tdtmvjournal.biomedcentral.com/track/pdf/10.1186/s40794-019-0096-6.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "5c15be467abd5df1921fab7e7b03708880771430",
"s2fieldsofstudy": [
"Medicine",
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
54002740 | pes2o/s2orc | v3-fos-license | Credit Expansion & Credit Limits
This article discusses the problem of credit expansion and credit limits. A distinction of the concepts of credit expansion and credit limits. The author proposes a classification of credit limits. The article discusses the features of credit booms. An importance of managing credit limits as necessary condition for the sustainable development of the credit sphere is highlighted.
Introduction
In today's economy, credit is again at the centre of events.The financial and economic crisis that began in the West in the credit sector, showed the need for more in-depth studies of credit and its impact on the economic environment.The companies are even more interested not only in overcoming the economic downturn, but also in finding a way to avoid previous mistakes and errors in the monetary regulation.The question of the new economic structures, measures to ensure stable and sustainable development, including credit instruments, have become the most popular in both academic and practical terms.
In theoretical terms credit expansion, known to the world for decades, is not ordinary, long explainable phenomenon, but a process, showing new sides of its development.Many things that seemed quite clear and understandable, has caused serious doubts, require new explanations.
For modern domestic financial and banking science fields this problem was particularly acute.It so happened that credit expansion was largely associated with monopolistic tendencies readily perceived as a phenomenon peculiar to developed economies, with a lot of inherent capital, seeking a more profitable use.It turned out, however, that credit expansion happens in both developed and developing countries, the consequences can be very diverse.This credit expansion brings not only troubles, causes crises of production and circulation, but also able to provide society with opportunities for economic development.
The question, therefore, is not how to deal with credit expansion, but how to guide the development of the monetary sector, creating imbalances and sharpen the contradictions.
Unfortunately, conventional economic theory can't answer these questions.It is possible that this is due to the fact that in general credit, as economic relations, has not been given due attention, explanations of cyclical development of the economy were more likely to be associated with money without credit as a special phenomenon, providing both the production and exchange with monetary circulation in specific effect.It seems that the actual content of the credit expansion due to the incompleteness of the research on credit was not disclosed in full.
Credit Expansion: Essence, Nature, Form and Types
In modern science and practice, the concept of expansion is widespread.According to its terminological interpretation, expansion comes from the Latin expansio means extension, distribution.From an economic point of view expansion is the expansion of economic impact, economic activities of a country, group, organization, political associations, groups, states, through displacement of other countries, firms, capital markets, acquisition of resources (for example, the export of capital).
In the literature on monetary relations credit expansion is treated as a kind of economic expansion, its content is also associated with the extension.The origins of this approach can be found even in works of the most eminent representative of the Austrian school of Economics, Ludwig von Mises, who believed that credit expansion and the extension of credit are synonymous.(von Mises, 1967) The following definition of credit expansion as the intensive expansion of credit transactions and operations of banks was adopted in modern literature.
Credit expansion as an economic process is not random, it acts as an expression of patterns of economic development, reflect the growth of material wealth and real savings.Spawn the interests of lenders and borrowers, credit expansion characterizes the development of the monetary system.
As part of the credit cycle, credit expansion occurs, exists and develops in the phase of economic recovery, manifests itself as an extension of credit to different subjects, the emergence of new fields of application of credit, as intensive and long-term expansion of credit supply.On the development horizon, credit expansion acts as a strategic process.
Of course credit expansion is not only a quantitative process of scaling up lending, but also, according to the characteristics of the International Monetary Fund, but also changes in money supply in circulation (Fjodorov, New English-Russian Dictionary of Economics & Banking, SPB, 2006, p. 221), which are the result of an increase in the credit supply through opening credit accounts in the chain of banks.
Credit expansion from an institutional point of view is not only a phenomenon affecting macro-economic relations, but also a process involving the activities of individual banks.We believe that credit expansion is reflected not only in the activities of the creditor, but the borrower who needs more capital, and in this case expresses the patterns of circulation and turnover at the macro and micro level of economic relations.
In modern Russian practice credit expansion visibly manifests itself in the expansion of the subjects and objects of lending terms and provision of credit resources and the payment for a loan, the benefits of lending, export of capital, export credit, geography of credit products.
With the development of banking practice, credit gets new inventory and production costs.Gradually, banks in both international and national practices were transferred from credit to the private objects to total objects, from lending for inventories to borrowers' costs, where one could guarantee repayment of the loan only in the future.It is known that in developed countries, some banks and specialized companies issue venture loans.
It is known that in the modern practice of collateralizing loans not only guarantees of wealth and property are used, but also a variety of types of securities and financial instruments (stocks, bonds, bills, certificates, insurance policies and others).In some cases, the collateral of a loan repayment are not only material object and various financial obligations, but also the personal assets of the borrower.
Unfortunately, in modern practice of collateral usage a range of credit derivatives takes a very special place, by volume substantially exceeding the volume of gross domestic product.Bubble of collateral and associated bubbles, as it is known, became a source of speculation, overheating and as a consequence the reason of the modern financial-economic crisis.
Credit expansion may be fueled by the expansion of the banks' resource base.It is known that internal sources of expanding the resource base of Russian commercial banks before the financial and economic crisis turned out to be significantly less than those of borrowing needs, which demonstrated enterprises and organizations.Using external sources (loans from foreign banks and companies), banks-residents were able to significantly increase their credit investments; credit expansion through external borrowing has received additional breathing.
The expansion of lending to economic entities can be done through interest rate policy of banks, stimulating demand for credit (declining payments for the loan) and the inflow of deposits for subsequent redistribution on a credit basis.It is known that the low interest rate on loans to purchase real estate in the United States led to the widespread development of mortgage loans, however, as it turned out, without proper assessment of the creditworthiness of borrowers, and later to significant losses of credit institutions.
Credit expansion especially during its boom, the critical point, may be accompanied by the liberalization of lending.Low line of borrowers cut-off, a more liberal approach to assessing the creditworthiness of the borrower causes an increase in lending, and in case of adverse shocks lead to a collapse of both the lender and the borrower.According to the testimony of A. Swiston (Swiston, 2008) adherence to lending standards has an impact on the business cycle.The tightening of lending by 20% reduces business activity by 0.75% in one year and 1.25% after two years.To similar findings come and other authors, marking the fall of GDP along with total contraction in credit.(Bayouni & Melander, 2008) In external economic relations credit expansion is manifested in the form of export of capital and the expansion of export credits.Fueled by an excess of capital within the country credit expansion overcomes national boundaries, allows lenders to conquer new frontiers and markets, increases the supply of credit in the new geographical boundaries.
For developing countries, credit expansion can occur within national borders, be distributed on the territory to a lesser extent covered by banking products and services, but feeling the need for the development of production and circulation through credit and payment services.Credit expansion in a similar situation allows one to explore new territory, to promote the development of individual industries and the regional economy, and socio-economic development of the country as a whole.
In all these cases, credit expansion manifests itself not as a mere expansion of credit, and as the penetration of credit in new spheres of human activity, intensive, large-scale development of credit, accompanied by an increase in the supply of credit, the expansion of what is being loaned.
The expansion of credit is one of the moments of credit expansion, however, it may occur out of credit expansion.In this sense, the expansion of credit is not yet a credit expansion.It is known, for example, that in times of economic depression, in the post-crisis economy there exist some recovery of loans, the repayment to the pre-crisis positions.We can say that following the economic recovery credit expansion reflects the recovery of the credit relations after the crisis.Full credit expansion begins to manifest itself after the depression, on the stage of recovery, during the accumulation of loan capital, seeking monetary support significantly expanding business activity of economic entities.
The source in the characteristic of credit expansion is the sources of its implementation.There are two types: expansion, developing based on real savings, and the so-called fiduciary expansion.In the first case, we are talking about a natural process occuring when free resources, reflecting really created and accumulated savings are accumulated in banks and redistributed on a returnable basis.In the second case, the credit expansion is carried out through the issuance of debt instruments, reflecting only future income, and is not an expression of real savings.As practice shows, such credit expansion is widely used for speculative purposes, generates bubbles, and requires perfect control mechanism, attention from the state and its respective regulatory frameworks.
The second type of credit expansion associated with the instruments of its implementation.It either occurs through lending or based on the loan (bond).As one knows, the European model of investment prefers banking (credit) form; it occurs based on the expansion in the supply of credit in the form of bank loans.This does not exclude borrowings of credit resources in the form of issuers of debt obligations (bonds).As practice shows, the second direction is less expensive (for the use of credit, as is the case for loans, the borrower does not need to pay the lending rate).Bank lending, in contrast to the bond loan, is in its consequences less risky form as the loan is accompanied by bank analysis of the creditworthiness of the borrower, which is impossible in the situation of the bonds.
Credit expansion can impact both the economy and individual economic actors.The expansion of credit at the macro level may not coincide with the trend of development of lending to individual borrowers.It may well be that the expansion of lending to the economy, the volume of loans granted to companies of different industries can be reduced.This situation is clearly observed in Russia, for example, in the period of reform processes, when the overall credit growth in the economy, the scale of lending to enterprises of light industry was significantly reduced.
Possible regulation of credit expansion can be both of managed and unmanaged nature.It is no secret that the credit for the modern economy is spontaneous.The state, in accordance with the selected doctrine does not regulate lending activities of each individual bank; each bank, based on the amount of accumulated resources and in accordance with their interests and their credit policy, provides decentralized redistribution of capital.This does not mean, however, that the role of the regulator of credit relations can be played only by the market.The government's goal is not only to show what is happening in the credit sector, but in a skillful and timely prevention and neutralization of the resulting imbalances.Unfortunately, this regulation is not always timely, but it is.It is known that during the current financial economic crisis in several countries, including Russia, the state, regulating the general economic environment, gave significant credit support to banks.
We believe that regulation of the credit must be carried out not only in critical situations, it is reasonable to implement in the context of structural, industrial policy, in the form of macroeconomic prudential requirements and regulations.Credit expansion, in our view, is not so much an uncertain phenomena in the economy, rather how to be a managed process, aimed at prevention of overheating of the economy, and promote its sustainable development.
Credit expansion in the pace of its expansion can be moderate and accelerated.Moderate credit expansion follows the economic growth by acceleration in lending ahead of this growth, causing different effects.
The effects of credit expansion may be destructive or positive.Predominant, unfortunately, is the idea that credit expansion has a negative effect on economic development.As it was shown, credit expansion as it turns out, is largely constructive, its negative impact only occurs if credit boundaries are violated.
Credit expansion can lead to very productive impact on the economy.It is known that in the middle of the XIX century in the world there was a wide development of railway construction of the railway network.Only in one decade in Europe the number of railways has doubled, in America-3.5 times.In 50-ies of the XIX century, there was also intense volumes' increase in coal, steel, cotton and other industries production.All this was accompanied by a significant increase in credit issuance.The banks issued short-term loans, and participated in financing industry.
Industrial growth in Germany in the early 70-ies of the XIX century was also accompanied by credit expansion, accelerated economic growth of the country.Unfortunately, credit expansion is often combined with speculation and subsequent fall in the price of securities that affect the loan and its repayment.Economic history and modern practice suggests that providing credit under inflated securities (promissory notes, shares and others), banks expanded credit opportunities to borrowers, adding in the form of imaginary deposits a lot of money, reflecting not available monetary resources, and divorced from the real value of money, which in turn brought hugh upward pressure on their prices.Sudden credit stress in the economy inevitably led to the loss of value of loan commitments and bank loan defaults rose, bank shares and other securities depreciated.
It is hardly possible to call credit expansion what is called a credit boom.The credit boom is, of course, the expansion of credit, however, due to its characteristics, target orientation and economic consequence it is significantly different from credit expansion.
Credit expansion is, in a sense, a strategic maneuver of lender, affecting his long-term plans.The credit boom is more of a tactical measure, the operational step in the development of credit services.
The credit expansion is of the broader goal.As noted, credit expansion is aimed at expanding the sphere of influence, gaining new opportunities, markets, gain new positions, including in a competitive environment.The credit boom pusrues more local goals-not to lag behind the other, taking advantage of the situation to get more profit.
An important difference is that credit expansion, as noted, expresses the process of constant development within the economic cycle.Credit boom as the expansion of credit is the only episode in credit activity, often associated with mass, high output by banks on the credit market, reflects herding pattern, reminiscent of the pre-sale, sale, the excitement of buyers willing to purchase the product at a cheaper price.
In a sense, the credit boom resembles a kind of coming fashion.Usually, it is associated with the extension to a certain direction of development of the banking activity, the expansion of lending to certain types of borrowers or types of credit, when the former traditional spheres of its functioning remain largely unchanged.Credit booms often act as the culmination (critical point) in rise of one of the lines of credit expansion.(Burakov, 2014a) The emergence of credit booms is usually associated with surplus capital seeking profitable manifestation, maintenance of speculative tendencies (for example, transactions in the real estate market, consumer lending).
Credit boom as a local splash is often characterized by excessive loan growth, extensive saturation of credit, which subsequently leads to a different drop in its volume, causing a decline in the financial stability of both borrowers and lenders.Studies of American authors, E. Mendoza, and Marco E. Terrones bring forth the evidence of the fact that the average duration of the credit boom of course is not measured by several days, and is 6-7 years.This fact allowed them to determine the credit boom as the episode in which the growth of credit to the private sector exceeds the usual extension of credit during the economic cycle.(Mendoza & Terrones, 2008) According to other authors, the average duration of one credit boom, as a rule, is 3.5 years, sometimes from two to five years.(World Economic Outlook, 2004) In any case, it is obvious that credit expansion expresses the continuous process, the long-term trend in the extension of credit, while credit boom-only episodic process that spans a relatively short historical period in credit activities.
Comparing credit expansion and credit boom it is impossible not to draw attention to the fact that credit expansion manifests itself in both domestic and foreign markets, while the credit boom is associated mainly with the internal movement of credit capital.
Credit expansion is hardly can be identified by normal acceleration of credit.The acceleration of credit growth (growth rate) during the period of credit expansion may occur repeatedly, this acceleration is not always can be described as the credit boom.Effects of credit booms can be very diverse.
Liberalization in the process of lending according to the experts of the International Monetary Fund is often associated with the rapid growth in housing prices.In the pursuit of clients and volume of loans banks often did not bother themselves with credit analysis, did not show proper requirements for loan documentation, which subsequently meant defaulted loans.The lending standards declined and in those cases, in those regions, where banks carried out the sale of debts to third parties (securitization).(IMF, 2008) Financial liberalization is the second most important factor accompanying a credit boom.It accounts for 1/6 of the cases, the formation of credit booms.
Credit expansion is largely connected with the complaince of the boundaries of credit.
Borders of Credit
The border of credit issuance and credit expansion is probably one of the most difficult.From a practical point of view, the accurate answer is not there yet.The society has no satisfactory recommendations about the scope of credit as at the macroeconomic and microeconomic level of credit relations.
The unresolved question about credit limits is also of quite significance from a theoretical point of view.
From the perspective of theoretical analysis it is somehow linked to the original and often differing scientific views about the nature of credit, its functions, purposes, causes.Knowledge of the boundaries of credit represents a synthesis of the most important categorical structures; without their prior understanding in the context of use of the credit our knowledge about the subject can be superficial, inadequate.(Burakov, 2014b) The limit as the boundary expresses the limit of the knowledge of things, its essence and purpose, as well as the measure.Measuring the limits is in turn converted to the limit manifestations of something, the definition of a sense of proportion, symmetry.Important prerequisites for the analysis gives the term norm characterizing typically the border to use as well as (for example, in philosophy) causative reason or another action.From the perspective of philosophy the border not only separates one object from another, but is the foundation of their relationship.
All these terminological interpretations from a theoretical point of view one way or another consider the border of a particular process through the prism of its essence.
If we turn to Aristotle, we will find his uncompromising statement that limit expresses the essence of each thing and its essence of being.
Departuring from the entity, including the entity's credit, ignoring the specific properties, using contrary to those of the entity within which it expresses, is inevitably in practice leads to the destruction of those relations, which are characteristics of the return movement of the loaned amount.
Let's not forget that the loan is not only the activity of the lender or the borrower, but the relationship between them, therefore, from a methodological point of view it is important that the boundaries of the loan express activity not one of them, but equally treat each of them, as well as to the third element of credit relations-what is being loaned amount.This means that the boundaries of the loan should be considered as the boundary of its functioning as a whole.This does not exclude, however, that each of the elements of a loan has its own limit (border) function.If, for example, we talk about the boundaries of the lender, although this will be the scope of its activities, but not the boundaries of the loan as a more comprehensive entity.
Borders of credit and borders of lending are different concepts.The first is a broader concept; it is addressed to the credit relations, affects the boundaries of the functioning of three components: the lender, the borrower and the loaned value.The second concept is associated with the activities of the lender providing the loan to the borrower.But here, it is always important to consider that the boundaries of the loan or credit are always a correlation with the position of the entity of the loan, its objective necessity and possibility of compliance with the terms of the credit transaction; it is always the problem of sizing the needs of alienation (receipt) of the loan and the calculation of participants' existing credit process relevant opportunities, for example in the form of resources.(Burakov, 2014c) Let's not forget that the transfer of what is being loaned by the creditor and its receipt by the borrower in committing these acts should be of interest to both sides.The creditor must be awared of the need for the loan, and the borrower of necessity of its receipt.However, in the credit transaction it is always important not only necessity, but also the possibility to comply with its norms.Such features of the lender and the borrower are always different, and without them, the credit relationship will not become a reality.Credit is a voluntary relationship, so its boundary is always the comparison of the need and feasibility of compliance with the loan agreement.
When exploring borders of credit corresponding to its specificity, it is important to connect it with those properties that separate this economic relationship from other monetary institutions.
The boundaries of credit is a kind of watershed (line, frame), beyond which the loan turns into its opposite, ceases to exist as a process with a specific entity.
Departuring from the essence of the loan, ignoring its specific properties, the use of the intrinsic limits, which he expresses, inevitably, in practice, leads to the destruction of those relations, which are characteristics of the return movement of the loaned amount.
Such a specific property that restricts the functioning of credit is its recoverability.Violation of recovery as a basic limit inevitably leads to the loss of credit of its specific qualities, turns a loan into another form of relations, not peculiar to its nature.Credit beyond the return movement, turns into its opposite, ceases to exist.
One should also not forget that the boundary is, however, not the only a thing that separates the subject from the other subjects, but also what connects it with them.Such a link, as is known, can be implemented through the laws that are expressing significant connections, reflect the relationship between entities.Such laws are the laws of repayment of the loan, the balance between released and redistributed resources based on return, the preservation of what is being loaned, the cost and time of operation.Violation of these laws is a violation of the borders of credit, which inevitably lead to the involvement of credit to cover the needs of not its nature, the emergence of imbalances and contradictions in credit sphere.
Border of credit relations can therefore be defined as the limit of the transmission by lender of loan value available for use by the borrower to meet its objective needs while preserving the essential properties and laws of credit.In conventional theory, credit boundaries, reflecting its essence, are most often associated not with credit in general, and limit the functioning of its individual forms, in particular the banking, commercial and consumer credit.
Borders of credit, however, do not depend on forms of credit.In whichever form the credit is not functioning (in the form of banking, commercial, consumer, and international credit) basic, essential features always permeate all stages of movement of the loaned amount.This does not exclude, however, that in each form of credit specific financial prerequisites exist.In each form, credit as materialistic entity becomes dependent and limited by well-defined flows.The expression boundaries of credit, so can be not only a baseline of compliance, but more, reflecting the peculiarities of the application of one or another form.It is therefore important not only to forget all that is inherent in the credit as an economic category, but also do not ignore what reflects the movement of credit in its concrete form.
The boundaries of crefdit can be divided into qualitative and quantitative; internal and external; time and space; due to supply and demand; starting and ending.
Qualitative border of credit is the most important thing in its essence, it is the basis of credit.Such basis, as noted earlier, is the return of what is being loaned.Practice shows that in a modern economy the boundaries of the loan in a number of areas are disturbed.This was manifested in violation of the basic principles of lending, the continuing growth of bad loans' issuance.Artificial prolongation of bank loans, highweight of non-performing loans, the maturity mismatch of resources to invest in long-term projects, the devaluation of deposits at an interest rate lower than inflation, etc. are not to be taken into account and present themselves a hidden violation of the boundaries of credit, when credit is drawn in covering unsustainable costs, the needs of borrowers, unsecured by actual savings.
Quantitative borders of credit are often linked to the specific needs of one of the participants of the credit process.Justified in this connection looks the treatment to the satisfaction of using the credit for needs of the production and circulation of additional capital, and not all of them, but only those the circuit of which provides the return of the loaned amount.We can say that this upper limit of credit that addresses the needs of the borrower is intended to highlight those needs, the limit of which is not contrary to the nature of the loan.In quantitative terms, it is important, therefore, not only to highlight what the borrower needs, but also something that can be returned to the creditor.
From a quantitative point of view, equally important is the lower boundary of credit.Typically, such a boundary is associated with the resources available, for example, for the creditor.We know that loaned is often can be only that what is with the lender, resources not employed in his own business.Resources obtained by the creditor, are the objective limit, which accounts for possibility of loan issuance.At the same time resources are the limit that defines not only the creditors' side, but also the borrower's one.
On a practical level, the limit of credit is often considered as gross domestic product, which creates the material preconditions for the return of what is being loaned.In practice, however, the loan amount is often more than you could expect to be used for gross domestic product.This situation is quite understandable, because the loan can repeatedly mediate different parts of the production and circulation and, therefore, exceed the amount of product produced.
Practice shows that the interpretation of the resources and/or their growth rate does not allow you to use this indicator as a reliable information about the observance of the boundaries of the loan.It is important to take into account the fact that the size of the manufactured product does not mean that those who borrowed, would have many opportunities for its repayment.It is not the fact that in the case of production of the respective gross domestic product borrowers will be able to return what was originally borrowed.
More productive information on the observance of boundaries have accumulations of capital created by society.Accumulation quantity as a limit of credit is equally important for the expression of the limit to meet the needs of the borrower in the use of additional capital, and for lender to grant the loan and its receive it back.
Conclusion
The disclosure of these boundaries is complicated because the boundaries are not an unchanged line, defining the limits of a certain subject.However, this does not give reason to believe that due to their volatility, such boundaries do not exist.In fact, the line as the limit may develop, however, the limits of this development and operation must exist under the stable entity.
From a theoretical and a practical point of view, it is important to consider the boundaries of credit, including their role in the economy.It is known that the purpose of credit is to ensure continuity of circulation and the turnover value in the form of circulating and fixed capital; acceleration of the circulation and circulation of capital; savings in public expenditure.
Each of these purposes, of course, is achieved in society, not only with credit.For example, to accelerate the speed one can use private capital.The feature, however, is that in this case is that other methods and tools will be used.Since credit can act as other monetary entities, its borders are movable, changeable, however, they always represent the essence of credit; deviation from this entity indicates the misalignment of credit use within its borders, the termination of its existence. | 2019-05-14T14:03:42.780Z | 2015-05-24T00:00:00.000 | {
"year": 2015,
"sha1": "839ffbb4758fec79aef95838ae5212fa08373840",
"oa_license": "CCBY",
"oa_url": "https://www.ccsenet.org/journal/index.php/res/article/download/49210/26412",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "839ffbb4758fec79aef95838ae5212fa08373840",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Economics"
]
} |
5031364 | pes2o/s2orc | v3-fos-license | Graphene hot-electron light bulb: incandescence from hBN-encapsulated graphene in air
The excellent electronic and mechanical properties of graphene allow it to sustain very large currents, enabling its incandescence through Joule heating in suspended devices. Although interesting scientifically and promising technologically, this process is unattainable in ambient environment, because graphene quickly oxidises at high temperatures. Here, we take the performance of graphene-based incandescent devices to the next level by encapsulating graphene with hexagonal boron nitride (hBN). Remarkably, we found that the hBN encapsulation provides an excellent protection for hot graphene filaments even at temperatures well above 2000 K. Unrivalled oxidation resistance of hBN combined with atomically clean graphene/hBN interface allows for a stable light emission from our devices in atmosphere for many hours of continuous operation. Furthermore, when confined in a simple photonic cavity, the thermal emission spectrum is modified by a cavity mode, shifting the emission to the visible range spectrum. We believe our results demonstrate that hBN/graphene heterostructures can be used to conveniently explore the technologically important high-temperature regime and to pave the way for future optoelectronic applications of graphene-based systems.
Introduction
The concept of van der Waals heterostructures has led to a new technology of layer-by-layer engineering of two-dimensional (2D) materials with atomic precision [1,2]. Out of the abundance of 2D materials, the two-graphene and hexagonal boron nitride (hBN), still remain the most unique couple, because of their outstanding synergetic properties. The hBN plays a vital role in ultra-high quality graphene devices [3,4], and as an enabling material for graphene superlattices [5]. Besides, it is also known to protect sensitive 2D materials from environment via encapsulation [6]. Graphene, among its many superlatives, has the highest breakdown current density j max ~ 10 12 A cm −2 (assuming graphene thickness 3.4 Å), as measured in transmission experiments of accelerated (180 keV) xenon ions [7]. In electronic transport devices, the breakdown current densities of graphene are more modest, reaching ~5 · 10 8 A cm −2 in a vacuum (and much lower values in air) [8].
The high current-carrying capacity of graphene has made it a simple source of thermal infrared emission [9,10]. Controlled by a bias voltage, a maximal emission temperature is limited to ~700 K due to heat dissipation to the metallic contacts and to the substrate [8,10]. The latter mechanism dominates for graphene devices longer than a few microns. In the case of hBN encapsulated graphene devices, heat can also be efficiently dissipated through radiative heat transfer to hBN via its hyperbolic phonon polaritons [11]. Heat transfer is substantially reduced for suspended graphene devices, which leads to much higher emission temperatures approaching 3000 K, and to the emission in the visible spectrum (incandescence) when biased by electrical pulses under high vacuum [12]. But the stability of these devices in air is severely limited due to the oxidation of graphene [8,12].
Metallic single-walled carbon nanotubes were also used as thermal light emitters, although, small saturation currents (<20 µA for a 3-4 nm diameter nanotube) and rapid oxidation in the air limit their use [13]. Interestingly, the electronic transitions between van Hove singularities led to a strong modification of the emitted spectra [13], which can be tuned even further by employing 1D photonic crystal cavities [14]. In this work, the encapsulation of graphene (Gr) with hBN allowed us to demonstrate robust incandescence of graphene devices in ambient conditions under continuous DC bias over extended periods of time. The hBN encapsulation provides excellent protection for graphene even at temperatures well above 2000 K, which we attribute to an unparalleled oxidation resistance of hBN [15][16][17][18]. Our study also reveals that dielectric cavity can support surface-guided waves at visible wavelength and tune the spectra of Planckian radiation. Figure 1(a) shows one of our hBN/Gr/hBN heterostructures fabricated on quartz substrate, using conventional dry peel technology followed by a standard e-beam lithography, dry etching and metallisation, as described elsewhere [4,19]. We designed our devices such that the distance between the contacts is long enough (L ~ 10 µm) to avoid any effects from the contacts while keeping the contact resistance low by increasing the metal-to-graphene contact length. To characterise the light emission we pass a large current I (current density, j = I/W, reaching 600 A m −1 ) through a narrow (W ~ 1 µm is the channel width) graphene channel by applying
Light emission from hBN-encapsulated graphene
where L is the channel length). The supplied electric energy is transformed into Joule heat and dissipated in graphene, encapsulating hBN, quartz substrate, and metallic contacts, leaving a small fraction to radiate into free space. Figure 1(b) shows the emission spectra recorded using Renishaw Raman spectrometer equipped with Si-based CCD detector (diffraction grating 1200/mm; edge filter and beam splitter were removed during measurements). The obtained spectra are mostly featureless and can be described using Planck's law, modified by energy independent emissivity (grey-body radiation) [9]: where hν is the photon energy, T is the temperature, c is the speed of light, k B and h are Boltzmann's and Planck's constants. Using equation (1) we obtained reasonable fits to the recorded emission spectra; extracted electronic temperatures and corresponding currentvoltage characteristics are presented in the inset of figure 1(b). We also evaluated the model for light emission from graphene proposed in [12] but found that the spectra are indistinguishable from greybody emission in the investigated spectral range. The emission efficiency (P emission /P electric ) of our devices was ~1.6 · 10 −5 , obtained using Stefan-Boltzmann law (P emission = AσT 4 , A-device area, σ-Stefan-Boltzmann constant) and electrical power (P electric = IV ) similar to [9,12]. For the emissivity (ε) we used 0.023, following literature [9], the details can be found in the supplementary information (stacks. iop.org/TDM/5/011006/mmedia) (section 3). Our devices demonstrated surprising long-term stability (many hours) at these very high current densities (>1.7 · 10 8 A cm −2 ), comparable with multilayer graphene-based infrared emitters but operated at much lower (7.1 · 10 6 A cm −2 ) current densities [10]. We further discuss a long-term stability and the mechanisms of breakdown in the supplementary information (section 1).
Enhancement of incandescence using photonic cavity
We prepared another set of hBN/Gr/hBN heterostructures but this time on SiO 2 /n-Si substrate and carried out the same high current light emission measurements, figure 2(a). We found that the emission spectra shift to the visible range and the graphene device incandesces (figures 2(b) and (c)). Considering this device is almost identical to that in figure 1, except for the presence of heavily doped silicon substrate, the drastic change in the emission spectra (see figures 1(b) and 2(c)) can be attributed to a resonance effect of a photonic cavity confined by the two interfaces: hBN/ air and SiO 2 /Si, figure 2(d).
It is instructive to analyse the parameters of this cavity using a simple model of a Fabry-Pérot resonator [20,21]. The optical path length of the cavity D = d i n i is ~740 nm for 290 nm SiO 2 and 150 nm hBN/Gr/hBN stack. The refractive indices n i of SiO 2 and hBN show negligible dispersion throughout the visible and near infrared spectral range, averaging at n SiO 2 ~ 1.49 and n hBN ~ 2.05 [20,22]. The reflectivities R of the 'mirrors' formed by hBN/air and SiO 2 /Si interfaces, are 0.12 and 0.19, estimated from Fresnel where n i and n j are the refractive indices of the materials forming interfaces (with n Si ~ 3.8 and n air = 1, from [20]). These relatively low reflectivities lead to the output coupling losses, typically parametrised by the mirror coupling coefficients, (2)) of the graphene device at different bias voltages (from 9 V to 23 V in 2 V steps, violet to dark red). The inset shows current-voltage characteristics of the device, as well as the electronic temperatures estimated from photonic cavity model, equation (2) We modelled the incandescence spectra by considering the grey-body radiation (equation (1)) modulated by a spectral line shape of the second mode of the cavity (using a Lorentzian, f (hν) = ( /τ ) 2 4(hν−2h/tRT) 2 +( /τ ) 2 , following [21]): Surprisingly, this simplistic approach provides reasonable quality fits for the data (dashed lines in figure 2(c)), using temperature as the only fitting parameter. The extracted temperatures are presented in the inset of figure 2(c), the values are higher than for the device on quartz (see inset of figure 1(b)) because of the larger electrical power applied in the former case. Our temperatures were somewhat lower (1500-2500 K) in comparison with the emission (2500-2900 K) reported for suspended graphene [12]. We attribute this behaviour to the dissipation of heat to the surrounding media through hBN [11]. The hyperbolic phonon polaritons of hBN [23,24] are predicted to efficiently (on a picosecond scale) cool hot graphene [11]. To reveal the interplay between the underlying physical mechanisms of cooling one need to perform time-dependent spatially-resolved thermal conductivity measurements, preferably on suspended hBN/Gr/hBN heterostructures, which is beyond the scope of this article. Another effect of the photonic cavity can be seen in figure 2(b)-two rows of bright dots on the opposite ends of the device (their positions coincide with the tips of the gold comb electrodes). The gold electrodes are in thermal equilibrium with environment due to their macroscopic size and very good thermal conductivity, hence the temperature would be too low for the incandescence of the gold tips. On the other hand, the incandescent light from the central area of the device could travel through the waveguide formed by a photonic cavity all the way to the contacts, where it radiates possibly via coupling to gold plasmon modes at the contact edges [25]. The waveguide modes guide the graphene light emission over a long distance (at least 5-10 µm) even in this simple cavity, while the performance can be further improved by optimising the dielectric material and thickness to enhance the light emission [26,27]. The frequency range of the photonic cavity modes can also be adjusted by optimising appropriate geometrical and material parameters.
Transport measurements in high-bias regime
To better understand the physics of graphene incandescence and to independently estimate the electronic temperatures in a high-bias high-current regime we performed four-terminal transport measurements on an hBN/Gr/hBN Hall bar device, figure 3. Electric field and current density values were obtained from measured 4-probe voltage V xx , applied current I, and the geometry of the Hall bar (L × W is 2 × 1 µm 2 ). At low biases, the current density increases linearly with the electric field and then deviates from the ohmic regime at |E| > 0.1 V µm −1 , figure 3(a). This saturation tendency is typically attributed to the presence of in-plane acoustic and optical phonons [28][29][30], and other effects, such as Auger scattering [31]. To estimate the charge carrier density n, we measured our Hall bar device in a perpendicular magnetic field B, following the procedure described in [33]. The transversal voltage V xy was linear with B within the measurement range of ± 2 T, allowing to estimate n = −IB/eV xy at different electric fields, figure 3(b). At a non-zero temperature (T) and in the presence of disorder (Δn < 10 10 cm −2 ) the measured carrier density is where n th (T) = π 6 kBT vF 2 is the thermally-induced carrier density, and v F is the Fermi velocity of electrons in graphene [34,35]. The gate induced carrier density n g = εε0 de ∆V g (εε 0 and d are the dielectric constant and the thickness of gate dielectric) was estimated from the corrected gate voltage ΔV g , where the self-gating effect (due to large bias voltage) was taken into account, red dashed line in figure 3(b). Using equation (3) we estimated the effective electron temperature of graphene as a function of electric field, figure 3(c). The obtained temperatures are in range with those estimated from the emission spectra (figures 1 and 2).
Knowing the current density j and the charge carrier density n we extracted the drift velocity v d = j/ne, and the mobility µ = v d /E at different electric fields (and, hence, temperatures). The v d slowly saturates starting at E ~ 1.5 V µm −1 reaching 4 × 10 5 m s −1 , almost 40% of the Fermi velocity. The mobility, high initially (~100 000 cm 2 V −1 ·s −1 ), drops to below 1000 cm 2 V −1 ·s −1 at high electric fields. From a simple Drude model we approximated the scattering time τ and the mean free path l mfp of the charge carriers from their mobility, µ √ πn/e = v F τ ≡ l mfp . For independent scattering processes, the scattering time τ is dominated by the shortest process, which at high bias fields is usually the emission of the optical phonons [28,36]. If the potential difference over the characteristic length l mfp is sufficiently high (eEl mfp Ω) a phonon of energy Ω can be emitted resulting in the dissipation (heat) of electric energy in graphene lattice, and surrounding hBN and SiO 2 . This process, in turn, limits the l mfp , and bounds the non-equilibrium Fermi distribution, as shown on the cartoon in figure 3(d) (depicted as the tilt of the Fermi level). Figure 3(d) compares the characteristic energy eEl mfp against the typical (at high density of states in the phonon dispersion spectra) energies of representative phonon branches of graphene [37,38]. Interestingly, the obtained characteristic energy values ~60 meV are sufficiently lower than graphene's in-plane phonon excitation energies, figure 3(d). Perhaps low energy acoustic phonons, longitudinal guided modes or other higher order acoustic modes at the interface of hBN and graphene limit the characteristic energy [37]. For a rather thin hBN substrate the characteristic energy might also be bound by optical phonons of SiO 2 substrate [28].
Raman spectroscopy and lattice temperature
To complement the light emission and transport study, we performed Raman spectroscopy on our hBN/Gr/ hBN heterostructures. To this end, we employed the same Renishaw spectrometer using 633 nm excitation laser line with incident power of 2 mW. The recorded spectra of the characteristic Raman peaks (G and 2D for graphene, E 2g for hBN) for different bias voltages are summarised in figure 4. The intensities of all three peaks monotonically decrease with temperature. There is also a downshift with increasing temperature for all three peaks: 5 cm −1 , 6 cm −1 , and 7.5 cm −1 for G and 2D peaks of graphene, and E 2g peak of hBN at the highest applied bias, respectively. The downshift of the G peak is attributed to the decay of Γ-point optical phonon into acoustic phonons through a range of phonon-phonon scattering processes [39]. Similar anharmonic decay processes (complicated by competing electron-electron scattering) of K-point optical phonons are also responsible for the downshift of the 2D peak of graphene [40,41]. However, both G and 2D peaks tend to blue-shift as a result of doping [42] and strain [43], thus making it difficult to extract the correct lattice temperature from the Raman peak shift. Furthermore, the linewidth of the 2D peak is also affected by strain, for example, the nanometre-scale strain variations contribute to broadening of the 2D peak [44]. We, therefore, focus below on a linewidth of G peak as the most robust method.
We observed a non-monotonic behaviour of the linewidth γ(T) of the Γ-point phonon mode (G peak) with temperature, figures 4(a) and (b). This behaviour is in agreement with previous experimental observations, where linewidth of G peak is also first decreased and then increased with temperature [45]. We follow a simple model, considering two processes by which a Γ-phonon acquires a finite linewidth γ G (T): by the anharmonic decay into two secondary acoustic phonons, γ ph-ph (T), or by creating an electron-hole pair (Landau damping), γ e-ph (T,µ) [39,45]. Here we ignore any higher-order anharmonic coupling processes. The total estimated linewidth of G peak γ G (T) = γ e−ph (T, µ) + γ ph−ph (T) is plotted in figure 4(b) (black line; the anharmonic contribution and the Landau damping are the blue and red lines, respectively). We use 10 cm −1 and 3 cm −1 for γ e-ph (0,0) and γ ph-ph (0), and include thermally-induced suppression of Landau damping, following procedures from [45,46]. The γ G values from figure 4(a) plotted as coloured circles versus electronic temperatures extracted from figure 2(c), are in a good agreement with the simple model suggesting the dominant anharmonic scattering mechanism.
Lastly, the hBN E 2g peak is also downshifting and broadening with temperature (figure 4(d)) due to similar anharmonic phonon decay processes in hBN [47]. This suggests there is a strong heat transfer from graphene to the encapsulating hBN. Despite the heat dissipation in the encapsulating hBN layers, the electronic temperatures of graphene were well above 2000 K, as estimated from the incandescence spectra. Understanding the detailed mechanism of heat dissipation through the encapsulating hBN requires further theoretical investigations.
Conclusions
We demonstrated an interesting region of a parameter space (high current and electric field, high electronic and lattice temperatures), accessible in simple graphene/ hBN heterostructures (and in the majority of existing devices) in ambient conditions. Electrically tunable light emission of graphene is modified by a photonic cavity formed at the interfaces, which can be further improved by optimising the parameters of the cavity. This atomically-thin light source is integrated with the cavity, which forms a waveguide, paving the way for new types of electrically-coupled photonic integrated circuits. This level of integration can be brought even further, by combining the photonic circuit elements with electrically-driven phononic and mechanical modes. We believe our study will stimulate further experimental and theoretical work in this fascinating research direction, where many interesting questions are now open, such as the understanding the mechanisms of heat dissipation in these devices, including the effects of hyperbolic phonon polaritons in hBN. | 2017-12-22T00:47:32.270Z | 2017-10-26T00:00:00.000 | {
"year": 2017,
"sha1": "b3f251b5a078e7a70cb4e0364b758f0513dd6512",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1088/2053-1583/aa97b5",
"oa_status": "HYBRID",
"pdf_src": "IOP",
"pdf_hash": "0af6046ce2b20ab7ab58587fc45b1e54dfe0c07f",
"s2fieldsofstudy": [
"Engineering",
"Materials Science",
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Materials Science"
]
} |
4734626 | pes2o/s2orc | v3-fos-license | The different pathogeneses of sporadic adenoma and adenocarcinoma in non-ampullary lesions of the proximal and distal duodenum
Non-ampullary duodenal adenoma with activation of Wnt/β-catenin signalling is common in familial adenomatous polyposis (FAP) patients, whereas sporadic non-ampullary adenoma is uncommon. The adenoma-carcinoma sequence similar to colon cancer is associated with duodenal tumors in FAP, but not always in sporadic tumors. We obtained 37 non-ampullary duodenal tumors, including 25 adenomas and 12 adenocarcinomas, were obtained from biopsies and endoscopic resections. We performed immunohistochemistry for β-catenin, the hallmark of Wnt activation, and aldehyde dehydrogenase 1 (ALDH1), a putative cancer stem cell marker. In non-ampullary lesions, abnormal nuclear localization of β-catenin was observed in 21 (84.0%) of 25 adenomas and 4 (33.3%) of 12 adenocarcinomas. In the proximal duodenum, nuclear β-catenin was less frequent in both adenomas and adenocarcinomas. Gastric duodenal metaplasia (GDM) was observed only in the proximal duodenum. All adenomas with GDM were the gastric foveolar and pyloric gland types, and showed only membranous β-catenin. The intestinal-type adenomas had nuclear β-catenin in the proximal and distal duodenum. ALDH1-positive cells were more frequent in adenocarcinomas than adenomas. Nuclear β-catenin accumulation frequently occurred in ALDH1-positive cells in adenoma, but not in adenocarcinoma. In the non-ampullary proximal duodenum, Wnt/β-catenin pathway activation was more closely associated with adenomas than adenocarcinomas, and while it might cooperate with ALDH1 in adenoma, it does not in adenocarcinoma. The pathogenesis thus may differ between sporadic adenoma and adenocarcinoma of non-ampullary duodenal lesions, especially in the proximal and distal duodenum.
INTRODUCTION
Most small intestine adenomas, including the ampulla of Vater, develop in the duodenum. Among these, sporadic non-ampullary (except the ampulla of Vater) duodenal adenomas are uncommon [1,2] Research Paper www.impactjournals.com/oncotarget (APC) gene. Approximately 60% of duodenal adenomas develop in patients with FAP, while the remaining 40% are sporadic [3]. Non-ampullary duodenal adenoma is thought to be a precancerous lesion, and the model of the adenoma-carcinoma sequence is predicted to involve the small intestine, including the duodenum and colorectum [4][5][6].
The Wnt/β-catenin pathway promotes organ development and tissue homeostasis [7], and disruption of the pathway causes various malignancies [8]. β-catenin, which binds to the product of the APC gene, is a multifunctional cytoplasmic protein with a molecular weight of 92 kDa [9]. The accumulation of β-catenin in the nucleus is the hallmark of Wnt pathway activation, and leads to cell proliferation by activating lymphoid enhancer factor/T-cell factor transcription factors [7]. Wnt pathway activation, particularly by APC and CTNNB1 gene mutation, is associated with early events in colon carcinogenesis in the model of the adenomacarcinoma sequence [9]. Nuclear β-catenin translocation increased with the progression of colorectal tissue from normal epithelial tissue, adenomas, to carcinomas in the colorectum [10].
The aldehyde dehydrogenase (ALDH) superfamily constitutes a group of NAD(P) + -dependent enzymes that metabolize endogenous and exogenous aliphatic and aromatic aldehyde molecules by oxidation to their corresponding carboxylic acids [13]. The human genome includes 19 known ALDH genes [14]. ALDH1 is a cytoplasmic member of the ALDH family of enzymes. ALDH1 expression was originally detected in a small subpopulation of hematopoietic stem/progenitor cells [15,16]. ALDH1 is thought to be a putative cancer stem cell marker in the stomach and colorectal cancers [17,18].
In the majority of colorectal cancers, the Wnt/βcatenin pathway is necessary for the self-renewal of cancer stem cells (CSCs) [19,20]. On the other hand, ALDH1 is a putative CSC marker, and ALDH1-positive cells have CSC potential in colorectal and gastric cancers [17,[21][22][23]. ALDH1-positive cells have shown higher levels of the regulatory components of Wnt/β-catenin signalling, such as β-catenin and TCF-4, compared with ALDH1-negative cells in breast cancer [24].
Gastric duodenal metaplasia (GDM) is a histologically detectable mucosal change thought to evolve following Helicobacter pylori infection, celiac disease, or Crohn's disease [25][26][27]. The regenerative stimuli are caused by an aberrant high production of gastric acid triggered by Helicobacter pylori infection [28]. GDM has been found even in the absence of these conditions and in duodenal adenomas [29]. Non-ampullary adenoma/carcinoma more often develops on the oral (proximal) side of the ampulla in Japan because of the higher rate of H. pylori infection compared with that in Western countries [30]. Further, the anal (distal) side of non-ampullary adenoma/carcinoma frequently occurs with colon adenoma/carcinoma, which may be influenced by bile acids [30].
In this study, we evaluated β-catenin and ALDH1 expression in non-ampullary duodenal adenomas and adenocarcinomas using immunohistochemistry (IHC), and investigated the clinicopathological characteristics and non-ampullary duodenal carcinogenesis to provide a foundation for therapy development.
Case pathology
Thirty-seven patients were included in this study ( Table 1), 29 of which were male and 8 were female. A total of 25 sporadic non-ampullary duodenal adenomas (25 patients) and 12 sporadic non-ampullary duodenal adenocarcinomas (9 patients) were obtained during the study period. Patient ages ranged from 26 to 77 years (mean, 40.5 years). The tumor position in the duodenum, except the ampulla of Vater, was separated into two parts. The upper part (proximal) was from the bulb to the periampulla, while the lower part (distal) was from the periampulla to the descending duodenum. Most adenocarcinomas were detected in the upper part (9 of 12 cases) because all adenocarcinoma cases were resected using the limiting technique, i.e. EMR/ESD. Biopsy was performed on 16 lesions (43.2%) and EMR/ESD was performed on 21 lesions (56.8%). Tumor size ranged from 4 to 70 mm (mean, 37.0 mm), and the tumor size of adenocarcinoma patients was higher than that of adenoma patients (P < 0.05). Additionally, no patients had a family history of Lynch syndrome.
Nuclear β-catenin expression was frequent in adenoma but not adenocarcinoma in the nonampullary duodenum
To investigate the expression of β-catenin in nonampullary duodenal adenoma and adenocarcinoma, we performed IHC with a β-catenin antibody. Normal epithelium in the duodenum showed strong membranous staining for β-catenin with a few nuclear β-cateninpositive cell in the bottom, and aberrant expression was observed in both adenoma and adenocarcinoma ( Figure 1A and Supplementary Figure 1). Abnormal nuclear localization of β-catenin was observed in 21 (86.4%) of 25 adenomas, which was higher than the 4 (22.2%) observed in 12 adenocarcinomas ( Figure 1B). These data suggest that Wnt/β-catenin activation is more closely associated with adenoma compared with adenocarcinoma in nonampullary lesions.
Non-ampullary duodenal adenomas with high-grade dysplasia and large size (≥20 mm in diameter) indicate a high risk of progression to adenocarcinoma [31]. In our cohort, 14 (56%) of 25 adenomas were more than 20 mm at their largest diameter, and nuclear β-catenin was frequent (93%) in these large adenomas (≥20 mm) ( Table 2). These results indicate that Wnt/β-catenin activation may contribute to the progression of sporadic non-ampullary duodenal adenoma similar to colorectal carcinogenesis. On the other hand, the relationship between nuclear β-catenin, i.e., Wnt activation, and tumor size may be associated with aberrant apoptosis of tumor cells. This may be supported by previous reports nuclear β-catenin contributes to increased tumor size [32,33]. Nuclear β-catenin is less frequent in adenocarcinoma than in adenoma in non-ampullary lesions. This suggests the conventional adenoma-carcinoma sequence associated with activation of Wnt/β-catenin pathway may be less frequent in non-ampullary lesions compared with colon carcinogenesis.
GDM was observed only in gastric type adenoma with membranous β-catenin
We focused on the positional differences and histological subtypes of adenomas and adenocarcinomas of non-ampullary duodenal lesions. In the initial manifestation of GDM of our cohort, small patches of gastric foveolar cells appear near the tip of a villus ( Figure 2A). The patchy GDM in the duodenal mucosa indicates that GDM might be a reactive metaplasia rather than a congenital lesion. GDM was found in 4 (16%) of 25 adenomas and all the cases were located in the proximal duodenum. It follows that GDM is commonly observed in the proximal duodenum and is associated with peptic ulceration caused by Helicobacter pylori infection in most cases [34].
All 4 of our cases showed only membranous β-catenin expression. The other 21 adenomasshowed nuclear β-catenin expression and these were all intestinaltype adenomas identified with expressions of CDX2(+), MUC5AC(-), and MUC6(-) ( Table 2). These data suggest that activation of the Wnt/β-catenin pathway is not involved in the development of gastric-type Data are presented as average ± SD (mean range). EMR, endoscopic mucosal resection; ESD, endoscopic submucosal dissection. * P < 0.05: adenoma versus adenocarcnoma in the tumor size duodenal adenoma with GDM in non-ampullary lesions. Consequently, total 25 adenomas were divided into 4 gastric-type adenomas (2 foveolar-type adenomas and 2 pyloric gland adenomas) and 21 intestinal-type adenomas. Next, we investigated the β-catenin expression pattern in different locations of adenomas and adenocarcinoma in the duodenum. We divided the part of non-ampullary duodenum; the upper (proximal) part from the bulb to the periampulla and the lower (distal) part from the periampulla to the descending duodenum. Nuclear β-catenin was frequent in adenoma and adenocarcinoma of the lower part and adenoma of the upper part ( Figure 3). However, in the upper part, adenomas and adenocarcinomas with membranous β-catenin accounted for 4 (25%) of 16 adenomas and 6 (66.7%) of 9 adenocarcinomas, respectively; this difference was statistically significant. Our data indicates that a subset of sporadic adenocarcinomas in the proximal duodenum might be independent of the Wnt/β-catenin pathway, while a subset of adenocarcinomas in the proximal duodenum is associated with the Wnt/β-catenin pathway as well as the distal duodenum.
ALDH1-positive cells showed significantly increased adenocarcinoma compared with adenoma
We investigated ALDH1 expression in nonampullary duodenal adenoma and adenocarcinoma by IHC analysis. ALDH-positive cells were not detected in normal duodenal epithelium; however, ALDH1-positive cells were observed in adenomas and adenocarcinomas ( Figure 4A). ALDH1 protein showed predominantly cytoplasmic expression in the tumor cell. Further, we analysed the signal intensity scores of ALDH1 immunostaining in adenoma and adenocarcinoma. The signal intensity was significantly higher in adenocarcinomas than adenomas ( Figure 4B), indicating that ALDH1 may associated with the progression from adenoma to adenocarcinoma in nonampullary lesions.
ALDH1-positive tumour cells were frequently co-localized with nuclear β-catenin in adenoma
To investigate whether Wnt/β-catenin and ALDH1 activation occur in the same tumor cells of adenoma and adenocarcinoma, we performed both IF and IHC staining for distinct conclusion. In the IF assay, the co-localization of nuclear β-catenin and ALDH1 strong expression was observed in tumor cells of adenoma and adenocarcinoma ( Figure 5A). In the IHC assay, we could more clearly detect the co-expressed cells. To evaluate the number Figure 5B). In adenomas, only ALDH1-positive cells, only nuclear β-catenin-positive cells, and double-positive cells of nuclear β-catenin and ALDH1 were 4.83%, 2.15%, and 4.14%, respectively, in the region in which nuclear β-catenin was observed most frequently. Inadenocarcinomas, only ALDH1-positive cells, only nuclear β-catenin-positive cells, and double positive cells were 11.5%, 0.06%, and 0.41%, respectively. Further, the positivity of nuclear β-catenin-positive cells in total ALDH1-posive cells between adenomas and adenocarcinomas, resulting that the positivity in adenomas was much more than in adenocarcinomas ( Figure 5B). These data suggest that Wnt/β-catenin signalling pathway might be associated with CSCs marked by ALDH1 in adenoma rather than adenocarcinoma.
DISCUSSION
In this study, we clarified the expression of β-catenin and ALDH1 in sporadic adenomas and adenocarcinomas in non-ampullary duodenal lesions by IHC. We have suggested the pathogenesis in sporadic non-ampullary duodenal adenoma and adenocarcinoma( Figure 6). We demonstrated that the Wnt/β-catenin pathway was associated with intestinal type adenoma in the proximal and distal duodenum, while the pathway was independent We did not include any cases of ampullary duodenal adenomas or adenocarcinomas because ampullary neoplasms may have combined features of neoplasms of the small intestine including duodenum, distal bile duct, and/or pancreas. Adenocarcinomas are particularly rare in non-ampullary lesions compared with adenomas.
Nuclear β-catenin accumulation, a key effector of the Wnt signalling pathway, was frequent in adenoma rather than in adenocarcinoma. There were more ALDH1positive cells with CSC potential in adenocarcinomas compared to adenomas. Further, nuclear β-catenin in these cells was more frequently co-expressed in adenomas rather than in adenocarcinomas. Our data suggest that Wnt/βcatenin activation might be associated with ALDH1expressing CSCs in a subset of non-ampullary duodenal adenomas, i.e. intestinal type adenomas. There are similar findings in breast cancer, where ALDH1-positive cells show higher levels of Wnt/β-catenin signalling compared with ALDH1-negative cells [24]. By contrast, in adenocarcinomas of the proximal duodenum, Wnt/βcatenin activation was less frequent and ALDH1-positive cells were more frequent than adenoma. This result may be caused by the development of diverse tumors in the proximal duodenum compared with the distal duodenum. Adenocarcinomas are difficult to categorize due to their heterogeneity, but adenomas are easily categorized into intestinal type, gastric foveolar type, and pyloric gland adenomas.
Our study showed nuclear β-catenin expression in 84% of adenomas and 33.3% of adenocarcinomas in nonampullary duodenal lesions. A previous study observed aberrant β-catenin expression and nuclear/cytoplasmic translocation in 38.1% of duodenal adenocarcinomas, including both non-ampulla and ampulla of Vater [37]. In contrast to our results in non-ampullary lesions, 8.2% of adenomas and 40.5% of adenocarcinomas showed nuclear β-catenin expression in ampulla (of Vater) lesions [38]. Alterations in β-catenin may occur late in neoplasm In non-ampullary lesions of the duodenum, Wnt/β-catenin pathway is associated with intestinal type adenoma in the proximal and distal duodenum, while the pathway was independent of gastric type adenoma in the proximal duodenum. The gastric type adenoma has two subtypes: gastric foveolar-type and pyloric gland adenoma. The gastric type adenoma is associated with GDM (gastric duodenal metaplasia), which may be injured from gastric acid or a Helicobacter pylori infection. In the intestinal type adenoma, Wnt/β-catenin activation occurs as well as in the majority of colon carcinogenesis. The ALDH1-positive cells, a putative cancer stem cell marker, is more expressed in adenocarcinoma than adenoma. However, the co-localization with nuclear β-catenin in adenoma is more frequent than adenocarcinoma. carcinogenesis in the ampulla of Vater. However, our data suggests that in non-ampullary lesions, alterations in β-catenin might be an early event.
In contrast to the typical adenoma-carcinoma sequence in the majority of colorectal carcinogenesis, the development of adenocarcinoma in non-ampullary duodenal lesions seems to have two types; 1) de novo type, i.e. gastric type, that occurs independent of the Wnt/β-catenin pathway, 2) adenoma-carcinoma sequence type, i.e. intestinal type, that develops dependent of the Wnt/β-catenin pathway like colon carcinogenesis. A possible difference between the nonampullary duodenum and colorectum can be observed in the developing gut. Developmentally, the duodenum appears to share characteristics with the stomach, particularly the antrum, rather than the colorectum [39]. In our study, all the gastric-type adenomas with GDM developed in the proximal duodenum, and were associated with membranous β-catenin. This suggests that the non-ampullary proximal duodenum is likely to have gastric characteristics injured by the environment, such as gastric acid and H. pylori.
Regarding colorectal cancers, the CSC population detected using ALDH1 enzymatic activity is typically very small, including 0.01-1% of the population of cancer cells [40]. In contrast, immunohistochemical studies examining various types of cancers have reported remarkably high fractions of positive cells [41,42]. Only one immunohistochemical study has examined small intestinal adenocarcinoma, and strong ALDH1 expression was observed in 33.3% of all cases [43]. In our study of nonampullary lesions, ALDH1-positive cells occurred more frequently in adenocarcinomas than in adenomas, and high ALDH1 expression (Score 3 and 4) was observed in 33% (3/9) of adenocarcinoma cases. This is similar to a colon adenocarcinoma study where ALDH1 positivity in IHC experiments demonstrated an increasing number of colonic stem cells over colon cancer progression [23]. Thus, the ALDH1-based CSC marker can be used to track CSCs in the sporadic non-ampullary duodenal adenoma/adenocarcinoma during the tumorigenesis as well as colon tumorigenesis.
In conclusion, nuclear β-catenin accumulation is more correlated with adenomas rather than adenocarcinomas in the non-ampullary proximal duodenum. The co-localization of nuclear β-catenin and ALDH1 is common in adenoma, and ALDH1 is highly expressed in adenocarcinoma rather than in adenoma. The pathogenesis may differ in sporadic adenoma and adenocarcinoma of the non-ampullary duodenum, especially the proximal and distal duodenum.
Patient and tissue samples
Tumor specimens were obtained from patients through biopsy, EMR, or ESD at Gifu University Hospital between 2008 and 2015, and Gifu Municipal Hospital between 2012 and 2015. Informed consent was obtained from all patients.
Diagnosis of adenoma and adenocarcinoma
Using hematoxylin and eosin-stained sections, adenoma was diagnosed when the tumor was circumscribed and found to be composed of tubular structures showing intraepithelial neoplasia. Adenocarcinoma was diagnosed when the tumor invaded into the lamina propria or through the muscularis mucosae. On biopsy, samples contained areas suggestive of true invasion. Adenocarcinoma without invasion was diagnosed when increased architectural distortion with glandular crowing and prominent cellular atypia and pleomorphic, hyperchromatic, pseudostratified nuclei were observed. The diagnoses were confirmed by at least two pathologists.
Immunohistochemical staining
All paraffin-embedded tissues were cut into 4 μmthick serial sections and deparaffinized. These sections were stained with hematoxylin and eosin or used for IHC. For IHC, the sections were placed in citrate buffer (pH 6.0), and then autoclaved at a chamber temperature of 121°C for 1 min to retrieve the antigen. The sections were then rinsed and blocked in 3% hydrogen peroxide in methanol for 10 min to remove endogenous peroxidase. Non-specific binding sites were blocked in 0.01 M phosphate-buffered saline (pH 7.4) containing 2% bovine serum albumin (Wako Pure Chemical, Osaka, Japan) for 40 min. They were incubated with β-catenin antibody (Mouse IgG, 1:500; BD Transduction Laboratories, San Jose, CA, USA) and ALDH1A1 antibody (Rabbit IgG, 1:500; Abcam, Cambridge, UK) overnight at 4°C. Primary antibodies were detected using a biotinylated anti-mouse and rabbit IgG (1:300; Dako, Glostrup, Denmark) for 30 min at room temperature, followed by incubation with avidin-coupled peroxidase (Vectastain ABC kit; Vector Laboratories, Burlingame, CA, USA) for 30 min. The bound complex was visualized using diaminobenzidine liquid chromogen (Sigma, St. Louis, MO, USA), and counterstained with hematoxylin. As a negative control, the primary antibody was replaced with normal serum IgG at a similar dilution.
For double-staining IHC, β-catenin and ALDH1 were used as primary antibodies. Signal amplification was performed using alkaline phosphatase-biotin complex (Vectastain), followed by a chromogenic reaction with an alkaline phosphatase substrate kit (Vecta Blue; Vector Laboratories) and diaminobenzidine.
In double staining, percentages of nuclear β-catenin-and/or ALDH1-positive cells in adenoma and adenocarcinoma were obtained by manually counting positively stained cells in 10 representative fields of adenomatous or cancerous regions, in which nuclear β-catenin translocation was most frequently observed under 400× high-power magnification.
Statistical analysis
Statistical analysis and graphical presentation were performed using the GraphPad Prism 6.0 package (Graph-Pad, San Diego, CA, USA). The Mann-Whitney U, Fisher's exact test, or chi-square test was used to test the relationship between categorical variables as appropriate. P < 0.05 was considered to indicate statistical significance. | 2018-04-03T05:12:12.899Z | 2017-04-12T00:00:00.000 | {
"year": 2017,
"sha1": "c689bdd8a62e283c49bd97a9616c0d6238697c9c",
"oa_license": "CCBY",
"oa_url": "http://www.oncotarget.com/index.php?journal=oncotarget&op=download&page=article&path[]=17051&path[]=58993",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c689bdd8a62e283c49bd97a9616c0d6238697c9c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
245976518 | pes2o/s2orc | v3-fos-license | Verifying Valve Lash in Engine Assembly Using Tolerance Interval Method for Monitoring Process Performance
: The process capability refers to the ability of a process to produce parts or provide services that are able to meet the given specification requirement. The process capability is commonly measured by process capability indices (PCIs) such as Cp and Cpk to evaluate process yield, process consistency, and statistical variation compared to the specification. Most of the component or hardware manufacturing industries require exhibiting a process capability index (Cpk) of 1.33 or above. This paper presents an alternate approach for monitoring a process to verify valve lash in engine assembly. Valve lash is the gap between the rocker arm and the valve spring in an engine and this gap must be tightly controlled due to the critical role valves play in engine performance. Achieving and maintaining process capability of valve lash of 1.33 is difficult and unrealistic due to the nature of the assembly process and multiple factors being involved. Using historical data collected from 1200 valves in 100 engines assembly, this paper illustrates the tolerance interval method for monitoring the process performance of valve lash in engine assembly is more practical and economical as compared to the conventional process capability.
Introduction
All processes show variations and the variations in manufacturing processes are due to common causes and special causes. The process capability refers to the ability of a process to produce parts or provide services that are able to meet the given specification requirement. Process capability was introduced in industry to monitor if the random or chance causes of variations as compared to the specification are at a certain level of acceptance. Process capability is evaluated when the process is under statistical control. A process that is operating with only chance causes of variation present is said to be in statistical control (Montgomery, 1997, p. 130). Cp and Cpk are the commonly defined process capability indices, to quantify if the process meets the requirements set by the designers or customers. Most of the automotive or component manufacturing industries require process capability index (Cpk) of 1.33 or higher (AIAG, 2009) (2) Where USL and LSL = upper and lower specification limit given by the designer $ = process performance average, usually estimated from the sample average = standard deviation of the process performance As defined in equation (1) and (2), Cp measures the relative ratio of given tolerance and the real process spread, but due to no process average location included it only tells the potential if the process can meet the specification. If the process average deviates too far from the target, the process may produce a lot of defective parts even when Cp is large. With Cpk considering where the process average actually is located and evaluating the process spread with respect to the given specifications, the magnitude of Cpk relative to Cp is a direct measurement of how far away it is from the target. As defined in equation (1) and (2), Cp measures the relative ratio of given tolerance and the real process spread, but due to no process average location included it only tells the potential if the process can meet the specification. If the process average deviates too far from the target, the process may produce a lot of defective parts even when Cp is large. With Cpk considering where the process average actually is located and evaluating the process spread with respect to the given specifications, the magnitude of Cpk relative to Cp is a direct measurement of how far away it is from the target.
The assumption for capability index evaluation is that a process is approximately normally distributed. If the process variation is aligned in the middle point between its specification limits, the calculated value for Cpk is equal to the calculated value for Cp. But with the process average deviating away from the specification center, the Cpk index degrades and is always smaller than Cp. Generally, a Cpk ≥1.33 indicates a process is capable of producing parts meeting specifications. Values less than 1.33 tell that the actual process variation is either too wide compared to the specification or that the location of the variation is off from the center of the specifications. It could be the combination of both spread and location measures that how far the process mean is from the nearer specification limit in terms of 3s distances. McCormick et al. (2000) mentioned that knowledge of the process mean and variance allows one to calculate with certainty where any value or range of values lies within the distribution. This is not necessarily the case for other distributions. The shapes of non-Normal distributions are usually affected by moments not expressible in terms of the mean and variance alone.
Valve Lash Control in Engine Assembly
Valves in internal combustion engines are to charge fresh air in the combustion chamber and discharge mixture of fired fuel and air out of engine. Because of the temperature change the valves experience, different materials are used in manufacturing engine valves. The engine contains rocker arms that rise and drop by the rotation of the camshaft and it is a key performance to evaluate the engine assembly quality due to valves' extreme working condition (Karasoy and Ebrinc, 2009;Magda, 2015). When a rocker arm is pushed down by the camshaft, it contacts a valve bridge, applying pressure on the valve spring and opening the valve stem at the correct time to allow pressure and gas either inside or outside of the system. This keeps the internal combustion engine at the right pressure level, preventing premature combustion and potential damage to internal components. Valve lash is the gap between the rocker arms and the valve springs, as shown in Figure 1. Due to the temperature range which engines experience, valve lash is required to react to changes from cold to hot as the metal expansion rates differ between the block, head, pushrods, valves, and other components. The valve lash must be tightly controlled, otherwise the engine will experience torque drop, noise, or excessive wear of the valve train components. If valve lash is set too loose (in bigger gap), the rocker arm acts like a hammer by pounding the valve tips, which can break the push rods. If the valve lash is too tight (in smaller gap), the valve may not close all the way, limiting engine performance.
Research has been conducted by Jafari et al. (2014) to detect valve damages in internal combustion engines using acoustic signal and artificial neural network. The result of this study could be used to identify the type of problem and its location after the engine has been assembled and put in use. In assembly process, optimum valve lash clearances are usually determined experimentally (Łukasz et al., 2015). Mechanics and manufacturers use various methods to set and verify valve lash to the best of their ability. Setting and measuring valve lash is a critical, yet not fully understood and no clear target specification for the valve lash in engine assembling is given because multiple parts mating together can easily cause tolerance stack issues. The most precise and cumbersome method for verifying valve lash would be to check each individual valve by rotating the crankshaft to the top dead center of each valve and verifying the valve lash with a feeler gauge. This time-consuming way is not ideal for a factory where the goal is to increase productivity and working efficiency.
The current method used in an engine manufacturing company to verify the valve lash for its six-cylinder diesel engine production without having to rotate the crankshaft multiple times is to measure the height of the stud protruding from the jam nut on the rocker arm assembly. If any one of the studs is out of the predetermined tolerance for the height gauge check, then further a feeler gauge is used to measure correlating valve to the failed stud(s) and the valve lash is adjusted accordingly. In this method, the upper and lower tolerance limits of the height gauge is critical and the proper limits can also catch other related failures such as valve bridge misalignment, missing valve bridges, missing rocker arm pushrods, pushrods out of place, etc. If these failures can be identified in advance prior to the engine reaching a customer, the cost on warranty claims could decrease.
When the stud height gauge method was originally established at this particular manufacturing facility, the valve lash tolerance was set as +/-0.5mm compared with a subgroup average from the six intake valves and six exhaust valves on one particular engine, which can be seen in quality check algorithm below. This tolerance (+/-0.5mm) was established from historical measurements and a mathematical stack-up of the related component dimensions and tolerances. In practice, the height gauge is measured from the top of the stud on the rocker arm assembly down to the nut, also located on the rocker arm assembly shown in Figure 2. The higher the measurement, the more the stud is protruding from the nut, thus resulting in looser valve lash. If the height is lower, the stud is tighter against the valve bridge, thus resulting in a tighter valve lash. The algorithm separates valves good from bad is as follows: Where 6 through & are the height measurement for each of the six-cylinder in one particular engine.
&
If 7 > $ + 0.5 then 1 (bad) Else 0 (good) If 7 < $ − 0.5 then 1 (bad) Else 0 (good) The same quality check is conducted both for intake and exhaust valves. The software then looks for any 1's and the corresponding valves. All rejected valves from this step must be verified with a feeler gauge ( Figure 3). The feeler gauge is a specific thickness determined by engineering depending on if it is an exhaust valve (0.66mm) or intake valve (0.31mm) and must go roughly one third of the way under the pushrod when the engine is at the top dead center. If the feeler gauge fails, the operator must loosen the nut, adjust the stud while using the feeler gauge to assess the appropriate valve lash, re-tighten the nut, and re-measure the valve lash with the feeler gauge to ensure nothing shifted.
The height gauge check is then performed again to see if the engine passes successfully. If it does, the engine is deemed good. If not, the failed valve(s) must be assessed again. If the operator and engineer are comfortable with the valve lash, regardless of the height gauge results, the engine can continue being built. With the current tolerance (+/-0.5mm) used for the stud height gauge check, roughly 25% of engines initially fail at least one valve in the first step. However, when checked with the feeler gauge, 95% of the failed valves are fine and do not actually need any adjustment. Ideally, the tolerance setup related to the height gauge should only catch outliers indicating a potential issue with the engine such as a valve bridge missing or not in place. Obviously, the current tolerance (+/-0.5mm) reported a lot of false errors, therefore this tolerance limits must be revised.
Statistical Tolerance Interval, an Alternate Approach
A statistical tolerance interval is an interval that one can claim to contain at least a specified proportion (p) of the distribution with a specified degree of confidence, 100 (1-a) %. Such an interval would be of particular interest in setting limits on the process capability of a product manufactured in large quantities (Hahn et al., 2017). According to international standard a statistical tolerance interval is an estimate interval, based on a sample, which can be asserted with confidence level (1-a), for example 0.95, to contain at least specified proportion p of the items in the populations (ISO, 2014). This international standard discusses two methods for determining the statistical tolerance intervals, a parametric method for the case where the characteristic being studied has normal distribution and a distribution-free method for the case where nothing is known about the distribution except that it is continuous.
Before using tolerance interval that depends heavily on the normality assumption, one should assess the adequacy of the normal distribution as a model, (Hahn and Meeker, 1991). As reported by Hahn and Meeker (1991), two-sided distribution-free statistical intervals from a random sample from a specified population one generally proceeds steps as follows: • Specify the desired confidence level for the interval; • Determine (from tabulations and calculations) the order statistics (ordered observations from smallest to largest) that provide the statistical interval with at least the desired confidence level for the sample size; • Use the selected order statistics as the end points of the distribution-free interval.
Tolerance interval method is useful in one sided tolerance (upper bound or lower bound) where process centering (targeting) is an issue and is less important than feature's conformance. Also, non-parametric (distribution-free) tolerance intervals methods can be used if the measured data do not follow normal distribution to assess the conforming proportions (McCormack et al., 2009). For the valve lash in the engine assembly, two-side intervals should be constructed as no single minimum or maximum value are specified.
Statistical Tolerance Interval Procedures
There are several procedures available in text to construct the statistical tolerance intervals depending upon the requirements. This paper will discuss the following statistical tolerance interval procedures.
Two-Sided Tolerance Interval when the Population is Normal Distribution with Known Mean (µ) and Standard Deviation (s)
The international standard (ISO, 2014) provides procedure when the values of the mean (µ), and the standard deviation (s), of a normal distribution are known and estimated, the distribution of the characteristic under investigation is fully determined. The two-sided tolerance interval to contain at least a specified proportion p of population, for p=0.90, 0.95, 0.99, is ± !
In this equation, µ is the population mean, s is the population standard deviation, and ! is the p-fractile of the standardized normal distribution, which can be found in an engineering handbook.
Tolerance Interval when the Population ss Normal Distribution with Known Mean and Unknown Standard Deviation
The The distribution of 9 / 9 is 9 /( − 1) with n-1 degrees of freedom. Thus, In practice, in vast majority cases sample means and sample standard deviations (instead of population mean and population standard deviation) can be estimated from collected data. The upper and lower limits for a two-sided tolerance interval are: Where k is called the factor < 9 ( − 1) is the a-fractile of the 9 (Chi-Square) distribution with n-1 degrees of freedom,
The Distribution-Free Statistical Tolerance Interval for any Type of Distribution
For establishing distribution-free tolerance intervals the order statistics (of sample is determined by solving binomial distribution function for smallest sample size n. International standard (ISO, 2014) describes the procedure for constructing distribution-free or for any type of distribution statistical tolerance bounds. Assume we have samples, 6 , 9 , 4, … . ; , of independent random observations on a population (continuous, discrete, or mixed) and let its order statistics (ordered observations from smallest to largest are called ordered statistics) be (6) ≤ (9) ≤ (4) ≤ ⋯ ≤ ; .
The interval with 100(1-a) % confidence that at least 100 P % of the population lie between the v th smallest observation and w th largest observation is: Where ≥ 0, ≥ 0, + ≥ 1, 0 ≤ ≤ 1 Hanson and Owen (1963) discussed that continuity requirement on cumulative distribution function (c.d.f.) is unnecessary. They noted that many c.d.f.'s which occur in practice are not continuous, and in many cases where distribution-free tolerance limits are applicable they are not being used because of an uncertainty as to whether the underlying distribution is or is not continuous or because of the certainty that it is not (Hanson and Owen, 1963). So, when the c.d.f. of the population characteristic X is not continuous, the statement will be slightly modified such that there is at least 100(1-a) % confidence that at least 100P% of the population is between x(v) and x(n-w+1) or equal to x(v) and x(n-w+1). When v + w = 1, formula (7) reduces to: The equation (10) shows the minimum sample size required for certain confidence level and proportion to be in conformance. Thus, to be 95% confident that at least 95% percent population lies below the largest value of the sample, the sample size must be n = log (1-0.95)/log (0.95) = (-1.3) / (-0.022) = 59. There are several tables available which gives minimum samples required for constructing statistical interval at certain proportion p and certain confidence level (1a)%. Hahn and Meeker (1991) have compiled and listed tables based on the selection of sample size n, confidence levels 100(1-a)% and at various levels of conformance proportion percentage P. With the help of statistical analysis software like Minitab® it is easily possible to calculate and construct distribution based or distribution-free statistical intervals.
Data Collection and Variable Designation
Ideally, because research findings based on an entire population are more accurate, this study would contain measurements from every engine over the past year or more to get total historical data. Unfortunately, collecting the population data is unrealistic. Thus, a smaller data set, or a sample, is being used to reflect the entire population accurately (Godwill, 2015). A sample size needs to be considered with great care. As explained by Hahn and Meeker (1991), with the increase of the sample size, the computed tolerance interval will approach the probability interval that actually contains the specified population proportion. Smaller sample size, however, result in a tolerance interval that is much bigger than this limiting probability interval. Referring to the sample size table in Hahn and Meeker (1991), the sample size should be large enough such that (1) with 95% confidence level at least 99% (p') of the population will be included within the tolerance interval and (2) with 1% probability that at least 99.7% (p*) of the population will be included (note: p* is larger than p'). The sample size selected according to the above criteria should be 390. The case study presented here contains 600 valve lash data from engine intake and exhaust valves, respectively, in total twelve hundred measurements.
These engines were tested using the height gauge during a one-week time frame and were all gauged by the same operator to ensure consistency in gauging method. These engines were not necessarily built consecutively due to two possible reasons: 1) a different operator gauged a portion of the engines and those results were not included, or 2) engines with valves that failed the stud height gauge check and needed the valve lash adjusted were not included. Of the one hundred engines included in this data, 79 engines passed the height gauge check algorithm on the first gauging attempt. The remaining 21 engines failed on one or more of the valves according to the tolerance (±0.5mm) setup in the height gauge software. Of the 21 engines that failed, a feeler gauge was used to measure the valve lash, and the valves were all deemed acceptable and not adjusted. This means all 100 engines are good; therefore a 100% acceptance. These failures of the 21 engines initially failed the height gauge check are important to be included to determine if the tolerance of +/-0.5mm should be increased.
As mentioned earlier, there are in total 100 engines with six intake valves and six exhaust valves on each engine, therefore in total 600 individual intake valve lash data and 600 individual exhaust valve lash data, respectively. The variable designations are as follows: X = any of the 600 intake valve lash values Y = any of the 600 exhaust valve lash values, $ = the average value of the six intake valve lash values taken from one engine.
Exam Data Normality of Intake and Exhaust Lash Value
First check the normality of individual intake/exhaust valve lash values and the normality of intake/exhaust valve lash subgroup averages using the Chi-square test. The normality plot shown in Figure 4 and Figure 5 indicates that both of the intake and exhaust valve lash individual data (X and Y) are normally distributed because the Chi-square test probability is larger than 0.05. The same normality test was conducted with the subgroup averages of the six intake valve lash data and the six exhaust valve lash data. The Chi-square test probability in Figure 6 and Figure 7 indicates that both of the subgroup averages follow normal distribution because the Chi-square test probability is larger than 0.05. However, the normality test found that ( − $ 7 ) and ( − $ 7 ) are not normally distributed for both the intake and exhaust valve lash data, as shown in Figure 8 and Figure 9. The basic statistic and normality test results are summarized in Table 1.
Tolerance Interval
Running analysis at 95% percent confidence (1-a) for intake valve lash values and exhaust valve lash values estimated the 99% of population (p%) tolerance interval with Minitab ® . The tolerance intervals of X, Y, ( − $ 7 ) and ( − $ 7 ) are summarized in Table 2 for intake valve and exhaust valve lash values. With the normal distribution assumption, the intake valve lash tolerance interval is (6.914, 8.466), which is ±0.776 from the average; the exhaust valve lash tolerance interval is (6.056, 7.971), which is ±0.958 from the average. Both of the two tolerance intervals are larger than the current testing allowance ±0.5mm.
Also, according to non-normal distribution using 95% confidence level, the 99% tolerance interval for ( − $ 7 ) and ( − $ 7 ) achieved confidence 98.3%, which are (-0.812, 0.612) and (-0.644, 0.717), respectively, for the intake valve and exhaust valve. These two tolerance intervals are also larger than the current testing allowance ±0.5mm. The tolerance calculations above along with the multiple false failure signals indicate the current testing allowance is not appropriate for inspecting the valve lash value and the false signal of nonconforming valve lash and caused a lot of unnecessary inspection waste. There was no definite target value for the stud height but only a tolerance range currently given for individual stud height compared with average stud height with respect to checking the intake valve lash and exhaust valve lash. The ideal situation would be all the lash values of six intake valves on one engine are exactly same, then the target difference of valve lash from average should be zero. Therefore, the capability indices Cp and Cpk are calculated by using the exiting ±0.5mm tolerance and tolerance intervals and their calculation results are shown in Table 3. Comparing the process capability evaluation, the tolerance interval method which more reflected the reality is more practical and economical to monitor the engine valve lash in the assembly process without sacrificing much for cost. The tolerance interval method provides greater than 95% confidence that 99% population of the valve assemblies are conforming to the specification.
Conclusion and Recommendation
This case study reviews the historical data of intake valve lash and exhaust valve lash collected from 100 engines assembled in a manufacturing facility. Above analysis indicates that the conventional capability measurement approach is inappropriate and can create false alarm instead. The tolerance interval method shows an advantage and provides ≥95% confidence that more than 98% population meets the tolerance specification which is more economic method.
In the data analysis the normality test indicates that individual X, Y and subgroup average $ , $ are all normally distributed, however, ( − $ 7 ) and ( − $ 7 ) are not normally distributed. This might be due to the data collection process. The data collected using convenient sampling method were within one-week period by one operator. If data collected from a random manor may provide insights that are more valuable. Companies can formulate randomization data collection schemes to keep recording and monitoring the process quality characteristic. If enough data were collected over a longer period, using tolerance interval would become an economical way of monitoring the process performance with a required degree of confidence. It can also be seen from this study that even with enlarged tolerance interval currently the process capability indices Cp ≥1, but Cpk ≤1, which clearly indicated that the quality characteristic has not reached the industry goal such as Cp and CPK ≥ 1.33. More efforts should be made to reduce the process variation. | 2022-01-16T16:23:02.272Z | 2022-01-01T00:00:00.000 | {
"year": 2022,
"sha1": "59673f1d3c57e7b9c085c27b26219a6683f3f22d",
"oa_license": "CCBY",
"oa_url": "https://researchleap.com/wp-content/uploads/2022/01/02_Verifying-Valve-Lash-in-Engine-Assembly-.pdf",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "cf6edeb5cf52289a0dc44a69b1ab6f9193ce728e",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": []
} |
233389874 | pes2o/s2orc | v3-fos-license | Sirtuin 1 and Skin: Implications in Intrinsic and Extrinsic Aging—A Systematic Review
Skin, as the outermost organ of the body, is constantly exposed to both intrinsic and extrinsic causative factors of aging. Intrinsic aging is related to compromised cellular proliferative capacity, and may be accelerated by harmful environmental influences with the greatest significance of ultraviolet radiation exposure, contributing not only to premature aging, but also to skin carcinogenesis. The overall skin cancer burden and steadily increasing global antiaging market provide an incentive for searching novel targets to improve skin resistance against external injury. Sirtuin 1, initially linked to extension of yeast and rodent lifespan, plays a key role in epigenetic modification of proteins, histones, and chromatin by which regulates the expression of genes implicated in the oxidative stress response and apoptosis. The spectrum of cellular pathways regulated by sirtuin 1 suggests its beneficial impact on skin aging. However, the data on its role in carcinogenesis remains controversial. The aim of this review was to discuss the relevance of sirtuin 1 in skin aging, in the context of intrinsic factors, related to genetic premature aging syndromes, as well as extrinsic modifiable ones, with the assessment of its future application. PubMed were searched from inception to 4 January 2021 for relevant papers with further search carried out on ClinicalTrials.gov. The systematic review included 46 eligible original articles. The evidence from numerous studies proves sirtuin 1 significance in both chronological and premature aging as well as its dual role in cancer development. Several botanical compounds hold the potential to improve skin aging symptoms.
Introduction
Aging is a continuous inherent process of gradual decline in bodily functions, and physical and mental capacity, accompanied by decreased fertility and increased susceptibility to diseases, eventually leading to death. Although the process is genetically determined, various internal and external stimuli can modify its course [1]. Thus, aging processes could be categorized as: intrinsic related to genetic, metabolic, and hormonal factors; or extrinsic, referring to external agents associated with environment and lifestyle. In most cases, extrinsic factors are modifiable, while intrinsic aging remains almost unalterable.
Skin, as the outermost organ of the body, is constantly exposed to hazardous chemical and physical agents, contributing to accelerated aging and cancer development. Aging of the world's population and rising emphasis on maintaining physical appearance results in increased demand for antiaging products, devices, and medical procedures [2]. On the other hand, in 2018 almost 300,000 people were diagnosed with melanoma, while the incidence of non-melanoma skin cancers was over 1 million cases [3]. The global public viewed.
Results
The search resulted in the retrieval of 589 records of which 545 were screened relevance and 46 ultimately included in qualitative synthesis. ClinicalTrials.gov identifi seven interventional studies, however, after screening none of them were eligible for t review. The flowchart presenting full search strategy is demonstrated in Figure 1. Tables 1-4. We have identified 37 experimental and 16 human observational stu ies of which seven were simultaneously conducted in vitro and in humans. Therefore, t assessment of risk of bias and applicability concerns were non-essential. The basic characteristics of the original papers on the researched topic are summarized in Tables 1-4. We have identified 37 experimental and 16 human observational studies of which seven were simultaneously conducted in vitro and in humans. Therefore, the assessment of risk of bias and applicability concerns were non-essential. Table 1. Summary of the studies on SIRT1 relevance in intrinsic aging.
Author
Year Population Key Observation
Intrinsic Aging-Basic Information
Sommer et al. [16] 2006 ∆Np63α transgenic mice Normal lung epithelial cells transfected with a vector containing∆Np63α transgenic mice exhibited an accelerated aging phenotype in the skin accompanied by a decrease in longevity correlated with levels of SIRT1. In cell culture beta-galactosidase accumulation and typical senescent morphology was rescued by SIRT1.
Yang et al. [17] 2011 HDFs, Hs68 cell culture exposed to 2-DG and DHEA 2-DG, but not DHEA, at non-cytotoxic concentrations extends lifespan in parallel with increased intracellular NAD + levels and SIRT1 activities Kalfalah et al. [18] 2014 Skin biopsies of females aged 20-67 Age-dependent decrease in SIRT1 Kim et al. [19] 2015 Passaged HDFs culture SIRT 1 is down regulated by increasing passage of HDFS.
Lee et al. [20] 2016 HaCaT keratinocytes Melatonin-induced autophagy play a protective role through SIRT1 pathway against skin cell damage as a result of hydrogen peroxide-induced cell death.
Golubtsova et al. [21] 2017 Skin biopsies retrieved from deceased donors: fetuses at pregnant age 20-40 week, people from birth to 85 years old Age-related decrease in sirtuin 1 content in HDFs is correlated with age-dependent decrease in fibroblasts proliferation. The highest level of SIRT1 is found between 20-to 40-week of pregnancy.
Systemic Sclerosis
Wei et al. [23] 2015 Skin biopsy samples of healthy adults and patients with SSc. Sirt1 −/− and wild-type mouse embryonic fibroblasts Reduced SIRT1 expression and protein level in SSc skin biopsy samples compared to healthy. Activation of Sirt1 attenuated fibrosis, while inhibition had profibrotic effects.
Zerr et al. [24] 2016 Skin biopsies of patients with SSc and healthy volunteers Sirt1 −/− and wild-type mice SIRT1 is decreased in TGF-β-dependent manner in patients with SSc and in experimental fibrosis. SIRT1 activation enhances the profibrotic effects of TGF-β with increased Smad reporter activity, elevated transcription of TGF-β target genes and raised release of collagen. Sirt1 KO inhibited TGF-β/SMAD signaling and reduced release of collagen in fibroblasts. Sirt1 −/− mice were less susceptible to fibrosis
Lim et al. [37] 2020 HaCaT, HDFs and B16F10 cells exposed to UVB irradiation AL has an antiwrinkle activity in damaged skin and can inhibit melanogenesis, regulates baseline MMP expression and induces collagen production in HDFs. AL inhibits elastase and MMP-1 and induces type 1 procollagen. AL increased the expression of SIRT1
Calapre et al. [41] 2017 ex vivo skin models, taken from non-sun exposed skin of healthy donors and NHEKs exposed to UVB plus heat SIRT1 mediates UVB plus heat induced survival of DNA damaged keratinocytes by: decrease in p-53 acetylation and downregulation of its downstream pathways, including: BAX, ERCC1, XPC Increase in Ki67.
Lei et al. [42] 2018 Primary HDFs obtained from foreskins of healthy human donors aged 5-20 years exposed to UVA irradiation Fluorofenidon alleviates HDFs senescence by inhibiting the mTOR and increasing SIRT1 Ding et al. [43] 2018 Epidermis isolated from skin biopsies obtained from the outer forearm and the buttock of healthy females Lower expression of HDAC1 and SIRT1 in sun-exposed skin compared with matched non-exposed skin Li et al. [44] 2020 HaCaT exposed to UVA irradiation α-l-Hexaguluroic acid hexasodium salt (G6) increases mitochondrial metabolism, alleviates oxidative stress, reverses the downregulation of SIRT1 and pGC-1a expression levels
UV-Related Carcinogenesis
Hida et al. [45] 2007 Immunohistochemical staining for SIRT1 expression in 87 cases of skin tumors and 20 normal skin samples. Sun-exposed and sun-protected skin regions did not differ in SIRT1 expression. SIRT1 was overexpressed in all samples of AK, BD, SCC and BCC. SIRT1 overexpression may have some relevance to the early stage of skin carcinogenesis.
Sirt1 cHET promotes UVB-induced skin tumorigenesis, whereas cKO Sirt1 suppresses skin tumor development but sensitizes the mice to chronic solar injury. In mouse skin, Sirt1 is haploinsufficient for UVB-induced DNA damage repair. SIRT1 is downregulated in parallel with XPC in human SCC. cKO deletion of Sirt1 augments p53 acetylation and sensitizes the epidermis to UVB-induced apoptosis in vivo, while heterozygous has no such effect. UVB induced tumor formation in Sirt1 WT and Sirt1 cHet mice but not in Sirt1 cKO mice.
Brandl et al. [47] 2019 human BCC human and murine normal skin Kim [57] 2016 UVB-exposed human keratinocytes, HaCaT cells Garlic Garlic pretreatment attenuated in a dose-dependent manner the production of ROS, proinflammatory cytokines and MMP-1 protein expressions. SA-β-gal and SIRT1 activity were ameliorated by garlic treatment.
Wahedi et al. [58] 2016 UVB-irradiated HaCaT and HDFs Juglone Juglone restored the expression of SIRT1 and Pin1 in almost dose dependent manner in irradiated cells. Juglone treatment upregulated SIRT1 in unirradiated skin cells.
Discussion
Since the discovery that silent information regulator 2 (Sir2) mediates the effect of caloric restriction on lifespan extension in yeast, many attempts have been made to establish if it translates to mammals [6,62]. Extensive research on sirtuin 1, the closest structural Sir2 homologue, suggest that it may extends human lifespan indirectly by the attenuation of age-related and metabolic diseases [63][64][65][66][67]. Structural and functional skin changes are the most apparent feature of human aging, and the molecular mechanisms underlying skin aging resembles that seen in internal organs. Histologically aged skin is characterized by epidermal and subcutaneous atrophy, flattened dermal-epidermal junction, degeneration of collagen and elastic fibers, and reduction of skin vasculature [68]. Other hallmarks of skin aging include wrinkles, roughness, decreased elasticity, and hyperpigmentation.
At the cellular level, senescence is a process of gradual decrease in proliferative capacity, eventually leading to irreversible arrest at the G1 phase of the cell cycle and altered biosynthetic phenotype of the senescent cell (SASP-senescence-associated secretory phenotype) [69]. Three major mechanisms cooperating together are involved in the process of cellular aging: telomere shortening, insufficient DNA repair system, and accumulation of cellular damage in the form of byproducts generated in each and every cellular process. The term replicative senescence is related to shortening of the telomere length, which limits the number of cell divisions. On the other hand, stress-induced premature senescence occurs in response to DNA and other macromolecules damage caused by oxidative stress, oncogene activity, or chemical substances. Finally, the imperfectness of all biological processes leads to the accumulation of damage in postmitotic cells [70]. All the processes together subsequently contribute to genomic instability, triggering switch to SASP and cessation of cell division. Serially passaged cells show signs of cellular aging [71].
Cutaneous Sirtuin 1 Expression Decreases with Age
First reports on sirtuin 1 relevance in skin aging dates back to 2006 [16,72]. SIRT1 is expressed in lower epidermis, upper dermis, and dermal-epidermal junction [72]. Sommer et al. [16], conducted a study on transgenic mice constitutively overexpressing ∆Np63 isoform, engaged in keratinocytes differentiation. The authors observed two main points: physical interaction between sirtuin 1 and ∆Np63, and correlation between its expression and senescent clinical phenotype of the rodents. The results suggest involvement of sirtuin 1 pathway in ∆Np63-induced aging model. Further studies on cell cultures and human tissues confirmed age-dependent downregulation of SIRT1 gene expression [18,19,21]. Serial passage of human dermal fibroblasts, which is a model of replicative senescence, revealed decrease in extracellular matrix components accompanied by SIRT1 reduction [19]. Another study on SIRT1 expression in human skin biopsies, retrieved from females aged 20-67, revealed age-dependent decrease in SIRT1 in dermal fibroblasts [18]. In line with this, Golubstova et al. [21]. performed an analysis of tissue specimens from human fetuses and people aged between 20-week gestation and 85 years. The highest SIRT1 expression was found in fetal skin specimens. There were two sharp declines in SIRT1 expression across all studied population. First, in people under 20 years old compared to fetuses and second after the age of 40. Sirtuin 1 expression correlates with total fibroblast amount, and its proliferative activity and may be responsible for the development of skin aging symptoms. Aged fibroblasts lose metabolic and replicative activity leading to imbalanced turnover of extracellular matrix with decreased collagen, elastin, and hyaluronic acid content [73]. Aged skin is characterized by disrupted epidermal barrier function, partially due to decreased filaggrin expression, and filaggrin was shown to be expressed in SIRT1-dependent manner [22,74]. In the future, SIRT1 may be used as a marker of function of active fibroblasts and to verify effectiveness of active ingredients of antiaging cosmetic products.
Premature Aging Related to Genetic Background
Extensive research on progeroid syndromes shed light on the molecular basis of aging processes. Progeroid syndromes are a group of clinically and genetically diversified disorders characterized by premature aging in more than one organ or tissue, and/or prominent susceptibility to harmful effects of ultraviolet radiation (UVR), resulting in increased inci-Cells 2021, 10, 813 9 of 21 dence of skin cancers [75]. We can distinguish progeroid syndromes related to alterations in the nuclear architecture, and resulting from the mutations in DNA repair system.
Hutchinson-Gilford progeria syndrome (HGPS) is a rare fatal early-onset premature aging syndrome, affecting approximately 1 in 20 million children worldwide [76]. The underlying cause of the disease is a mutation-mediated alternative splicing of mRNAs issued from the LMNA gene, leading to the production of truncated lamin A precursors termed progerin [77]. Lamin A is a component of the nuclear matrix, and by regulation of the chromatin structure, is involved in replication, transcription, and DNA repair. The acetylation of histones and transcriptional factors acetylation is implicated in the regulation of gene expression. Given the fact that sirtuin 1 possesses histone deacetylase activity and progerin expression was found to raise during normal aging, gave researchers an impulse to investigate the interactions between those proteins [78]. Indeed, lamin A had been identified as a direct potent activator of SIRT1 [33]. In contrast, SIRT1 association with progerin was considerable reduced [33]. Interestingly, the study brought evidence that resveratrol, beyond direct SIRT1 activation, enhances lamin's A stimulating effect. Moreover, mice harboring the LMNA mutations treated with resveratrol for four months had significantly increased survival, and displayed reversal of the progeroid phenotype [33]. Of note, in November 2020, the FDA approved the first drug for HGSP-lonafarnib-an inhibitor of farnesyltransferase which prevent the buildup of truncated protein [79].
Werner syndrome, also called adulthood-progeria, is caused by autosomal recessive mutations in WRN encoding helicase engaged in DNA repair [80]. Clinically it is characterized by the absence of the growth spurt seen early in puberty. In the further course of the disease several others hallmarks of aging develops, including accelerated atherosclerosis which is the leading cause of death in those patients. Small experimental study demonstrated involvement of sirtuin 1 in transcriptional regulation of WRN [26]. Sirtuin 1-dependent decline in WRN protein activity contributes to impaired DNA repair system during aging.
The efficiency of DNA repair system decreases with age [81], possibly due to reduced constitutive protein levels, contributing to chronical aging. Apart from that, mutations in the components of the repair system lead to multi-systemic premature aging with increased risk for cancer, and in some cases neurodegeneration [77,80,82]. It was proposed that neurogenerative phenotype is related specifically to mitochondrial dysfunction [28,32,33]. Noteworthy, the activation of SIRT1 by poly-ADP-ribose polymerase-1 (PARP1) inhibitors or NAD + precursors were able to rescue mitochondrial defects, attenuate neurodegeneration, and extend lifespan in experimental models [28,32,33]. Moreover, a literature review revealed implication of sirtuin 1 in response to DNA damage by interaction with nucleotide excision repair (NER) system. NER system consists of two subpathways: global genomic (GG-NER) and transcription-coupled repair (TC-NER) [83]. The system is essential for repair damage related to chemical and physical agents, especially UV radiation. Direct absorption of UV light induces the formation of covalent bonds between DNA strands that distorts the helical structure [84]. The process of DNA repair occurs in sequential steps, beginning with the recognition of the site of helical distortion, followed by dual incision, repair synthesis, and ligation [83]. The first step is dependent on the XPA protein, which beyond recognition, is responsible for binding of the DNA excision repair 1 protein complex (ERCC1), and its recruitment to the DNA damage site [83]. The protein is regulated by posttranslational modifications by checkpoint kinase ATR or SIRT1 deacetylase under UV exposure [27,85]. SIRT1 augments NER pathway through XPA deacetylation, and promote interaction between XPA and ATR [27,29]. Transcription of XPC protein, a component of the preincision complex, was demonstrated to be regulated in SIRT1-dependent manner [30]. The disturbed interplay between sirtuin 1 and NER pathway is not only one of the mechanisms underlying cell vulnerability in genetic defects of the NER system, but probably also participates in normal aging and skin carcinogenesis.
Progressive Tissue Fibrosis as a Hallmark of Aging
Several recent studies gave a new perspective on the pathophysiology of systemic sclerosis (SSc) which determines its perception as a premature aging syndrome. The main features of the disease involve fibrosis, vascular abnormalities, and dysregulation of the immune system [86]. TGFβ signaling plays a crucial role in the development of internal organ fibrosis, a hallmark of aging, and may be attenuated by SIRT1 activation in kidney and heart [87,88]. In line with this, recent findings confirmed the significance of sirtuin 1 in TGFβ-related fibrosis in SSc [23][24][25]. SIRT1 expression and protein level in skin samples obtained from patients with SSc and animal models were considerably decreased compared to healthy controls. As genetic and pharmacologic activation of SIRT1 substantially reduced skin fibrosis, it has the potential to introduce novel therapies in SSc and morphea.
Sirtuin 1 Affects Ultraviolet Irradiation-Related Skin Changes by Numerous Targets
Chronological aging is eminently influenced by several environmental factors that according to Krutmann et al. [89] may be divided into following major categories: sun radiation, air pollution, tobacco smoke, stress, nutrition, lack of sleep, and temperature. It is estimated that up to 90% of symptoms of premature facial aging is related to cumulative exposure to sunlight [90]. A recent study investigating the epidermal expression of SIRT1 in samples obtained from healthy volunteers revealed significantly reduced expression in sun-exposed area compared to non-exposed skin [43]. The results suggest relevance of epigenetic regulation of gene expression in the pathogenesis of photoaging.
Photodamage is attributed to UVB (280-315 nm), UVA (315-400 nm), visible light (400-740 nm), and infrared radiation (740 nm-1 mm), with deeper skin penetration with the increasing wavelength [89]. UV radiation acts on cells through several mechanisms that interfere with each other and aggravate detrimental effects. The mechanisms involve release of proinflammatory cytokines, upregulation of collagenases, direct DNA oxidation, and indirect modifications of DNA, proteins, and lipids related to ROS production [91] ( Figure 2). be regulated in SIRT1-dependent manner [30]. The disturbed interplay between sirtuin 1 and NER pathway is not only one of the mechanisms underlying cell vulnerability in genetic defects of the NER system, but probably also participates in normal aging and skin carcinogenesis.
Progressive Tissue Fibrosis as a Hallmark of Aging
Several recent studies gave a new perspective on the pathophysiology of systemic sclerosis (SSc) which determines its perception as a premature aging syndrome. The main features of the disease involve fibrosis, vascular abnormalities, and dysregulation of the immune system [86]. TGFβ signaling plays a crucial role in the development of internal organ fibrosis, a hallmark of aging, and may be attenuated by SIRT1 activation in kidney and heart [87,88]. In line with this, recent findings confirmed the significance of sirtuin 1 in TGFβ-related fibrosis in SSc [23][24][25]. SIRT1 expression and protein level in skin samples obtained from patients with SSc and animal models were considerably decreased compared to healthy controls. As genetic and pharmacologic activation of SIRT1 substantially reduced skin fibrosis, it has the potential to introduce novel therapies in SSc and morphea.
Sirtuin 1 Affects Ultraviolet Irradiation-Related Skin Changes by Numerous Targets
Chronological aging is eminently influenced by several environmental factors that according to Krutmann et al. [89] may be divided into following major categories: sun radiation, air pollution, tobacco smoke, stress, nutrition, lack of sleep, and temperature. It is estimated that up to 90% of symptoms of premature facial aging is related to cumulative exposure to sunlight [90]. A recent study investigating the epidermal expression of SIRT1 in samples obtained from healthy volunteers revealed significantly reduced expression in sun-exposed area compared to non-exposed skin [43]. The results suggest relevance of epigenetic regulation of gene expression in the pathogenesis of photoaging.
Photodamage is attributed to UVB (280-315 nm), UVA (315-400 nm), visible light (400-740 nm), and infrared radiation (740 nm-1 mm), with deeper skin penetration with the increasing wavelength [89]. UV radiation acts on cells through several mechanisms that interfere with each other and aggravate detrimental effects. The mechanisms involve release of proinflammatory cytokines, upregulation of collagenases, direct DNA oxidation, and indirect modifications of DNA, proteins, and lipids related to ROS production [91] (Figure 2). Photoaging is commonly attributed to the UVA rays which penetrate to the dermis and are primarily responsible for alternations in human dermal fibroblasts and extracellular matrix. Although photoaging affects all three layers of the skin tissue, UV-related changes in the epidermis seems to be secondary to those in the dermal compartment [89]. Extracellular matrix (ECM) constitutes a highly dynamic tridimensional structure, filling intercellular spaces, built up of structural and functional proteins, glycoproteins and proteoglycans. It undergoes constant reconstruction, exhibiting significant alterations with the passage of time [92]. Decreased content of collagen, associated with aging, is a result of the imbalance between production and proteolytic degradation of the fibers [93]. Fragmentation and disorganization of the dermal ECM attenuates interaction between fibroblasts and collagen I [94]. The deprivation is the direct cause of the acquisition of senescent phenotype by fibroblast and may be in a vast majority reversed by enhancing structural support of the ECM. The theory is supported by the observation on skin biopsies retrieved from healthy individuals ≥70 years of age, treated with cross-linked hyaluronic acid injection [95]. The procedure stimulated fibroblast proliferation, expanded vasculature, and increased epidermal thickness. It highlights the pivotal role of the ECM turnover in the maintenance of skin tissue function.
Matrix metalloproteinases (MMPs) represent a group of 23 proteolytic enzymes that play a key role in physiologic and pathologic processes in human body [96]. They are essential for proper angiogenesis, apoptosis, wound healing, and scar formation. On the other hand, they facilitate tumor cell invasion and metastasis, or contribute to photoaging [73,96]. Of the 19 MMPs presented in human skin, only three were shown to be induced in response to UV radiation: MMP-1, MMP-3, and MMP-9 of which MMP-1 is responsible for initiation of collagen degradation [97]. The source of MMPs in the skin can be both epidermal keratinocytes and dermal fibroblasts [73]. The UV-related damage to the collagen and elastin ECM fibers is the cause of wrinkles as well as laxity and loss of skin firmness [73]. A study conducted by Ohuguchi et al. [35] demonstrated regulation of basal and stimulated expression of MMP-1 and MMP-3 by sirtuin 1 which proves its impact on intrinsic and extrinsic aging. Recent findings from experimental study on human keratinocytes and dermal fibroblasts revealed the ability of Lactobacillus acidophilus (KCCM12625P) to induce mRNA expression of SIRT1 in parallel with decline in MMP-1 and elastase expression [37]. Similar results were obtained with the use of pyrroloquinoline quinone (PQQ), a potent antioxidant, known as longevity vitamin and activator of sirtuin 1 and 3 [39]. Specifically, UVA-irradiated HDFs preconditioned with PQQ exhibited reduced expression of β-galactosidase and MMPs (1 and 3).
Sirtuin 1 overexpression was found to enhance oxidative stress resistance via FOXO3αpromoted upregulation of the ROS superoxide dismutase 2 (SOD2) and catalase [40]. Experimental evidence showed considerably decreased sirtuin 1 in human dermal fibroblasts exposed to hydrogen peroxide [36,38]. Consistently with previous studies on other tissues, SIRT1 activators were able to inhibit apoptosis, and increase cell survival [36,38,98]. Taniguchi et al. [36] compared pretreatment with ascorbic acid and its stable derivative 2-O-α-glucopyranosyl-l-ascorbic acid (AA-2G) with regard to its protective effect against cellular damage and senescence [36]. Only AA-2G prevented the decrease in SIRT1 expression under cellular stress which rationalize the widespread use of stable vitamin C derivatives in the cosmetic industry. However, it must be noticed that in the described research the ascorbic acid also was able to retain fibroblasts proliferative capacity, but its effect was short-lasting and insufficient to obtain the desired results in clinical settings.
Recently, considerable effort has been put into exploring molecular mechanisms underlying the effect of SIRT1 on the UV-induced alterations in skin. Beyond increased oxidative stress resistance and enhanced NER repair system, sirtuin 1 reduces transcriptional activity of p53 and p16 proteins and its downstream pathways [40][41][42]44]. While generally cell survival is advantageous in terms of skin aging, chronic accumulation of damage and attenuation of p53-related cell cycle arrest may contribute to the survival of DNA damaged keratinocytes, and thus, skin carcinogenesis. UVB radiation is almost completely absorbed by the epidermis, and triggers generation of reactive oxygen species [99]. Epidermal SIRT1 expression is downregulated upon UVB exposure, and consequently the acetylated p53 protein activates apoptotic cell death [100]. Sublethal chronic exposure to UVB is insufficient to initiate apoptosis, but induces keratinocytes differentiation [101]. On the one hand, it promotes cancer cell survival, but on the other, is responsible for therapeutic effect of phototherapy on abnormal hyperproliferation of keratinocytes seen in psoriasis. Under natural conditions, the skin is exposed to UVR and high temperature at the same time, but those factors act oppositely on sirtuin 1 expression [100,102]. Several studies have shown that heat stress promotes skin carcinogenesis [102][103][104]. The issue of subsequent action of UVB and heat on skin cancer was raised in the study conducted by Calapre et al. [41]. The findings from the study are directly in line with previous findings on regulation of SIRT1 expression by UVR and high temperature [38,105]. Their experiment also found clear support for the SIRT1-mediated cell survival under heat, and concomitant heat and UVB exposure. In the presence of SIRT1 inhibitor the level of acetylated p53 protein, as well as apoptosis, were substantially elevated. To sum up, the posttranslational modification of p53 plays a role in UVB and heat induced cell survival and carcinogenesis.
Based on human cancer-associated data, previously reported in the literature, SIRT1 serves as a tumor promoter [45]. Immunohistochemical analysis of sirtuin 1 expression in samples obtained from non-melanoma skin cancers and actinic keratosis have shown its overexpression in comparison to normal skin and benign tumors [45]. However, the SIRT1 gene dosage also have a prominent impact on skin response to UVR [46]. The wild type and heterozygous for Sirt1 mice are prone to skin cancer development due to reduced DNA repair and cell survival, whereas Sirt1-homodeficiency increases apoptosis signaling and decreases tumorigenesis, but sensitizes skin to solar injury [46]. Furthermore, SIRT1, through MMP2 regulation, is implicated in tumor metastasis [50,56]. Sirtuin 1 biological role in cancer is complex and depends on its targets in specific signaling pathways.
Sirtuin 1 as a Promising 'Weapon' against Pollutant-Related Premature Aging
According to World Health Organization, 91% of the world's population live in places where air quality exceeds guideline limits for major pollutants [106]. Advances in understanding air pollution on human health prompted global public health institutions to raise awareness of governments and individuals about environmental protection. The skin is directly exposed to air pollutants which may be smoothly absorbed to subcutaneous tissue, contributing to aggravation of several chronic inflammatory skin diseases and acceleration of skin aging [107]. Major air pollutants include polycyclic aromatic hydrocarbons, particulate matter, ozone, carbon monoxide, nitrogen, and sulfur dioxide [106]. All of them are associated with increased oxidative stress and extrinsic aging. Tobacco smoke remains major source of hazardous indoor air pollutant for both: active and secondhand smokers. Cigarette smoke is composed of over 7000 chemicals of which, at least, 70 are carcinogenic [108]. In addition to the obvious impact on the development of lung cancer and chronic obstructive pulmonary disease (COPD), the dramatic acceleration of natural skin aging by cigarette smoke is well documented [109]. The characteristic pattern of cigarette-related facial aging includes deep periorbital and perioral wrinkling, accompanied by orange skin discoloration, and occasionally large comedones [109]. The detrimental effect of smoking is driven by massive oxidative stress generation, antioxidant mechanisms impairment, and upregulation of matrix metalloproteinases [110,111]. So far, there are only indirect evidence for sirtuin 1 regulation of skin changes related to air pollutants. The assumptions about the role of SIRT1 might be driven from studies on COPD. Chronic inflammation, excessive oxidative stress, and increased activity of metalloproteinases underlie the pathogenesis of COPD. In details, SIRT1 overexpression may prevent airway inflammation, and oxidative stress induced by particulate matter [112][113][114]. Moreover, SIRT1 overexpression, by downregulation of MMP-1 and MMP-3, has beneficial effect on mice emphysema and human COPD [115]. Of note, it regulates exactly the same metalloproteinases that are critically involved in the imbalance of extracellular matrix turnover in the skin [116,117]. However, further research is needed to investigate the role of SIRT1 in skin aging related cigarette smoke and ambient pollutants.
Sirtuin 1 Mediates the Benefits of Caloric Restriction on Healthspan
The most valuable non-pharmacologic intervention to improve lifespan is caloric restriction, and its beneficial effects are mediated by raised sirtuin 1 level [62]. The advantageous impact of CR is explained by the hormesis dose-response phenomenon [118,119]. The hypothesis assumes that low-intensity stressor stimulates cellular processes to enhance defense capacity, while high-intensity one, exceeding the threshold dose, leads to deleterious biological effects [120]. Doses within hormetic zone induce adaptive cellular responses which further improve tolerance to high-dose factors. The term should be equated with the pre-and postconditioning, previously reported in medical literature [120]. Cellular stress response, stimulated by fasting or its polyphenolic mimetics, upregulates the expression of vitagenes, a family of genes, consisting of heat shock proteins (HSPs), thioredoxin system, sirtuins, and superoxide dismutase (SOD). The equilibrium between long-term benefit from fasting compared with harm from insufficient caloric intake may vary across the population, and precise recommendations on energy intake restriction presumably cannot be ascertained.
Sirtuin 1 switches the energy source from glucose to lipids during fasting. Inhibition of glycolysis increases NAD + /NADH ratio [121], and sirtuin 1 activity in keratinocytes and dermal fibroblasts which translates into augmented differentiation of keratinocytes and prolonged life of the latter [17,22]. So far, direct assessment of CR on human skin has not been performed, however, long-term CR decreases glycation of extracellular proteins in rodents [122]. On the other hand, replicative senescence is not affected by fasting [123]. A study performed by De Cabo et al. [56] on cultured human fibroblast, treated with serum obtained from rodents on various dietary regimens, investigated alterations in several aging biomarkers. The researchers found that serum from rats fed on CR significantly delays the downregulation of SIRT1 and increase in β-galactosidase associated with senescence phenotype. However, the impact of fasting on human skin aging has not been evaluated so far.
Sirtuin 1 Activators for the Topical Treatment of Premature Aging Symptoms
From the standpoint of applicability of sirtuin 1 anti-aging properties in dermatology and cosmetology, the most profitable would be to use topical activators of the protein. In 2007, a study on biopeptides derived from yeast Kluyveromyces revealed increase in SIRT1 expression in human dermal fibroblasts and keratinocytes with concomitant decrease in β-galactosidase [48]. The results were confirmed by the objective improvement in skin aging symptoms after four-week application of a topical formula enriched in the yeast biopeptides [48]. However, among studied so far activators, resveratrol has drawn the most attention. It is a phytoalexin from the stilbene family of polyphenols synthetized in the skin of grapes, blueberries, raspberries, mulberries, and peanuts in response to infection or physical injury [124]. Whereas the phytoalexin concentration in plants is noxious for microorganisms, it is harmless to humans, and may induce adaptive stress response in compliance with hormesis concept. A large amount of data demonstrated health benefits of resveratrol, including the maintenance of cutaneous functions and delay of senescence symptoms [125][126][127][128][129]. Furthermore, it protects skin from various tissue damaging stimuli: UV radiation [130] or excessive oxidative stress, generated by tobacco smoke [131]. The biological effects of resveratrol mentioned above were also reported in vivo with topical application. Numerous lines of evidence proved protective effect against UVR-induced damage [132][133][134] and increased antioxidant capacity [135,136]. The major limitation of topical resveratrol applicability is fast isomerization and poor water solubility [137,138] what prompted the researchers to investigate a novel stable and potent resveratrol derivative [139,140]. In regards of skin aging, the data on the results of oral application are conflicting. A placebo-controlled trial of a dietary supplement containing polyphenols demonstrated visual improvement in multiple age-related skin changes as well as decreased systemic oxidative stress [141]. In contrast, an antioxidant cocktail, composed of trans-resveratrol, selenium, and vitamin E and C, increased antioxidant capacity without obvious skin improvement in a study on 60 healthy volunteers [142]. The discrepancies may be partially explained by multicomponent composition of the supplements in both studies. The same interpretation problem may arise from the study on topical formulation containing resveratrol, retinol, niacinamide, and hexylresorcinol [143]. Although the study confirmed effectiveness in combating numerous skin aging symptoms, it cannot attribute this effect directly to resveratrol. Similarly, due to involvement of multiple mechanisms, not all effects of the resveratrol are associated with SIRT1 activation. In reference to sirtuin 1 pathway, resveratrol regulates keratinocytes differentiation and proliferation, promotes wound healing, decreases keloid formation, participates in antioxidant defense, and UVR protection [144].
Garlic (Allium sativum) is the source of many valuable bioactive compounds and has been used as natural remedy since ancient times. Traditionally, it is considered as natural antibiotic due to the content of allicin that is responsible for the majority of its antimicrobial activity [145]. Moreover, garlic extract is also a potent antioxidant what made it applicable in dermatology. Topical application of garlic extract is beneficial for the treatment of inflammatory skin diseases, keloids, wound healing, as well as skin aging and photoprotection [146]. The study on human keratinocytes exposed to UVB confirmed that some of these properties are mediated by SIRT1 pathway [57]. Garlic pretreatment reduced oxidative stress, inflammation, and aging biomarkers [57]. Several other studies investigated the effect of pretreatment with multiple biological compounds on UVR-related alterations in human dermal fibroblasts [49,52,58]. The examined botanical compounds, including resveratrol and Epilobium angustifolium extract, were able to restore UV-induced downregulation of SIRT1 and decrease the activity of metalloproteinases [49,52].
Several other Chinese herbal medicine-derived compounds were found to attenuate oxidative injury by the upregulation of the SIRT1 pathway [59,60]. Kumar et al. [59] pointed out the potential antiaging application of Antrodia cinamomea, an endemic medicinal mushroom from Taiwan, possessing hepatoprotective effects. The fungus contains several bioactive compounds of which antcin M have been reported to reduce hyperglycemiainduced oxidative stress in dermal fibroblasts. Tremella fuciformis, another example of edible mushroom, has numerous beneficial features including: anti-inflammatory, antioxidant, antitumor, and antiaging [60]. It has been used for a long time as an important part of Chinese cuisine, dietary supplement, and antiaging ingredient of topical creams, serums, and facial preparations; however, the involvement of SIRT1 mechanism was described recently [60].
In contrast, some bioactive compounds, isolated from plants and fungi, delayed cellular senescence of HDFs without affecting the SIRT1 expression and activity [51,54]. For instance, epigallocatechin gallate (EGCG), a polyphenol extracted from green tea, significantly suppressed the p53 acetylation but had no effect on SIRT1 [51]. It is worthy to mention that experimental concentration of the EGCG, providing antiaging effects, greatly exceeded normal dietary consumption. Similar results were obtained by Takata et al. [54] in a study aimed to determine the effect of saikokeishito (TJ-10) on HDFs senescence. Saikokeishito is a traditional Japanese herbal medicine (Kampo) consisting of nine herbal components: Bupleurum root, Pinellia tuber, Scutellaria root, Jujube fruit, Ginger rhizome, ginseng root, cinnamon bark, peony root, and Glycyrrhiza root [54]. TJ-10 has been widely prescribed against infectious diseases, chronic pancreatitis (due to suppression of the TGFβ expression), and other gastrointestinal diseases as well as an antiepileptic drug (activation of the GABA-A receptors). The results presented by Takata et al. suggests that it also participates in dermal senescence; however, it is unclear at the time if oral intake of the TJ-10 may bring about objective improvement in skin aging.
The conflicting results on oxidative stress reduction by food-derived compounds may be due to polyphenols instability in incubation medium that enhances the production of hydrogen peroxide and affect SIRT1 activity [147]. Because the dietary polyphenols undergo rapid metabolism, form glucuronide and sulfate conjugates in the liver, the bioactivity of botanical compounds in the human body may differ from the results obtained in the in vitro studies [147]. However, from the standpoint of hormesis response, it is not essential to reach experimental concentrations, exceeding dietary intake, and sometimes high doses of phytochemicals may be toxic [148]. Growing body of evidence indicates that hormetic mechanisms explain similar health benefits achieved by numerous phytochemicals intake. The key element of cellular stress response is activation of the nuclear factor-erythroid 2 p45-related factor 2 (Nrf2) which binds to antioxidant response element (ARE), a promoter of several genes crucial for redox homeostasis. ARE encodes cytoprotective proteins with diversified functions: antioxidant, anti-inflammatory, repair or removal of damaged macromolecules, conjugation, and glutathione homeostasis [120]. Menendez et al. [53] demonstrated that anticancer and antiaging activity of the complex of polyphenols naturally presented in extra virgin olive oil is related to Nrf2 signaling. Hydroxytyrosol, one of the main polyphenols of the olive oil, has been recently shown to prevent neurodegeneration in nematodes and rodents [149,150]. The authors of the studies postulated that Nrf2 signaling pathway and hormesis response may be responsible for the observed effects. Considering common pathophysiologic background of neurodegenerative disorders, age-related diseases, and skin aging nutraceuticals may be effective and safe therapeutic targets to improve healthy aging.
Conclusions
The past few decades have witnessed rapidly growing interest in the topic of skin aging. Research advances in the field of molecular mechanisms involved in both intrinsic and extrinsic aging may translate into novel effective therapeutic antiaging interventions. SIRT1 regulates the proliferation and differentiation of the major cell types of the skin. It mediates antiaging effects through the target molecules, participating in the maintenance of genome stability, oxidative stress response, extracellular matrix turnover, and regulation of the cell cycle. The same mechanisms are involved in the skin carcinogenesis, however, SIRT1 may serve as tumor suppressor or oncogene, depending on the gene dosage and downstream pathways. Experimental studies support the conclusion that SIRT1 may prevent UVR-related premature aging and skin cancer development. Modulation of SIRT1 activity may constitute a novel approach to photoprotection, and hold a great potential for improving outcomes in patients with skin cancers. In addition, small interventional studies in humans demonstrated encouraging results of topical SIRT1 activators in reducing premature aging symptoms. Although SIRT1 has been extensively researched, further studies are needed to explore the topic of SIRT1 activity in terms of skin changes, induced by environmental pollutants, which help to exploit the full potential of this longevity protein. | 2021-04-26T05:15:13.979Z | 2021-04-01T00:00:00.000 | {
"year": 2021,
"sha1": "42bccb5327014d28f910921d49f99bab90c3316c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4409/10/4/813/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "42bccb5327014d28f910921d49f99bab90c3316c",
"s2fieldsofstudy": [
"Biology",
"Environmental Science",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
255034461 | pes2o/s2orc | v3-fos-license | Gluten Degradation by the Gut Microbiota of Ulcerative Colitis Patients
Several studies have reported improved disease symptomatology in ulcerative colitis (UC) patients consuming a gluten free diet. This observation coupled with diversity depletion in the gut microbiota of UC patients led us to hypothesize that UC-associated enteric microbes differentially metabolize dietary gluten to produce immunogenic products that promote inflammation. Gluten concentration in stool was determined using gluten-specific ELISA, and gluten intake was assessed by food frequency questionnaire (FFQ) in UC (n = 12) and healthy controls (HC; n = 13). Gluten-metabolizing bacteria were isolated on minimal media supplemented with 1% gluten from UC and HC and identified by 16S rRNA profiling. Cell-free culture media from gluten metabolizing gut bacterial isolates was assessed for immunogenicity in vitro using HT29 colonocytes. Compared to HC, UC patients did not consume gluten differently (Mann–Whitney; p > 0.10) and exhibited equivalent levels of gluten in their feces (Mann–Whitney; p = 0.163). The profile of gluten-degrading bacteria isolated from UC stool was distinct (Chi-square; p ≤ 0.0001). Compared with Enterococcus isolates, products of gluten degradation by Bacillus strains induced higher IL8 and lower occludin (Mann–Whitney; p = 0.002 and p = 0.059, respectively) gene expression in colonocytes irrespective of whether they originated from UC or healthy gut. Members of HC and UC microbiota exhibit gluten-degrading ability, metabolites of which influence genes involved in inflammation and barrier function in enteric colonocyte cultures. Preliminary findings of this study warrant further investigations into the mechanisms by which gut microbiota contribute to UC pathogenesis through gluten degradation.
Introduction
Ulcerative colitis (UC) is the most common form of inflammatory bowel disease (IBD) particularly in westernized nations, where diets and lifestyle have dramatically changed over the last several decades towards, amongst other transitions, a high intake of refined grains containing gluten [1][2][3][4].
UC is a debilitating, idiopathic gastrointestinal inflammatory disease that can lead to life-threatening complications including a higher risk for cancer [5]. Current treatment options include anti-inflammatories, immunomodulating biologics and/or surgery. Though pharmaceutical interventions may manage disease symptoms, they are not universal in their efficacy and there is a major need to develop a better understanding of factors that lead to maintenance of chronic inflammation in this patient population so that novel therapies may be developed.
A consistent feature of UC patients is microbiome perturbation [6]. Patients with UC characteristically exhibit compositionally distinct fecal and intestinal mucosal microbiomes Microorganisms 2023, 11, 12 2 of 9 depleted of microbial diversity commonly observed in age-matched healthy subjects [7,8]. Loss of gut microbial diversity results in loss of functional gene pathways encoded by these microbes, particularly metabolic pathways, resulting in altered metabolism of dietary substrates [9]. In addition to inflammation, increased intestinal permeability is found in 40-50% of patients with IBD, with the highest values for those with active disease [10]. Increased permeability can be assessed using occludin [11] and the overexpression of this tight junction protein considered a mucosal response to the increased permeability [12]. The inflammation in the gut of IBD patients can be assessed using IL-8 as a biomarker [13].
Gluten, a dominant component of Western diets [14], is more commonly associated with celiac disease and almost all knowledge of how gluten influences the immune system stems from these patients. In celiac disease patients, both undegraded gluten and gluten degraded to specific peptides, i.e., gliadin and glutenin, promote inflammatory responses in the gastrointestinal tract [15,16]. Other research has shown that gluten is largely resistant to cleavage by human gastrointestinal digestive enzymes [17], and that it is primarily intestinal microbes that metabolize gluten in the human gut [18]. Support to examine the relationship between UC and gluten comes from recent observations in patients with IBD, rheumatoid arthritis and psoriasis, who report reduced disease symptomology whilst consuming a gluten free diet (GFD) [19,20]. A recent study reported improvement in gastrointestinal (GI) symptoms such as abdominal pain, diarrhea and headaches in 70 UC patients on GFD [21]. Additionally, a study examining 1647 adults found that 65.6% of patients with IBD who implemented a GFD experienced an improvement of at least one clinical symptom, 38.3% reported fewer and less severe flares while abstaining from gluten and 23.6% required less medication to control disease [22].
Several studies have been conducted to investigate a possible connection between celiac disease and IBD with various results. Two studies did not find a higher prevalence of human leukocyte antigen (HLA) DQ2/8 [23,24] in patients with UC compared to healthy subjects, the necessary antecedent for developing celiac disease [25]. Elevated concentrations of anti-tissue-transglutaminase, anti-tTG antibodies, characteristic of celiac disease [15], have also been reported in IBD patients [26], suggesting that similar disease pathology may exist in UC patients.
These observations have led us to hypothesize that due to gut microbial species depletion, patients with UC are enriched for species that differentially metabolize dietary gluten to produce immunogenic products that promote inflammation. Our aim is to isolate gluten-degrading bacteria from the gut of UC patients and healthy subjects, transfer the degraded gluten products to a colonic cell-line and test the cells expression of occludin, to assess barrier function, and IL8 to assess inflammatory response.
Study Population and Sample Collection
Stool samples were collected from age-matched healthy participants (n = 13) and physician-diagnosed ulcerative colitis patients (n = 32) as previously described [7]. The University of California San Francisco Committee on Human Research approved the protocol for collection of stool samples used in this study. Subjects provided written informed consent. In this study, the participants were grouped based on disease severity and associated fecal bacterial community composition, previously classified [7]. Samples from the group with highest disease activity (UC n = 12, HC n = 13) (see Supplementary Table S1 for patient characteristics) were used for gluten and microbiological assays performed in this study. Home stool samples were collected and maintained on ice until stored at −80 • C.
Assessment of Gluten Intake in Study Participants:
All subjects completed a Block 2005 Food Frequency Questionnaire (FFQ) which was used to assess gluten intake as combined consumption of white bread, bagels, English muffins and biscuits.
Quantification of Fecal Gluten Content
Amount of gluten in fecal samples was determined using the gluten ELISA (Allertek, Gainesville, FL, USA) kit. Protocol was followed per manufacturer's instructions, except for dilution with Sample Dilution Buffer, which was reduced from 10× to 2×, due to anticipated lower concentrations of gluten in feces compared to food samples. Briefly, frozen stool (0.25 g) was weighed and resuspended in Sample Dilution buffer at 1:1 w/v ratio. Samples, standards and controls were loaded into a 96-well plate containing the 401.21 antibody, which recognizes both the gliadin and the glutenin fractions of gluten. Gluten concentrations were calculated against a standard curve based on 450 nm absorbance.
Isolation of Gluten Degrading Bacteria
Pooled healthy or UC fecal samples were cultured on M9 Minimal Salts X5 (M9; Sigma-Aldrich, Saint Louis, MO, USA) agar plates with added trace elements, MgSO 4 , CaCl 2 and 1% gluten (Sigma-Aldrich, Saint Louis, MO, USA) under aerobic (AE), microaerophilic (MA) or anaerobic environments (AN) for 24 h. Anaerobic conditions were maintained in an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA). Microaerophilic atmosphere was generated using CampyGen and CO 2 Gen gas generating kits (BD, Franklin Lakes, NJ, USA). Putative gluten degrading isolates were validated by sub-culture onto M9 media with or without addition of filter-sterilized (0.2 µm) (Thermo Fisher, Waltham, MA, USA) 1% gluten for 72-hour under the oxygen availability conditions used for initial isolation.
Identification of Gluten-Degrading Microorganisms
Confirmed gluten metabolizing isolates were plated on Luria-Bertani (LB) agar (Sigma-Aldrich, Saint Louis, MO, USA) and incubated aerobically for 48 h at 37 • C. Species identification was determined by Sanger sequencing (QuintaraBio, San Francisco, CA, USA) of the full-length 16S rRNA gene amplified using 27F/1492R primers, and by using gDNA purified followed by QIAquick PCR Purification Kit (Qiagen, Hilden, Germany) from each isolate. The resulting sequences were classified using the ribosomal rRNA SILVA (https://www.arb-silva.de/, accessed on 1 May 2019) database [27]. Isolates were cryopreserved at −80 • C in LB broth containing 50% (v/v) sterile glycerol.
Epithelial Cell Assessment of Bacterial Products of Gluten Degradation
Gluten degrading bacteria isolated from UC patient and healthy subject feces were cultured overnight and diluted to an OD600 = 0.1 in M9 media + 1% gluten, prior to growth for 8 h under aerobic conditions. Cellular material was removed by centrifugation (13,000 rpm for 1 min) and supernatants were passed through a 0. 50 • C for 2 min, 95 • C for 10 min (1 cycle); 95 • C for 15 s and 60 • C for 1 min (40 cycles); and a final melting curve cycle of 95 • C for 15 s, 60 • C for 1 min and 95 • C for 15 s. Gene expression was normalized to β-actin expression and relative fold change in expression determined by the 2-∆∆CT (where CT is threshold cycle) method using PBS exposed cells as control. β-actin primers were AAGATGACCCAGATCATGTTTGAGACC [forward] and AGCCAGTCCAGACGCAGGAT [reverse].
Statistical Analysis
Mann-Whitney test was used for group comparisons. GraphPad Prism version 6 (GraphPad Software. La Jolla California, CA, USA, www.graphpad.com, accessed on 1 February 2019) software was used for statistical analyses. Differences in gluten-degrading bacterial profiles were compared using Chi-square test.
The Gut of UC Patients Retains Gluten Degrading Capability
Using data captured in the FFQ, gluten intake was assessed based on subject reported consumption of gluten containing foods and compared between healthy subjects and UC patients. Gluten consumption was not found to significantly differ between groups (Supplementary Table S2). A comparative analysis revealed that the concentration of gliadin and glutenin (the metabolites of gluten) in feces of subjects, as assessed by gluten-specific ELISA, did not differ between the two groups (Mann-Whitney; p > 0.05) (Figures 1 and S1), suggesting that UC patient retain the ability to degrade gluten despite loss of microbial function in their gut microbiome.
Real-time quantitative PCR was performed in triplicate with SYBR Green master mix (Life Technologies, Carlsbad, USA) in the QuantStudio™ 6 Flex Real-Time PCR System (Applied Biosystems, Waltham, MA, USA) using the primers for occludin (GATGAG-CAGCCCCCCAAT [forward], GGTGAAGGCACGTCCTGTGT [reverse]) to assess barrier function and IL-8 (CTGAGAGTGATTGAGAGTGGAC [forward], AACCCTCTG-CACCCAGTTTTC [reverse]) to assess inflammatory response. PCR reaction conditions were as follows: 50 °C for 2 min, 95 °C for 10 min (1 cycle); 95 °C for 15 s and 60 °C for 1 min (40 cycles); and a final melting curve cycle of 95 °C for 15 s, 60 °C for 1 min and 95 °C for 15 s. Gene expression was normalized to β-actin expression and relative fold change in expression determined by the 2-∆∆CT (where CT is threshold cycle) method using PBS exposed cells as control. β-actin primers were AAGATGACCCAGATCATGTTTGA-GACC [forward] and AGCCAGTCCAGACGCAGGAT [reverse].
Statistical Analysis
Mann-Whitney test was used for group comparisons. GraphPad Prism version 6 (GraphPad Software. La Jolla California, CA, USA, www.graphpad.com, accessed on 1 February 2019) software was used for statistical analyses. Differences in gluten-degrading bacterial profiles were compared using Chi-square test.
The Gut of UC Patients Retains Gluten Degrading Capability
Using data captured in the FFQ, gluten intake was assessed based on subject reported consumption of gluten containing foods and compared between healthy subjects and UC patients. Gluten consumption was not found to significantly differ between groups (Supplementary Table S2). A comparative analysis revealed that the concentration of gliadin and glutenin (the metabolites of gluten) in feces of subjects, as assessed by gluten-specific ELISA, did not differ between the two groups (Mann-Whitney; p > 0.05) (Figures 1 and S1), suggesting that UC patient retain the ability to degrade gluten despite loss of microbial function in their gut microbiome.
Distinct Profiles of Gluten Degrading Bacteria Are Evident in UC and Healthy Feces
Bacteria capable of growth on minimal media supplemented with gluten as the sole carbon source were evident in feces from both healthy subjects and UC patients. Sequencing of full length 16S rRNA was performed for 37 isolates to determine whether distinct bacterial species were capable of gluten degradation in UC patients compared with healthy subjects. Classification using the SILVA database indicated that the profile of bacterial capable of degrading gluten in the gut microbiome of UC patients was significantly distinct from that observed in healthy subjects (Chi-square; p ≤ 0.0001; Figure 2). Bacterial isolates from patients with UC predominantly belonged to the Enterococcus. In comparison a greater variety of bacterial species were isolated from the gut microbiome of healthy subjects, including species belonging to Escherichia, Bacillus and Enterococcus. These data suggest that the healthy gut microbiome possesses a greater breadth of species capable of gluten degradation.
ing of full length 16S rRNA was performed for 37 isolates to determine whether distinct bacterial species were capable of gluten degradation in UC patients compared with healthy subjects. Classification using the SILVA database indicated that the profile of bacterial capable of degrading gluten in the gut microbiome of UC patients was significantly distinct from that observed in healthy subjects (Chi-square; p 0.0001; Figure 2). Bacterial isolates from patients with UC predominantly belonged to the Enterococcus. In comparison a greater variety of bacterial species were isolated from the gut microbiome of healthy subjects, including species belonging to Escherichia, Bacillus and Enterococcus. These data suggest that the healthy gut microbiome possesses a greater breadth of species capable of gluten degradation. Figure 2. Distribution of gluten degrading bacteria isolated from stool of UC and HC subjects on minimal media supplemented with 1% gluten, Chi-square; *** p 0.0001.
Products of Gluten Degradation by Enterococcus and Bacillus isolates Induce Distinct Immune Response in Colonocytes Irrespective of Their Source
We hypothesized that gluten-degrading bacterial isolates from UC patients induce barrier dysfunction and pro-inflammatory response compared to HC isolates. To assess this, the cell-free products of gluten degradation by various UC and HC fecal bacterial isolates was compared. In a proof of principal experiment, we used IL8 (a neutrophil chemoattractant [28]) and occludin (an epithelial tight junction protein [29]) expression as a proxy for immune activation and barrier dysfunction following exposure of HT29 colonocytes to culture supernatants from Bacillus or Enterococcus isolated from stool of UC or HC subjects.
Higher IL8 expression was evident following exposure of colonocytes to the cell-free products of Bacillus isolates grown on gluten containing M9 media when compared with Enterococcus isolates (Mann-Whitney; p = 0.002; Figure 3A). However, cell-free products of gluten degradation by Enterococcus isolates trended towards increased occludin expression (Mann-Whitney; p = 0.059; Figure 3B). This finding was consistent irrespective of the source of gluten metabolizing strains (HC or UC; Supplementary Figure S2A,B) and indicated that these genera may contribute to inflammation if actively engaged in gluten metabolism in the gut. Contrary to our hypothesis we did not observe that UC isolates
Products of Gluten Degradation by Enterococcus and Bacillus Isolates Induce Distinct Immune Response in Colonocytes Irrespective of Their Source
We hypothesized that gluten-degrading bacterial isolates from UC patients induce barrier dysfunction and pro-inflammatory response compared to HC isolates. To assess this, the cell-free products of gluten degradation by various UC and HC fecal bacterial isolates was compared. In a proof of principal experiment, we used IL8 (a neutrophil chemoattractant [28]) and occludin (an epithelial tight junction protein [29]) expression as a proxy for immune activation and barrier dysfunction following exposure of HT29 colonocytes to culture supernatants from Bacillus or Enterococcus isolated from stool of UC or HC subjects.
Higher IL8 expression was evident following exposure of colonocytes to the cell-free products of Bacillus isolates grown on gluten containing M9 media when compared with Enterococcus isolates (Mann-Whitney; p = 0.002; Figure 3A). However, cell-free products of gluten degradation by Enterococcus isolates trended towards increased occludin expression (Mann-Whitney; p = 0.059; Figure 3B). This finding was consistent irrespective of the source of gluten metabolizing strains (HC or UC; Supplementary Figure S2A,B) and indicated that these genera may contribute to inflammation if actively engaged in gluten metabolism in the gut. Contrary to our hypothesis we did not observe that UC isolates capable of growth on gluten as the sole carbon source were more inflammatory or capable of greater barrier disruption compared with their HC phylogenetic counterparts (Mann-Whitney; p > 0.05; Supplementary Figure S2A,B).
Microorganisms 2023, 11, x FOR PEER REVIEW 6 of 9 capable of growth on gluten as the sole carbon source were more inflammatory or capable of greater barrier disruption compared with their HC phylogenetic counterparts (Mann-Whitney; p > 0.05; Supplementary Figure S2A,B). Figure 3. Expression of (A) IL8 and (B) occludin genes in colonic epithelial cells exposed to gluten degradation products produced by Bacillus and Enterococcus isolates. Results were obtained from two independent experiments and based on combined observations from 2 strains for each genera. Statistical significance was determined using Mann-Whitney test.
Discussion
Our observations suggest that UC patients do not differ in their consumption of gluten and retain the capacity to degrade ingested gluten. However, we note that the diversity of bacteria capable of gluten degradation is diminished in UC patients and consists primarily of Enterococcus and Bacillus species. This suggests that the products of these species' metabolism of gluten predominate the luminal contents of UC patients. This is a first report, to our knowledge, linking the gut microbiome of UC patients to gluten degradation and the possible mechanism of gluten sensitivity. Depletion of bacterial diversity is a well-documented signature of the UC gut microbiome [6]; this is reflected in the reduced diversity of gluten degrading bacteria isolated from UC subjects. Gluten degradation by enteric microbiota has been observed in celiac disease [30], and the byproducts of bacterial metabolism of gluten are known to contribute to the immune activation [19].
Bacillus species have been shown to hydrolyze gluten well [31], and inactivated Bacillus species have been shown to improve the symptoms of UC [32] and alter the production of both immune activating and anti-inflammatory cytokines and chemokines [33]. Our observations in this study suggest a higher frequency of gluten-degrading Bacillus in healthy subjects compared to UC patients. In vitro isolates from both subject groups exhibit comparable capacity to induce IL8 and occludin expression in cultured HT29 colonocytes. On the contrary, members of genus Enterococcus have been shown to contribute to development of intestinal inflammation in mice [34,35] and have been found increased in feces of UC patients [36]. Our findings indicate a higher frequency of gluten-degrading Enterococcus in the stool of UC patients. However, in vitro findings indicated no difference in the capacity of gluten metabolism products derived from a UC-associated Enterococcus strain to induce either IL8 or occluding when compared with an isolate from a healthy subject. Increased occludin expression was observed following exposure to the gluten metabolism products of the Enterococcus isolates, compared to those from Bacillus, though whether this reflects increased barrier disruption is unclear. Overall, our cell-culture Figure 3. Expression of (A) IL8 and (B) occludin genes in colonic epithelial cells exposed to gluten degradation products produced by Bacillus and Enterococcus isolates. Results were obtained from two independent experiments and based on combined observations from 2 strains for each genera. Statistical significance was determined using Mann-Whitney test.
Discussion
Our observations suggest that UC patients do not differ in their consumption of gluten and retain the capacity to degrade ingested gluten. However, we note that the diversity of bacteria capable of gluten degradation is diminished in UC patients and consists primarily of Enterococcus and Bacillus species. This suggests that the products of these species' metabolism of gluten predominate the luminal contents of UC patients. This is a first report, to our knowledge, linking the gut microbiome of UC patients to gluten degradation and the possible mechanism of gluten sensitivity. Depletion of bacterial diversity is a well-documented signature of the UC gut microbiome [6]; this is reflected in the reduced diversity of gluten degrading bacteria isolated from UC subjects. Gluten degradation by enteric microbiota has been observed in celiac disease [30], and the byproducts of bacterial metabolism of gluten are known to contribute to the immune activation [19].
Bacillus species have been shown to hydrolyze gluten well [31], and inactivated Bacillus species have been shown to improve the symptoms of UC [32] and alter the production of both immune activating and anti-inflammatory cytokines and chemokines [33]. Our observations in this study suggest a higher frequency of gluten-degrading Bacillus in healthy subjects compared to UC patients. In vitro isolates from both subject groups exhibit comparable capacity to induce IL8 and occludin expression in cultured HT29 colonocytes. On the contrary, members of genus Enterococcus have been shown to contribute to development of intestinal inflammation in mice [34,35] and have been found increased in feces of UC patients [36]. Our findings indicate a higher frequency of gluten-degrading Enterococcus in the stool of UC patients. However, in vitro findings indicated no difference in the capacity of gluten metabolism products derived from a UC-associated Enterococcus strain to induce either IL8 or occluding when compared with an isolate from a healthy subject. Increased occludin expression was observed following exposure to the gluten metabolism products of the Enterococcus isolates, compared to those from Bacillus, though whether this reflects increased barrier disruption is unclear. Overall, our cell-culture based screen for pro-inflammatory potential of gluten metabolism by enteric bacteria does not exclude the possibility that UC isolates may elicit distinct epithelial responses compared to isolates from similar genera derived from healthy subjects in an already inflamed gut of UC patients or in association with other UC microbiota community members. It does indicate that the products of gluten degradation by distinct enteric bacteria are capable of inducing expression of genes associated with inflammatory immune activation and barrier dysfunction that may contribute to inflammation in UC patients.
Our observations are encouraging and, although very preliminary, these warrant further investigation into the mechanisms by which diet interacts with the gut microbiome to influence intestinal inflammation in UC patients. Limitations of this study include sample size, culture-based methods, and the use of an in vitro cell line assay. The findings of this study need to be further evaluated in a larger cohort of subjects. Additionally, investigation using community-based approaches such as functional metagenomics and metatranscriptomics will need to be incorporated into future studies for more "in depth" evaluation of gluten degrading members of the gut microbiome and how they relate to UC pathogenesis. Even though the results are preliminary, they show that the gut microbiome of IBD patients respond differently to gluten in their diet, potentially contributing to inflammation associated with the disease. Patients with IBD are very motivated to diet interventions as this gives them influence in their otherwise very disease-controlled life. Preliminary findings of this study warrant further investigations into the mechanisms by which gut microbiota contribute to UC pathogenesis through gluten degradation to provide patients with better understanding of ways to manage their disease. Figure S1. Repeated ELISA measuring undegraded gluten in stool samples from patients with ulcerative colitis and healthy subjects. P-value obtained using Mann-Whitney test; Supplementary Figure S2. Expression of (A) IL8 and (B) occludin genes in colonic epithelial cells exposed to gluten degradation products produced by Bacillus and Enterococcus isolates from UC and HC subjects. Results were obtained from two independent experiments using three biological replicates. Statistical significance was determined using Mann-Whitney test.
Author Contributions: S.V.L., V.A., J.D., Y.P. and E.O.S.H. contributed to the study conception, study design and protocol, interpretation of data, drafting of the paper or revising it critically for important intellectual content. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement:
The data underlying this article will be shared on reasonable request to the corresponding author. | 2022-12-24T16:26:23.269Z | 2022-12-21T00:00:00.000 | {
"year": 2022,
"sha1": "7ecaea8fa09ad87c494cabb436e398a046bd553c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2076-2607/11/1/12/pdf?version=1671613633",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "5d83f1168e31354e70d4154cad2d652af12d24dc",
"s2fieldsofstudy": [
"Medicine",
"Agricultural And Food Sciences",
"Biology"
],
"extfieldsofstudy": []
} |
9578042 | pes2o/s2orc | v3-fos-license | Linear polarization dependence of microwave-induced magnetoresistance oscillations in high-mobility two-dimensional systems
We examine the effect of changing the linear polarization angle $\theta$ of incident microwaves with respect to the dc current on radiation-induced magnetoresistance oscillations in a two-dimensional (2D) system within the balance-equation formulation of the photon-assisted magnetotransport model, considering the radiative decay as the sole damping mechanism. At an extremum the amplitude of oscillatory magnetoresistance $R_{xx}$ exhibits a sinusoidal, up to a factor of 5, magnitude variation with rotating the polarization angle $\theta$. The maximal amplitude shows up generally at a nonzero $\theta$, which is dependent upon the extremum in question, the 2D electron setup, the radiation frequency and the magnetic field orientation. These results provide a natural explanation for the experimental observations by Mani {\it et al.} [Phys. Rev. B {\bf 84}, 085308 (2011)], and Ramanayaka {\it et al.} [Phys. Rev. B {\bf 85}, 205315 (2012)].
Early measurement on L-shaped Hall bars indicated that the period and phase of radiation-induced magnetoresistance oscillations are insensitive to the relative orientation between the microwave polarization and the current. 9A later experiment 13 carried out on specimens with a square geometry in a quasioptical setup reported a striking result that not only the frequency and phase but also the amplitude of radiation-induced resistance oscillations and the zero resistance regions are notably immune to the sense of circular and linear polarizations of microwaves.This influential result raised a big challenge to the existing theoretical models, in which, though no detailed investigation was reported, some kind of polarization dependence was believed to exist, 19,22,23 and thus expedited the emergence of different scenarios capable of showing polarization immunity of microwave magnetoresistance response. 33,34In a recent study, Mani et al. 17 found a strong sensitivity in the amplitude of radiationinduced magnetoresistance oscillations to the relative orientation of the linear polarization with respect to the Hall bar axis.Particularly, more detailed measurement 18 by rotating, by an angle θ, the polarization direction of linearly polarized microwaves with respect to the long axis of the Hall bar electron devices, showed, at relatively low microwave power, a strong sinusoidal variation in the diagonal resistance R xx vs θ at the oscillatory extrema.And, unexpectedly, the angle for the maximal oscillatory R xx response under a given-power linear-polarized microwave, which is not at 0 o or 90 o , appears to depend upon the radiation frequency, the extremum in question, and the magnetic field orientation.So far, there has been no theoretical explanation for these interesting observations.There is an urgent need to analyze the detailed polarization dependence from a theoretical model for radiation-induced magnetoresistance oscillations.
II. MAGNETORESISTANCE UNDER POLARIZED RADIATION
We deal with an isotropic 2D system of short thermalization time, consisting of N s electrons in a unit area of the x-y plane.These electrons, scattered by random impurities and by phonons in the lattice, are subjected to a uniform magnetic field B = (0, 0, B) in the z direction.When an electromagnetic wave of angular frequency ω illuminates perpendicularly onto the 2D plane with the incident electric field at z = 0 and a dc current flows within the plane, the electric field inside the 2D system involves a dc component, E 0 , and an ac component The steady-transport state of this electron system can be described by the drift velocity of the electron integrative (the center of mass) motion, consisting of a dc part, v, and a stationary time-dependent part, in the 2D plane, together with an average temperature T e , characterizing the isotropic thermal distribution of electrons in the reference frame moving with the center of mass. 36They satisfy the following force and energy balance equations: 22 Here, is the time-averaged damping force against the electron drift motion due to impurity scattering, and is the time-averaged rate of the electron energyabsorption from the radiation field.In Eqs. ( 8) and ( 9), U (q ) is the effective impurity potential, Π 2 (q , Ω ) is the imaginary part of the electron density-correlation function at temperature T e in the presence of the magnetic field without the electric field, ω 0 ≡ q • v, and /ω.Note that, although contributions of phonon scattering to F and S p are neglected in comparison with those of impurity scattering at the considered low lattice temperature, it provides the main channel for electron energy dissipation to the lattice with a time-averaged energy-loss rate W , having an expression as given in Ref. 22.
The ac components v s and v c of electron drift velocity should be determined selfconsistently in terms of the incident ac field E i (t) by the electrodynamic equations connecting both sides of the 2D system, taking into account the scattering-related damping forces F s and F c . 22owever, for high-mobility systems at low temperatures, the effects of these scattering-related damping forces are much weaker in comparison to those of radiative decay 31 and thus negligible, whence v s and v c are in fact directly given from Eqs. ( 5) and ( 6) by the high-frequency electric field E(t) inside the 2D electron system.On the other hand, by solving the Maxwell equations connecting both sides of the 2D electron gas which is carrying the sheet current density N s ev(t), E(t) is determined by the incident fields E is and E ic based on the setup of the 2D system in the sample substrate. 22f the 2D electron gas locates within a thin layer under the surface plane at z = 0 of a thick (treated as semi-infinite) semiconductor substrate having a refractive index, n s , the ac field E(t) driving the 2D electrons, which equals the sum of the incident and the reflected fields at z = 0 and equals the transmitted field (the field just passes through the 2D layer), can be expressed as 37 Here n 0 is the refractive index of the air and c and ǫ 0 are, respectively, the light speed and the dielectric constant in vacuum.If the 2D electron gas is contained in a thin layer suspended in vacuum at the plane z = 0, then By combining Eq. (10) or Eq. ( 11) with balance Eqs. ( 5) and ( 6), the ac drift-velocity components v s and v c , and thus the argument ξ, can be determined by the incident field components E is and E ic .
The imaginary part of 2D electron density correlation function in a magnetic field, Π 2 (q , Ω ), can be written in the Landau representation as 38 where The Landau-level broadening, which results from impurity, phonon, and electron-electron scatterings, is assumed to have a Gaussian form [ε n = (n + 1 2 )ω c is the center of the nth Landau level, n = 0, 1, 2, ...], with a B 1 2 -dependent half width expressed as where τ s , the single-particle lifetime or quantum scattering time in the zero magnetic field, is assumed to be related to the transport relaxation time τ or the zerofield linear mobility µ 0 using an empirical parameter α by 22 1/τ s = 4α/τ tr = 4αe/mµ 0 .
For an isotropic system where the frictional force F is in the opposite direction of the drift velocity v and the magnitudes of both the frictional force and the energydissipation rate depend only on v ≡ |v|, we can write In the Hall configuration with velocity v in the x direction v = (v, 0, 0) or the current density J x = J = N s ev, and J y = 0, the longitudinal linear (v → 0) resistivity under the incident radiation of Eq. ( 1) due to impurity scattering, can be written in the form where Π ′ 2 (q , Ω ) ≡ ∂Π 2 (q , Ω )/∂Ω .
III. OSCILLATORY RESISTIVITY VERSUS POLARIZATION DIRECTION
We consider an incident radiation of form (1) linearly polarized along the direction having an angle θ with respect to the x-axis: E is = (E iω cos θ, E iω sin θ) and E ic = 0.The 2D electron gas, having carrier sheet density N s = 2.2×10 15 m −2 and zero-temperature linear mobility µ 0 = 800 m 2 /V s from short-range impurity scattering in the absence of the magnetic field, is assumed to locate within a thin layer under the surface plane at z = 0 of a thick GaAs-based substrate with a refractive index of n s = 3.59.
Figure 1 shows the calculated energy absorption rate S p , the electron temperature T e and the longitudinal linear resistivity R xx of this system irradiated by linearly polarized (LP) incident microwaves having frequency ω/2π = 40 GHz and amplitude E iω = 1.15 V/cm (i.e., incident power P iω = 17.5 W/m 2 ) at different polarization directions (θ = 0 o , 30 o , 60 o , 90 o , 120 o , and 150 o ), together with the dark resistivity, as functions of the normalized inverse magnetic field ω/ω c (ω c = eB/m).The lattice temperature is assumed to be T = 1.5 K and the Landau level broadening parameter is taken to be α = 3.We see that the electron energy absorption S p and thus the electron temperature T e , both showing a marked main peak at cyclotron resonance ω/ω c = 1 and secondary peaks around its harmonics ω/ω c = 2, 3, 4, ..., are essentially the same for all different polarization directions of radiation.This can be understood directly from expression (9) of S p in the case of relatively low strength of incident microwave when the dominant contribution to it comes from the terms n = ±1 and J 2 ±1 (ξ) ∼ ξ 2 .Writing explicitly the θ-dependent expression of ξ 2 in the case of incident plane-polarized radiation having an angle θ with respect to the x-axis we can see that, after summing over all the directions of q , the angle dependence disappears due to sin 2 θ + cos 2 θ = 1.Furthermore, the energy-loss rate W has θ-dependent behavior similar to that S p .Thus the electron temperature T e , determined by the energy-balance equation (7) involving only S p and W , has similar θ-dependent behavior.The situation is different for the frictional force F and the resistivity R xx , because of additional q -direction dependent weighted factors showing up inside the q summation in the expressions ( 8) and (17), leading to sensitive θ-dependence of R xx as seen from the numerical results shown in the upper part of Fig. 1.The linear-polarized microwave excited magnetoresistivity R xx oscillates strongly, having the same oscillatory period and nearly the same phase and, at cyclotron resonance ω/ω c = 1 or its harmonics ω/ω c = 2, 3, 4, ..., taking equal values for all different polarization directions θ.We can label, for all θ, the main peaks and valleys of the oscillatory resistivity respectively as P 1, P 2, and V 1, V 2, etc.The amplitude of resistivity oscillation, however, varies sensitively with changing the polarization direction of the microwave.The maximal or minimal amplitude of a peak or a valley shows up at different polarization angles θ for different peaks or valleys.For instance, the maximal and minimal amplitudes show up, respectively, at θ ≈ 59 o and θ ≈ 149 o for P 1, at θ ≈ 25 o and θ ≈ 115 o for V 1, at θ ≈ 12 o and θ ≈ 102 o for P 2, at θ ≈ 8 o and θ ≈ 98 o for V 2, and show up at polarization angles closer to θ = 0 o and θ = 90 o for higher order peaks and valleys.The exact polarization angle for the maximal or minimal amplitude of an extremum appears to depend on the radiation frequency and the 2D electron setup in the substrate.In obtaining the above results the sample setup is so assumed that 2D electrons locate within a thin layer under the surface plane at z = 0 of a thick semiconductor substrate having a refractive index of n s = 3.59, and the relevant damping in question, i.e., the radiative decay, is fully determined by this sample setup from the electrodynamic equation (10).Except for this, no other damping parameter nor any mechanism capable of producing polarization rotation is introduced in the present model.
Next we present the results obtained under magnetic field reversal.Figure 2(a) plots the longitudinal resistivity R xx vs ω c /ω over the range −2.5 ≤ ω c /ω ≤ 2.5 for the system irradiated by a linearly polarized microwave along the θ = 30 o direction with frequency ω/2π = 40 GHz and incident power P iω = 17.5 W/m 2 .The extrema of interest here are labeled as P + 1, V + 1, P + 2, and V + 2 for those in the domain B > 0, and P − 1, V − 1, P − 2, and V − 2 for those in the domain B < 0. The values of longitudinal resistivity R xx at P + 1, V + 1, P + 2, and V + 2 are plotted in Fig. 2(b), and those at P − 1, V − 1, P − 2, and V − 2 qre plotted in Fig. 2(c), as functions of the microwave polarization direction θ.The present treatment is for an isotropic system.By symmetry we always have P + 1(θ) = P + 1(π + θ), P − 1(θ) = P − 1(π + θ), P + 1(θ) = P − 1(π − θ), etc.If the maximal amplitude of an extremum, e.g., P + 1, shows up at θ = 0, the maximal amplitude of the corresponding extremum at the reverse magnetic field, P − 1, must be at a different polarization angle.The effect of asymmetry in a real sample itself would produce further complexity.
IV. SUMMARY
We have examined the effect of changing the polarization angle θ of the incident linearly-polarized microwaves with respect to the dc current on radiation-induced magnetoresistance oscillations within the balance-equation formulation of the photon-assisted magnetotransport model.
The present investigation takes the incident microwave field, rather than the ac field inside the 2D electron system, as the input quantity, allowing a direct determination of the dominant damping mechanism in the highmobility system, the radiative decay, from the experimental sample setup without introducing any artificial damping parameter, thus enabling a direct comparison with experimental measurement.
We find that the amplitude of radiation-induced magnetoresistance oscillation varies sensitively with changing θ.At an extremum the amplitude of oscillatory magnetoresistance R xx exhibits a sinusoidal, up to a factor of 5, magnitude variation with rotating the polarization angle θ.The maximal amplitude shows up generally at a nonzero θ, which is dependent upon the extremum in question, the 2D electron setup, the radiation frequency and the magnetic field orientation.
being the associate Laguerre polynomial, f (ε) = {exp[(ε−µ)/T e ]+1} −1 is the Fermi function at electron temperature T e , and ImG n (ε) is the density-of-states (DOS) function of the broadened Landau level n.
FIG. 1 :
FIG. 1: (Color online) Magnetoresistivity Rxx, electron temperature Te and energy absorption Sp are plotted versus ω/ωc under the irradiation of linearly polarized (LP) microwaves of frequency 40-GHz and incident power Piω = 17.5 W/m 2 at several different polarization directions (θ = 0 o , 30 o , 60 o , 90 o , 120 o and 150 o ) for a system described in the text at temperature T = 1.5 K. | 2012-11-06T07:40:57.000Z | 2012-07-09T00:00:00.000 | {
"year": 2012,
"sha1": "ac6d3ed73aefd92204b0b6743843f751e2973c95",
"oa_license": null,
"oa_url": "https://epub.uni-regensburg.de/26694/1/PhysRevB.86.045303.pdf",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "ac6d3ed73aefd92204b0b6743843f751e2973c95",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Materials Science"
]
} |
7665900 | pes2o/s2orc | v3-fos-license | First record of Naineris setosa ( Verrill , 1900 ) ( Annelida : Polychaeta : Orbiniidae ) in the Western Mediterranean Sea
The Polychaeta Naineris setosa (Verrill, 1900) is reported for the first time in the Western Mediterranen Sea. Several specimens were collected in 2012 and 2013 in Boughrara Lagoon (Tunisia). The vector of introduction of N. setosa likely commercial shipping because Gabes is an international harbour with dense shipping traffic. A detailed description of specimens from Boughrara Lagoon is provided in this paper.
Introduction
Introduction of new species is one the major factor affecting biodiversity in the Mediterranean Sea in the past few decades.At least 986 alien species are known in the Mediterranean, 775 in the eastern basin, 308 in the western basin, 249 in the central basin, and 190 in the Adriatic Sea.These alien species represent 5.9% of the total biodiversity (Zenetos et al. 2010(Zenetos et al. , 2011(Zenetos et al. , 2012)).In the light of its geographic position, between western and eastern basins, the coasts of Tunisia appear to be very sensitive to colonization by exotic species.To date, 124 marine invasive alien species have been recorded along the Tunisian coastline, of which 11 are polychaetes (UNEP-MAP-RAC/SPA 2013).
The orbiniid Polychaete Naineris setosa (Verrill, 1900) is considered to be a subtropicaltropical American species (Blake and Giangrande 2011).The species described as Aricia setosa by Verrill (1900) from an intertidal beach in Bermuda is distributed in Puerto Rico (Treadwell 1901), Florida (Hartman 1951) and the Gulf of Mexico (Rioja 1960;Perkins and Savage 1975;Hernandez-Alcantara and Solis-Weiss 1989).The first Eastern Pacific record was from Acapulco (Mexico) (Hartman 1957).N. setosa is also present in Costa Rica, Ecuador (Galapagos Islands) (Blake andGiangrande 2011), Belize (Solis-Weiss andFauchald 1989) and Brazil (Rizzo and Amaral 2001).Recently, N. setosa was reported for the first time outside its native area, in aquaculture facility near Brindisi (Adriatic Sea, Italy) (Black and Giangrande 2011).However, this species was not included in the updated list of alien species for the Mediterranean Sea because, after being one of the most abundant polychaetes (500 ind.m -2 ), this species disappeared from the area (Zenetos et al. 2012).The present paper reports the first record of N. setosa along the Tunisian coasts.
Study area
The Lagoon of Boughrara is located on the south-eastern coast of Tunisia.This small bay (about 500 km 2 ) is bounded on the north by the Djerba Island and on the south by the mainland.The bay connects with the Gulf of Gabes by the Ajim-Jorf channel (2.2 km wide) and by the El Kantra Channel (160 m wide) (Figure 1).The average depth is about 4 m with a maximum of 16 m in the center of the lagoon.The salinity is very high especially in summer with an average value up to 42 (Ben Aoun et al. 2007).Boughrara Lagoon is a vulnerable environment with limited access to open sea and is submitted to negative effects of aquaculture activities, fishing ports, sewage outfalls, and industrial wastes (organic matter, fluorides, cadmium, acids, etc.) (Ben Aoun et al. 2007).This study was conducted within the framework of a National Institute of the Sciences and the Technologies of the Sea (INSTM) project and aimed to evaluate the general ecological status of the Lagoon of Boughrara.
Material examined
Benthic macrofauna was sampled from 15 stations in Boughrara Lagoon between 2012 and 2013 during summer (August) and winter (January) seasons (Figure 1, Appendix 1).Samples were collected by scuba divers with a quadrat (1 m², 10 cm depth) and sieved through 1 mm mesh.The remaining fraction was fixed in 7% formalin and later transferred into 70% ethanol.All specimens were examined under a Nikon SMZ1500 stereomicroscope and one specimen photographed with a Nikon DS-Fi 2 camera.Specimens were measured with the NIS-Elements Analysis software (Nikon Instruments Inc., Melville, New York, USA).
Results
Fifty specimens of Naineris setosa were collected in 2012-2013 from four different sites in the Boughrara Lagoon (Figure 1, Appendix 1).This species displayed highest abundances in summer at stations "Ajim Harbour" and "Gulf of Gabes" with 21 and 19 individuals m -2 , respectively.Abundances were lower in winter (max 6 ind.m -2 ).N. setosa was always collected from in muddy sediments.
Abdomen from setiger 25 with narrow notopodial postsetal lobes, becoming short in the posterior abdomen.In parallel, size of neuropodial postsetal lobe sharply reduced and finally tapering to narrow tips.
Thoracic notopodia with long crenulated capillaries setae arranged in 3 rows.Neuropodial thoracic setae are all crenulated capillaries arranged in two bundles: one, with about 8 rows; and two, the lower group, about 4 rows.
In abdomen, the number of setae is more reduced than in the thorax.The notopodia have crenulated capillaries and 1-2 furcate setae (hard to see).The neuropodia with crenulated capillaries and 2-3 straight, bluntly pointed uncini more posteriorly.
Remarks:
In Mediterranean Sea, only two species of Naineris are recorded: N. laevigata (Grube, 1855) distributed apparently from Italy to south of France (Fauvel, 1927), and N. setosa (Verrill, 1900) from the Adriatic Sea, Italy (Blake and Giangrande, 2011).The main difference between the two species is that N. laevigata presents subuluncini in thoracic neuropodia, whereas N. setosa presents only crenulated capillaries (Fauvel 1927;Blake 1996;Solis-Weiss and Fauchald 1989;Blake and Giangrande 2011).Furthermore, the prostomium of N. laevigata is rounded and the branchiae appears often from setigers 7 or 8 (Fauvel, 1927).Specimens found in this study agree with description of N. setosa from Solis-Weiss and Fauchald (1989) and Blake and Giangrande (2011) with presence of prostomium broadly truncated, branchiae appearing on setiger 6 and presence of crenulated capillaries only in thoracic neuropodia.
Among the material examined by Solis-Weiss and Fauchald (1989), variations were found: prostomium rounded or T-shaped; eyesposts scattered or ranged in two or four sickles; 13 or 23 thoracic setigers; and thoracic neurosetae ranged in one or two bundles.All these characteristics appear to be correlated with animal age (Solis-Weiss and Fauchald 1989; Blake and Giangrande 2011).In accordance with these characteristics, animals found in this study are adults: prostomium Tshaped, few eyespots ranged in two sickles, 21 thoracic setigers and thoracic neurosetae ranged in two bundles.
Discussion
The vector of introduction of Naineris setosa in the Lagoon of Boughrara cannot be easily determined.According to the Mediterranean Action Plan for Invasive species (UNEP-MAP-RAC/SPA 2005) and in agreement with recent evaluation of pathways at Pan European Level (Katsanevakis et al. 2013), shipping (ballast water and sediments, anchoring, and fouling) is the second main vector of species introduction into the Mediterranean Sea after migration across the Suez Canal.N. setosa is probably not a Lesseptian species as this polychaeta is a priori not present in the Indo-Pacific region.Blake and Giangrande (2011) suggested that N. setosa was introduced in Italy via aquaculture.Despite the presence of two fish farms in the Boughrara Lagoon, this way of introduction may be dismissed because of the origin of the cultivated fishes: Tunisia, Republic of the Philippines, and France.As N. setosa was collected in four sites directly connected to the Gulf of Gabes, transport via ship hulls or ballast waters has been hypothesized.Indeed, Gabes is an international harbour with a favoured place in the middle of the Mediterranean Sea, between the oriental and occidental basins.It receives a great deal of traffic from Europe (mainly from Italy), Asia, the USA, and Mexico and is classified among the main harbours for ships carrying crude oil in the Mediterranean Sea.Indeed, there is extensive traffic of crude oil between Italy and Tunisia frequent contact between Brindisi in Italy (where N. setosa occurs; Blake and Giangrande 2011) and Skrika in the Gulf of Gabes.Even if maritime transport seems to be the most likely vector of introduction of this species in the Lagoon of Boughrara, it is difficult to identify the exact origin of N. setosa which could arrive either directly from the western Atlantic (USA, Mexico) or as a result of secondary spread from Italy.
Naineris setosa is clearly tolerant to severely degraded environmental conditions.During this study, specimens were found in four sites classified as polluted: Ajim Harbour supports many fishing activities; the Gulf of Gabes receives industrial waters charged in phosphate; the fish farming site has a heavy aquaculture influence; and El Kantra 2 is perturbed by extension of the pass of El Kantra.These four stations are characterized by high salinity (between 38.5 and 43.1) for both winter and summer seasons, muddy sediments, some areas of anoxic mud (e.g., the Ajim Harbour and fish farming stations), and by polluted waters.Blake and Giangrande (2011) indicated that the existence of the dorsal crests on the abdomen of N. setosa is atypical and is perhaps an adaptation, with the very vascularised branchiae, to low oxygen levels in the sediments.
With the extensive levels of intra-Mediterranean shipping traffic, Naineris setosa might be present in other Mediterranean harbours.As the Mediterranean Sea connects naturally with the Atlantic Ocean through the Strait of Gibraltar and with the Black Sea by the straits of Bosphorus and the Dardanelles, the presence of this species in other adjacent areas should be watched closely.Finally, the impact of nonindigenous species on the native ecosystem can be significant and may be extremely damaging to native species, local environment, and human activities (Lavesque et al. 2010).Further studies will be necessary to evaluate the major factors driving the spread of N. setosa in the Gulf of Gabes and to understand the consequences of its presence on the structure of benthic communities. | 2017-10-11T00:19:39.765Z | 2014-06-01T00:00:00.000 | {
"year": 2014,
"sha1": "a97beebd2d7abbdcf30808c68f88ac463ec3fddc",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3391/bir.2014.3.2.05",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "4b5cb4c67b548a7e18f50feacf55eca4a8e2a657",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Biology"
]
} |
263835755 | pes2o/s2orc | v3-fos-license | Dissecting Overheads of Service Mesh Sidecars
Service meshes play a central role in the modern application ecosystem by providing an easy and flexible way to connect microservices of a distributed application. However, because of how they interpose on application traffic, they can substantially increase application latency and its resource consumption. We develop a tool called MeshInsight to help developers quantify the overhead of service meshes in deployment scenarios of interest and make informed trade-offs about their functionality vs. overhead. Using MeshInsight, we confirm that service meshes can have high overhead—up to 269% higher latency and up to 163% more virtual CPU cores for our benchmark applications—but the severity is intimately tied to how they are configured and the application workload. IPC (inter-process communication) and socket writes dominate when the service mesh operates as a TCP proxy, but protocol parsing dominates when it operates as an HTTP proxy. MeshInsight also enables us to study the end-to-end impact of optimizations to service meshes. We show that not all seemingly-promising optimizations lead to a notable overhead reduction in realistic settings.
INTRODUCTION
Service meshes are becoming the de facto communication substrate for cloud applications.A survey by the Cloud Native Computing Foundation (CNCF) found that 90% of the organizations are using or evaluating them [29].Service meshes solve important problems related to communication among loosely-coupled microservices-the dominant paradigm for cloud applications [1,42,50]-including discovering where services are located, establishing secure connections, and handling communication failures.They also offer advanced capabilities such as rate limiting, load balancing, and telemetry, via additional message processing filters.
However, the overhead of service meshes is a key concern for both their users and developers [4,23,35,68].In most common scenarios where service meshes are employed, application traffic traverses software proxies called sidecars, which increases request latency and consumes more resources.Service meshes can add tens of milliseconds to request latency in some settings [4], and they can consume multiple CPU cores even at moderate load [23].These overheads can degrade the end-user experience, increase operational costs, and decrease revenue [40,52].
Today, the only way users can learn service mesh overhead is by running their application with and without service meshes and measuring the increase in latency or CPU usage.This black box approach, employed by some studies [4,23,28,68] and the MeshMark tool [13], has significant limitations.There are many ways to configure a service mesh, each with different performance implications, and it is infeasible to measure them all.Thus, application developers cannot evaluate functionality-performance trade-offs and discover the best way to configure the service mesh for their specific application and deployment environment.
The concerns around the performance of service meshes and the lack of alternatives to black box measurements is apparent in the documentation of Envoy (a popular sidecar) [20]: We are frequently asked how fast is Envoy?or how much latency will Envoy add to my requests?The answer is: it depends.Performance depends a great deal on which Envoy features are being used and the environment in which Envoy is run.• • • We encourage users to benchmark Envoy in their own environments with a configuration similar to what they plan on using in production.
The black box measurement approach has another limitation.There are several ongoing efforts on optimizing service mesh overhead [3,12].Without a detailed accounting of the underlying contributors to overhead, it is hard to quantify the effectiveness of such optimizations, especially as it relates to the end-to-end impact on a wide range of applications.
Instead of a blackbox approach, MeshInsight models the sidecar's operation as a combination of independent components (e.g., read, write, and IPC) and key aspects of the workload.By characterizing individual components, we can estimate the overhead of a sidecar based on the components used in a given configuration.Then, given workload characteristics such as the call graph, request rate, and message sizes, MeshInsight can estimate the end-to-end overhead for the application without needing to deploy the application under that configuration.Similarly, if a service mesh developer wants to evaluate an optimization that reduces the overhead of some component(s), MeshInsight can estimate its end-toend performance impact across a wide variety of application workloads.
Using MeshInsight, we study the overhead of Envoy.Envoy is the dominant sidecar, used by many service meshes [9, 10,14,26,27,45] including Istio [24], one of the most popular service meshes today [2].Our experiments confirm that service meshes can have substantial performance penalties.Across two popular benchmark applications [8,42], depending on the configuration, request latency increases by 27-269% and CPU usage increases by 42-163%.In a large dataset of microservice-based applications [58], we find that using Envoy increases latency by up to 100 ms and consumes 200 more virtual CPU cores for a quarter of the call graphs.We also find that, for a given service mesh configuration, the overhead for different applications varies by multiple orders of magnitude.Such high variation based on service mesh configuration and on application characteristics validates the need for a tool that developers can use for their specific deployment scenarios.
MeshInsight's compositional approach provides insight into the sources of overheads.We find that when configured as an HTTP or gRPC proxy, protocol parsing alone represents 63-77% of the total overhead.When configured as a TCP proxy, most overhead stems from inter-process communication (IPC) and socket write operations.The overhead of individual filters (i.e., network functions) varies significantly.Some increase latency by as little as 3% atop the baseline overhead, while others increase it by as much as 85%.
We also use MeshInsight to evaluate the end-to-end overhead reduction for service meshes using two Linux kernel features: 1) Unix domain sockets in place of TCP connections for IPC, and 2) zero-copy writes for TCP sockets.Given that IPC and socket writes are substantial contributors to overhead, we wanted to understand if these features help reduce overhead.We model the performance of our components in the presence of these two features and evaluate their end-to-end impact using the large microservices dataset [58].We find that Unix domain sockets are helpful, reducing the average latency overheads by 27% and the CPU overheads by 18%, for TCP proxy.But, surprisingly, zero-copy socket writes have negligible improvements.For small message sizes that are common to microservice workloads, the additional system calls in implementation negate the savings from avoiding the copy operation.
We make two key contributions in our work: • We develop a tool that application and service mesh developers can use to quantify overhead.It allows application developers to make informed decisions regarding trade-offs between performance and functionality while enabling service mesh developers to assess the overall impact of their optimizations.MeshInsight accomplishes this without requiring the deployment of each setting in a production environment.
• We conduct the first systematic study of service mesh latency and CPU overheads across a wide range of realistic microservices.It confirms that the overhead is generally high, but it also shows that it varies significantly based on the sidecar configuration and application characteristics.Our study also provides a detailed accounting of performance overheads and informs future optimizations.
Our tool and the results of our study will inform work on improving the performance and reducing the resource consumption of service meshes, which now play a central role in the modern application ecosystem.MeshInsight is available at https://github.com/UWNetworksLab/meshinsight
BACKGROUND
Service meshes emerged to solve challenges that arose when applications moved from monoliths to the microservices architecture.Instead of being a large binary, applications are composed of multiple microservices.Figure 1 shows an example where the BookInfo application from Istio [11] uses five microservices.Each microservice is run as an independent process (or container), often on different hosts, and can be scaled independently.Such decomposition enables agile application lifecycle management, fault-tolerance, scalability, and reuse of building blocks across applications [44].
It is common for modern applications to use tens of services [1,44,50].But decomposing applications into multiple microservices creates new problems as well.What used to be a function call now becomes an RPC over the network.Developers must figure out how services discover, communicate, and authenticate with each other.They must also figure out how to monitor and secure inter-service communication and how to handle failures.Early adopters of microservices built custom communication frameworks to solve these problems [19,21].Service meshes emerged to solve them in a reusable manner, and they further evolved to provide other functions such as rate limiting and load balancing.
The convenience of service meshes comes at the cost of performance and resource overhead.The overhead could be in the control plane or the data plane.The control plane handles service discovery, metric collection, and certificate management, and it appropriately configures the data plane.For example, in Figure 1, to balance load across multiple instances of the Product service, the Frontend data plane can spread Frontend-to-Product connections across different Product instances (multiple instances not shown in the figure).The service mesh control plane tracks where all the instances are running and configures the Frontend data plane accordingly.We focus on the data plane overhead as it impacts every request and is on the critical path of user experience.
The most common implementations of the service mesh data plane use software proxies called sidecars.The sidecar is a separate process that is co-located with each instance of an application service, which enables the sidecar to mediate all of the service's network access to apply network policies, enforce encryption, and log statistics.Recent service mesh proposals (e.g., Cilium [12] and Ambient Mesh [22]) aim to inline (in the kernel) some data plane functionality and propose per-host proxies.However, kernel acceleration does not eliminate sidecars, which are still needed for complex functionality, and sidecars have some significant benefits over per-host proxies.Thus, sidecars are likely to persist [16,17] for the foreseeable future.We focus on the sidecar model and leave the analysis of hybrid models to future work.
Sidecar Data Path Figure 2 shows the data path for both outbound and inbound traffic.In this paper, while we apply our modeling approach and analysis to Envoy [18], other sidecar proxies [31] share this architecture.During initialization, the control plane adds iptable rules that redirect all inbound and outbound traffic to the sidecar.As a result, logical connections between microservices are broken into three separate connections: two connections between the sidecars and their associated microservices, and one connection between the sidecars.When the application sends a message, it executes a write system call.The kernel network stack and the loopback device process the message and notify the sidecar.The sidecar then reads the data from the kernel, processes it, and writes it back to the kernel.Finally, the NIC driver transmits the message.Similar to the inbound traffic, upon receiving a message, the sidecar intercepts the message before it is passed to the application.The exact processing done by the sidecar depends on its configuration; Figure 3 shows the general data flow.When a message arrives, it is parsed based on the protocol of choice (e.g., TCP, HTTP, gRPC).In TCP mode, traffic is treated as an opaque TCP stream; in HTTP and gRPC mode, messages are parsed as per the protocol, which enables additional functionality specific to the protocol.After the message is parsed, it is processed by one or more filters.Filters are short programs that process individual messages and implement network functions like traffic monitoring, rate limiting and fault injection.Envoy filters can be written using 1) C++ code, 2) Lua scripts, or 3) WebAssembly modules.All of Envoy's 40+ built-in filters are C++-based, while application developers tend to write custom filters using Lua or WebAssembly.
Given the data paths in Figures 2 and 3, we can see a number of sources of overhead.First, without a sidecar, to send a buffer, the kernel copies the application buffer into a kernel buffer, which the NIC can subsequently access through Direct Memory Access (DMA).With a sidecar, the buffer must additionally be copied to the sidecar buffer and then back into a kernel buffer (resulting in two extra copies).Second, there are many additional system call invocations, such as the sidecar waiting for data through epoll and reading and writing buffer from/to kernel.Finally, using sidecars incurs extra IPC invocations (e.g., the loopback interface in Istio).In addition, sidecars may need excessive computation on the buffer, including parsing the data stream into data structures for HTTP, JSON, and RPC data formats.
MODELING SERVICE MESH OVERHEAD
Our goal is to characterize the overheads of service meshes in a way that supports two classes of developers.First, we want to enable application developers who are looking to deploy service meshes to understand overhead as a function of service mesh configuration (e.g., proxy and filter configurations), so they can appropriately trade off functionality and overhead.Second, we want to enable service mesh developers to understand the impact of their optimizations (e.g., zero-copy writes) on a wide range of real-world applications without having to necessarily implement them fully.Neither class of developers is well-served by currently available tools and techniques.
A key challenge we face is the large operational space of service meshes.An application may be running Envoy in one of many possible ways, each with different performance implications.There are at least three dimensions of variations: ) type of proxy (e.g., TCP, HTTP, and gRPC); ) which filters are used; ) application workload, where the salient characteristics are message sizes and rates.These variations, and their combinatorial combinations, mean that a black-box approach to measuring overhead is a non-starter.
We must instead model the overhead of finer-grained components and then compose the individual component overheads to predict overhead for a given deployment scenario.Our models are a concise representation of performance overheads over the entire operation space, including both the service mesh configuration and application workloads.This modeling approach allows us to reason about a service mesh's large operation space.For example, for an application developer to choose an appropriate configuration for a service mesh, instead of benchmarking every possible configuration, we can use our models to quickly predict the performance overheads for any service mesh configuration for their workloads.
This approach is necessarily an approximation.It characterizes each component in isolation, ignoring interactions that may arise when they contend for resources (under high utilization).The overhead of contention is hard to model and sensitive to co-located workloads and background system activity [41,63].Consequently, MeshInsight predicts the best-case overhead (i.e., a lower bound) and cannot be used to predict tail latency.As we show later, MeshInsight's predictions track actual overhead well enough to effectively distinguish between the configurations with different overheads.
MeshInsight overview To be practical, we must carefully choose the granularity and nature of the components.If they are too fine-grained, accurately characterizing their overhead (and composing them) will be difficult; if they are too coarse-grained, we'll suffer from the same challenge as with black-box measurements.The overheads of the chosen components should also be largely independent, allowing us to easily compose their overheads to estimate total overhead.The total performance overheads of a service mesh are the sum of these components.
Table 1 shows the component-level breakdown we use.The first four represent the sidecar's interactions with the application and the kernel: 1) inter-process communication (e.g., loopback) between the application and its sidecar proxy, 2-3) writes and reads from the sidecar proxy to the kernel, which also include the data copies (note that write also includes the TX's TCP/IP processing), and 4) blocking waits on socket ready notifications (e.g., from epoll).The next three breakdown message processing inside the sidecar: 5) parsing the messaging protocol (e.g., HTTP), and 6) processing done by user-configured filters; and 7) baseline processing by the sidecar to move packets between the two components above.
Figure 4 shows the workflow of MeshInsight.It has an offline profiling phase and an online prediction phase.The offline phase generates performance profiles of components, and the online phase predicts the service mesh overhead based on these profiles, sidecar configuration, and application workload.
MeshInsight models the performance of a component as a function of message size and request rate because these two workload properties are the primary determiners of performance.In our experience, simple linear functions of these two properties suffice (see next section for details).The profile of a component is specific to the platform, which includes the hardware (e.g., CPU, memory), OS, and Envoy version, but independent of individual microservices.Overhead predictions are made only for previously profiled platforms. 1hile we focus on Envoy in this work, our approach can be extended to other existing service mesh proxies (e.g., linkerd) with minor configuration changes.
The online prediction phase uses performance profiles from the offline phase to provide performance predictions in the context of a specific application deployment scenario.These predictions are based on an extended call graph (described in Section 4.2) that captures application behavior.
MESHINSIGHT DESIGN
We now describe the design of MeshInsight in more detail.
Building Component Profiles
In the offline phase, for each component in Table 1, MeshInsight builds performance profiles that characterize components' message processing latency and CPU usage as a function of message size and rate.The profiler exercises the components under a few different settings and then extrapolates their performance to other settings.
We conduct two types of profiling runs: ) sidecar configured as TCP, HTTP, or gRPC proxy, with no additional filters; and ) sidecar configured with only the filter(s) of interest.Profiling uses an echo server paired with a sidecar.We use wrk [32] and wrk2 [33] for load generation as well as high-precision measurement of end-to-end latency.To measure request processing latency and CPU without contention effects, the wrk client generates requests such that at most one request is outstanding at any time.
To quantify the overhead of individual components during a run, we exploit the fact that all components (except filterssee below) correspond to specific kernel-or user-space functions.We measure the latency of each component using a modified version of BCC's funclatency [15], an eBPF-based tool that uses Kprobe and Uprobe to monitor time spent on a function.We measure CPU usage of each component using the standard sampling technique [46], which allows us to quantify the CPU consumption of any function.We identify the function for a component using a mix of the name, the input/output, and process ID.
Filters present two wrinkles in this process.First, most of them do not have a known function.We measure the overhead of a filter by subtracting the overhead of a setup without the filter from an otherwise identical setup with it.Second, the overhead of some filters depends significantly on certain configuration parameters.For instance, the overhead of the Rate Limit filter, which limits the network traffic to a service, depends on whether the developer needs to limit traffic rate on the entire service mesh (global rate limiting) or on per service instance basis (local rate limiting).Likewise, the overhead of the Tap filter, which logs traffic, depends on whether the log is written to a file or sent over the network.In our profiling, we treat such filters as different components.
Data from the profiling runs and the following assumptions (empirically validated in §5) enable us to estimate the sidecar's overhead in any setting.
• A1: The total latency and CPU overhead of the sidecar is the sum of components' overhead.
• A2: Latency overhead is a linear function of message size (see below).
• A3: CPU overhead is a linear function of message size and proportional to message rate.Thus, to estimate the overhead of the sidecar in a configuration with multiple filters, which has not been measured directly, we can add the overhead of the base configuration without filters and the overheads of filters that are employed.
To estimate the latency of a component for a message size , per prior work [54,72], we use this linear function: + × , where is the base message processing latency and is the per-byte latency.We assume that latency for processing a message does not vary based on the request rate.
To estimate CPU consumption of a component for request rate , we use the following linear function: × ( + × ), where denotes the baseline per-message CPU usage and denotes the per-byte CPU usage.
The impact of message size captured in the two equations above assume that a message is processed by each component (e.g., read or written) as one unit.This assumption may be violated for huge messages that are split into multiple units.Size threshold at which a message is split may be overridden by applications or Envoy, but is typically at least a few KB; it was always above 4KB for platforms that we have experimented with.The implication for MeshInsight is that it will underestimate the overhead for messages that are split; the actual latency and CPU cost are higher for such messages.We quantify this underestimation in §5.3.Fortunately, the vast majority of message sizes are small [61,69,74], and the impact of our modeling approximation is therefore minimal.
To estimate ( , ) and ( , ), MeshInsight profiles components using five different message sizes (100B, 1KB, 2KB, 3KB, 4KB) and four different request rates (25%, 50%, 75%, and 100% of the application's maximum achievable throughput).We then apply linear regression to the collected data.These two tuples represent the latency and CPU profile of a component for a particular platform.The message size and rate are chosen empirically; we find that these choices have minimal impact on profile accuracy given sufficient data points.
Predicting Overhead
Application developers can use MeshInsight to estimate service mesh overhead in any deployment scenario of interest by providing an extended call graph (ECG).The ECG captures the details of the deployment and interactions among microservices in response to a request.
Formally, an ECG is a tuple ( , , , ), where = { 1 , 2 , ..., } is a set of vertices representing microservice instances; is a map from microservice instances to platforms; and is map from microservices instances to configurations (i.e., protocol and filters). is a DAG (directly acyclic graph) of microservice invocations.Each node in the graph is an invocation, and the edge represents an invoked-after relationship.An invocation is a tuple ( , , , , ), where is the calling microservice (empty if called externally), is the microservice invoked, , are the expected size (in bytes) and rate of messages (in requests/second) along this invocation, and is the timestamp of the invocation without a service mesh.ECGs are extracted automatically with an openTelemetry-standard distributed tracing tool like Jaeger [25] and a trace analysis tool CRISP [76].Since Jaeger traces do not contain the message size and rate for each invocation, we collect these using custom scripts.MeshInsight can also handle dynamic call graphs by using all available Jaeger traces and generates a probability distribution of the overhead for an API. Figure 5 shows an example of for the Bookinfo application in Figure 1.An external client calls Frontend, which in turn calls Product.The Product service calls Reviews and Details in parallel.Reviews calls Ratings and responds to Product after getting the response.Product responds to Frontend after both Details and Reviews respond.Finally, Frontend responds to the external client.
Generating Predictions.Given an ECG, MeshInsight estimates the latency and CPU overhead.It starts by computing the overhead of each invocation.For an invocation ( , , , , ), the overhead is based on messages of the given size and rate leaving the service at and entering the service at .The sidecar configuration for these services tell us which components are exercised.We compute the component-level overhead (using and ) and then sidecar-level overhead by summing component overheads.In some configurations, not all components are subjected to the same and .For instance, if a fault injection filter, which can be configured to drop some messages, is present, downstream filters will see a lower message rate.Currently, we ignore such intra-sidecar variations, though our model can be extended to account for them if needed.
MeshInsight computes end-to-end request-level overheads using sidecar-level overheads computed above.For the CPU, it simply adds all the sidecar-level overheads.For latency, simple addition does not work because computations can happen in parallel, and we thus need critical path analysis.In the Bookinfo example above, the end-to-end latency depends on which of Details or Reviews is slower to respond to Product, and the latency of the faster one is not critical.To this end, MeshInsight uses to compute the critical path of the application and only report the end-to-end latency overheads along the critical path.It is important to note that introducing a service mesh might alter an application's critical path.To accommodate such changes, MeshInsight executes a critical path analysis (i.e., CRISP) after adding the overhead predictions for each hop to the ECG.
Quantifying the impact of service mesh optimizations.The prediction techniques described above can be used by service mesh developers to estimate the end-to-end impact of their optimizations.To enable this estimation, service mesh developers need to provide information on the impact of their optimization for the component(s) they have optimized.That is, they need to update the performance profiles.This update may be based on the estimated impact of their planned optimization (which has not been implemented yet).For example, the developer may estimate that their optimization will lower the baseline write overhead by 50%.Alternatively, new performance profiles may be based on running the MeshInsight profiler after implementing the optimization.
Once information on new performance profiles is provided, MeshInsight can estimate the overhead of the new system and how much improvement it brings compared to the original.
CHARACTERIZING SERVICE MESH OVERHEAD
We now use MeshInsight to characterize the overhead of service meshes for benchmark applications and validate the accuracy of its predictions.We study overhead for both basic deployments and when filters are added.We also study how individual components contribute to the overhead and the impact of message size and rate.In the next section, we demonstrate how MeshInsight can help application and service mesh developers.Our experiments use Cloudlab [37] machines with two 16-core Intel Xeon Gold 6142 CPUs (2.6 GHz) and 384GB RAM, Ubuntu 20.04 LTS (Linux kernel v5.4.0),Kubernetes v1.12.5, Istio v1.13.0, and Envoy 1.21.0.We disable Turbo-Boost, CPU C-states, and dynamic CPU frequency scaling to reduce measurement variance.
Application Benchmarks.To characterize the overhead of service meshes on realistic applications, we consider two popular microservices benchmarks: Online Boutique [8] and Hotel Reservation [42].Online Boutique (Figure 6) has 11 microservices.It is a web-based e-commerce application where users can browse items, add them to the cart, and purchase them.Microservices are written in different languages (Python, C#, Java, and Go) that communicate using gRPC.Hotel Reservation (Figure 7) has 17 microservices and supports searching for hotels using geolocation, making reservations, and providing hotel recommendations.All microservices are implemented using Go and communicate using gRPC.
We deploy these applications on multiple Cloudlab hosts and consider two deployment scenarios: TCP and gRPC.(HTTP cannot be used because the applications are gRPCbased.)In TCP mode, all sidecars are configured as a TCP proxy that relays the message to the application service.In gRPC mode, the sidecars parse the gRPC stream and collect basic application-level metrics.
We consider three queries for each benchmark.For hotel reservations, the queries are: (1) (Q1) User: Checks the username and password.Calls User and its MongoDB.
(2) (Q2) Search: Returns available hotels based on location and check-in/check-out dates.Calls Search, Rates, Geo, Profile, Reserve and their MongoDB and Memcached storage.
For online boutique, the queries are: (1) (Q1) Index: Returns the home page.Calls Currency, Products, Cart and Ad.
We derive the ECG (extended call graph) for each query and intentially force a cache miss on each memcached access for a deterministic call graph.The message sizes in each ECG are based on the actual traffic of the application; we find that most messages are small (a few hundred bytes).We set the request rates close to the maximum the machine can sustain in gRPC mode for each query.
Figure 8 shows the base (without the service mesh) latency and CPU usage of the application along with predicted and measured overheads.The measured overhead is the latency or CPU usage of running the application with the service mesh minus that of running it without the service mesh.Latency is end-to-end, measured at the client.
We see that service meshes can be a significant source of overheads.When operating in gRPC mode, it can increase latency by up to 269% and consume 163% more CPU.In TCP mode, the overhead is lower but still substantial-latency increases by up to 61% and CPU usage by up to 92%.The next section sheds light on why these two modes behave differently.
Service mesh overhead will increase further as filters are added (see below).These high overheads, and differences in overheads in different settings, is why MeshInsight is needed to enable developers to appropriately trade-off performance and functionality.For instance, TCP mode might be enough for most services, and gRPC mode is limited only to services where extra control or visibility is required.
In Figure 8, we can also see that MeshInsight predictions track well the overhead for all queries across both benchmarks even though their individual performance varies significantly.Predicted overhead tends to be lower than measured overhead.Underestimation is more pronounced for gRPC.The primary factor is that in gRPC (and HTTP) mode Envoy adds extra headers and increases the response size.Larger responses slightly increase the latency and CPU use of downstream components that are outside of the service mesh (e.g., client-side processing).We currently do not model this effect.
Breaking down the Overhead
To shed light on sources of overhead, Table 2 shows how much each component contributes when Envoy is run in different protocol modes and without any filters.This experiment uses a synthetic application (echo server) with 100-byte messages at 30K requests per second.We use a synthetic application so that we can completely control the workload and can study the HTTP mode as well (which benchmark applications do not allow).For benchmark applications, relative contributions of different components in TCP and gRPC modes are roughly similar to those in Table 2.
We can draw several conclusions from this data.First, using HTTP and gRPC is substantially more expensive than TCP.The additional overhead of HTTP is roughly 4x for latency and 3x for CPU; for gRPC it is 5x for latency and 4x for CPU.The bulk of this additional overhead stems from protocol parsing, accounting for 63-77 % of the total overhead.
Parsing overhead is unfortunate because the application code will spend resources on parsing as well.Because of the way sidecar data path is organized, its parsing effort cannot be shared with the application.If enabled, such reuse will have a notable impact on the architecture of service meshes.
Second, we see that IPC overhead is notable (30% for TCP) and notification overhead is small (3% for TCP).This observation implies that asynchronous processing between the application and sidecar is not expensive by itself, but the default IPC mechanism in Envoy is expensive.We can tackle this overhead by either putting the sidecar in the same process as the application.However, this can have security implications because a malicious application may circumvent the network policies.In addition, upgrading a sidecar more complicated would require recompiling the application.Another option is to use a more lightweight IPC mechanism.We will consider the second option in the next section.
Third, some components have disparate impacts on latency and CPU.This disparity is most pronounced for "Sidecar Other", where its contribution to CPU overhead is far greater than its contribution to latency, but occurs for other components as well (e.g., Write).It implies that some optimizations may impact one type of overhead but not the other, and developers need to be careful that optimizing for latency does not hurt CPU and vice versa.
Impact of Filters
We find that sidecar filters can be quite expensive, even when configured as a no-op.We also empirically confirm that, as assumed by MeshInsight's model, their overhead is additive.
We study five different filters, covering all three ways to write an Envoy filter: 1) Fault Injection: a built-in, C++ filter that helps test the resilience to communication failures; 2) (Local) Rate Limit: a built-in, C++ filter that rate limits traffic to a service instance.3) Tap (File): a built-in, C++ that records traffic and is configured to log to a file; 4) Lua: a custom, noop filter written as a Lua script; 5) WebAssemtly: a custom, no-op filter written as a WebAssembly module.We add these filters on Envoy configured in HTTP mode.Table 3 shows the overhead of each filter inferred by MeshInsight when subjected to the same workload as the previous section (100 byte messages, 30K requests per second).We see that different filters have widely different overheads.The baseline overhead of C++ filter is low, as evidenced by the low overhead of Fault Injection and Rate Limit filters.The overhead of Tap (file) is high because of its interaction with the file system.On the other hand, even no-op Lua or WebAssembly filters have substantial overheads, with Lua being 3x more expensive for latency and nearly 5x more expensive for CPU.
To study the composability of filters' overheads, we consider five different filter configurations, each with a different way to combine filter types: 1) C : combines all three types of C++ filters; 2) C combines the Lua and WebAssembly filters; 3) C : combines the Lua filter with all three C++ filters; 4) C : combines the WebAssembly filter with all three C++ filters; and 5) : combines all five filters.
Figure 9 shows both predicted and measured overheads of each of these combinations.The measured overhead denotes latency and CPU usage with the filters minus that without the filters.We see that filter combination overheads can be quite high when multiple expensive filters are employed (something that the developers must avoid).We also see the predictions of MeshInsight, based on adding individual overheads on top of base HTTP proxy overhead in Table 3, are quite accurate.
Impact of Message Size and Rate
We now show how sidecar overhead increases with message size and rate.We vary message sizes from 100 bytes to 16KB.The upper end of this range is well beyond the maximum size that we directly profile (4KB).We vary message rates from 10K to 50K requests per second.
Figure 10 plots latency overhead for HTTP proxy without filters.The latency increase is similar for other protocols.We see that latency overhead increases slowly with message size.Going from 100 bytes to 16 KB (which represents a very large message), the latency overhead increases by 53 ms.This increase represents only a 30% increase for HTTP.The presence of filters does not significantly change the impact of message size on latency, as most filters operate on message headers, not payload (which has the most bytes).
We also see in Figure 10 that MeshInsight models the impact of latency increase well, though its prediction accuracy drops for very large messages (which are uncommon [61,69]).
As mentioned earlier ( §4.1), the reason for this lower accuracy is that messages larger than a few KB are split into multiple units, which have higher latency and CPU costs.
Figure 11 shows CPU usage for HTTP proxy.We see that the CPU overhead increases linearly with message sizes (for a fixed rate) and linearly with message rate (for a fixed size), and that MeshInsight tracks this increase well.Large messages have a relatively higher impact on CPU usage than they did on latency.CPU usage increases by 62% when the message size increases from 100 bytes to 16 KB.
HELPING DEVELOPERS PREDICT OVERHEAD
In addition to characterizing service mesh overhead in detail, MeshInsight has two more use cases: (1) helping application developers determine how to configure service meshes, and (2) helping service mesh developers evaluate potential optimizations.
Application Developers
Given the application call graph, MeshInsight can predict service mesh overhead for different configurations.With this knowledge, application developers can make the right trade-off between desired functionality and overhead.We demonstrate this use case using the Alibaba microservice traces [58].These contain over 20M call graphs from microservices-based applications, collected over 7 days in an Alibaba cluster.While most call graphs have a small number of microservices, 10% of them have over 40 microservices and the largest ones have thousands of microservices.We randomly select 1M call graphs for our experiments.The traces do not contain message sizes or rates; we assume these to be 100 bytes (because service mesh messages tend to be small) and 30K requests per second, which represents a moderate load.
Figure 12 shows the latency and CPU overheads for five different service mesh configurations: three protocol modes without filters, HTTP with filters, and gRPC with filters (all five filters described in §5.2).Since a call graph can have multiple paths, the latency overhead is computed for the critical paths, which we extract from the Alibaba trace based on both invocation's timestamp and response time.We see in Figure 12 that in any given configuration, the sidecar overhead across applications varies by multiple orders of magnitude.Even for the simplest configuration (TCP), latency overhead varies from 0.2 to 100 ms and CPU from 3 to 1000 virtual cores.Thus, different applications are impacted disparately by service meshes.
We also see that the overhead of different service mesh configurations varies by an order of magnitude for both latency and CPU.The median latency overhead is 0.2 ṁs in TCP mode but it is 2 ms in gRPC mode with filters, and the 75th percentile varies from 1 to 10 ms.Similarly, the median CPU overhead varies from 20 virtual CPU cores in TCP mode to over 200 in gRPC mode with filters.
These massive variations based on service mesh configuration and application characteristics are why we need a tool like MeshInsight using which application developers can learn the overhead of their specific deployment scenarios of interest.
Service Mesh Developers
MeshInsight enables service mesh developers to judge the impact of potential optimizations.There are several ongoing efforts to optimize service meshes [3,12].While such optimizations can be benchmarked in isolation, it is difficult to understand the end-to-end impact on real-world applications.MeshInsight allows service mesh developers to estimate the potential performance changes without excessive software prototyping effort.
To demonstrate this use case, we consider the potential use of two Linux kernel features in Envoy.Porting Envoy to use new kernel features is a big effort and may also introduce some functional limitations, so the Envoy developers may want to estimate the impact even before taking on this work.All they need to provide MeshInsight is an estimate of new performance profiles for various components.To estimate these speedups we enable these features in the context of a simple sidecar (with 676 lines of C++) that has the same data path architecture as Envoy.We then profile the components to learn their new performance profiles when these features are active.The Linux features we study are: Unix domain sockets In §5.1, we saw that IPC overhead is a significant contributor to overhead, adding at least 11 microseconds to latency and consuming 0.5 virtual cores, representing 30% of latency overhead and 15% of CPU overhead in TCP mode.By default, Envoy-to-application IPC uses a TCP connection, which traverses the TCP/IP network stack for the loopback interface.One potential option to reduce overhead is via Unix domain sockets [30], which is lighter weight than TCP sockets while providing the same API.When using a Unix domain socket, the kernel copies the data to kernel space and directly puts the constructed socket buffer on the receiving side's socket queue, avoiding the expensive network stack processing.
Reducing Write Overhead A second significant source of Envoy overhead is the latency and CPU usage of copying data, which is embedded in the Read and Write component.It is already noticeable with 100 byte messages (Table 2) and gets worse with larger messages.Linux kernel supports zero-copy TCP sockets starting from version 4.14.For write calls, applications can pin memory buffers in userspace; the kernel signals to the application after sending the buffers to enable garbage collection or re-use of the buffers.This eliminates the need for copying the data from the userspace to the kernel.Linux does not support zero-copy for read calls.
We use MeshInsight to evaluate the performance implications of these optimizations using the same Alibaba workloads as above.We consider TCP mode sidecars.The absolute savings for HTTP or gRPC mode will be similar but the relative advantage in those modes will be small because their performance is bottlenecked by parsing rather than IPC or data copy.
Figure 13 shows the latency overheads when using Unix domain socket and zero-copy write.We observe substantial improvement from the domain socket when message size is 100 Bytes: the average speedup is 0.71 ms.For 4KB messages, the average speedup 0.78 ms.However, because the latency prior to the optimization is higher (4.01 ms versus 2.88 ms), the relative decrease in latency is lower with 4KB.This result is in line with IPC overheads constituting less to the total latency overheads when the message size is large.
We also see that zero-copy writes bring negligible performance improvement in our setting.This optimization has additional performance cost for buffer lifecycle management [7].For the message size regime of most interest for microservices, the gains of avoiding a copy is negated by this additional cost.
Figure 14 shows the CPU overheads with the two optimizations.When messages are 100 bytes, the average CPU overhead reduces from 86 to 71 virtual cores with Unix domain sockets.The difference is quite substantial.When messages are 4KB, Unix domain sockets reduce the CPU overheads from 107 to 91 virtual cores.As for latency, zero-copy writes do not improve the CPU overheads notably, reducing the CPU overheads from 86 ms to 84 ms for 100-byte messages.
DISCUSSION
MeshInsight provides a systematic way to understand the overhead of service meshes and confirms that they can significantly increase latency and CPU usage of distributed, microservices applications.While we focused on the currentlydominant sidecar architecture, our work has implications also for alternative architectures being considered in the industry.
In-Kernel Proxies The Linux kernel has increasing support for extensibility using eBPF.Researchers have also proposed using safe languages, such as Rust, for kernel extension development [60].One approach to reducing the system call overheads and the data copy overheads (across userspace and kernel) is to implement a sidecar's functionality inside the kernel.Katran [6] offloads layer-4 load balancers into the kernel using BPF.It is now possible for Envoy to run a limited set of filters directly in Cillium [5], a popular framework for using eBPF on the network data path.However, our study shows that while removing the system call and data copy overhead can be useful for TCP proxies, it will offer limited performance improvement for HTTP and gRPC proxies because protocol parsing is the major overhead in those configurations.
Hardware Offloading Another direction being currently explored is to offload the sidecar logic to programmable network hardware [36].While this is a promising direction, there are substantial challenges.For example, it is still questionable if programmable network hardware can do complex layer-7 protocol processing efficiently.There are a variety of layer-7 protocols (e.g., HTTP, gRPC, MySQL).These protocols are more complex than the fixed functions people typically offload to programmable network hardware, such as firewalls, NATs, and layer-4 load balancers [55,59,75].In addition, many filters in sidecars (e.g., encryption) require reconstructing the original data stream, and this means programmable network hardware also needs to run TCP/IP packet processing (e.g., packet loss recovery, congestion control).Prior works on offloading application logic to programmable network hardware use UDP as the transport to circumvent this issue [66].
New Directions on Reducing Service Mesh Performance Overheads.Our work shed light on some new directions that are worth exploring.We observe that protocol parsing is a major overhead for HTTP and gRPC proxies.Unfortunately, using a sidecar today means that there is duplicated protocol processing.A sidecar parses an HTTP request from TCP streams, optionally modifies it, and serializes it into a TCP stream again.The application service receiving this stream has to parse the HTTP request yet again.
There are two potential methods to eliminate this double parsing.The first is by linking sidecars to libraries that applications use to parse various protocols.Another is to create a new transport protocol for efficient parsing by sidecars, which is possible in this context where both ends of the communication are sidecars.Both of these methods, however, have limitations.The first one does not work with unmodified applications, and the second one does not help when filters need to inspect HTTP headers added by applications.In future work, we will investigate these and other methods.
RELATED WORK
Our work is related to several threads of prior work.
Microservices.The microservice architecture has introduced many additional challenges to the research community, and many recent works have been trying to address various aspects of the microservice architecture, including reducing tail latencies [51,70,71], fault-tolerance [50], debugging [43], isolation mechanisms [34], energy efficiency [57], monitoring [53], and ensuring correctness [64].Because the application logic is decomposed into many independent microservices, this increases the amount of communication between various microservices within an application.
Today, operators increasingly use service mesh to deploy microservice applications for enabling better visibility and control over the communication between microservices.Unfortunately, the networking aspect of the microservice architecture and the service mesh has received little attention.We provide the first systematic study of the performance aspects of a service mesh, and we show that the service mesh can lead to major performance overheads.
Performance of the host network stack.Understanding the host networking stack performance is a common goal in many previous works.Peter et al. [65] breaks down the latency overheads of the Linux network stack.Neugebauer et al. [62] studies how PCIe affects network performance in host networking.Farshin et al. [39] examines how Intel Data Direct I/O technology (for NIC to access CPU's lastlevel cache directly) speeds up host networking performance.More recently, NSight [47] uses Intel Processor Tracing to diagnose latency in network stacks.
While we share data gathering primitives from these works, our focus is on the data path of service meshes (which traverses the network stack multiple times and has a substantial userspace processing component).We decompose the overhead of a sidecar proxy in the datapath of host networking, and we identify the key contributors to high overhead such as IPC and protocol parsing.
Performance of online services and network proxies.Many works have investigated the performance of network proxies and developed improvements such as hardware offloading [59,67,75], kernel offloading [12,68] and re-homing TCP connections [48].However, these works mostly focus on layer-4 proxies, while sidecars are layer-7 proxies.Our measurements of sidecars provide insights into performance bottlenecks of layer-7 proxies.We plan to combine insights from our study and techniques for improving the performance of layer-4 proxies to develop high-performance layer-7 proxies.
Similar to our approach, Stewart et al. [72] uses linear models to profile the performance characteristics of online services, treating them as black boxes.Conversely, our model breaks down the overhead of sidecars into distinct components, highlighting the main contributors to service mesh overhead.This not only spotlights the primary contributors but also empowers service mesh developers to assess the cumulative impact of their improvements.
Reducing inter-process communication overheads.Reducing IPC overheads is one of the oldest research topics in the operating system community.Immich et al. [49] and Venkataraman et al. [73] study the existing IPC mechanisms' performance on Linux.IPC performance is a critical design aspect for microkernels [38,56].The goal of our work is not to develop techniques to lower IPC overhead but to build a tool that helps evaluate the impact of such techniques on the end-to-end performance of service meshes.
CONCLUSION
This paper presents a tool, MeshInsight, that systematically quantifies the overhead of service mesh sidecars.Its compositional approach can analyze a wide range of deployment scenarios (i.e., the combination of service mesh configuration and application characteristics), without the need to directly measure them.This ability can help application developers pick the appropriate service mesh configuration for their specific application needs; as we showed using a large dataset of microservice applications, the overhead of service meshes can vary by orders of magnitude based on the configuration, and in a given configuration, the overhead can again vary by orders of magnitude across applications.
MeshInsight can also identify the primary contributors to the overhead in any scenario.We find, for instance, that IPC and socket writes are the main contributors when the service mesh is configured in TCP mode but protocol parsing dominates in other modes.This ability can help service mesh developers as they work to lower the overhead of service meshes.
Figure 3 :
Figure 3: Message processing inside a sidecar.Filters operate on individual messages and form a chain.
Figure 4 :
Figure4: Overview of MeshInsight (MI) workflow.The MeshInsight Database stores performance profiles associated with a hardware and software platform configuration, which are generated by the MeshInsight Profiler during the offline profiling stage §4.1.In the online phase §4.2, these profiles are used by the MeshInsight Predictor, in conjunction with Jaeger Traces provided by the user, to compute latency and CPU performance predictions for application deployment.The user can optionally provide a speedup profile, which MeshInsight uses to adjust the predictions accordingly.
Figure 5 :
Figure 5: An example call graph for the Bookinfo application.
Figure 8 :
Figure 8: Predicted and measured overhead for Online Boutique and Hotel Reservation applications.Base denotes latency and CPU usage when the application is run without a service mesh.The error bars for measured overhead are standard deviations.
Figure 9 :
Figure 9: Prediction results of different filters configurations.
Figure 10 :
Figure10: Impact of message size on service mesh latency overhead.X-axis is on log scale.
Figure 11 :
Figure 11: Impact of message size and rate on service mesh CPU overhead of service meshes.
Figure 12 :
Figure 12: Latency and CPU overhead for application call graphs in the Alibaba trace.
Figure 13 :
Figure 13: End-to-end latency reduction in the Alibaba trace with Linux features.
Figure 14 :
Figure 14: End-to-end CPU usage reduction in the Alibaba trace with Linux features.
Table 2 :
Contribution of different components to the overhead of a single sidecar instance in different protocol modes.The numbers report both inbound and outbound overheads.
Table 3 :
Latency overhead of five filters.The percentage in parentheses denotes the additional overhead atop baseline HTTP mode (without any filters). | 2023-10-12T12:38:04.464Z | 2023-10-30T00:00:00.000 | {
"year": 2023,
"sha1": "7a65d13bce291a9e3bd2aebf322bcbfe4b33d157",
"oa_license": "CCBY",
"oa_url": "https://dl.acm.org/doi/pdf/10.1145/3620678.3624652",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "7a65d13bce291a9e3bd2aebf322bcbfe4b33d157",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
219061221 | pes2o/s2orc | v3-fos-license | The Effects of Student’s Attitudes and Self-Efficacy on Science Achievement
One of the critical subjects in school that needs to be assessed is a science subject. Without a science subject, students cannot observe and understand a phenomenon on earth. However, results from an international study such as Trends International in Mathematics and Science (TIMSS), students in Indonesia performed poorly compared to students from another country. Furthermore, science is one of the essential education for children as it included in the STEM Education (Science, Technology, Engineering and Mathematics). From some empirical evidence, student’s attitude and self-efficacy (beliefs about their ability and skill) were found to be dominant predictors of student’s achievement, not excluded, science achievement. However, most of the research analyses the data under conventional regression analysis. Instead of under the structural modelling, and so the results can be considered carefully. This research will analyze a science achievement of Indonesian cohort, and the predictors would be self-efficacy, student’s attitudes toward science, school and teaching. Five hundred seventy-six data of students would be examined path analysis to answer the research questions. The results were found that both student’s attitude and self-efficacy had a significant direct role in determining student achievement in science. To be specific, attitude towards science had the most significant impact on science achievement, over self-efficacy. However, interestingly, the pattern of the effect from those predictors was different toward Science achievement. The practical aspects of the results of this study will be discussed in the discussion section.
Introduction
Science is critical knowledge in many aspects of children's daily lives. Through science education, children can understand the living, physical, material and technological components of the environment. This subject also can improve a student's logical and creative thinking. An understanding of science is necessary as society grapples with significant challenges and opportunities that confront our world.
In Indonesia, however, science is not the most favourite subject for Indonesian students. Students tend to avoid this subject in school. It is merely because most students think that science is strongly related to mathematics subject. Whenever the subject is a computational demanding and or require higher logical thinking, then students will hate it. According to the TIMSS study in 2015, student achievement in Indonesia was lower compared to other participants in the study, either in mathematics and science subjects. Even though both subjects required for human life, but Indonesian students seldom did well in these areas. The student often using mathematics function to solve a problem in the science subject. For that reason, the researcher will focus on science subject alone. Given such poor results, considerable research has sought to identify students' in-and out-of-school experiences that influence achievement and related outcomes, especially those that are alterable by policy intervention.
Science is an outstanding school subject, essential for school students to take. It is, in fact, the most important of all subjects as it allows students to question their environment to discover new things. Firstly, science is an essential school subject because it allows students to question their environment to discover new things. In science classes students are encouraged to question why things work and how they work, often carrying out experiments to find out first hand. When students come across new information, they cannot but ask and question how a particular thing works. Carefully using chemicals and observing reactions, students discover for themselves the effects of different scientific ideas. After experimenting students often set up their experiments to discover their "what if's…" putting into action their ideas and questions.
Secondly, science itself contains many aspects of the Math curriculum. Biology contains large vocabulary lists and terms that a student must have a good understanding. They use these to answer general questions just like in English. Physics takes part the mathematical side of science, incorporating numbers, calculating rates of reactions, time, motion and speed. Chemistry develops logical thinking, as students need to grasp the chemistry ideas and work out reactions to obtain a product.
By knowing the importance of science subject, several studies have been conducted to find out what the most critical aspects of increasing student achievement in science. For example, Aru (2014) found that the typical reasons that contribute to poor performance in secondary school are a poor methodology in science education, negative attitude towards science subjects among students and lack of resources such as textbooks and well-equipped laboratories. He continued, teachers should regularly attend the in-service training to equip themselves with a new and modern methodology of teaching science subjects.
Despite the critical importance of early interest and success in science in the early grades, limited research exists on individual factors that are related to children academic achievement in the science subject. Research in science achievement has focused on the high school level. Given the significance of school performance in science in early grades, one should examine the variables such as school motivation, academic engagement, and attitude toward science and how these are related to academic success in science subjects. By the time students are in primary school, they have already made implicit decisions about pursuing or not pursuing advanced mathematics and science courses, and earlier success determines these choices. A serious examination of the factors that affect mathematics and science achievement in middle grades is critical because it is in those years that students contemplate and negotiate future trajectories.
3-10
This is an open access article under CC-BY-SA license (https://creativecommons.org/licenses/by-sa/4.0/) From the background, as mentioned earlier, the author offers some research questions as follows: 1. How is the structural relationship of science achievement with self-efficacy, attitude towards the teacher, attitude towards school, attitude towards science?
2. What is variable that have the highest impact on science achievement?
The author will investigate the relationships between science achievement and some variables such as attitude towards science, self-efficacy, under path analysis approach. Some research on science achievement had been done through non-structural regression analysis (for example, a multiple regression analysis). Under as such analysis, it is hard to understand student's performance in science comprehensively. Therefore, this research will focus on the path analysis of variables.
Literature Review
Students' Attitudes and Science Achievement Koballa & Glynn (2007), in their research, said that attitudes as both the facilitators and products of science learning and research efforts focused on documenting student attitudes and their relationship to science achievement. The investigation of Weinburgh & Steele (2000) stated that attitudes are not reflections of what humans are pre-thought, but the attitudes are inferred from behaviours. In other words, a good attitude comes from right action, and conversely. According to Baldwin et al. (1999), prior knowledge and experience shape the learning process, which in turn affects a student's attitude. As a result, attitudes and beliefs of the student have become one of the targeted areas of study because some researchers have demonstrated that they play a role in how students benefit from their academic experiences, for example (Redish, Saul & Steinberg, 1998). Newhouse (1990) mentioned in his study that attitude was a critical factor in influencing human behaviour, including performance or achievement. Hacieminoglu (2016) said that the attitude variable was affected by personal opinion, and it caused by personal life experiences and education. Along with it, Anderson (1988) stated clearly that attitude and cognitive variables were psychological constructs that were considered to be a significant predictor of human behaviour, in this context, achievement. Related to that statement, Erkus (2003) pointed out that measuring the attitude variable and knowing the level of attitudes of individuals related to objects or circumstances are mandatory in many fields.
In this study, there are three types of attitudinal variables, namely, student attitudes towards science, student attitudes towards school and student attitudes towards the science teacher. In general, an attitude is a person's feeling positively or negatively to the enterprise of science. While Simpson & Oliver (1990), mentioned that attitude is whether a person likes or dislikes science. That definition can be applied to any attitudinal variables. Papanastasiou & Zembylas (2002) found that both Science achievement and attitude toward science had a non-recursive effect. It means that both variables can affect each other, depends on the characteristics of the educational system in the country. Nasr & Soltani (2011) found that there was a significant and little correlation between attitude towards biology and achievements in biology (r = 0.12; p < 0.05). Baker (1985) found exciting results from his study that students who earned grade A and B in Science achievement had a negative attitude towards science. Meanwhile, students who earned low scores in Science achievement had a more positive attitude towards science.
Furthermore, Baker (1985) stated that these unexpected results could be happened due to two reasons. First, higher ability students may find science was boring at school and affected their attitude. Secondly, the instrument he used was not measuring attitude towards science. Instead, it measured another construct.
4-10
This is an open access article under CC-BY-SA license (https://creativecommons.org/licenses/by-sa/4.0/) Cannon and Simpson (1985) found that attitude accounted for 1% of student achievement in science and the ability accounted for 17% of the achievement. However, later, by the end of the semester, the ability accounted for only 10% of the achievement, and attitude contributed 5% slightly higher than before. Nevertheless, according to Wang & Staver (1996), the relationships between student's attitude and Science achievement might vary from across countries, depending on their social systems and educational systems such as schools. Kan & Akbas (2006) found that both attitude and self-efficacy had a significant impact on science achievement. However, from their study, attitude towards science had a more significant impact on science achievement than self-efficacy itself. In another study, Anderman & Young (1994, in Hacieminoglu, 2016 found that from Hierarchical Linear Modelling (HLM) analysis indicated there was a positive yet small correlation between students' self-efficacy and mastery goal learning (r=.19, p < .01). Similar results with a higher index of correlation were found in the study by Middleton & Midgley (1997) that academic selfefficacy was related to the mastery goal learning and performance (r = .43, p < .01). In general, those studies found that there were positive relationships between academic self-efficacy and academic achievement. However, it is not specific in science achievement.
Self-Efficacy and Science Achievement Bandura (1986) defined self-efficacy as people's judgement of their capabilities to organise and perform a task required to achieve particular performances. Furthermore, Pintrinch & Schunk (2002) stated that selfefficacy could be described in the question such as "Can I do this task in this situation?". Bong & Skaalvik (2003), in their research, stated that self -efficacy represents what an individual perceives about his / her abilities and skills. Schruba (2008) identified that the value of self-efficacy lies in the students' evaluation of their views, the effectiveness of teaching and learning strategies, and the development of higher self-reported confidence in understanding science, particularly biology.
Related to science, the author identifies there is no specific research explicitly states relationship these two variables. However, Baldwin et al. (1999) noted that self-efficacy might affect a student's desire to become scientifically literate. A self-efficacy plays a vital role in student-level achievement. Joo et al. (2000) found that students' academic self-efficacy predict their performance on the given written test. Zeldin et al., (2007) stated that people are more likely perform well in the tasks they feel they can be successful and will be more likely to have higher self-efficacy when they have successful experiences.
In a general context, Pintrich & Schraben (1992) stated that students with high self-efficacy would initiate their activities that related to study and develop learning strategies, so in turn, will likely achieve more progress and better results in science achievement. Meanwhile, students with low self-efficacy tend to not interested to involve in the learning process and consequently, they will have shallow knowledge and low academic achievement. Different results obtained from the study by Jamil & Mahmud (2019), where they found selfefficacy has a small correlation with science achievement (r = 0.124, p > 0.05). Furthermore, female students have higher self-efficacy (M = 3.49, SD = 0.363) compared to male students (M = 3.28, SD -0.384).
Due to an emphasis on feelings and beliefs, the author considers attitude and self-efficacy affect student achievement, in this study is science subject. Schruba (2008) posits that attitudes relate to feelings; it is how student feel about science itself, either positively or negatively. Self-efficacy involves student beliefs about how "good" or "bad" they are in a particular subject. Baldwin et al. (1999) have stated about the relationship between belief, attitude and behaviour. Furthermore, Baldwin said, if a college student judges his/her ability to be lacking in science (opinion), that lack of confidence may lead to a dislike for science (attitude) and subsequent avoidance of science education (behaviour).
Instrument
Data for this research were drawn from the Trends in International Mathematics and Science Study (TIMSS) 4 th -grade student Indonesian cohort 2015. The TIMSS provides reliable and timely data on the mathematics and science achievement from the 47 participating countries, including Indonesia. TIMSS data have been collected from students at grades 4 and 8 since 1995 every four years, generally. TIMSS is sponsored by the International Association for the Evaluation of Educational Achievement (IEA) (https://timssandpirls.bc.edu/).
Initially, for Indonesia cohort, the cases were 4025. However, since the research was only using booklet 11 and 12, the sample size reduces become 580 (n = 580). Variables that include the analysis are science achievement, self-efficacy, students' attitude towards science, students' attitude towards a teacher, students' attitude towards school. Even though the TIMSS Study has many variable levels (student, parent, teacher and school background), but due to time limitation, this study will only focus on the student level.
There were 28 items in total for measuring science achievement. Those 28 items were total of two booklets (booklet 11 and 12). Self-efficacy was measured by seven items with four ordinal categories (agree a lotdisagree a lot). Students' attitude towards science was measured by nine items and four ordinal categories as well. Student's attitude towards teacher measured by ten items and with the same number of categories. Lastly, students' attitude towards school measured by seven items, also with four ordinal response categories.
Since the instrument used in this research from TIMSS data, the author will not check item validity. The instrument used in the TIMSS study has been well validated. Therefore, no need to further check whether the item valid or not. Even if it is needed, the detail of this application will be discussed in another article.
As cited in Pedazhur & Kerlinger (1982), path analysis was developed by Sewall Wright as a method for studying direct and indirect effects of variables hypothesised as causes of variables treated as effects. A path analysis was not intended to discover whether some variables are causing another. Instead, the path analysis was drawn from two bases that are knowledge and theoretical considerations. At least, there were two kinds of variables that can be identified in path analysis. First, an exogenous variable is one whose variation is assumed to be determined by causes outside the hypothesised model (Pedazhur & Kerlinger, 1982). This variable is used primarily to explain the endogenous variable. The endogenous variable means a variable whose variation is defined by exogenous or other endogenous variables in the model. Likewise, in regression analysis, a path analysis has some underlying assumptions. Pedazhur & Kerlinger (1982) made the assumptions into five as follows: 1) Relations among variables in the model are linear, additive, and causal. Consequently, curvilinear, multiplicative, or interaction relations are excluded; 2)Each residual is not correlated with the variables that precede it in the model; 3) There is only a one-way causal flow in the system; 4) The variables are measured on an interval scale; 5) The variables are measured without error.
In this study, science achievement was put as an outcome variable, where self-efficacy and students' attitude towards science have a direct effect on the outcome. Furthermore, students' attitude towards school affects the students' attitude towards a teacher and then affecting both students' attitude towards science and students' self-efficacy. In other words, both students' attitude towards science and self-efficacy have a direct effect. Meanwhile, students' attitude towards school and teacher have an indirect effect on the outcome variable.
Results
Based on the results, the hypothesized model was shown fit with the data (chi-square=20, df = 5, RMSEA = 0.072, SRMR = 0.029, p-value = > .05). It means that the hypothesised model was not rejected (see Figure 1). The Table 1 provides information about the regression coefficient. Based on the tables above, the most substantial regression coefficient comes from the attitude towards the teacher goes to student self-efficacy, that is .859. This coefficient was significant. Furthermore, the lowest regression coefficient was student self-efficacy towards science achievement (-.126). However, this effect was still significant at .05. Interestingly, the effect of self-efficacy on science achievement was negative.
Even though two independent variables were having a substantial impact on science achievement, the rsquare of science achievement relatively small (2.6% or 0.026). It means that 2.6% variability in science achievement can be explained by the attitude towards science, students' self-efficacy, attitude towards schools and attitude towards teaching. Another variable should define the rest of the variability in science achievement.
Attitude towards teacher on attitude towards science was significant, with .396 of the regression coefficient. It can be said that the higher student attitude towards a teacher, the more elevated student attitude towards science. This regression was resulting in a 23% variability in student attitude towards science. Attitude towards school had a significant effect on student self-efficacy with r-square 12%. Moreover, the latter path of direct impact was the attitude towards the teacher on the attitude towards school. This path was significant with .457 of regression coefficient and 91% of r-squared.
In this research, the author did not examine the direct effect of attitude towards school and teacher on science achievement. Instead, the author examines the role of attitude towards science and self-efficacy as a mediator effect of attitude towards school and teacher on science achievement. The following information is the total effect (direct and indirect effect) of all variables that go to science achievement (see Table 2 Basically, there were two full main effects. Firstly, it starts from the attitude towards school goes to attitude towards a teacher then goes to attitude towards science and finally at science achievement. The second was the attitude towards school goes to attitude towards the teacher and shift to self-efficacy and finally to science achievement. Based on the results, the second path was having the highest total effect (-.05). This coefficient was significant at .05.
Furthermore, the first scheme was having .026 of the regression coefficient, and this coefficient was also significant at .05. Again, interestingly, the path comes from self-efficacy had a negative coefficient effect on science achievement. However, a different result was found from the path of students' attitude towards science. This path had a positive coefficient.
Based on these two conditions, even though the second path had a slightly more substantial impact on science achievement, yet the first path had a more rational effect on science achievement.
Discussion
From this study, it was found that attitude towards science was influencing student achievement in science positively. This finding was corresponding to previous research. For example, Carey & Shavelson (1988) in their study was found that a positive attitude will have an impact on student achievement in science. In details, the more positive student attitude towards science, the higher student achievement in science. Moreover, eventually, performance will cause positive attitudes. Schibeci & Riley (1986) was examined a recursive model between science achievement and student attitude towards science. The results were supported by the model in which attitudes towards science influence performance and conversely. However, Reynolds and Walberg (1992) were found different results in their study. They have concluded that science achievement affects science attitude but not the reverse. However, in this study was only used a non-recursive model which is an attitude towards science influencing achievement in science.
Moreover, similar results were found from the study in Narmadha & Chamundeswari (2013), that there was a positive and moderate correlation between students' attitude towards science and their academic achievement in science (r = .39, p < .01). Also as cited earlier, Nasr & Soltani (2011) found a weak positive correlation between students' attitude and science achievement (r = .12, p < .05). These two last studies had slightly different results compared to this study, especially the way to examine the relationship among variables. These two were using a correlational analysis to test the hypothesis. However, in this study, the author used the path analysis to examine the effect of students' attitude towards science.
Another focus variable was the effect of self-efficacy on science achievement. In many studies, self-efficacy was found to be the essential factor that influences student achievement (see Reynolds & Walberg, 1992). Self-efficacy will have a positive impact on student achievement. However, in this study, self-efficacy was found having a negative direction on the regression coefficient towards student achievement in science. This result means that the higher self-efficacy on the student, the lower student achievement in science. This finding is contradictory to what Bandura's wrote in his article, "perceived self-efficacy in cognitive development and functioning." (1993). He mentioned that student who believes in his/her efficacy to manage their learning and to master academic activities due to their aspirations, motivational level and academic accomplishments. Ahmad & Triantoro (2013) found concurrent results from their study with Bandura's statements that students with higher self-efficacy will most likely obtain higher scores on the achievement test.
Similar results to the previous study, Liu, et al. (2006). found that students' self-efficacy in middle school is increasing students' achievement in science and affecting students' engagement in a computer-enhanced problem-based learning environment. However, Liu et al. found that students' attitude towards science was not a significant predictor of science achievement.
From the author's perspective, self-efficacy means that students are confident about their ability in terms of academic achievement. The worst is if the student sometimes believes that their knowledge or ability is higher than it really is. Moreover, eventually, lead them into believing that they know everything and can anything, but actually they not. In shorter words, they are just too confident about their ability.
The findings of this study that attitudes of the student towards science were influencing student achievement in science positively. However, a student's self-efficacy was found having a negative effect on a student's science achievement. The first finding was consistent with many researchers. However, the second one was somewhat different from many articles.
Even though those two variables were proven to be a predictor of science achievement in students, however, the effect was relatively small. It can be said that a positive attitude may increase students' achievement in science, but it does not mean so much. The author believes that most likely teachers in Indonesia are teaching their students to like or to enjoy science activities. In fact, that kind of principle does not affect so much students' performance in science.
Moreover, students should not be overestimated about their ability in science performance. In this study, most students too confident about their ability in science. However, in reality, that perception was completely wrong. In fact, the more they are confident about their ability in science, the less they performed in science. Probably, what a teacher needs to do is they should focus on mastery learning of science instead of increasing students' self-efficacy.
The main limitation of this study did not explore more the relationships between teacher and parental variables level and their impact on student achievement. In future research, it is expected that these three variables will be combined jointly to see the effects on science achievement. Also, the score of science achievement derived from all the book tests in science achievement. Hopefully, the results of this study to be considered when the stakeholders want to take decision making or even to create a new policy in education that related to student achievement. | 2020-04-30T09:09:27.943Z | 2020-04-27T00:00:00.000 | {
"year": 2020,
"sha1": "34eba74c919ea449aae9bd84994e1bca47b9aa65",
"oa_license": "CCBYSA",
"oa_url": "http://journal.uinjkt.ac.id/index.php/jp3i/article/download/14490/pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "eeb2431e9d31a59dad4646207cd0cf4f9a0fd86e",
"s2fieldsofstudy": [],
"extfieldsofstudy": []
} |
55930743 | pes2o/s2orc | v3-fos-license | The application of biomass ashes in sewage sludge management – a SWOT analysis
Biomass plays an important role amongst renewable energy sources and for this reason, it is sometimes called as “future fuel”. According to different statistics, the share of biomass in renewable energy production is in excess of 50% and rising. The main disadvantage of biomass combustion in power plants is the production of combustion by-products, namely fly and bottom ash. In line with the sustainable development and waste hierarchy, biomass combustion by-products should be managed in different sectors of economy. Due to the specific properties, biomass ash might be applied as an effective reagent in sewage sludge management. This article shows the SWOT analysis (strengths, weaknesses, opportunities and threats) associated with the use of biomass combustion by-products in sewage sludge treatment. The presented analysis could constitute the initial stage in implementation of new method in a big way in treatment plants.
Introduction
Due to the increasing consumption of biomass in final energy production, the amount of generated combustion by-products is systematically growing.According to Vassilev et al. [1], the worldwide production of biomass ashes is estimated on 476 million tonnes every year.As long as coal fly ashes might be applied in many sectors of economy, biomass combustion by-products (BCBP) are relatively problematic waste.Specific physical and chemical characteristics, especially the content of alkaline elements, cause that biomass ashes do not meet requirements as an addition in cement and asphalt production.It is worth highlighting that cement production uses approximately 2 million tonnes of coal fly ashes in Poland every year.So far, the main receiver of biomass ashes is agriculture.Different references show that the use of BCBP as a fertilizer improves the growth and yielding of crops [2][3][4].However the application of biomass ashes in agricultural practices constitutes only several per cent of worldwide production of ashes.For this reason, the recycling of BCBP in other purposes is a necessity.
Advantageous properties of biomass ashes enable their application in sewage sludge management.Research carried out by Wójcik et al. [5,6] confirmed the positive impact of BCBP on sewage sludge treatment, especially on sludge thickening and dewatering.
Additionally, the content of alkaline elements in biomass ashes enables their application in sewage sludge higienization.However the use of combustion by-products in sewage sludge treatment on a big way demands to take into account different factors (Table 1), which determine the effectiveness and profitability of the whole investment.Taking into account the interdependence between particular factors ensures the achievement of the solution (Fig. 1).
Table 1.Factors affected on the application of biomass ashes in sewage sludge management in treatment plants.The application of new treatment method on an industrial scale demands to prepare the SWOT analysis (strengths, weaknesses, opportunities & threats).On the basis of the aforementioned analysis, the decision concerning the implementation of a new solution might be taken (Fig. 2).
Methodology
The increasing amount of generated combustion by-products causes that the preparation of SWOT analysis associated with the application of biomass ashes in sewage sludge management is essential.The SWOT analysis included strengths, weaknesses, opportunities and treats of presented solution and was prepared on the basis of results obtained both in laboratory tests and on an industrial scale.Environmental, legal, social and economical factors were considered in presented SWOT analysis and were divided into four main categories according to the scheme below (Fig. 3).
SWOT analysis
Weaknesses Threats Strengths Opportunities
Results
SWOT analysis of the application of biomass ashes in sewage sludge management helps to identify both external and internal conditions that might appear during the implementation of new solution in treatment plants.The internal conditions are particularly related to strengths and weaknesses.However, the external factors refer to opportunities and threats.Detailed juxtaposition of all conditions associated with the use of BCBP is presented below (Table 2).T1: afraid of treatment plant's leadership related with the implementation of new sludge treatment method T2: small amount of experiments concerning the influence of biomass ashes on the effectiveness of sewage sludge processing.
S1 & S2 possibility to manage BCBP and the reduction or even elimination of ash landfilling:
As previously mentioned, the increasing share of biomass in final energy production results in the production of combustion by-products, such as fly and bottom ashes.Although biomass ashes have valuable properties which enable their application in many sectors of economy, the use of BCBP is limited by stringent standards.According to Polish standard PN-EN 197-1 [7], only silica fly ashes from coal combustion might be added to cement.Then the Polish norm PN-EN 450-1 [8] enables to apply only silica fly ash in concrete production.The impediment in applying of biomass ashes in the construction industry is also the high content of alkaline elements and the high losses of ignition (LOI).According to Rajczyk et al. [9], the losses of ignition for biomass ashes usually exceed 10%.For ashes with a high value of LOI, Giergiczny [10] indicated the influence of unburned materials on concrete surface, what impedes the application of concrete with ash in construction and architecture.Additionally, the high value of the aforementioned parameter influences unfavourably on puzzolanic activity.Due to the content of nutrients for plants, biomass ashes are utilized in agriculture.The literature review confirmed the positive impact of BCBP on the growth and yield of plants.But available data indicate that only 30% of generated biomass ashes are recycled in Poland every year.The low rate of recycling requires new utilization methods of biomass ashes.The use of biomass ashes in sewage sludge treatment could manage aforementioned waste in line with legal, environmental and social requirements.By means of that, the storage of BCBP might be reduced in a significant way.For this reason, the proposed solution fits into the overall principles concerning the waste hierarchy and sustainable development.
S3 & S4 improvement of the effectiveness of sewage sludge treatment and reduction of the use of chemical reagents in treatment plants:
Due to the high rate of dispersion of flocs, raw sewage sludge is characterized by low-thickening and low-dewatering capacity.In order to improve the effectiveness of sludge dewatering, chemical conditioning is commonly used.Specific properties of biomass combustion by-products are strong reasons to make them use in sewage sludge management.Biomass ash particles are incorporated in sewage sludge matrix and that is the reason why the shape, size and structure of flocs are altered [11].It results in the improvement of sludge dewaterability.The positive impact of biomass ashes on sewage sludge treatment was confirmed by laboratory tests.Wójcik et al. [5,6,12] proved that the addition of biomass ashes into sludge could improve the effectiveness of sludge dewatering even by several per cent in comparison to raw sludge.
Additionally, sewage sludge conditioning by means of biomass ashes results in the increase of pH of sewage sludge and the reduction of total number of bacteria.The effectiveness of biomass ashes could result in the reduction of consumption of chemical reagents in the future.By means of that, it will be possible to reduce the operational costs of treatment plants in a significant way.
S5 & S6 possibility to use the mixture of sewage sludge and ash in agricultural practices and improvement of fertilizing properties of sewage sludge:
One of the main criteria associated with the implementation of new solution on an industrial scale is the possibility of recycling of obtained product.Sewage sludge is a source of nutrients, such as: N, P, Ca, Mg and microelements.Biomass ashes delivered, for example: Ca, Mg, Na and K.The application of both sewage sludge and biomass ash enriches the soil in valuable elements as well as improves the growth and yield of plants.Kiper et al. [13] proved the positive effect of the mixture of sludge and ash on the plant's growth and the reclamation of degraded lands.Additionally, the application of the mixture of sewage sludge and biomass ash creates a closed circuit in the local ecosystem.
S7 improvement of the consistence of sewage sludge at convenient to further application and transport:
The addition of biomass ashes results in the improvement of elastic properties of sewage sludge.Not only is such consistence convenient for transport and land-applying but also eliminates the danger of possible leakage of mixture.In comparison to the application of lime, the phenomenon associated with the secondary liquefaction is not observed.
S8 & S9 prevalence of waste and low costs of obtaining ashes:
The use of biomass ashes in sewage sludge management is supported by the fact that BCBP is relatively easy-obtained materials and occurs in large amount, especially in industrial areas.
W1 public awareness associated with the use of BCBP in treatment plants:
The addition of biomass ashes into sewage sludge and further application of obtained mixture in agricultural practices might result in social discontent.There are known examples of social protests associated with the application of sewage sludge in agricultural practices.Similarly, the agricultural utilisation of mixture of sewage sludge and biomass ash might meet with social disapproval.For this reason, one of most important steps before the implementation of new solution on an industrial scale is the environmental education.
W2 high profitability of investment for treatment plants located near power plants: As previously mentioned, the cost of the application of biomass ashes in treatment plants is limited to the transport cost.For this reason, the whole investment is particularly profitable for treatment plants located near power plants.Even despite the long distance between treatment plant and power plant, the replacement of chemical reagent with the use of biomass ashes could decrease the operating costs of treatment plants in a significant way.
W3 increase the mass of sewage sludge as a result of the application of ashes: Similarly to the application of other chemical reagents, dosing of biomass ash into sewage sludge results in the increase of the volume of waste.According to Liu et al. [14], the total mass of generated sewage sludge containing BCBP might increase even by 30%.However, the possibility to use of generated product in agricultural practices causes that the aforementioned weakness has secondary matters.
W4 necessity to buy new devices dedicated to dosing and storage of ash in sewage sludge treatment:
The implementation of sewage sludge processing technology with the use of biomass ashes demands the small modernization of treatment plants.In costeffectiveness analysis, it is necessary to include additional cost associated with the acquisition of new for dosing and storage of ash, such as: silo, feeder and metering device of biomass ash.However, this cost will occur in the first year of investment.
W5 & W6 requirements associated with the content of heavy metals and microorganisms in the mixture of sludge and ash before the application in the environment: Both sewage sludge and biomass ashes are waste product which might contain heavy metals.But research carried out by different authors proved that most of BCBP is characterized by the low concentration of heavy metals and meet requirements included in the Act of 10 July 2007 on Fertilizers and Fertilizers Application (Table 3) [15].The influence of ashes on the reduction of microorganisms should also be assessed.Detailed tests should protect the environment against the contamination.
O1 & O2 soil quality improvement and reclamation of degraded lands after application of mixture of sludge and ash and the replacement of traditional fertilizers:
As previously mentioned, the application of obtained mixture of sewage sludge and ash has a positive impact on the improvement of soil quality.Both sewage sludge and biomass ashes could contribute to the change of soil class in terms of the content of different elements, such as: K, Mg or Na.Additionally, the application of mixture of sewage sludge and biomass ash could recreate soil conditions and restore the balance between nutrients.Traditional fertilizing might be reduced or even eliminated by the use of aforementioned mixture.
O3 & O4 development of domestic biomass market and the improvement of energy security and supporting the requirements associated with EU climate and energy package (3 x 20%):
Energy consumption is one of the main factors which stimulate social and economic development.For this reason, the energy demand is systematically growing.The new utilization method of combustion by-products could contribute to the popularization of biomass amongst renewable energy sources and the creating of local perennial plant's plantations.The development of biomass market might contribute to the achievement of 20% share of renewable energy sources in final energy production to 2020.
T1 & T2 afraid of treatment plants related with the implementation of new sludge treatment method and a small amount of experiments concerning the influence of biomass ash on the effectiveness of sewage sludge processing:
Because the sewage sludge treatment with the use of biomass ashes is a new technology, this solution might meet with the disapproval of treatment plant's leadership.Additionally, the small amount of research concerning the influence of ashes on sewage sludge treatment might be a limiting factor.
Strategies for promoting the use of biomass ashes in sewage sludge management
Based on the presented SWOT analysis, strategies for the development of sewage sludge treatment with the use of biomass ashes could be proposed (Fig. 4).When establishing strategies, the principle "maximizing strengths and opportunities, transforming weaknesses to strengths and minimizing threats" should be applied.
Summary
Based on the aforementioned SWOT analysis, internal and external conditions of the application of biomass ashes in sewage sludge treatment were presented.The obtained results show that the aforementioned utilization of biomass ashes is a method for which strengths are dominant.The main advantages are the improvement of effectiveness of sludge processing and the reduction of biomass ashes landfilling.Furthermore, biomass ash is inexpensive, effective and easy-obtained reagent.In the meanwhile, such utilization of BCBP has some weaknesses concerning the content of heavy metals in ashes and associated with this social awareness of applying mixture of sludge and ash in agricultural practices.
In addition to that, the findings present some major opportunities, as well as the improvement of soil quality and the reclamation of degraded lands.Due to the content of nutrients in the mixture of sewage sludge and biomass ash, the aforementioned product might replace the use of traditional fertilizers.Additionally, the application of obtained mixture in perennial plants plantations contributes to the development of biomass market.In contrast, it is necessary to consider some threats associated mainly with the small amount of research concerning the use of biomass ashes in sewage sludge treatment.
Taking everything into consideration, the use of biomass combustion by-products in different steps of sludge treatment is the method for which advantages are dominant.
Details technological Fig. 1 .
Fig. 1.Relationship between particular factors affecting on the investment.
Fig. 2 .
Fig. 2. Procedure of the implementation of sewage sludge treatment technology with the use of biomass ashes.
Table 2 .
SWOT analysis for the application of biomass ashes in sewage sludge treatment.
S4: reduction of the use of chemical reagents in treatment plants S5: improvement of fertilizing properties of sewage sludge S6: possibility to use the mixture of sewage sludge and ash in agricultural and reclamation practices S7: improvement of the consistence of sewage sludge at convenient to further application and transport S8: prevalence of waste, especially in industrial areas S9: low costs of obtaining ashes limited to the transport costs W1: public awareness associated with the use of BCBP in treatment plants W2: high profitability of investment only for treatment plants located near power plants W3: increase the mass of sewage sludge as a result of the application of ashes W4: necessity to buy new devices dedicated to dosing and storage of ash in sewage sludge treatment, W5: requirements associated with the content of heavy metals, W6: need for control of the content of microorganisms in the mixture of sludge and ash before the application in the environmentOpportunities ThreatsO1: soil quality improvement and reclamation of degraded lands after the application of mixture of sludge and ash O2: replacement of traditional fertilizers O3: development of domestic biomass market O4: improvement of energy security and supporting the requirements associated with EU climate and energy package (3 x 20%)
Table 3 .
Content of heavy metals in biomass ashes and in sludge according to different authors. | 2018-12-05T08:18:48.991Z | 2018-01-01T00:00:00.000 | {
"year": 2018,
"sha1": "278101b60234d1c8276a931ddb5cc3f57f955f9a",
"oa_license": "CCBY",
"oa_url": "https://www.e3s-conferences.org/articles/e3sconf/pdf/2018/19/e3sconf_eko-dok2018_00196.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "278101b60234d1c8276a931ddb5cc3f57f955f9a",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
55486017 | pes2o/s2orc | v3-fos-license | Seven Stimuli to Identify Opportunities of Innovation : A Practice of Training Innovative Engineers and Some Findings in China
To transfer knowledge to companies by training engineers is directly related to identifying opportunities of innovation. This paper introduces the knowledge system of four levels and an interactive training model for innovative engineers. Training outputs of two classes as cases are analyzed in order to find some factors to affect the training activities. Seven stimuli to identify opportunities of innovation, which are implied in the knowledge system and the training process, are concluded from many face-to-face discussions with the engineers joining our classes. The application of the stimuli is also described, which should be applied in the future training process to improve the possibility to identify opportunities of innovation for innovative engineers.
Introduction
Innovation, the implementation of new ideas is viewed by researchers as the key to both sustaining a competitive advantage [1] and the lifeblood or the best hope for their future growth for companies [2].In China, the innovation capability to both leading domestic companies [3] and other companies [4] is also a key factor for their survival in facing the competition in the market.But most innovations result from a conscious, purposeful search for opportunities-within the company and the industry as well as in the larger social and intellectual environment [5].Understanding the opportunities and their identification represents one of the core intellectual questions for creativity and innovation management [6][7][8].
Training or external learning for companies is directly related to identifying opportunities of innovation.Bozeman [9] shows that training as a kind of external learning is becoming more widely recognized for improvement of human resources of companies.Bauernschuster et al. [10] argue that if innovation is the weapon, education or, especially, training is the ammunition that renders it use-ful and effective.Bao et al. [11] find that external learning increases the opportunities of innovation for Chinese companies, which includes technical and administrative learning, learning technical knowledge or learning administrative knowledge.
In 2008, China government made a training plan to transfer the knowledge of creativity and innovation to the companies nationwide in order to increase their innovation capabilities.Our center was selected as one of the major institutions for the training program.We have carried out several classes to train the engineers from companies in the past years.The training activities are going on now.
One of objectives of the training program is to train many engineers for various industries and make them become innovative.But there is no definition for an innovative engineer in China.In the literature [12], inventtors are classified into five categories related to the innovation process, namely entrepreneurs with technology, industry-specific inventors, professional inventors, grantsmen, and inveterate inventors.We define that an innovative engineer is an industry-specific inventor, who has specific technical improvements for product designs or processes in their workplaces.The improvements include ideations and inventions.The engineers to join the calsses should generate new ideas and push the ideas into inventions.The companies may form development teams to transform the inventions into innovations.As a result, the training becomes an activity to increase innovation capability of the companies.
In the past five years, we have carried out 20 classes.More than 150 companies from manufacturing, energy, materials, food and other industries in different regions of China joined the program.More than 500 engineers did follow the process and have been certificated to innovative engineers.Most of them have applied patents and some have developed new products which are being sold in markets now.A number of innovations of different types are emerging as a result of our training program.
Many discussions face to face with the engineers in our classes show that to identify opportunities for innovation is the most important step to follow the training classes for engineers.But we should analyze what stimulates them to identify an opportunity.This paper first presents the practice of our training process for classes.Then we make analysis for two training cases and try to locate some factors to affect training.Third, we present the seven stimuli for engineers to identify opportunities of innovation, which are the main findings from our training practices.Last, a new model of the stimuli application is outlined.
How to Identify Opportunities for Innovation
There are many studies related to the opportunity identification for innovation.Drucker [6] gives out seven key areas looking for innovation opportunities, and he argues that most innovative business ideas come from methodic analysis of these seven areas.Detienne and Chandle [8] indicate four ways in which opportunities are identified: active search, passive search, fortuitous discovery, and creation of opportunities.Koen et al. [13] show that formal or informal processes may be utilized for opportunity identification.Toubia [14] recommends four sources of opportunities: observational research, blue ocean strategy [15], disruptive technology [16] and lead users.The view of David [17] is that various tools and methods help to identify existing opportunities: lateral thinking; metaphoric thinking; positive thinking; association trigger; capturing and interpreting dreams.Robert [18] finds that pattern recognition is used for opportunity identifycation.Gregoire [19] explores that variations in the su-perficial and structural similarities characterized new technology-market combinations systematically influence the formation of opportunities.The studies show that there is some ways or processes for identifying opportunities of innovation, but the research in this area is going on.
Training and Identifying Opportunities
The research studies in the literatures [20,21] establish a positive linkage between training and innovation in companies.Vichet [22] reveals that majority of the training participants perceive that training contributed moderately, highly, or very highly to the company's innovation.Frazis et al. [23] analyze the data obtained from US companies and find that companies with more innovative workplace practices have a tendency to offer more training.Steven [24] finds that team training accelerates the pace of change in GE.Christian and Uschi [25] examine some companies in German speaking countries and find that high quality, curriculum-based training at the workplace is positively associated with general innovation, product innovation, process innovation, and patent applications, which makes the companies more innovative.Izyani [26] investigates some knowledge-based companies in Malaysia and shows that training activities positively influence innovation of them.Shohreh et al. [27] confirm that there is a statistically significant relationship between participation in training courses and numbers of innovations in food firms in rural Iran.Anja and Igor [28] show that in-house learning is not sufficient for generating innovation and that companies need to supplement internal knowledge with knowledge acquired outside the companies.Yannis et al. [29] support that both internal capabilities and openness towards knowledge sharing among companies are important for upgrading innovative performance.
All the studies show that training is positively relative to the innovations in companies by new or external knowledge which implies opportunities for innovations.To identify the opportunities is a start-up of innovations.
The Knowledge to Be Transferred by Training
Many companies have organized the training programs to uplift creative capabilities of their engineers [30].In these programs creativity techniques are the knowledge to be transferred to the companies.There existed many creativity techniques [31,32], which are divided into two types, intuitive and logical [33].TRIZ, theory of inventtive problem solving, is one of them, which is developed by Altshuller [34] in former USSR.Several countries, including China, have established National TRIZ Asso-ciation, the main motivation in which is transfer TRIZ to companies.TRIZ possesses considerable advantages over other techniques in identifying problems and offering direct solutions to them with confidence [35].TRIZ has been transferred to many companies in the world through training classes in the past years.Kamal et al. [36] study the impact of TRIZ training on creativity and innovation of engineers in companies, and indicate that participation in TRIZ training led to short-term improvements in both the creative problem solving skills and motivation to innovate, and these are associated with longer term improvements in their idea suggestion in the workplace.Nakagawa [37] shows the experiences training engineers successfully to solve real unsolved problems using TRIZ in Japan.In the practice, he integrates the TRIZ into Unified Structured Inventive Thinking (USIT) [38] in order to make the TRIZ simpler.Jun and Shin [39] state that TRIZ is used as an innovation tool more aggressively in SAMSUNG than any other company in the world and the training has been carried out in this company for years.Through a survey, Imoh et al. [40] conclude that the application of TRIZ leads to more effective inventive teamwork, faster ideation, foreseeing how technical systems and technologies develop, but there are some challenges associated with TRIZ, and understanding TRIZ an "inordinate time requirement".
Integrating TRIZ with other techniques, methods, and processes shows a trend for the easy application in the literatures.Axiomatic design (AD) is applied as a complementarily of TRIZ to find contradictions [41,42].Stratton and Mann [43] show that TOC may be used together with TRIZ to find and solve contradictions.Cascini et al. [44] integrate TRIZ and optimization tools to form a systematic design.Tan et al. [45] apply TRIZ to the Pahl and Beitz's model [46] to form a new concept process model.Sun and Tan [47] connect the TRIZ with disruptive innovation process [16] to forecast.The integration may make up for some deficiencies of TRIZ, such as finding a problem.
Training as an external learning process has a positive linkage with the innovation in companies.But the literatures show that its process should be carefully designed for engineers of companies to overcome the difficulties in the study.The following factors must be considered in our training program in order to gain better results.
1) Select appropriate engineers from the companies to join the classes.
2) Design a knowledge system to be transferred to the companies, in which TRIZ is the core.
3) Design a training process to help the engineers to overcome the difficulties in the study.
4) Try to find stimuli for opportunity identification of innovation in order to improve the quality for training program nationwide.
Practice to Train Engineers into Innovative Ones in China
The objects for engineers to join our training classes are as follows: 1) Learn the new knowledge related to creativity, invention and innovations.
2) Identify new or unsolved problems in their workplaces.
3) Generate new ideas from solving the problems in order to improve designs, processes, or develop new products that are new in the companies or in the markets.
4) Transform the new ideas into inventions whose viabilities are proofed.
5) Push the team members to put the inventions into innovations in their companies.
As Figure 1 shows the activities during, before and after a training class.The engineers to join a class from different companies or one company should be selected at first place.Then there will be the training activities: lecturing, identifying problems, discussing, generating ideas and making inventions.The last step is the activity after training in which the inventions may be transformed into innovations.
Selection of the Engineers to Join the Classes
The engineers to join the classes should identify new problems or unsolved problems in their workplaces, where the engineers work.There are diverse workplaces for engineers who work at different stages of the innovation in a company.
Figure 2 shows an entire innovation process in a manufacturing company typically in China.The process is divided into three stages: fuzzy front end (FFE), new product development (NPD), and commercialization [48].The fuzzy front end is considered as the first stage of the innovation process and covers the sub-processes includeing from the opportunity identification, opportunity analysis, idea generation, idea selection, and concept definition [48].The outputs of FFE are the ideas evaluated and as the input of NPD.In the NPD stage, the ideas from FFE are transformed into products.There are two sub-processes in NPD, design and manufacturing.In the design process there are four sub-processes, namely design specification, conceptual design, embodiment design, and detailed design [46].In the manufacturing process, the first is to design the process and then actual manufacturing.The commercialization is the last stage, in which the products are put into markets.The environments for fuzzy front end, new product development, ommercialization and the sub-processes are the work-c
The Knowledge System for Training
places, where the engineers are supposed to find new or unsolved problems.
Training is a process to transfer external knowledge into companies.The knowledge to be transferred with the knowledge existed in companies will inspire the engineers to generate new ideas in workplaces [56].A key attribute of a new idea is novelty, which is the first statement of something not previously known or demonstrated.That a selected idea is embodied in a tangible yet provisional form-a proof of its viability is an invention worth a key attribute of feasibility [57].That inventions will be further refined and reach some final form with commercial intent, such as a functional device or service, are innovations with the key attribute of utility, in addition to the novelty and feasibility.Transferring external knowledge into companies, as a result of training, is an object of our training classes.
Figure 2 also shows that the engineers at different stages have corresponding responsibilities.The job for design engineers is to make design specification, conceptual design, embodiment design and detailed design.The process engineer will design the process for manufacturing, control the quality, test for the products, etc. Chief engineer/R & D engineers may generate new ideas in fuzzy front end and solve difficult problems in design, manufacturing or marketing.The entrepreneurs may pay more attention to all the activities happen in the innovation process.Ideation [49,50] in fuzzy front end, conceptual design [51,52] and embodiment design [2,53], process design [54,55] in NPD are key activities in the innovation process.The engineers related to these activeties which are the core for innovation processes should be selected to join our classes.In practice, there are indeed some entrepreneurs working in middle or small companies joining our classes.So we should also consider their needs for the classes.
Figure 3 shows the knowledge system for training, which will be transferred to the companies.This system is divided into two parts, the knowledge of four levels and application cases.The four levels are basic concepts, basic methods, systematic methods and computer-aided innovation (CAI).The first level knowledge is about the definitions, such as creation, invention and innovation, well structured and ill structured problems [58,59], rouine and inventive problems [60], process of innovation The companies to join the classes should also be selected.They must have strong demands for innovation and will assist the engineers to join the classes for the whole training process.t
Process design:
Development for new processes, Improvement for existing processes.
Embodiment design:
Development for new structures, Improvement for existing structures.
Conceptual design:
Development for new working principles, Improvement for existing working principle.
Knowledge levels
Application Cases [61,62], etc.The second level is the basic methods, such as the methods in TRIZ [63,64]: contradiction solving, standard solution, effect, technical evolution, and some other creativity techniques.The third level is the systematic methods, such as a method for incremental innovation, radical innovation [65], disruptive innovation [47,66,67], patent round innovation [68], analogy-based design [69], etc.The fourth level is about the computer-aided innovation, including CAI tools and application process [45,70,71].In the training process we present many cases which show the applications of the knowledge step by step in different level as the application cases.It is a significant training activity that the engineers imitate the processes in those cases.
A Training Process Model
New or external knowledge should be transferred into a company which is integrated with the knowledge existed in the company in order to support the process of innovation.The knowledge system in Figure 3 is new to Chinese companies and should be transferred into them.These knowledge must also be integrated with the knowledge existed in these Chinese companies.The training process should be designed to satisfy the need for knowledge transfer and integration.TRIZ is the core in the knowledge system in Figure 3.The training processes for TRIZ in companies have been studied for years.Rantanen and Domb [72] have developed a flowchart or a model for TRIZ training in a com-pany.Jun and Shin [39] have also developed a flow chart for training TRIZ in Samsung.These flowcharts show that the TRIZ experts outside the organization should carry out a TRIZ pilot project to show that it is powerful for innovation in a company.This process is not suitable for the situations faced in China.We have no time to carry out a pilot project for every company to test TRIZ because the local governments push many companies to join a training class at the same time.Before the beginning of the training, most of the companies do not know what TRIZ is.A new model for this situation is needed in China.
We put forward an interactive model for training innovative engineers, as in Figure 4.There are four main parts, an innovation process, a training process, an interface between the two parts, and the companies to join the program.The innovation process includes fuzzy front end, new product development and commercialization.The training process are seven steps which are selecting companies, selecting engineers, training stage-1, finding problems, training stage-2, finding solutions and summing up. in the middle of the two parts is an interface, which includes opportunities and solutions for innovation, also the contained problems.The companies selected to join the class may be one or more.A class lasts 6 to 15 months accordingly.
Step 1: Selecting companies The companies to join the class are selected.Some institution of a local government, or an organizer, is reponsible for the organization and selection of the com-s Step 2: Selecting engineers The companies selected make recommendation for a list of engineers to join the class and the organizer of a class is responsible for the final selection of the engineers.We make suggestions that the engineers with bachelor degrees should have almost ten years work experiences and with master or doctor degree may be unrestricted.
Step 3: Training stage-1 Our teacher team gives lessons to the engineers.In this stage, the major knowledge is the level one and two, that are basic concepts, basic methods.The most methods in TRIZ, such as contradiction solving, standard solution, effects and technological evolution etc will be taught.Many cases applying these methods are also demonstrated.The knowledge transferred in this stage will provide a background to identify opportunities, problems and solving them for engineers.
Step 4: Finding a problem Every engineer joining the class must find an inventive problem from the innovation process or the workplace of the companies for innovative activities.The problems are implied in opportunities found.An engineer needs to understand the theories or methods in depth and to connect them to the situations of workplaces and identify an opportunity and a problem implied.
Step 5: Training stage-2 Again, our teacher team gives lessons to the engineers for the level three and four knowledge, which are systematic methods and CAI.The major methods in this stage are extended methods of TRIZ.Such as anticipatory failure determination (AFD) [73] is a methods to be trained, which is an application of I-TRIZ to risk analysis and prediction developed by Ideation International Inc in USA.Some cases applying these methods are also demonstrated.The knowledge transferred in this stage will provide a background to solving problems for engineers.
Step 6: Solving problem In this phase, every engineer must develop at least one accessible technical solution for the problem in a few months, at the same time they should work.At the beginning, the solutions are ideas.After that ideas should be transformed into inventions, which may be a new design prototype, new process in the form of a patent application, or a new concept accepted by the company.
Step 7: Summing up Summing Up is the last phase, in which the final oral examination is made and engineers will present their results with slides.Members of a committee specific in charge of the examination make discussions with them.An evaluation is made and a certificate is presented to some qualified engineers who are innovative.
Finding a problem or an obstacle in the innovation process in step 4 is a key activity to follow the training process for an engineer.If he or she does find a problem the knowledge system studied in step 3 and 5 are very useful for them to solving it and to get the solutions.The solutions are returned to the innovation process, in which the obstacle is eliminated.If an engineer cannot find a problem he or she will be sifted out in the middle of the training process.
In the past five years, we have carried out 20 classes for more than 150 companies nationwide.More than 500 engineers did follow our training process shown in Figure 4 and have been certificated as innovative engineers.Most of the engineers who finished the processes did have applied patents and some of them have been developed to new products which are being sold in markets now.Several innovations in companies are introduced as a result of our training program.
Analysis of the Training Outcomes
Every engineer who finishes the training process does have one or more inventions.Some of the inventions have been developed to innovations in companies.The first step in the process for an engineer is to identify an opportunity for innovation.More than half of the engineers joining the classes do pass successfully the step.The major factors for them to identify opportunities should be studied for the future training activities.75 engineers were selected from 19 companies in Guangdong province, including BYD, BROAD-OCEAN and GAC et al. 52 passed the final examination.30 of them were certificated as Innovative Engineers Level 2, while the others were Level 1. 23 engineers did not follow the training process and dropped out in the middle.All the 52 engineers found out 52 inventive problems from the innovation processes of 17 different companies and solved them at last.As a result, 52 inventions were formed, in which 36 had patent applications, which were 22 patents for inventions and 14 patents for utility models.They have been sent to SIPO (State Intellectual Property Office of the PRC) and all application numbers have been got the end of the class.Problems in different stages and the methods to solve them for this class are shown in Figure 5.
Figure 5(a) is the relationship between the numbers of the problems and innovation stages or sub-processes.The most problems, which are 18, are found from manufacturing.The least problems, which are 4, are in fuzzy front end.The problems from design process are total 30, in which the problems to the conceptual design, embodiment design and detailed design are 10, 15, and 5 respectively.Figure 5(b) shows the relationship between the numbers of the problems and the methods to solve them.The method for solving contradictions is mostly used by 42 engineers.Standard solutions and trimming are used 3 times respectively.Four methods, ideal final result, resources, AFD (anticipate possible failures) and technology evolution (evolution lines), are applied once.
Most companies to join the class are belonging to manufacturing industry.The problems found by engineers are mainly in manufacturing, embodiment design and conceptual design.The major methods to solve these problems are contradiction solving, trimming and stan-dard solution of TRIZ. Figure 5 shows that to find contradictions in manufacturing processes, embodiment design or conceptual design process, is certainly an opportunity of innovation for engineers.
Case 2: LF-1
LF-1 was the first training class organized from July of 2012 to January of 2013 by Hebei Science and Technology Department in Langfang, a city between Beijing and Tianjin.Our center carried out the training process.41 engineers were selected from 17 companies in Langfang to join the class.22 passed the final examination and certificated as Innovative Engineers.19 engineers did not follow the training process and dropped out in the middle.All the 22 engineers found 22 inventive problems from the innovation processes of different companies and solved them at last.
Figure 6 shows the results.10 problems are found from conceptual design process in Figure 6(a), while the only 1 occurs in detailed design process.The problem numbers from design processes are 17.This also illustrates that there are the most opportunities for innovation from designs processes.Figure 6(b) presents that the methods for solving contradictions are also mostly used by 10 engineers.S-field analyses and standard solutions are used 6 times.Ideal final result and trimming are applied twice.
The companies in this class are from different field, such as geology.Most problems are found in conceptual and embodiment design.The contradiction solving and standard solutions are still the major methods to solve problems.To find problems in design stages is also an opportunity of innovation for engineers.
Major Factors for Successful Training
The training activities for trainers mainly include giving lectures, discussions of all possible opportunities and problems, making suggestions and debating, etc. Discussions can be carried out in classrooms, in workplaces, by e-mail, on telephones, and by text massages etc. From several years training activities we summarize that 4 factors are major for successful training, which are the experiences in workplace for engineers, the knowledge system to be transferred, the pressure and responsibility for engineers, the stimuli for identifying opportunities of innovation.
Factor 1: The experiences in workplace for engineers Figures 5(a) and 6(a) show that the inventive problems are directly related innovation stages or sub-processes.The experience in workplaces is the first factor for ngineers to identify an opportunity and find the prob-e Factor 4: Stimuli for identifying opportunities of innovation lems.Several years work in one or similar workplaces make the engineers to understand the situation in depth.They know that there are certainly some problems because of chaotic situation.They estimate that something should be changed.
How to define a problem from a chaotic situation of a workplace is the most important step to follow our classes. Why do the engineers not define a problem for a long time in the workplaces?
For many discussions with engineers face to face we find that the stimuli for opportunity identification are an important factor.The stimuli are contained in the knowledge system that is transferred to the engineers in our training.Such as, Figures 5(b) and 6(b) show that engineers frequently apply the method of contradiction solving.They try to find a contradiction which is an inventive problem in workplace.So to find a contradiction is a stimulus for identifying opportunity of innovation.We need to identify other stimuli for the future application in the training classes.
Factor 2: the knowledge system to be transferred to the engineers
Why have some changes not happened for a long time?One reason is that the engineers lack the suitable knowledge to push the change to happen.So the knowledge system to be transferred to the engineers by training should be carefully designed in the creativity and innovation domain.We select TRIZ as the core knowledge and others as periphery one.The core and periphery knowledge need to be integrated and developed carefully in order to be accepted easily by the engineers.
Factor 3: the pressure and responsibility for engineers
Seven Stimuli for Identifying Opportunities
The engineers to join our classes are selected as excellent ones from companies.They all have pressure and responsibility to follow the training process.The pressure makes the engineers in tension states.They must concentrate all the vigor on study and research activities during the training process.The responsibility makes the engineers consider seriously what kind of problems should be found and solved for innovation of the companies.
The engineers selected to join our classes do have indepth experience and long-term focus in product or process designs in different workplaces, which are the playground for creativity and innovation in the domain.But stimuli are also needed for them to identify opportunities for innovation.The knowledge system shown in Figure 3 and the training process in Figure 4 imply some stimuli.We find that there are seven stimuli, which excite the engineers to identify opportunities of innovation in our classes.
Stimuli 1: Be close to an Ideal System Any system, whether it is a technology, product or process, is in evolution to the direction of ideality.One way to define ideality is the ideal system, which occupies no space, has no weight, requires no labor or maintenance, etc.The ideal system delivers benefit without harm and solves its own problems.The current state of any system is not an ideal system, but it is final state for the system evolution.So an innovative engineer's job is to push a system to be close to the ideal system at least a little step.The step will result in an innovative solution for a product or a process.Making an imaginary ideal system and considering how to be close to the system for an innovative engineer is opportunities for innovation.
Case 1: A new switching power supply Golden Field Industrial, Located in Dongguan, Guangzhou, China, is a company producing computer gadgets and accessories -PC case, switching power supply, multimedia speakers, mouse and keyboard, etc.Four engineers in this company joined the first training class conducted in Guangzhou from August, 2010 to March, 2011.One of the engineers has gotten a patent for a new switching power supply which reduces the power loss to zero in the standby mode for computer gadgets.The new supply will help computers to save energy which is meaningful for industries, offices, families.The opportunity for this new technology is stimulated by Stimuli 1.
In the standby mode for a computer the power loss is from 0.4 W to 6 W.This is a huge energy loss for the whole China or the world.The opportunity for a producer of computer peripherals is to develop an energy saving supply, which is a small step to the direction of an ideal system.Being stimulated the engineer does develop a new supply to reduce the loss to zero.
Stimuli 2: Use unexpected resources
A resource is anything in and around a system that is not being used to its maximum potential.Substances, fields, functions, information, time and space are all possible resources.Some resources are explicit but others are tacit.One of the key concepts in TRIZ is that the strongest solutions transform the unwanted or even harmful elements in a system into useful resources.The suitable application of any resource might lead to discovery of an opportunity for innovation.
Engineers working in a company for several years may be familiar with the surroundings and all the resources but they are not used to applying some resources in innovation process especially tacit once.When they stud-ied the different viewpoints for resources and some application cases, some resources are unexpected treasure and may stimulate some opportunities for them.
Case 2: Fresh keeping wolfberries Qinghaiqing, a company located in Qinghai province, China, produces wolfberry and buckthorn based products.Three engineers in this company joined the first training class conducted in Xining, Qinghai, from January, 2011 to May, 2012.One of the engineers has gotten a patent for a fresh keeping technology for wolfberries and also a new product for the company.The people outside Qinghai, such as Shanghai or Beijing, can taste the fresh wolfberry produced in Qinghai in the near future.The wolfberries are traditionally dried, packed and transported to different places to be sold.Fresh wolfberries are not only tasted good but also full of nutrition.The new product has made the company to have competitive ability in this market.The opportunity for this new technology is stimulated by Stimuli 2.
There is a kind of resources called evolutionary resources, including the knowledge developed in the given area, or other areas, sociology, marketing and psychology etc.According to the concept of the evolutionary resources, the storing technologies in low temperature, the storing film, and the storing technologies using controlled air in other areas may be resources for storing the fresh wolfberries.But resources need integrating into new system for the specific application.This is stimulation for the engineer to identify an opportunity for development of a new technology and new product for fresh keeping wolfberries.
Stimuli 3: Find a contradiction
In TRIZ, contradiction is one of the core concepts, which is used to formulate problems and guide towards innovative ideas.A contradiction arises when two mutually conflicting demands are put on in the same system or a situation.This happens quite often in product design and manufacturing processes.To evolve a system or technology further contradictions should be resolved.TRIZ offers 40 inventive principles, a matrix and 39 parameters to solve contradictions.Using these principles engineers may come up breakthrough solutions but not compromises, or trade-offs.The solutions always result in an incremental innovation.
The engineers in china have studied the concept of universality of contradiction in their educations from middle schools to universities.But Chinese way of dealing with contradictions is seeking a "middle way" that is retaining basic elements of opposing perspectives.In our class we pay more attention to find and solve contradicttions in the specific process for innovation.Solving a contradiction means to eliminate it but not to find a middle way.Most of the engineers could use this stimulus more easily for opportunity identification.
Case 3: A new processing method for chestnuts Liyuan, a company located in Tangshan, Hebei province, China, produces chestnuts, cereals and other agricultural by-products, by industrial deep processing methods.One engineer in this company joined the training class conducted in Baoding, Hebei province, from May, 2011 to January, 2012.The engineer has gotten a patent for a new processing method of chestnuts which have a constant mouth feel.The opportunity for this new technology is stimulated by Stimuli 3.
Some consumers complain that the sweetness level of the chestnuts produced in this company is lower and the hardness level is higher.In the past several years the company has not made improvements for the product to meet the needs.The engineer made an analysis and found that there were two contradictions for the quality problems of the product.The matrix and inventive principles were applied to solve the contradictions.At last a new processing method was formed and the experiment for the improved chestnuts showed that the processing was right.An improved product has been put into the market.The Stimuli 3 inspired the engineer to make this innovation happen.
Stimuli 4: Trim some elements A product consists of several elements and links among them.An element is defined as a physically distinct portion of the product that could not be divided further for analysis.A link is kind of relationship between two elements, which is an action.Two elements and a link between them make up a function under TRIZ concept.All the functions for the product form the function model which is a function net.There are four kinds of actions which are useful, harmful, excessive and inadequate.If one of harmful, excessive or inadequate actions is identified the link and the two elements are a problematic function.There may be one or more problematic functions in the function model for an existing product.One or some problematic functions should be eliminated for the improvement of the product performance.
Trimming in TRIZ is a kind of operation to cut off some elements relating to the problematic functions by some rules.Basic principle of trimming is simplification for an existing product which reduces the cost, size, weight, or simplifies the operations that make the products easy to be used.As a result, trimming is an opportunity of innovation for innovative engineers.
Case 4: A new structured motor Broad-Ocean, a company of Guangzhou, China, is a producer of micro-motors.Two engineers in this company joined the first training class conducted in Guangzhou from August, 2010 to March, 2011.One of the engineers has gotten a patent for a new structural design for the one-way asynchronous motor produced in this company.The efficiency assembling process for one motor is improved and the cost is reduced by the new design.The opportunity for this engineer is stimulated by Stimuli 4.
In the traditional design the rotating magnetic field in one-way asynchronous motor is produced by a capacity connected outside the shell of the motor.The module of the capacity is located to the shell by screws that lead to some harmful results.That the module may fall off in the operation is a clear weakness.According to the basic idea of the Stimuli 4, the capacity should be trimmed off.This is an opportunity for the engineer.The engineer did take advantage of the opportunity to develop a new structure that the capacity is not outside the shell.New products using the structure have been put out.
Stimuli 5: Anticipate possible failures
To anticipate possible failures for products or processes are certainly a kind of opportunities for innovation.Anticipatory failure determination (AFD) may be applied for this purpose, which is an application of TRIZ to risk analysis and prediction.There are two templates, AFD-1 and AFD-2, for applications.AFD-1, failure analysis, applies to find the cause of a failure that has already occurred but is not yet understood.AFD-2, failure predicttion, is to identify possible failures that have not yet occurred for a new system design or for any system in which negative effects or drawbacks have not yet manifested themselves.The basic concepts and methods of TRIZ, such as resources and contradictions solving, may be used in the process of AFD.The application of AFD for both existing and being designed products is an opportunity of innovation for the engineers in the classes.
Case 5: New currency recognition modules GRGBanking, a company located in Guangzhou, is a provider of currency recognition and cash processing solutions in the market.Automatic Teller Machine (ATM) is a major kind of product in this company.More than 20 engineers in this company joined our two training classes from August, 2010 to December, 2012.One of the engineers has gotten several patents for different currency recognition objects in ATM.The new ATMs using the patents have been operated in several cities.The opportunity for this engineer is stimulated by Stimuli 5.
One challenge for the company is to recognize the counterfeit cash in or out ATM in high precision.Precisions for recognition used to be a problem in this company.The engineer applied the AFD-1 and AFD-2 to identify the root causes of old designs and put forward new principles and formed new modules for ATM.The AFD stimulated the engineer to find problems and solutions.
Stimuli 6: Add another purpose function
Open Access AJIBM Seven Stimuli to Identify Opportunities of Innovation: A Practice of Training Innovative Engineers and Some Findings in China 735 A function for a product is what does.There are two kinds of functions, purpose and operation.The purpose function is a description of a users' intention or the purpose of a design, and the operation function is a description of the intended operation of the design.Users can understand the usefulness of the purpose function for an existing product.That is the reason why users buy this product.Maybe some users hope to buy products with more purpose functions.Multi-function is an evolution trend for some products.So to add another purpose function is an opportunity for innovative engineers.
Case 6: New structured headstock and tailstock of EMU Tangshan Railway Vehicle (TRC) located in Tangshan, Hebei province, China, produces electric multiple units (EMU), passenger coaches, etc.We organized a class for this company from May, 2012 to December, 2012.At first 60 engineers in different workplaces were joined the class and 28 engineers followed the whole training process.One engineer has gotten a few patents for the new structured headstocks and tailstocks of EMU.The opportunity for the new technologies is stimulated by Stimuli 6.
The headstock and the tailstock in an EMU are of the same structure.The headstock in one running direction is the tailstock for the reveres running.The air dynamics for the headstock and the tailstock are different because of the state change for running condition.The air resistances are not in optimal condition for the same structures for both sides.Adding a function to reduce resistance for headstock and tailstock is an opportunity for the new design of EMU.Stimuli 6 makes the engineer to design new dynamic structures for headstock and tailstock to adapted different conditions.The new technologies reduce energy loss when the EMU is running.The inventions of the engineer are quite important for the company, also for the industry.
Stimuli 7: Change behaviors A behavior for a product is how does.Behaviors can be regarded as actions or physical state transitions among the elements of a product; or it can be regarded as the physical interactions including the input actions and output actions to or from elements.These actions can be both the intended and unintended, such as side-effects.If one or some physical state transitions are substituted by new ones the performance of the product may be better.This will result in a kind of invention and innovation.So to change some behaviors for selected products is an opportunity for innovative engineers.
Case 7: A new testing instrument for solar cell modules
Qinghai Tianpu Solar Energy Company, located in Qinghai province, China, is a producer of photovoltaic products for west China.Three engineers in this company joined the first training class conducted in Xining, Qianhai, from January, 2011 to May, 2012.One of the engineers has developed a new product, a testing instrument for solar cell modules for the company.The opportunity for this new product is the application of changing behaviors.
The voltage, current, and peak power for a solar cell module should be tested by instruments for evaluating the performance.The instruments existing in the market now are used in house of module producers.But the modules are operated in open countries, which are outside the workshops of the producers.The existing instruments are not suitable to some customers' new needs for operations in some locations.Changing the behaviors of the existing instruments is an opportunity for development of a new one which can serve better in different locations.The LabVIEW and Matlab are used and the reliability is increased in the new design.The new product has been developed and tested successfully in this company.
Possibility in Applying the Stimuli
The stimuli are concluded from the knowledge system in Figure 3 with the help of training practices.One stimulus may or may not be used in a stage of innovation.Table 1 shows the possibility of the application for every stimulus in different stages or sub-processes of innovation.There are three types of possibilities, high, middle or low.High or low possibility means that the result to find an opportunity could happen or almost could not happen.Middle possibility is between high and low.
The table also shows the following features: 1) Every stimulus may be used to identify opportunities in fuzzy front end.
2) Every stimulus may be used to identify opportunities in conceptual design.
3) To find a contradiction is an important activity in opportunity identification for fuzzy front end and new product development.
There is no any symbol in the right commercialization column of the table.In practice we find that a few entrepreneurs, chief engineers or R&D engineers do find opportunities in commercialization stage.But we have not concluded one or two stimuli for them to use.This will be a research topic in the future.
Figure 7 is a process model for applying the 7 stimuli.First, the engineer selects one or more stimuli and then applies them to a stage, such as fuzzy front end.If an opportunity is identified the engineers find the implied problems.There are two possible paths to manage the problems which are called self-circled or passing on.In he first path, the engineers solve the problems in their t workplace.In the second the problems are passed on other engineers who are working in relevant workplaces.
The second path shows that some problems in one stage or a sub-process should be solved in another stage or sub-process.7 stimuli are not many enough to make a difficult process of being chosen one by one for engineers.Now we have added the 7 stimuli as new knowledge in the training stage-1 and stage-2 of Figure 4. We hope that they will assist the engineers to join the classes to identify opportunities a little easier.
Discussions
An interactive model is developed in this study for training innovative engineers for a variety of companies in China.The specific feature for this model is that the engineers must find and solve inventive problems in their workplaces during the training process.The solutions from the problems are new ideas which are improved to form some inventions.Experiences show that engineers and their companies make high evaluation about the training process.
To organize the companies and engineers to join the classes are the job of local governments but the lecturing and other activities are the job of our center.The advantages of two sides are brought into play.This is feasible model for transferring creativity and innovation knowledge to companies in China.
The knowledge system transferred in the training process is specially organized at four levels, which are basic concepts, basic methods, systematic methods and computer-aided innovation.TRIZ is selected as the core knowledge and the others dealing with creativity and innovation are supplementary in this system.TRIZ is strong in solving difficulties or inventive problems but other techniques are needed to find problems and integrated into an innovation processes.The integration of two kinds of knowledge forms a whole knowledge system from finding to solving a problem for an innovation process.The knowledge system is different from only TRIZ and is a key factor for successful training.
From many discussions face to face with the engineers in the training process, 7 stimuli are found, which excite them to identify opportunities for innovation in our classes.The knowledge system of four levels is pregnant with all the stimuli.The stimuli are directly related with fuzzy front end, new product design and manufacturing.That the engineers apply them one by one to workplaces may help them to identify opportunities and the problems contained.
The knowledge system is an open system, to which new relevant knowledge could be enriched.The classes are organized in more and more provinces in China and the experiences are accumulated.Some new stimuli may be concluded and should be added to the stimulus set of this study.
It needs our attention whether the stimuli may excite the engineers to identify opportunities in the stage of commercialization.
The stimuli are suitable for adding to the interactive training process for engineers.But for a long term consideration they should be applied in the process of creativity and innovation management for companies.We are trying to make experiences in one or two companies now in this direction and hope to have some results in the future.
Conclusions
The practices show that improving the innovation capability is possible through training engineers in China.More than half of the engineers having joined our classes do pass the training process though some engineers drop out in the middle.
TRIZ, which is the core technique in knowledge system with four levels, is particularly useful for the engineers to make inventions.Because learning TRIZ is not easy, the training process should be developed carefully.The interactive training model for training innovative engineers in this study is a possible selection.
The knowledge to be transferred and the process used imply some stimuli to identify opportunities of innovation, in which problems are contained.Seven stimuli are concluded by many face to face discussions with the engineers to join our training classes.A possible process to apply them is also described.
There will also be some researches for future work, such as adding new knowledge to the system, finding new stimuli and applying them into commercialization in companies for a long term management.
Figure 2 .
Figure 2.An innovation process and workplaces for engineers in a manufacturing company.
Figure 3 .
Figure 3. Knowledge system for transfer in the training process.
4. 1 .
Case 1: GD-1 GD-1 was the first training class organized by Guangdong Science and Technology Department in Guangdong province from August of 2010 to March of 2011, which is the organization of local government for the development of sciences and technologies.Our center carried out the training process.
Figure 7 .
Figure 7.A process applying the stimuli.
An interactive training model for innovative engineers in China.
panies to join the class for a region.For example, the Productivity Promotion Center of Guangdong (PPCG), which is an institution supported by Guangdong Science and Technology Department in Guangdong province, was responsible for the selection of the 19 companies from the province to join the first training class held from August of 2010 to March of 2011. | 2018-12-12T03:30:22.489Z | 2013-12-12T00:00:00.000 | {
"year": 2013,
"sha1": "8dbdcece6a6a6cb1384bdfc16e63bce32c9001a3",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=41082",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "8dbdcece6a6a6cb1384bdfc16e63bce32c9001a3",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
267295806 | pes2o/s2orc | v3-fos-license | Introduction: The New Face of Trans Visual Culture
: Transness throws into question how many so-called Western cultures—i.e., those ideologi-cally descended from the colonial project—have sutured “reality” to the “privileging of sight”. At the crux of trans-visual culture is a need to be understood outside current modes of visual apprehension. As a methodology rooted in trans-embodied experiences, trans provides a mode for decolonizing the privileging of sight and moving toward a new understanding of bodies, identity, representation, and visual culture. It is imperative to explore such methods in today’s political climate, and it is advantageous to apply them to trans-visual culture, as exponential innovations can be discerned. In this article, I will deploy a trans visual studies methodology to the work of contemporary trans masculine artist and photographer Wynne Neilly to explore how his work engages a praxis of transing identity. I will discuss how his work shifts the understanding of identity and representation to one decoupled from optical ontology and how he works to unseat White masculinity as the center of Western art and visual culture.
Introduction
In the spring of 2021, TIME magazine featured Elliot Page on its cover in a piece by Katy Steinmetz entitled, "'I'm fully who I am': Actor Elliot Page and the fight for trans equality."The feature on Page reflects a groundbreaking moment-a cultural highlighting of trans masculinity.Elliot Page, with his newly released book Pageboy: A Memoir (Page 2023), has cemented his position as one of the newest icons of the trans movement.At the least, as Steinmetz puts it, Page is "now one of the most visible trans people in the world" (Steinmetz 2021).Seven years prior, TIME magazine ran the now iconic Laverne Cox cover story wherein Steinmetz dubbed the era the "Trans Tipping Point".I mention this to point out that Steinmetz and TIME have been responsible for shaping important cultural discourse around the trans movement and trans representation and to show that even the seemingly recent fight for trans equality has been long and arduous.Even in the recent pop culture arena, the Trans Tipping Point has been less of a critical juncture and more of an ongoing battle that is continuously unfolding.Furthermore, a brief look at the two cover stories highlights that trans masculinity is significantly underrepresented, given that it took seven years after the announcement of the "Trans Tipping Point" for a trans male actor to grace the cover of a major magazine and enter the public spotlight.
The portrait photographs (Steinmetz 2021) accompanying the feature reveal Page looking vulnerably at the camera, posing in T-shirts and long-sleeved thermal undershirts, evocative of the revival of 90s aesthetics donned by many in the early 2020s.The cover image frames the actor within the iconic red border and the prominent title, TIME, in grey letters floating behind Page's somewhat forlorn face, as he looks into the camera with bedroom eyes.He leans gently against a light grey wall, sporting a black sweatshirt, jeans, and white sneakers, flanked by a small dog (Figure 1).His poses suggest he is comfortable with the photographer, which translates into a feeling of openness with the reader.The images depict several intimate photographs which cast the actor as eager yet trepidatious about publicly announcing his transness.
Arts 2024, 13, 22 2 of 14 The quietly contemplative photographs were taken by trans male artist Wynne Neilly.On 16 March 2021, Neilly posted the above cover image on his Instagram feed accompanied by the hashtags #transrights and #protecttranskids and the following caption: I don't even have the words at this moment to describe my immense gratitude for this experience.This is my dream assignment.I have been wanting this and working towards this for so many years.I'm so proud of Elliot.And I am so grateful to the trans elders who risked everything to make this moment happen.I will probably have to add more to this caption later when I have the ability to form more words but for now I am just grateful, honoured and proud to be trans in this moment of time.
Thank you, Elliot.
The enthusiasm, pride, and respect articulated in the sentiment are likely due to the fact that the actor and the photographer are both transmen and reflective of the state of trans rights in Spring 2021 when the issue launched.The weight of the cultural moment must have been palpable to both the actor and the photographer, and it marked a moment when global culture could no longer ignore the relevance of trans people.In the movement for trans rights, trans masculine people have historically been less visible than trans feminine people and largely occluded from cultural recognition.Laverne Cox appeared on the cover of TIME magazine in June 2015, and Caitlin Jenner appeared on Vanity Fair's cover later that year.While Cox and Jenner may be the most famous trans people to appear on magazine covers, numerous trans feminine people and trans women have appeared in mainstream TV, cinema, and media over the recent decades, including Alexandra Billings, Candis Cayne, Jamie Clayton, Gia Gunn, Dominique Jackson, Janet Mock, Indya Moore, Peppermint, Jen Richards, Michaela Jae Rodriguez, Hunter Schafer, The Wachowski sisters, and many more.Trans masculine visibility has remained scant, however; thus, Page's high-profile transition in mainstream visual culture wields major cultural significance.In recent decades, the field of trans representations has been growing in all sectors of culture globally.Facilitated by the advent of social media, political progress protecting the rights of trans folks, advances in accessibility for trans medical care, and inroads in media, visual culture, and contemporary reflecting LGTQ+ perspectives, trans representation is becoming a diverse field of study, and in many spaces a divisive and contested issue.The complexities and nuances of trans representations continue to unfold and are increasingly discussed in various sectors of cultural studies, cinema, academia, and beyond. 1 The defining trans representations in the decades preceding the Trans Tipping Point included a handful of tragic stereotypes sensationalizing trans femininity and erasing trans masculinity.While some inroads have been made into thinking about how masculinity has been constructed in visual culture, trans masculinity has yet to be substantially visualized, recognized, or theorized.One of the foremost trans masculine activists and culture producers of the last decades (and founder of the highly influential magazine Original Plumbing: Trans Male Quarterly), Rocco Kayiatos, has spoken about how, as a young trans man, it took him many years to realize his trans identity, because he saw no examples of trans masculinity in the world around him, save for a chance encounter with Loren Cameron's book, Body Alchemy, in his late teens. 2 Cameron's influential 1996 book, which portrays numerous trans masculine experiences in photographic form, was a radical intervention into visual culture and has been a touchstone for many transmasculine people in the ensuing years (Cameron 1996).Neilly's image of Page, is significant in that it is a trans masculine photographer imaging a trans masculine actor featured on the cover of a prevalent pop cultural magazine and it will surely resonate far beyond what we can now imagine.
Cover stories such as Page's are vital in that they make clear that knowledge about trans people for trans and cis people alike is not readily available.Seeing trans people in visual culture is often the only way most folks know anything about trans people.For this reason, trans representations carry significant cultural weight.Sam Feder's 2020 feature film Disclosure is an apt example.It takes a nuanced look at the deployment of trans stereotypes in mainstream cinema and visual culture.There is a scene in which Laverne Cox discusses how 80% of Americans do not know any trans people, so everything they know about trans people has been gleaned from media (Feder 2020).Bearing this in mind, representations of trans folks can be extremely damaging when the construction of trans characters in media and the rhetoric attached to trans people are bound up with stereotypes and vitriol.Such representations often reinforce the gender binary while also inspiring hate crimes and anti-LGBTQIA+ legislation.Especially when anti-trans sentiment is coupled with racism.
While writing this article, I have anxiously followed the state of anti-LGBTQ+ legislation and the escalation of hate crimes globally.In the US in July 2023, for example, there have already been 269 anti-LGBTQ bills introduced in nearly all 50 states.The ACLU has a live map tracking the bills on the home page of their website using increasingly saturated tones to correspond to increasingly aggressive numbers of bills introduced in each state.The map's highlighting of anti-LGBTQ+ sentiment across the nation is visually arresting (ACLU 2023).The US is not the only country in the news currently demonstrating increased hostility toward LGBTQ+ people, however.As I write this, Ugandan president Yoweri Museveni is currently considering supporting a bill that will criminalize identifying as gay, lesbian, trans, or queer, with sentences of up to 20 years in prison (Nocholis 2023).
Viewing the ACLU map of the US, one is also viewing the binary, essentializing, and linear ideologies of colonialism applied to land.Such mapping strategies are ever-present in US renderings of this section of North America, a location more aptly known as Turtle Island.Not only does the US outline represent a truncated portion of Turtle Island, but it is also indicative of the racist, colonialist, gender-oppressive history of the genocide of indigenous people, along with which came binary gender paradigms as part and parcel of White supremacy.The essentialized, binarized conception of land into discrete and closed categories (i.e., countries and states) reflects the same colonialist logic that seeks to define and rigidly boundary reductive ideas about gender categories.
Backlash against LGBTQIA+ liberation is often aimed at trans and queer people who visibly disrupt dominant binary, cisgender, and White supremacist paradigms.Trans people who embody aesthetics outside established binary cis-supremacist paradigms are the most frequent targets of violence.Bearing this in mind, it is compelling to consider how drag performance has been a target of much of the recent backlash.Within LGBTQIA+ life and discourse, trans people and drag performances are understood as distinct from one another (albeit often overlapping).To oversimplify, for example, trans is an identity, and drag is an art.Nevertheless, right-wing attacks on LGBTQIA+ liberation have targeted drag performances and trans identities.From the viewpoints of heteropatriarchy, White supremacy, and cis supremacy, drag performances and trans people are disrupting the dominant world order.Both trans identities and drag performances disrupt the false belief in the essentialized collapsing of binary sex and binary gender and disrupt the pseudo logic that identity is hermetically sealed.Trans identities and drag performances exponentially explode the myriad possible configurations of gender, biological sex, and sexuality and the infinite expressions and configurations thereof.Significantly drag performances and trans identities are often targeted when corporeal presentations visually disrupt dominant aesthetic categories of binary gender.
Visual culture is part of the negotiation of identity formations, and currently, a significant shift is transpiring due to the wave of trans visual culture.While there is an ongoing increase in complex and nuanced trans representations due to growing numbers of trans and queer people across all aspects of culture, media, and art production, the radical transformation is not occurring as one may expect via inroads made by creating positive representations of trans people.Positive representations do not disrupt the problematic and troubling pseudo-logics of White supremacy and heteropatriarchy.Such an approach actually works with the same logic of oppressive schemas, reinforcing the false pretext that visual representation is the key to knowledge or liberation, thus staying invested in binary oppositions and reifying colonialist logic.The methodologic intervention of trans visual culture is a debunking of the pseudo-logic that seeing is indexically related to knowing and that in looking at people, one can know someone's identity.The methodologic break with this outmoded, colonialist pseudo-logic is the transing of identity.
Briefly on Identity and Representation
Living in locations where dominant culture seeks to maintain culture as White supremacist, and heteropatriarchal affords Elliot Page what Devon W. Carbado has formulated as "negative identity signification" (Carbado 2005).Carbado has observed that White men live on the "white side of race" and "the male side of gender"."He is, in this sense, the norm: Mankind; the baseline.He is our reference.We are all defined with him in mind."(ibid., pp. 190-212).Yet, as a transmasculine person, Page's relationship to this position of "negative identity signification" and this position of "norm" is not that of a cis man.Page lived decades of his life treated as female, a gender that he was not, while he was wrestling internally with who he was in a world where the examples of trans lives seen in public view demonstrated that to be trans was to live a life unlivable.Arguably one of the most notable public-facing examples of trans masculinity, Page represents a shift in the schema of how White masculine identity is understood.Trans lives and trans methodologies developed from trans lived experiences demonstrate that identity is innate not tied to visual apprehension and malleable not static, thus defying the stoic logic of White supremacy and cis supremacy.As trans visual culture continues to proliferate and trans discourse moves forward, understandings of how visual culture is bound up with the negotiation of identities continue to unfold and shift.Still, the logic of the dominant order persists, positioning Elliot Page as an incorporable example of trans masculinity and affording him a place as "one of the most visible trans people in the word" (Steinmetz 2021).
Via the lens of the pseudo-logic of White supremacist and cis supremacist ideologies, processes of gendering and racializing are part of the dominant order and enmeshed with visual culture.Identities and values are assigned to bodies via visual encounters based on an ideological matrix: a complex system of processes about expectations of how people can look and how identities can acceptably be performed.Cultural representations of identity with regard to who and what are incorporable are proposed via visual culture and the pseudo logic that looking and knowledge are inextricably linked, and stereotypes become part of culturally specific understandings of various groups.Judith Butler's research investigating regulatory practices that govern gender and culturally intelligible notions of identity is useful here.The intellectual work I am referring to dates from the 1990s and early 2000s, a time before the so-called Trans Tipping Point, and yet the reductive ideologies Butler observes still hold firm decades later.Butler noted that how people embody norms is often intimately enmeshed with survival.Butler noted that for some, survival is a "burning issue" and how norms are embodied for some folks precisely relates to whether life is livable.Butler describes those who live outside the domain of which she describes as an exclusionary matrix as abject and describes their lives as unlivable (Butler 1993).For the binary gender system to remain intact, those whose identities are viewed as challenging the system must be made examples of as offenders of the system.Butler's observations are clearly demonstrated in visual culture when we see trans subjects who most fully reflect dominant cultural ideologies such as Elliot Page as being (marginally) accepted, while subjects existing too radically outside sanctioned genders are viewed as not only expendable but necessarily eradicable, as exemplified in particular by the countless acts of violence and aggression perpetrated against working class trans femmes of color. 3Zackary Drucker and Kristen Lovell's 2023 feature film The Stroll for example, gives trans perspective and voice to the lives of trans femmes in New York's Meatpacking District, exploring topics such as working to survive, gentrification, discrimination, and the trans rights movement (Drucker and Lovell 2023).
To explore how representation is enmeshed with identity formations, it is critical to consider how the ideological matrix that governs gender is the same as the apparatus of racialization.The two are inextricably linked, with countless intersectional mobile and culturally specific points on intersecting schemas.Writing about the problematic centering of White bodies in art history and visual culture in the West, Devon W. Carbado points out that the situating of the specific identity group as invisible or as lacking signifiers inaccurately positions White subjects as the default.As many scholars have observed, there are no neutral positions--not in terms of identity, ideas, or visual culture.Everything is uniquely situated, and any suggestion otherwise needs to be challenged and de-centered.Profound discourses on identity and representations that challenge the exclusionary practices of art history perpetrated under the guise of the neutrality veil can be found in Derek Conrad Murray and Soraya Murray's Uneasy Bedfellows: Canonical Art Theory and the Politics of Identity (Murray and Murray 2006), and Alpesh Kantilal's Productive Failure: Writing Queer Transnational South Asian Art Histories (Patel 2017).The powerful work accomplished by scholars such as these who deploy methods that bring together critical methods of exploring identity, representations, and the exclusionary practices of academia, the art world, art history, and visual culture, is a model for the work that may be conducted via deploying trans visual studies and the study of trans visual culture.
White supremacist, cis supremacist ideologies are rooted in the legacy of the pseudologic of colonialism and rely on visual culture to attach systems of beliefs to various groups of people through visible differences.The Black, queer feminist insights of bell hooks reveal that, "From slavery on, White supremacists have recognized that control over images is central to the maintenance of any system of racial domination."(Hooks 1992).This notion is then tied to photography, as photography emerged during a period deeply bound to colonialism and the rationalist belief that seeing was the highest of all the senses.Via visual assessment, colonialist logic emboldens the ascribing of identity and value to others.For Preciado, the emergence of photography marked an essential stage in producing a new sexual subject via "visual truth" (Preciado 2013).Preciado has observed that the discursive framing of photography as indexical and the suturing of beliefs about photography's "technical production" endowed it with "the merit of visual realism" (Preciado 2013, p. 111), which, in turn, has tied photography to a significant stage in the production of gender via the belief in visual truth.Preciado writes, "The truth of sex takes on the nature of visual discourse, a process in which photography participates like an ontological catalyst, making explicit a reality that would not be able to emerge any other way."(Preciado 2013, p. 112).
The cultural conception of photographs, gender, and racialization (in locations ideologically descended from the colonial project) remains linked to the notion that via visual assessment, the truth about who people are may be ascertained.Although photography scholars have noted that we are beyond the digital turn, making the current state of photography "post-photographic", photographs are often still viewed as a "window into a world". 4However, no assumption about who someone is based on looking at them can be trusted.Moreover, such a practice can be highly inaccurate, and, more to the point, this schema is deeply problematic, for it upholds a belief that looking is equivalent to knowing and disregards both that looking is ideologically informed and that subjects inherently know who they are more intimately than anyone looking at a given subject.
While photography has often been mobilized as an extension of the colonialist project and in the service of reifying essentialized and reductive identities, photography is also an excellent vehicle for challenging the legacy of oppressive conceptions of visuality, identity, and representation.Transing identity provides a new methodology for understanding representations that move beyond the outmoded colonialist logic that prefers to equate ascertaining information from a surface as inherently linked to interior truth or, in this case, identity.Transing identity prioritizes embodied self-knowledge.Transing identity with regard to visual culture and photographic portraiture provides a pivotal intervention into the understanding of representation.Nearly a decade before making the cover photo of Page, Neilly was beginning his medical transition and photographing his experience via Polaroid every Friday (the day of his testosterone injection).The ongoing self-portrait work Female to "Male" images Neilly's changing corporeality as he undergoes medical transition (Figure 2).In a rich lineage of artists using photography to chronicle transforming their corporeality as durational performance, including Adrian Piper, Eleanor Antin, Tehching Hsieh, Cassils, and others, Neilly's project inserted a popular vernacular process of self-documenting one's medical transition with the use of hormones as a trans person into fine art discourse.The specificity of picturing his medical gender transition lends itself to distinctly different insights than those ascertained from studying any of the above-mentioned similar projects.
Wynne Neilly
Arts 2024, 13, x FOR PEER REVIEW 7 of 15 equate ascertaining information from a surface as inherently linked to interior truth or, in this case, identity.Transing identity prioritizes embodied self-knowledge.Transing identity with regard to visual culture and photographic portraiture provides a pivotal intervention into the understanding of representation.
Wynne Neilly
Wynne Nearly a decade before making the cover photo of Page, Neilly was beginning his medical transition and photographing his experience via Polaroid every Friday (the day of his testosterone injection).The ongoing self-portrait work Female to "Male" images Neilly's changing corporeality as he undergoes medical transition (Figure 3).In a rich lineage of artists using photography to chronicle transforming their corporeality as durational performance, including Adrian Piper, Eleanor Antin, Tehching Hsieh, Cassils, and others, Neilly's project inserted a popular vernacular process of self-documenting one's medical transition with the use of hormones as a trans person into fine art discourse (Steinmetz 2018).The specificity of picturing his medical gender transition lends itself to distinctly different insights than those ascertained from studying any of the above-mentioned similar projects.Assigning gender to people in locations ideologically entrenched in the logic of colonialism is currently predominantly achieved via assessing and interpreting corporeal aesthetics, and assigning a gender to the subject being viewed.A deceptively benign and often unconscious occurrence, the process of looking is, as scholar Nicole Archer poignantly described, "ideologically saturated" (Archer 2017).Looking at the self-portrait photographs of Female to "Male", the highly ideological process of interpreting the surface aesthetics of the figure vis-a-vis the photograph comes into focus.The conceptual flattening of Wynne Neilly the photograph and Wynne Neilly the person invites us to recognize the phenotypic judgments we are constantly making about people's visual morphology.The repeated aesthetics of the photographs of the subject slows the process of making assumptions about the subject.The numerous images of Neilly invite viewers to reflect on who they think the subject is.With each new picture, the subject appears slightly different in body shape, yet the tonality of their skin tone remains, by and large, the same.Based on subtle shifts in body thickness, muscle tone, body hair, face structure, shoulder breadth, chest definition, facial hair, and posture, one may assign a variety of gender assumptions to each image of the subject.Yet based on processes of racializing, the subject remains white, highlighting the consistent racializing of the subject regardless of gender.
Processes of gendering are always racialized.What comes into view with Female to "Male" is that gender is always bound by aesthetics, and when gender is assigned to others, it is conducted via an arbitrary and ideological process of fixing expectations to people based on looking at them.Via Neilly's Polaroids, one is able to see how the color of the subject's skin relates to this process of gendering.In each image, standing in front of a white wall, Neilly brings into relief the color of his skin, highlighting how interpretations of a given physicality are always necessarily bound up with expectations of gender and processes of racialization.For example, in Figure 2 one may identify the figure in the top left image as a soft butch, in the top right image as a little bit butcher, in the bottom left image as a young man, and in the bottom right image as a man.Ascribing gender is contingent on situating a subject within a field of similar-looking bodies and interpretations of them in relation to said field.As such, this process is always culturally and sub-culturally specific.Aesthetic signifiers such as body posture, hairstyle, piercings, jewelry, and tattoos connote specific subcultural affiliations and situate the racialized subject within these locations, suggesting an interpretation of the embodiment of and performance of gender.For example, the bottom right image, where one might assume the figure is male, reflects a specific subset of twenty-something, queer, White masculinity where a subtle mustache coupled with slim upper body musculature connotes masculinity more than the longish hair and slim waist might connote femininity.In differently racialized bodies, the aesthetics of body shape and hair might be interpreted differently, resulting in a variety of disparate assumptions about gender and racialized aesthetics.
Due to the way gender necessarily intersects with corporeality as signifier, as one engages in the process of looking at Female to "Male", the figure's gender embodiment is interpreted within the pre-existing matrix of gendered expectation viewable within visual culture, which, in the "West", is inherently racialized.Repeatedly imaging himself, ostensibly under the guise of making his shifting gender morphology visible, the artist is also bringing his White identity into view.Being careful not to conflate heterosexuality, Whiteness, and cis-normativity, I would like to suggest that Female to "Male" as praxis invites a reflection on the unseating of cis-normativity while also working to make Whiteness visible.This intervention is imperative when one recalls the work of queer of color critique, which has compellingly outlined that without attending to racialization and naming Whiteness, one limits and homogenizes queer theory under Whiteness as the default.In addition, continues to center heteropatriarchal White supremacy (Ferguson 2004;Chen 2017;Johnson and Henderson 2005).
Have/Hold and the Transing of Identity
Wynne Neilly recently showcased an unprecedented body of work, Have/Hold, at We Buy Gold, Gallery TPW, Toronto, Canada curated by Michèle Pearson Clarke.A collaboration with fellow artist Kyle Lasky, the project explores the intimate friendship between the two White transmen in photographic form (Figure 3).Playing with the language of marriage vows and bonds of matrimony, the title signals the affection, vulnerability, and love explored in the work.Neilly describes the project as "challenging traditional narratives and stigma around masculine intimacy" (Neilly and Lasky n.d.).Artist friends Neilly and Lasky met in art school at a time when both were beginning to transition and developing a deep friendship.Now that the collaborators are regularly interpreted as male in daily life and find themselves dating women, their relationship with one another strikes the artists as an "anomaly" among representations of male friendship, and they find their closeness is often misinterpreted as that of lovers.Neilly attributes their ability to be so intimate to their friendship being rooted in queerness and not in heteropatriarchal cis masculinity (Neilly and Lasky n.d.).The images in Have/Hold transmit these sentiments, intimacies, and complexities, pushing the boundaries of what a masculine friendship can look like, and challenging assumptions viewers may be inclined to make about the work and the artists.Whiteness, one limits and homogenizes queer theory under Whiteness as the default.In addition, continues to center heteropatriarchal White supremacy (Ferguson 2004;Chen 2017;Johnson and Henderson 2005).
Have/Hold and the Transing of Identity
Wynne Neilly recently showcased a new body of work, Have/Hold, at Arsenal Contemporary Art in Toronto, ON, as part of the 2023 New Generation Photography award, an honor he holds alongside Hannah Doucet and Gonzalo Reyes Rodríguez.A collaboration with fellow artist Kyle Lasky, the project explores the intimate friendship between the two White transmen in photographic form (Figure 4).Playing with the language of marriage vows and bonds of matrimony, the title signals the affection, vulnerability, and love explored in the work.Neilly describes the project as "challenging traditional narratives and stigma around masculine intimacy" (Neilly and Lasky n.d.).Artist friends Neilly and Lasky met in art school at a time when both were beginning to transition and developing a deep friendship.Now that the collaborators are regularly interpreted as male in daily life and find themselves dating women, their relationship with one another strikes the artists as an "anomaly" among representations of male friendship, and they find their closeness is often misinterpreted as that of lovers.Neilly attributes their ability to be so intimate to their friendship being rooted in queerness and not in heteropatriarchal cis masculinity (Neilly, N.D.) The images in Have/Hold transmit these sentiments, intimacies, and complexities, pushing the boundaries of what a masculine friendship can look like, and challenging assumptions viewers may be inclined to make about the work and the artists.In a wall-size installation of collaborative, photographic self-portraits, the artists wallpaper the gallery with two floor-to-ceiling, black and white photobooth images of themselves.In the left image, the two faces look out of the picture plane with bright eyes and broad smiles beneath mustaches, their faces exuding glee and their dark t-shirts blending into one amorphous torso before a rumpled backdrop curtain.In the image on the right, the two subjects have turned their buzzed heads toward each other to engage in In a wall-size installation of collaborative, photographic self-portraits, the artists wallpaper the gallery with two floor-to-ceiling, black and white photobooth images of themselves.In the left image, the two faces look out of the picture plane with bright eyes and broad smiles beneath mustaches, their faces exuding glee and their dark t-shirts blending into one amorphous torso before a rumpled backdrop curtain.In the image on the right, the two subjects have turned their buzzed heads toward each other to engage in a tender kiss.The expression of love between the two close friends is palpable.Functioning as a mural behind the other, smaller pictures, the massive photographs ground the installation in feelings of intimacy, humor, and tenderness.Their matching buzz cuts and facial hair, coupled with the kiss, draw references to various queer subcultures and photographs found in the oeuvres of Wolfgang Tillmans and Cathy Opie, where buzzcuts appear donned by queers of various genders in reclamation and queering of an aesthetic oft deployed to signal white supremacist homophobia.The series of images, hung in a loose but thoughtful composition of overlapping and undulating rectangles, shows a story of two friends who enjoy travel, adventures, leisure, and intimacy.The figures inhabit a world where Neilly and Lasky move from camping to hotel rooms to haircuts by windows, from bathing to embracing in a world populated only by them.It is a portrait of the pair and a love letter to friendship.The configuration includes a smaller, more formal, double portrait of the two, each in a plaid shirt and backward baseball cap in what appears to be a studio setting, where Neilly stands in front and Lasky hangs his arm over Neilly's shoulder.Hanging in the installation is the red plaid shirt worn by Lasky in the photograph, bringing the image into the present through tactility and sentimentality.The plaid flannel shirt creates aesthetic allusions to dyke culture of the 1990s, especially in the US and Canada, cultural touchstones that the trans men who once inhabited queer dyke circles make fond and playful reference to by including an image of the collaborators wearing plaid flannel earlier on in their medical transitions where a viewer might want to categorize Neilly and Lasky's gender presentations as masculine presenting women, or as Neilly describes them, "young, butch women" (Neilly and Lasky n.d.).The inclusion of this image and the shirt as sculptural object highlight the origin of their friendship in queer female assumed identities.This visual point rings as pertinent in that it aestheticizes a transing of identity.One cannot assume who the subjects are based on a visual read of a single given image.Identity is far more complex than that which may be viewed in a singular photograph.
In a framed, central photograph, the two appear nude in what looks to be a hot spring in a rugged landscape (Figure 4).Snowcapped mountains rest in the distance, and grassy marshland dapples the midground of the picture.Centered and gazing directly outward through the picture plane, Neilly sits behind and above Lasky on the edge of the pool, with his legs flanking Lasky's sides, one hand placed gently on Lasky's shoulder and the other on Lasky's chest.Their pale, tattooed bodies blend into one another.At first glance, one might not differentiate between Neilly's leg and Lasky's arm; both sport black and grey tattoos, pale skin and body hair.The positioning of their bodies both conceals and reveals signs of transness.Lasky's body blocks places where one might see markings of Neilly's transness, and Lasky's right hand rests in a gentle fist beneath the water, obscuring any view of his genitals.Nevertheless, on Lasky's hairy chest, appear faint red, uneven lines of top surgery scars, soft markings of his transness, faintly echoing the rugged horizon line.
Sofia Cordova, Torreya Cummings, Laura Aguilar, Elle Perez, Tyler Mitchell, and others, work that considers the complex histories of landscapes in relation to colonization and queer and trans utopia and solace.Disidentification is about recycling and rethinking encoded meaning.The process of disidentification scrambles and reconstructs the encoded messages of a cultural text in a fashion that both exposes the encoded message's universalizing and exclusionary machinations and recruits its workings to account for, include, and empower minority identities and identifications.Thus, disidentification is a step further than cracking open the code of the majority; it proceeds to use this code as raw material for representing a disempowered politics of positionality that has been rendered unthinkable by the dominant culture.(Muñoz 1999) Disidentifying with dominant cultural ideas about White masculinity, Have/Hold works on and against essentialized notions of heteropatriarchal, cis-supremacist, Whitesupremacist, binary masculinity.Rather than replicating toxic traits of White masculinity, including machismo, hostility, aggression, homophobia, hyper-independence, emotional repression, and feminist aversion, for example, Have/Hold displays images of trans queer White masculine tenderness, vulnerability, intimacy, and platonic love.Neilly's and Lasky's expressions of identity in Have/Hold reside not in their overt performance of transness but in their transing of the expectations of White masculinity as ideologically constructed in Anglo-European and North American contemporary culture.The picture of the two friends, nude and bathing close together in the wild landscape, recalls images of Greco-Roman baths, gay cruising grounds, spaghetti westerns, and formal portraiture.The soft and gentle gesture of Neilly resting his limbs on Lasky and Lasky resting his left arm in return over Neilly's knee speaks volumes about the beautiful tenderness of friendship and the ability to be vulnerable-things familiar in queer circles but seldom seen in mainstream depictions of White, heterosexual, cis masculinity.
When the figures are interpreted as cis White men, the two central figures, selfpossessed and dominating the image, open questions about their relationship to the landscape that recall histories of colonization, stories of conquest, and colonialist ideologies that painted and pictured Turtle Island as an empty wilderness for the taking. 5When viewed as trans, the tender depiction of masculinity references histories of queer and trans people seeking solace in locations populated only by LGBTQ+ community and practices of image-making by and for LGBTQ+ people in landscapes with the intent of freely being oneself and the photographic work of artists such as Tammy Rae Carland, Ryan McGinley, Sofia Cordova, Torreya Cummings, Laura Aguilar, Elle Perez, Tyler Mitchell, and others, work that considers the complex histories of landscapes in relation to colonization and queer and trans utopia and solace.
Strategically deploying and transforming dominant cultural modes of White masculinity, Have/Hold reflects Neilly and Lasky's identities and relationship, playfully and skillfully breaking open nuances of White masculinity.The deployment and transformation of White masculinity one encounters in the images that constitute Have/Hold can be understood in conversation with José Esteban Muñoz's concept of disidentification.As José Esteban Muñoz puts it, Disidentification is about recycling and rethinking encoded meaning.The process of disidentification scrambles and reconstructs the encoded messages of a cultural text in a fashion that both exposes the encoded message's universalizing and exclusionary machinations and recruits its workings to account for, include, and empower minority identities and identifications.Thus, disidentification is a step further than cracking open the code of the majority; it proceeds to use this code as raw material for representing a disempowered politics of positionality that has been rendered unthinkable by the dominant culture.(Muñoz 1999) Disidentifying with dominant cultural ideas about White masculinity, Have/Hold works on and against essentialized notions of heteropatriarchal, cis-supremacist, Whitesupremacist, binary masculinity.Rather than replicating toxic traits of White masculinity, including machismo, hostility, aggression, homophobia, hyper-independence, emotional repression, and feminist aversion, for example, Have/Hold displays images of trans queer White masculine tenderness, vulnerability, intimacy, and platonic love.Neilly's and Lasky's expressions of identity in Have/Hold reside not in their overt performance of transness but in their transing of the expectations of White masculinity as ideologically constructed in Anglo-European and North American contemporary culture.
Transing identity in this way can be understood as a methodological intervention and lived maneuver that de-sutures understandings of identity from the surface aesthetics of physical bodies, highlighting the complex and slippery relationship between corporeality and identity.While simultaneously cracking open, disidentifying with, and transforming dominant cultural expectations of identity categories.In other words, the pair's facial hair and flat chests do not equate to any given type of identity.Nevertheless, these aesthetics play with circulating cultural associations about body hair and physiognomy.Flat chests and mustaches are often assumed to signal masculinity or are ascribed masculine gender attributes, and yet this is not necessary or given.Neilly, for example, is more connected with being trans than being male.In Neilly's own words: I very strongly identify with being trans.My trans identity is not binary in the ways that society probably expects it to be.When heteronormative or mainstream society imagines a female born body transiting to a body that is perceived as masculine, there is an automatic reading of that person being "female to male" or FTM.This FTM experience might be very relatable and true for many trans people, but it is also completely wrong for others.(Neilly 2015) Identity for trans people is rooted in who we know ourselves to be and who we tell people we are.In articulating one's transness, one often says, "This is who I am" in the face of a dominant cultural paradigm (and often actual people saying), "I know who you are based on what I see and how I assign categories and values to what I see."The trans person rejects the logic of the colonizing gaze.In this way, trans representations create a crisis of Western conceptions of visibility and visuality.As Micha Cárdenas notes, For trans visibility to be a reality, there would have to be an essential trans identity to make visible, but there is not.How could one make visible an identity that begins with the claim: "I am not what I appear to be; I know this because of a feeling that I have; I am my vision of my future self."(Cárdenas 2017) Transing identity is a radical shift at odds with dominant Western ideologies about identity and visuality.When Wynne Neilly and Kyle Lasky created their collaborative photographic project, they were visually creating a rejection of the pseudo-logic that a viewer can know who they are by looking at them.Moreover, when a viewer accepts the figures as they articulate themselves to be, even if it seems at odds with how they appear, then the viewer is rethinking the colonial ideologies that undergird how people residing in locations ideologically entrenched in colonial pseudo-logics have been trained to look at people and portraits.As trans scholar B. Preciado notes, male and female exist only because they are continuously produced.Gender is dependent on visuality and tied to photography.Emerging during the era of colonization, photography marked a significant stage in the production of sexed and gendered subjects via the notion of "visual truth" Preciado 2013).Have/Hold disidentifies with White masculinity, breaking with the conceptualized collapsing of index and indexed or photograph and corporeality.Have/Hold highlights precisely how meaning and value are assigned to bodies based primarily on ideological interpretations of visual information.The work also proposes the unfixing of identity and the de-suturing of Whiteness and masculinity as given and neutral positions.
Part of the colonialist project was the formation and deployment of the concept of the stereotype.The stereotype has since endured as an integral component of various identity-based oppressions in locations ideologically descended from colonial projects.The pseudo-logic of the stereotype is dependent on the ocular-centric cultural belief in valuing looking.According to post-colonial scholar Homi K. Bhabha, a stereotype is "a fixed reality" which is at once an "other" and yet knowable in that it is visible.(Bhabha 2012).Bhabha's observations are invaluable to the discussion at hand.Stereotypes are not factual, but they become endowed with cultural currency when they are anxiously repeated in visual culture to such an extent that the reductive iconic image and the (often damaging) ideas sutured to representations become synonymous with a group of people.Bhabha observes that stereotypes are created about people in an attempt to malign and oppress.Stereotypes produce and cement troubled, problematic, and reductive beliefs about groups of people, essentializing identities and perpetuating ill-treatment of those deemed worthy of mistreatment.However, the deeper issue with stereotypes is that they reinforce and are reinforced by a perpetuated belief in the false equivalence of seeing and knowing.Transing identity makes apparent that this pseudo logic is inaccurate and works on and against such belief systems, mobilizing other ways of engaging visuality and identity.
Have/Hold pushes the limits of White masculinity, making plain that, like all identities, masculinity, in the current colonial schema, is both constructed by and co-constituted by representations.Have/Hold reflects what Jack Halberstam has called a "repudiation of the veracity of the visual" (Halberstam 2018, p. 96).The praxis of transing identity that Neilly's work engages highlights that the surface of the picture and the aesthetics of a person in an image and a person's corporeality in daily life are not necessarily correlative to any assumption one may make based on looking.The rejection of the problematic belief that one may understand how another person identifies based on looking at them is a challenge to a foundational concept of a particular Western colonialist worldview and opens a new way of rethinking visuality, representation, and identity.This suggests that if one unfixes the idea of reality being sutured to the privileging of sight as articulated by C. Riley Snorton new understandings, methodologies, and frameworks for inquiry become possible (Snorton 2017).Recognizing that identity resides within and is transmitted outward via complex embodied, referential, gestural, and performative aesthetics, trans people and trans methodologies shift the discourse on identity to one that originates in self-articulation and embodied feeling and away from an ocular-centric worldview, calling into question the meanings that are made between the gendered subject and those interpreting gender via nuanced ideologically informed processes rooted in colonialist logics.Specifically, Have/Hold suggests that binary, heterosexual, White supremacist, cis-supremacist masculinity is a stereotype, a reduced and essentialized fiction, and that masculinity as gender is, in fact, a concept that is contingent on location, temporality, and culture, and is porous, transformable, intersectional, and ever-evolving.
Concluding Thoughts
Trans-self-representations such as those found in Wynne Neilly's oeuvre challenge the dominant cultural logics of White supremacy and cis supremacy, debunking an inaccurate and troubling belief that seeing has an uncomplicated and indexical relationship with the ascertaining of knowledge.Studying trans self-imaging practices such as Neilly's provides a methodology that moves beyond binary structures, de-essentializes how we think about representation and identity, and encourages continually malleable, self-reflexive methods of knowledge formation.Transing identity intervenes in visual culture theory and, in this instance, discourses on photography and the complex relationship between representations and ideas of truth and authenticity, revealing that the indexicality often associated with photographs is similar to the essentialist ways in which one may assume the exteriority of a subject matches their identification, thus opening up nuanced and often creative modes of engaging these ideological legacies while undoing their presupposition.
Transing identity as a methodology rejects the belief that by looking at someone, one can know their identity.Transing identity shifts the understanding of identity that privileges that which is embodied and experiential.Identity comes from within, not from someone else looking at us and telling us who we are.Transing identity provides a methodological way out of colonialist and White supremacist logic.This logic dictates that the viewer ascribes one's identity via a process beyond reproach and void of agenda or ideology.Transing identity highlights that identity is malleable, fluctuating, intersectional, and collaborative, created in community, and communicated in a field of visual and performed aesthetics, where we all are both interpellated and disidentifying to varying degrees, with numerous aesthetics and gendered and raced frameworks circulating in given temporal, social, and geopolitical locations.Transing identity underscores the humanity of people, the infinitely complex play of visual interpretation, and opens up nuanced possibilities for further investigation into questions of identity, representation, and visuality.
Arts 2024 ,
13, x FOR PEER REVIEW 2 of 15 several intimate photographs which cast the actor as eager yet trepidatious about publicly announcing his transness.
Wynne Neilly is a queer, White trans artist who has created trans and queer representations for over a decade.He earned a BFA in Image Arts from Ryerson University in Toronto in 2012 and has gone on to do commissioned work for publications such as TIME and creative work that uses portraiture to explore trans and queer identity.Neilly has had solo and group exhibitions throughout the U.S. and Canada at the International Center of Photography, New York; Joseph Gross Gallery, Tucson, Arizona; and Toronto Image Works Gallery, Toronto.His work has been featured in Aperture, Vision Magazine, Unbound, and Flash Forward, among others.The artist's oeuvre can be situated in the legacy of Loren Cameron and in discourse with artists such as Robert Girard, Robert Mapplethorpe, Catherine Opie, Del la Grace Volcano, and Andy Warhol, as well as with younger artists including Chris Berntsen, Cassils, Zackary Drucker, Jess Dugan, Rhys Ernst, Ren Hang, Texas Isaiah, Amos Mac, Tyler Mitchell, Zanele Muholi, Cayote Park, Elle Perez, Lorenzo Triburgo, and others who use portrait photographs to explore the complexities of queer and trans identities.
Neilly is a queer, White trans artist who has created trans and queer representations for over a decade.He earned a BFA in Image Arts from Ryerson University in Toronto in 2012 and has gone on to do commissioned work for publications such as TIME and creative work that uses portraiture to explore trans and queer identity.Neilly has had solo and group exhibitions throughout the U.S. and Canada at the International Center of Photography, New York; Joseph Gross Gallery, Tucson, Arizona; and Toronto Image Works Gallery, Toronto.His work has been featured in Aperture, Vision Magazine, Unbound, and Flash Forward, among others.The artist's oeuvre can be situated in the legacy of Loren Cameron and in discourse with artists such as Robert Girard, Robert Mapplethorpe, Catherine Opie, Del la Grace Volcano, and Andy Warhol, as well as with younger artists including Chris Berntsen, Cassils, Zackary Drucker, Jess Dugan, Rhys Ernst, Ren Hang, Texas Isaiah, Amos Mac, Tyler Mitchell, Zanele Muholi, Cayote Park, Elle Perez, Lorenzo Triburgo, and others who use portrait photographs to explore the complexities of queer and trans identities.
Figure 4 .
Figure 4. Installation Image of Have/Hold, 2021, Work by Wynne Neilly and Kyle Lasky.Courtesy of Stephen Bulger Gallery.Installation image shot by Toni Hafkenscheid.
Figure 3 .
Figure 3. Installation Image of Have/Hold, 2021, Work by Wynne Neilly and Kyle Lasky.Courtesy of Stephen Bulger Gallery.Installation image shot by Toni Hafkenscheid.
Figure 5 .
Figure 5.Our Favourite Spot, 2018, Work by Wynne Neilly and Kyle Lasky.Courtesy of Stephen Bulger Gallery.Strategically deploying and transforming dominant cultural modes of White masculinity, Have/Hold reflects Neilly and Lasky's identities and relationship, playfully and skillfully breaking open nuances of White masculinity.The deployment and transformation of White masculinity one encounters in the images that constitute Have/Hold can be understood in conversation with José Esteban Muñoz's concept of disidentification.As José Esteban Muñoz puts it,
Figure 4 .
Figure 4. Our Favourite Spot, 2018, Work by Wynne Neilly and Kyle Lasky.Courtesy of Stephen Bulger Gallery. | 2024-01-28T16:43:29.231Z | 2024-01-25T00:00:00.000 | {
"year": 2024,
"sha1": "5960354a628a70a72fae9f061f87c03324ccb43c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2076-0752/13/1/22/pdf?version=1706181665",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "3f180209889ddcde6db1aba9e097c9fa7eb58981",
"s2fieldsofstudy": [
"Sociology"
],
"extfieldsofstudy": []
} |
219041846 | pes2o/s2orc | v3-fos-license | Kinetic Study of Simultaneous Ethanol Decomposition and Reduction of Low-grade Iron ore at Transient Temperature
An ethanol-assisted ironmaking method had previously been proposed, resulting in a significantly lower temperature of iron reduction. However, the kinetic mechanism remains unknown. A kinetic model for transient temperature condition was introduced, then numerically solved and curve-fitted to the experimental data using MATLAB software. There were 8 main reactions involved: (1) decomposition of ethanol, (2) decomposition of methane, (3) steam reforming of methane, (4) water-gas shift, (5) Boudoard reaction, (6) direct reduction by C, (7) indirect reduction of iron oxide by CO, and (8) iron reduction by H2. Curve fitting methods were successfully conducted with satisfying results (R2 > 0.90) calculating three parameters consisting of reaction rate constants, activation energies, and diffusion factors of each reaction. The interpretation of the calculated diffusion factors reveals that the reaction of gases inside pore might be inhibited by carbon deposition on the pore surface of iron ore. It revealed that the direct reduction by C predominates the overall reduction process.
Introduction
The light alcohol, such as ethanol is a prominent candidate as a reducing agent for ironmaking [1]. Not only is it less hazardous, but it can be produced from a variety of biomass sources. Its use could lead to a decrease in CO2 emissions. Currently, bioethanol is produced by the fermentation of sugar cane, corn grains, and other starch-rich materials, and is a mature technology [2]. Recent researchers have focused on producing bioethanol from lignocellulosic biomass, the so-called second-generation bioethanol. Ethanol production from lignocellulose would open the possibility for the utilization of different and low-cost biomass, such as agricultural wastes and forestry residues (e.g., straw, grasses, corn stover, and wood); in this way, a competition of energy-food production from agricultural sources would be avoided [3]. Another remarkable finding of ethanol production from gas fermentation, recently, is also getting more attractive to be implemented. Ethanol recovery like Lanzatech process, utilize the tail gas of iron and steelmaking processes might be one approach to realize the idea to ensure ethanol security as ironmaking fuel [4]. More approaches are required to evaluate the benefit of ethanol utilization as a reduction agent for ironmaking.
On the other hand, the abundance of goethite over hematite and magnetite ore needs more attention to be effectively utilized. Goethite, is a low-grade iron ore containing high combined water, requires more energy for converting it to hematite through a dehydration process before reduction [5]. For compensating the energy requirement for a dehydration process, the iron reduction reaction should be faster or proceeded at a lower temperature. Figure 1 shows the proposed process of ethanol-assisted ironmaking utilizing the porous iron ore from goethite dehydration as raw material. In the system of ethanol decomposition and iron reduction, ethanol acts as a reducing agent for iron oxide, meanwhile the iron oxide act as a catalyst for ethanol decomposition producing H2 and CO. As reported in [1], ethanol reduces the iron oxide at a significantly lower temperature compared to coke-iron oxide reaction. However, the reaction mechanisms are still unclear. As one approach, the porous iron ores is submitted to a programmed transient temperature condition (with a constant heating rate) while a reducing gas mixture flows over it. This method has been commonly known as a temperature-programmed reduction (TPR) [7]. The current study is aimed to reveal the mechanism of the reaction during the simultaneous ethanol decomposition and iron reduction process through kinetic simulation and data curve-fitting.
Ethanol decomposition and iron reduction in temperature-programmed reaction
Australian goethite (GE) ore with the total Fe content of 57wt% and combined water (CW) of 8.6wt% was employed in this study. The original ore was sieved so that the particle size range from 1 to 2 mm. The porous ore was already prepared by dehydrating the original one at 300 o C with a heating rate of 3 o C min −1 and kept for 24 h under air atmosphere, as mentioned in the previous study [1]. The reduction experiment was conducted by reducing 0.3 g of dehydrated low-grade iron ore in a quartz tube reactor (ID of 6 mm) under an inert Ar atmosphere. The reactor was heated in a vertical infrared furnace to reach final temperature (900 o C) at a heating rate of 10 o C min −1 . Ethanol reagent (99.5 vol%) at 0.04 ml min −1 was drop-wisely introduced into the iron ore bed in the reactor during heating up. The outlet reactor gases were analyzed in-line using quadrupole mass spectroscopy (QMS200, Pfeiffer).
Sample Analysis and Characterizations
2.2.1. Gas Analysis. The outlet reactor gases, such as C2H5OH, H2, CH4, CO, CO2, and H2O were analyzed using quadrupole mass spectroscopy (QMS200, Pfeiffer). The evolution of total outlet gas flowrate was also monitored and recorded using a gas flowmeter. The molar flowrates of each gas were calculated as in our previous study [1] using (1).
̇= 41571.74 (1) The factor of 41571.74 is the conversion factor of gas volumetric flowrate to molar flowrates at NTP -Normal Temperature and Pressure -condition [µmol Nml -1 ]. The NTP condition is defined as 20 o C and 1 atm.
Reduction degree (RD) calculation.
Characteristics of the ore structures and compositions before and after the reactions were analyzed using X-ray powder diffractometry (XRD; Miniflex, Rigaku). The reduction degree (RD) can be estimated as the molar percentage of the removed oxygen to the total removable oxygen in the sample. For instance, the RD of pure substances of Fe2O3, Fe3O4, FeO, Fe, and Fe3C were considered as 0%, 11%, 30%, 100%, and 100%, respectively. As considering the presence of some gangue materials such as SiO2, Al2O3, and other oxides within the ore samples. The RD of each sample was then calculated based on the composition of iron oxides in the sample using equation (2) = ∑
Carbon content analysis.
The carbon content in the iron ore was calculated by the weight change of the iron ore after reduction using ethanol and after combustion at 1000 °C for 30 min under 500 ml of airflow. Considering the reoxidation of reduced iron ore to hematite, the carbon content was calculated using a similar method in our previous study [1].
Microscopic Analysis.
To observe the deposition of carbon from ethanol in the pore of iron ore, carbon mapping cross-sectional analysis was conducted using a scanning electron microscope (SEM) equipped with energy-dispersive X-ray spectroscopy (EDS) JEOL JSM-7400F. In addition, to confirm the nanoscale contacts between the deposited carbon and iron ore, transmission electron microscope (TEM) was taken using a JEOL JEM-2010 microscope at an operating voltage of 200 kV at the highest magnifications.
The kinetic equation with the second-order reaction at each reaction in term of the generation of component is expressed as in (3).
=
(3) The mechanism includes the mass transfers of both reactants and products, in term of external diffusion and internal diffusion in the pore, the adsorption/desorption into the pore surface, and the surface reactions. In overall terms, introducing a resistance concept in reaction model would simplify the overall kinetic term as in (4) Mass transfer phenomena of gases into pore including internal and external diffusion are evaluated by the coefficient of apparent mass transfer effect on the overall kinetic, , for each component . Thus, the overall kinetic term should be Some assumptions were taken into consideration: (i) was assumed as a constant. (ii) Since the iron ore bed (0.3 g) in the reactor (ID = 6 mm; H = 7 mm) is quite small, and with a preliminary experiment confirming the temperature is uniform at any axial position around the ore bed. In other meaning, reactions were assumed uniform at any positions in bed. In addition, since in the transient temperature condition, the equilibrium is not reached. Thus, the equilibrium constant was not taken into consideration.
In the complex system, the overall rate generation of component can be expressed The term of ( 1 − 1 0 ) in the overall kinetic equation was introduced to improve the independency of each parameter during curve fitting.
Under the transient temperature state with a heating rate = (i.e., 10 K min -1 ], the mole balance of each components can be expressed as (8) In the case of ethanol, initial molar amount, ̇2 5 0 is 686 µmol min -1 , while for other product gases, the ̇0 = 0, respectively. (8), 24 parameters were estimated, consists of eight kinetics constants of reactions at T0 (= 750 o C), ~ , eight activation energies, ~ , and eight coefficients of the mass transfer effect, ~ . These parameters were estimated by a curve fitting method using the sum square error ( ) as the objective function. Considering the SSE should be not containing any units. Thus, the unitless SSE was introduced in the fitting as in (9).
Simultaneous ethanol decomposition and iron reduction
This study utilized the data of the simultaneous ethanol decomposition with iron reduction reaction that previously reported in [1]. The raw data was then reorganized to find a correlation between kinetic activity and oxide structure in the ore as depicted in Figure 2. These data can be interpreted briefly that significant ethanol decomposition is started at 300 o C, together with iron reduction by CO (Fe2O3 + CO → Fe3O4 + CO2). At 400 o C, Iron reduction by H2 is started (Fe2O3 + H2 → Fe3O4 + H2O, and Fe3O4 + H2 → FeO + H2O). At 600 ~ 700 o C, the ethanol decomposition seems slower when iron ore structure change from Fe3O4 to FeO. Above 700 o C, significant iron reduction by deposited-C is started (FeO + C = Fe + CO). It is remarked by as C content increasing, the RD and CO increase significantly. While there is no significant generation in H2O from reduction by H2 and CO2 from reduction by CO. Figure 2. Simultaneous ethanol decomposition-iron reduction of dehydrated goethite ore in [1]. Data were reorganized for kinetic analysis.
Curve fitting results
RD and carbon contents in the ore are only evaluated at 100 o C intervals. Since the RD is expressed in molar percentage, it is difficult to use in the kinetic calculation. Thus, the term of removed oxygen (RO) was introduced to describe the remaining amount of removable oxygen in the iron oxide component. The initial RO (as hematite content in the ore) and carbon amount of the iron ore were 718 and 0 µmol, respectively. The RO and carbon content are both expressed as the accumulated amount (i.e., in molar unit). The gas flowrate data were obtained in the expression of the mole amount per time unit (e.g., in the unit of µmol min -1 ). To arrange the same unit order of all measured component amount, the molar flowrate data of the generated gases j were reformed into the accumulated molar amount of gas using (10) , = , −1 + ∑ +̇, −1 ) with = index of data point (10) Figure 3 shows the curve fitting result of each component in the accumulated molar amounts. In addition, to demonstrate the reduction behavior, the simulated RO was then converted back to RD. Based on the goodness of fit, R 2 , the proposed model was satisfactorily fit to the data for H2O, C, RD, CO, H2, and C2H5OH. However, it was quite fit CH4 and CO2. Based on the curve shape of RD compared to the amount of C, H2, and CO, as the reducing agent. The RD is found strongly correspond with the carbon amount, especially at the temperature higher than 700 o C, and slightly with H2 and CO, even if the H2 and CO are abundantly present during the ethanol decomposition process. A previous study [8] mentioned that the increase of RD corresponds to the existence of H2 and CO, which are dominate the reduction process. However, the reduction by H2 is tended to a constant rate at the temperature higher than 450 o C [9]. The reduction by H2 and CO seems to have an effect at the lower temperature only. Understanding the reduction behavior under simultaneous process is complicated to observe. In our approach, the simulation calculated the rate constant of the proposed reaction model obtaining the parameters of the rate constant 0 and activation energy of each reaction were listed in Table 1. Figure 3. Component molar accumulated data and the model curve fitting results. The R 2 closes to 1 means the model almost fits to experimental data. The 0 is evaluated for T0 = 750 o C, meaning that all 0 simulate the rate of each reaction at 750 o C. These values describe the predominant reactions occur during the simultaneous ethanol decomposition and iron reduction process. The higher values of 0 , the more dominating reaction is. As in the ethanol decomposition, ethanol decomposed into CH4, H2O, and deposited C. The reaction then immediately proceed to the steam reforming generating H2 and CO, and water gas shift reaction generating more H2 and CO2. However, a near-zero value of 0 for Boudoard reaction indicates that Boudoard reaction does not occur in this process since the reaction temperature for Boudoard reaction requires higher than 800 o C [12]. Meanwhile, upon the iron reduction, it is obvious that at 750 o C the iron reduction by carbon predominates the overall reduction process. It seems that the iron reduction by H2 and CO have no significant effect above this temperature. The activation energies obtained from this study were then found mostly agreed in comparison to the other references data, which were evaluated from other non-simultaneous reaction studies. 3.3 Interpretation of parameter . Table 2 shows the obtained parameters of from the curve fittings. The parameter of was introduced to compensate for the unknown effect of ore particle size, surface area, etc. as well as mass transfer effect through simultaneous surface reaction occurs.
in this term correlates with the mass transfer properties of each reacted components.
could be also introduced as Thiele modulus describing the relationship between diffusion and reaction rate in porous ore with no mass transfer limitations. Thiele modulus is generally used in determining the effectiveness factor for catalyst pellets [18]. This approach is set according to assumptions of the diffusion through a surface film is very fast compared to diffusion into the grain interior.
The overall generation of RO, H2O, and CO are controlled by the reaction, which corresponds to the reduction process by C and H2. Meanwhile, the generation of other components, specifically the deposited carbon and CH4 are controlled by diffusion that might correspond to the ethanol decomposition reaction, steam reforming, and water gas shift taking place inside the ore. Since it was difficult to distinguish the contribution of each reaction in the component generation, the interpretation of might be limited to the overall generation of the component only. The factors of mass transfer is only considered as a constant, however, it might be a function of temperature as well to correlate with the effect of sintering phenomena at higher temperature reaction. pore. This will be left for further investigations. In addition, using cross-sectional SEM-EDS observation (Figure 4), it was confirmed that carbon from ethanol decomposition deposited inside iron ore through diffusion in its pore. At the lower temperature (400 o C), the deposited carbon seems only concentrates on some points. Then, as temperature elevated (700 o C), it continues to be accumulated and homogenously distributed inside the iron ore, meaning that the contact between carbon and iron oxide for reduction can proceed in all part of iron ore. TEM observation at the iron ore sample after reduction with ethanol at 750 o C is shown in Figure 5. The TEM image consists two contrast colors: the dark grey (or close to black) one corresponds to the higher density component (i.e., Fe), and the transparent one corresponds to carbon. It confirms that the surface of iron ore is already covered by amorphous carbon in a nano-scaled contact that allows the significant iron reduction by carbon (direct reduction) causing the iron reduction by carbon predominates the overall reduction process. It can be also interpreted that the deposited carbon also might inhibit H2 and CO to effectively reduce the iron ore. This phenomenon corresponds with the reaction rate of the iron reduction by H2 and CO found lower than the reduction by C; nevertheless, the activation energy of reduction by C is the highest among other reaction. As an overall evaluation, the transient state kinetic study can provide more details information about the main mechanisms of simultaneous ethanol decomposition and iron reduction over low-grade iron ore. This result contributes to a significant improvement in our understanding of what we found in our previous study in [1]. It reveals that even if H2 and CO are significantly generated from a complex reaction during ethanol decomposition, the iron reduction by H2 and CO seems only act at the temperature lower than 750 o C. At above the temperature, the reduction by carbon (direct reduction) predominates the overall reduction process. However, further studies need to be done in order to investigate the other parameters to extend the accuracy of the proposed reaction model. This study confirms the one benefit from kinetic aspect to implement the ethanol-assisted iron reduction process for future ironmaking.
Conclusions
Ethanol, counted as a carbon-free product from biomass, is a prominent candidate to substitute the reducing agent for ironmaking, instead of the depleting coal-based coke. The kinetic study is necessary for understanding the mechanism of the simultaneous ethanol decomposition -iron reduction process. Curve fitting method was successfully demonstrated with satisfying results (R 2 >0.90) calculating 24 parameters consisting 0 , , and . There were 8 main reactions involved: (i) decomposition of ethanol, (ii) decomposition of methane, (iii) steam reforming of methane, (iv) water gas shift, (v) Boudoard reactions, (vi) iron reduction by C, (vii) iron oxide reduction by CO, and (vii) iron reduction by H2. However, it seems Bourdoard reaction does not occur during the process. The reduction by H2 and CO proceeds from lower temperature. Meanwhile reduction by deposited C (direct reduction) occurs at higher temperature as well as the deposited carbon is accumulated covering the pore surface. The interpretation of the obtained parameters shows that the reaction of gases inside pore might be inhibited by carbon deposition on the pore surface of iron ore. | 2020-05-07T09:13:39.172Z | 2020-05-01T00:00:00.000 | {
"year": 2020,
"sha1": "10d0a8fd4fa080cab4f519a29a42a85f88e07198",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1757-899x/778/1/012056",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "84e483e2ef1bcf36538b010d667631aad51883a8",
"s2fieldsofstudy": [
"Engineering",
"Chemistry",
"Materials Science"
],
"extfieldsofstudy": [
"Physics",
"Chemistry"
]
} |
208033367 | pes2o/s2orc | v3-fos-license | Experiences of internet-delivered and work-focused cognitive behavioral therapy for stress: A qualitative study
Introduction Stress is one of the major challenges of modern society, causing significant costs and personal problems. In the recent decade a growing body of research has provided support for the efficacy of internet interventions for stress. However, few studies have focused on how participants experience internet interventions for stress. Method The current study was a qualitative follow-up study of an internet-delivered and work-focused cognitive behavioral treatment for stress. The aim was to capture participants' experiences of the treatment and their views on effects on health and well-being. Participants were selected from a controlled study (n = 27), using a criterion-based sampling approach selecting those participants who had completed all treatment modules. Nine semi-structured interviews were held, and the material was analyzed using Thematic Analysis. Results The results indicated that most of the participants experienced positive effects on their mental health and well-being in both life and at work. All participants emphasized the importance of having access to therapist support. In line with previous research, participants found the intervention to be extensive and demanding and expressed the need for extended treatment time and therapist support. Conclusions Considering the limitations of the present study, future research could examine the feasibility of reducing the length of each session, extending deadlines and increasing therapist support. This could improve treatment efficacy and further enhance utilization in the target population.
Theoretical background
Stress and mental health disorders constitute major challenges in today's society. This is especially the case within the working population. For example, every fourth employee has experienced stress during most of their working time (Eurofound, 2017). The Global Burden of Disease Survey estimated that mental disease, including stress-related disorders, will be the second leading cause of disabilities by the year 2020 (World Health Organization, 2008).
Stress also contributes to the onset of many psychiatric disorders, such as anxiety, depression and insomnia (Åkerstedt, 2006;Melchior et al., 2007). In addition, it can result in impaired work functioning and problems in work participation, in the form of absenteeism and longterm sick leave (Joyce et al., 2016;Kessler et al., 2006;Sanderson and Andrews, 2006). Decreased work participation due to stress and common mental health disorders is problematic as it leads to large costs for society. This points out the need for further development, evaluation and implementation of interventions that improve health and wellbeing among employees (Arends et al., 2012).
Despite the well-known health implications of stress and the benefits of mental health interventions, access to treatment is still relatively limited (Chisholm et al., 2016;Goetzel et al., 2014;Hilton et al., 2008;Richardson and Rothstein, 2008). An effective way to increase this access is through internet-based treatment (Andersson et al., 2019). During the last two decades, a growing body of internet-and computerbased interventions has been developed to increase an individual's psychological resources and resilience to a wide range of mental disorders (Andersson et al., 2019), including stress (Heber et al., 2017). Results show that internet-based interventions for stress have both short-and long-term effects on mental health outcomes for a range of participants, including employees (Ebert et al., 2016;Ly et al., 2015;Persson Asplund et al., 2018;Zetterqvist et al., 2003).
Although the efficacy of internet-based interventions is well-established, we still have fairly limited knowledge of how participants experience internet-based interventions (Fleischmann et al., 2017;Halmetoja et al., 2014). Dimidjian and Hollon (2010) stated that both quantitative and qualitative methods should be used in psychological treatment research, and Bendelin et al. (2011) emphasized that qualitative research may help to identify the unique aspects of internet-based treatment.
Participants in previous qualitative studies have reported both positive and negative experiences of internet-based interventions. Positive statements concerning the availability, flexibility and anonymity of the intervention are common among participants (Alberts et al., 2018;Bendelin et al., 2011;Halmetoja et al., 2014;O'Mahen et al., 2015;Pugh et al., 2015). However, participants have suggested more individualized interventions and express the importance of customizing both feedback and therapist support (Gerhards et al., 2011;Ly et al., 2015;O'Mahen et al., 2015;Rozental et al., 2015b;Svartvatten et al., 2015). Furthermore, participants request more content-related therapist support, more background information and better understanding of their own situation (Macdonald et al., 2007;Gerhards et al., 2011). Regarding interventional feedback, participants expressed the importance of tailoring both the amount and the type of feedback to their individual needs (Svartvatten et al., 2015).
To our knowledge, no previous study has examined the experiences of internet-delivered treatment that focuses in both stress and the work situation. We have only found one previous study that has examined participants' experiences of internet-based treatment for stress. Fleischmann et al. (2017) evaluated an internet and app-based stress management intervention used among distressed college students. The students expressed the need for more individualized content, background information, content-related support and a better understanding of their situation. Aligned with previous research (Rozental et al., 2014), participants also described that session-length and frequency of reminders were demanding and constituted a possible stress factor. Internet-based treatments usually contain a large amount of information and instructions, which places high demands on cognitive and executive abilities. Current research suggests that cognitive disturbances are common in chronic stress disorders. For instance, Grossi et al. (2015) and Savic et al. (2017) reported changes in cortical and limbic volume, areas responsible for integration of emotional, cognitive and physiological responses. This corresponds to several of the symptoms that appear in chronic stress disorder. Therefore, experiences of high demands and stress caused by the intervention should be taken under special consideration in the development of internet interventions addressing distressed and exhausted individuals.
Aims
The main purpose of this study was to examine the experiences for participants in an internet-delivered and work-focused cognitive behavioral treatment for stress. As far as we know, no previous studies have examined participants' experiences in an internet-based intervention, addressing both mental health and the work situation. The study was a qualitative follow-up of a randomized controlled pilot trial (manuscript in preparation).
Research trial
The present study was conducted as a follow-up to a randomized controlled pilot trial examining the efficacy of work-focused and internet-based cognitive behavioral treatment (iCBT) for stress, among employees. In the primary study, 54 participants were randomized to either a guided iCBT (n = 27) or a waitlist control group (n = 27) who received unguided self-help after three months. Assessments were conducted at baseline (T1), at post-treatment (T2, ten weeks after randomization) and at three-month follow up (T3). The present qualitative study was conducted after the three-month follow up.
Selection of participants
Twenty-four participants were selected from the treatment group of the original study (n = 27), using a criterion-based sampling approach. Participants who had completed all treatment modules and the threemonth follow-up assessment were selected. The rationale for only including completers was that they were able to evaluate the entire concept of the intervention. Initially, eleven participants reported an interest in participating in interviews, of whom two declined due to lack of time and not responding to invitation. Nine participants remained, i.e. 33% of the original treatment group.
Participants were recruited nationally through advertisement in newspapers, social media and the Swedish Association of Local Authorities and Regions, the principal organization for municipalities, education and healthcare in the country. To be eligible for the study, participants had to fulfil the criteria for an adjustment disorder described in the subdivision F43 Reaction to severe stress, and adjustment disorders of the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10;World Health Organization, 1992). Diagnoses were established through telephone interviews, using the Mini International Neuropsychiatric Interview (Sheehan et al., 1998), and additional criteria from national diagnostic guidelines (Socialstyrelsen, 2003). In addition to adjustment disorder, participants had to fulfil the following criteria: (i) a minimum age of 18 years, (ii) mastering Swedish, (iii) have access to a computer or a tablet computer with internet access and (iv) currently employed. Mild to moderate forms of comorbid conditions were accepted as long as these were considered secondary to the primary adjustment disorder. Participants' on full-or part-time sick leave were also included.
Participants were excluded from the study if they (i) recently had participated in a stress management program, (ii) currently were suffering from bipolar disorder, psychosis, post-traumatic stress syndrome (PTSD), eating disorder, substance abuse, severe forms of depression, anxiety disorder or personality disorders, or (iv) were showing suicidal ideation based on item 9 of the Montgomery Åsberg Depression Scale-Self Rated (Montgomery and Asberg, 1979). Participants taking medication (e.g., antidepressants or sleep medication) were not excluded from the study but were requested to keep their medication constant during the study period. In total, 84 individuals were screened in the original study. 30 were excluded according to the inclusion and exclusion criteria specified above.
Intervention
The iCBT evaluated in the present pilot trial was based on our previous studies (Ly et al., 2015;Persson Asplund et al., 2018;Zetterqvist et al., 2003), showing moderate to large effect sizes (Cohen's d) for stress and burnout (0.50 to 0.95), depression (0.60 to 0.86) and small effect sizes for insomnia (0.34). The iCBT program included contemporary CBT techniques adapted for stress, and recovery from work training inspired by Hahn et al. (2011), combined with workfocused CBT adapted from a manual by Lagerveld et al. (2012). All components were integrated and represented in each module.
The iCBT consisted of ten modules distributed over ten weeks, with sessions lasting 60-120 min each week (Table 1). Each module contained between 5 and 16 regular pages of information. All participants were requested to complete each module and homework assignment before being able to continue. Delayed participants were able to catch up during the last modules of the program. All participants had access to the treatment one year after the post treatment assessment (T2). Each module contained information, exercises, worksheets, images, examples, audio-and video files and homework exercises.
In the first module (introduction), the participants received information about the layout of the program, and they defined their individual goals (e.g., "I would like to be more assertive"). The first week also contained information and exercises on stress physiology, stressors (e.g., workload, pace, social support) and how to manage those stressors. During the second week (balance) the participants were introduced to different recovery techniques (Hahn et al., 2011) and applied relaxation (Öst, 2009). These reoccurred throughout the program. The third and fourth week contained exercises related to behavioral activation (Martell et al., 2001) and the work-home interface and values-based action skills applied from Acceptance Commitment Therapy (Flaxman and Bond, 2010). Between weeks five and ten, participants were able to choose and practice in a stress-related area, assertiveness, perfectionism, procrastination or worry. In addition, participants could choose to focus on sleep management, physical activity and/or time management.
In accordance with the CBT-based program described above, participants also received work-focused modules. These modules were adapted for iCBT with permission from Lagerveld et al. (2012). Workfocused CBT is built on the same conceptual framework as regular CBT. For example, CBT principles are used to change the appraisal of work stressors (e.g., "it is ok although the task is not 100% complete or perfect"), change the dysfunctional behavior (e.g., working late close to bedtime, accept more work despite heavy workload), or increase work and health promoting behaviors (e.g. recreational activities, assertive behavior, return-to-work activates). Exposure is given special attention. Gradual exposure can help individuals develop more effective coping skills when dealing with work-related stressors (e.g., assertiveness). It can also stimulate gradual return to the work setting for individuals on a long-term sick leave (Blonk et al., 2006).
Following the work-focused CBT manual and the module-based layout of the program, participants outlined their (1) Work content, including tasks, demands and workload, (2), Work conditions, comprising employment type, status and security, (3) Work relations, involving cooperation, leadership, organizational justice, feedback, social support, conflicts, harassment and bullying, (4) Job satisfaction, including, meaningfulness and the importance of work for contentment. In module four, participants were asked to summarize and reflect on their entire work situation (e.g., Do you enjoy work? Do you want to stay? What do you need to thrive and cope?). In the consecutive modules (5-10), the participants were encouraged to create a plan for workplace adjustments together with their supervisor, based on their needs and health conditions. Similarly, participants who were on sick leave were asked to make a gradual return-to-work plan together with their general practitioner.
Every week, participants in the iCBT group received personalized written e-mail from a therapist with feedback on the exercises. Guidance was given to the CBT and work focused modules simultaneously. The therapists were Psychologists or Counselors with extensive experience on stress-related disorders who were specifically trained to perform feedback according to a standardized manual. The aim of the feedback was to provide support and encouragement, monitor homework assignments and adherence to the intervention. The iCBT group had full access to regular healthcare during the trial. The therapists were requested to minimize their support to one e-mail per week covering one third of a regular letter and a maximum of 15 min of conversation per week and participant.
Participants
Tables 2 and 3 present the demographics of the nine selected participants original treatment group (TG). In general, the participants did not differ significantly from the participants in the. All participants were female, in contrast to 92% female participants in the TG. Age ranged from 28 to 61 years with a mean of 48.50 years (SD = 10.38) being representative of the TG (M = 47.67, SD = 8.42). Similarly to the TG, all participants had a stress or burnout diagnosis and seven of nine participants had additional diagnoses (depression, anxiety or insomnia).
Interview procedure
Qualitative data was collected through semi-structured interviews, Recovery in everyday life Work/life balance, applied relaxation (step 2) stress diary, coping with stressors, sleep management (step 1) and values Exercise involving job satisfaction, meaningfulness and the importance of working for contentment 5 Challenges Gradual exposure in a stress-related area, recovery through mastering new skills, applied relaxation (step 3) and sleep management (step 2) Summary of the entire work situation. Plan for workplace adjustments and gradual return-to-work plan (only for participants on sick leave) 6 Challenges, continued Continued exposure, values, applied relaxation (step 4) and sleep management (step 3) Effort, appreciation and reward and continued plan for workplace adjustments and gradual return-to-work plan 7 Physical exercise Recovery through physical exercise, applied relaxation (step 5) and sleep management (step 4) Control, responsibility and autonomy and continued plan for workplace adjustments and gradual return-to-work plan 8 Planning Time management, values, applied relaxation (step 6) and sleep management (step 5) Social support, instrumental, emotional and informative support and continued plan for workplace adjustments and gradual return-towork plan 9 Stressed and cognitive functioning Stress, burnout and cognitive functioning, how to manage distractions and temporary memory problems and applied relaxation (step 7) Recovery in the workplace and continued plan for workplace adjustments and gradual return-to-work plan (Kvale and Birkmann, 2014). After completion of the three-month follow-up (T3), the selected participants were contacted via e-mail. If the participants responded positively, an interview appointment was made. Telephone interviews were conducted between December 2016 and January 2017 by two master's degree Counselors (AJ, IHB). After receiving informed consent, the interviews were recorded and anonymized during transcription. Interviewers and coders were not blinded to the participants' treatment outcomes. The average duration of an interview was 52 min (range, 38 min to 1 h 17 min).
Preconceptions
Interviewers and coders were experienced master's degree Counselors and healthcare professionals, with CBT orientation. They had previous experience of CBT of stress and mental health disorders. The interviewers and coders had provided support as therapist in the iCBT group. However, therapist did not interview their assigned participants.
Interview structure
Interview questions were selected based on previous research (Bendelin et al., 2011;Fleischmann et al., 2017;Rozental et al., 2014;Svartvatten et al., 2015). The interview questions aimed at capturing the participants' experiences of the intervention regarding usability, user-friendliness, feedback and support, motivation, session length, extent, work effort and effects on mental health and wellbeing Abbreviations: PSS-10 = perceived stress scale, 10-item version; SMBQ = Shirom Melamed Burnout Questionnaire, 22-item version; T1 and T2 = Pre-and postmeasurement (10 weeks after randomization). Abbreviations: PSS-10 = perceived stress scale, 10-item version; SMBQ = Shirom Melamed Burnout Questionnaire, 22-item version; MADRS-S = Montgomery Åsberg Depression Rating Scale; GAD-7 = generalized anxiety disorder 7-item scale; T1 and T2 = pre-and post-measurement (10 weeks after randomization). Table 4 Interview guide. Questions 1. On average, how many hours each week did you use the program? 2. What sections, exercises or tasks did you find most useful? In what way were they useful? 3. What sections, exercises or tasks did you find less useful? In what way were they less useful? 4. Did you think something was missing in the program? 5. How did you experience the extent and workload (e.g. text, exercises and homework) of the program? 6. How did you experience the level of difficulty of the program? 7. How did you experience your work effort in relation to what you think is required to get the most out of the program? 8. What is your opinion regarding the content of the program? What could be improved? 9. How user-friendly did you find the program? What could be improved? 10. What recommendations would you give to someone who is considering the program? 11. How did you experience having a contact person? 12. How quickly did you get a response from the contact person? How satisfactory was the response? 13. What do you think about the contact person and their ability to give support? 14. What has been most helpful with the contact person? 15. Reflect on whether there was anything you lacked in the contact or not, for example, if there was something that the contact person could do more or less? 16. How would you describe your mental health today? Has it changed? If so, how? 17. How do you apply the knowledge you have gained in the program in your daily life? (Table 4).
Analysis
The interview material was analyzed through thematic analysis according to recommendations from Braun and Clarke (2006). Thematic analysis is a practical and flexible method in which collected material is identified, analyzed and reported on the basis of themes (Joffe, 2011). In the analysis, a combination of deductive and inductive thematic analyses was used. A combination of empirically-controlled and theory-driven themes is common in qualitative research and is superior to a validity point of view (Braun and Clarke, 2013).
Coding procedure
In the analysis process, the authors followed Braun and Clarke's (2006) six-step coding procedure, including (1) Transcription and repeated reading of all material, (2) Coding of keywords (3) Organizing keywords in themes and sub-themes (4) Reviewing and revising themes, (5) Defining, analyzing and naming each theme and sub-themes, (6) Compiling of results and implementing the final analysis. In the first stage, two researchers worked individually to encode all material with repeated reading and coding of keywords and phrases to search for patterns and themes. In the second stage, coding of keywords and themes was compared, reviewed and refined. Disagreements were discussed, and revisions were made in an iterative process until a consensus was achieved. The keywords were then encoded and processed in the qualitative computer program NVivo Software (version 11.4.1.). The advantage of a computer-aided analysis of qualitative data was that all transcripts could remain digital and that large amounts of material could be sorted and grouped based on a coding procedure.
Ethics
The current study was part of a clinical trial, which received ethical approval by the Ethics Committee of Linköping University, Sweden (ref. number 2016/11-31). The anonymity of the participants was ensured using automatically generated identification codes. In addition, the information given by the participants during the interview was anonymized during the transcription process.
Content
The theme content describes the participants' experience of the treatment program and the IT-platform. The content of the program was conceptualized as the level of difficulty and the work-effort in relation to the participants' resources. The experiences of the IT-platform included the user-friendliness of the layout and functions (e.g. login, navigation, video-files).
User-friendliness
Most of the participants perceived the intervention and IT-platform (Vlaescu et al., 2016) straightforward, easy to navigate, easy to understand and functioned without interference. Typical comments were: "There were no interruptions, one could easily move back and forth" (Participant 5), and "There was no problem, it was terrific" (Participant 3). Two participants wished that the program had contained more movies and audio files, in order to facilitate learning and variation: "It could have been a little easier and varied, instead of having a lot of text" (Participant 8). Three participants found the login process difficult because of the two-factor authentication. Three participants would have preferred to print all the modules in one document, instead of printing each module separately. One suggestion was that all participants should have had access to a group forum, where they could have asked each other's questions. This participant also suggested a more smartphone-friendly version.
Module extent
All participants described the program as extensive and demanding: "It was easy to get a little, what should I say, frustrated or stressed over the extent of the program" (Participant 4). However, none of the participants expressed that they wished to reduce the extent of the program: "I don't think anything should be removed" (Participant 2). A majority of the participants perceived the intervention as informative, relevant and well balanced between theories and exercises. As a strategy, not being overwhelmed by the extent of the program, many participants reported that they chose to focus on the most relevant sections: "It was like a large smorgasbord and I choose to work with those sections most central to me" (Participant 8). Five participants expressed the need to extend the treatment after the ten-week period: "If you are really distressed you might not have the strength to finalize the program in ten weeks" (Participant 1) and wished for the "opportunity to extend to twenty weeks" (Participant 7).
Level of difficulty
Three participants expressed that the treatment was demanding: "It requires both motivation and an ability to comprehend and express yourself in writing. It's not for everyone" (Participant 4). However, none of the participants described the material as too theoretical. Four participants expressed that their stress-related difficulties, with symptoms of both emotional and physical fatigue and cognitive disturbances, made it difficult to utilize some of the material: "Sometimes I had to read a section several times to understand, and it was not because they were badly written, it was just because I didn't manage to concentrate" (Participant 1). Accordingly, several participants expressed that, in future development, more consideration should be taken to severe symptom of stress and burnout, (e.g. fatigue and cognitive disturbances). For example, examine the feasibility of reducing sessions to a length that does not drain the participants' resources.
Advantages
Advantages describe the participants' overall satisfaction with the intervention and the flexibility of the internet-based format. It also captures a deeper understanding of the participants' experiences of the sections and exercises most use-and helpful to them.
Overall satisfaction with the intervention
All participants expressed an overall satisfaction with the intervention. Several participants also expressed their gratitude for getting . A majority of the participants stated the advantage of the internet-based format and the flexibility to read, reflect and answer wherever and whenever suited: "I just think in a situation, where you have to commute a long way … this is a way to access treatment" (Participant 5). Two participants expressed the advantage of avoiding their fear of having to meet a therapist face-to-face: "In fact, it can be nice to avoid seeing people in the eyes" (Participant 2). Another advantage mentioned was the opportunity to go back and repeat what was written in previous modules.
Sections perceived as useful
In general, participants expressed that the intervention was useful. Most useful were the sections regarding psychoeducation, managing stress factors in both work and private life, and modules on sustainable working life: "Recovery and sustainability in my working life, that's what I was missing and needed to work with every day" (Participant 6). Several participants emphasized that they had previous knowledge about stress. However, the section on psychoeducation had given them deeper knowledge about physical and mental symptoms of stress and burnout and how they could manage these through behavioral change: "There was a lot of information about stress, what was happening in the body and so on, I think… it was a essential to understand what stress actually is and how it affects our behaviors and symptoms" (Participant 9). Six participants found it useful to reflect on their work situation. For example, one participant expressed it as, "Previously I worked alone… it was the most fun job I've ever had as a nurse… I was well-liked, it was fantastic… but I realized that this loneliness didn't work for me, I hadn't realized that before" (Participant 3).
Disadvantages
This theme describes the participants' experiences of the sections and exercises not use-or helpful to them, and disadvantages of ending the treatment and therapist support after ten weeks.
Sections perceived as less useful
The participants had different opinions about which sections or exercises they perceived as less useful. The task that the majority of the participants mentioned as less useful was the stress diary. The stress diary was described as too extensive and was often skipped: "I thought it was difficult, it takes time because it is often a hundred small things that stresses you and if you have to write it down, … it becomes more stressful" (Participant 7). In addition, there were several heterogeneous statements about which sections or exercises they perceived as less useful. Typically, the participants mentioned individual exercises that they did not consider relevant or useful, such as sleep management or return-to-work plan, when not having insomnia nor being on long-term sick leave.
Disadvantages of ending treatment
Four participants expressed disadvantages of ending the treatment and therapist support after ten weeks. They would have preferred a follow-up session and continued support to maintain positive change: "I probably would have needed a follow-up after a month or two, just to stay on track… I feel that I need help from someone kicking me in the butt by asking, how's it going for you, have you tried this, what's your thought about this … it surely would have been helpful" (Participant 8). Some experienced the end of the treatment as abrupt: "It felt so empty when it was finished, because then there was nothing more… I missed it" (Participant 5).
Guidance
Guidance describes how the participants' perceived the therapist support. The theme refers to the quantitative (e.g. frequency, usage) as well as qualitative aspects of the therapist feedback and support (e.g. authenticity, individualization) and how it meet the participants needs and preferences.
Perceived support
The need for support varied between the participants. Some stated that they had limited, or neutral contact. Others, a more close and personal contact. "I didn't need him so much. A few times I received some hints and ideas and I contacted him when I was ready with one module and he replied" (Participant 9). However, all respondents emphasized the importance of having access to therapist support. Some even stated that the contact was essential for completing the treatment: "If I had done it myself without support, I would not have managed it. Because I procrastinate, I'll do it later" (Participant 4). According to a majority of the participants, the main function of the therapist was to monitor progress and adherence and provide support and encouragement: "Otherwise, I think, if there hadn't been a person who expected feedback, the ten-week program had taken at least thirty weeks" (Participant 1). Another important function was to help participants adjust their ambition: "My therapist urged me to lower my ambition, that I couldn't expect to manage everything the same way as before" (Participant 6). All participants described that they were satisfied with the feedback from the therapist, both in terms of frequency (response time) and quality (content). Support was perceived as individualized with a deep understanding of their situation: "She was super at grasping what she read and maybe between the lines too" (Participant 8) and "The support gave me more than I had expected. I didn't think it was possible to get it this way. When you didn't even talk to each other" (Participant 2). Response time ranged from the same day to three days: "I almost thought, is she never off duty? Because she responded very quickly" (Participant 1).
Need for extended therapist support
Three participants expressed a need for telephone or video support when necessary. E-mail contact was sometimes perceived as inconvenient, formal and limiting and telephone or video support was presented as a more relaxed and spontaneous alternative: "Sometimes it was difficult, yes, to ask a question by e-mail, I am used to dialog, so sometimes, I think, I really would have wanted to discuss things over the phone" (Participant 5).
Commitment
This theme was conceptualized as the time, effort and active use of the treatment.
Time spent on treatment
The participants described a large variation in the time spent on the program. Time ranged between 1 1/2-2 h and 8-10 h per week and with an average of 2-4 h: "Somewhere between two to three hours. I'd say two hours on the actual tasks and one hour when you practice some techniques" (Participant 3).
Effort
Eight participants stated that they would have spent more time on the program to maximize its full potential, if they would have had the time and energy. However, this was considered too extensive in relation to their capacity. Some also expressed the ambition to return to the program after the treatment was completed: "I think I did a good job, but if I would have had more energy, I can imagine I would have put even more time into the program." (Participant 5).
Commitment to treatment
A majority of the participants expressed an active use of the treatment; that the program was present in their thoughts between sessions: "I was thinking a lot about the questions and the exercises in the program. Even when I wasn't using it. I was reflecting about myself, my values, and how I feel about work." (Participant 8). Despite symptoms of fatigue and cognitive disturbances with concentration and memory, the participants made efforts to implement new knowledge to improve their life situation: "I really think that this was hard work, but I managed to do it anyway…" (Participant 4).
Effects
Effects describe the participants' experiences of effects in terms of mental health and well-being as well as insights, in both life and at work.
Mental health and well-being
Six participants explicitly described that their mental health and well-being had improved by participating: "It felt like a comeback, I'm much happier and more positive. It's fun to talk to people, hang out with friends and stuff." (Participant 2). Participants expressed that these effects were also perceived by co-workers: "Even my colleagues said, welcome back! It was as if I was back in the game" (Participant 6). Three participants stated that although their perceived stress had not diminished, they felt that they had learned how to manage stress with new strategies: "I'm still very stressed but now I listen more to my body, I listen more to my signals. Now I can recover better during the day and I sleep well at night. So, in a way my mental health is better, even though, I still have many symptoms… but not in the same way as before." (Participant 7). Five participants described that they were more aware of their stress behaviors: "Previously I got really stressed in queues as well, I was always annoyed about it, but now I think in a different way." (Participant 9).
Insights
Several participants described that the treatment gave rise to new insights, in both life and at work. For example, four participants stated that they now understood how much stress affected their life and the exercises on valued living and life balance brought them deeper understanding of the need for change. Two participants described that they felt sad when they realized the priorities they had made: "I remember that when I saw the result, it was like a punch in the face. It was probably the hardest part of all, just that." (Participant 1). A majority of respondents commented on the need for re-prioritizing their life situation and that behavioral changes should be seen as an ongoing process: "I think this is a work in progress… because you want to feel good, and just like someone who works with a physical demanding job has to work out in a gym in order to cope with their work, I who… have a mentally demanding job… maybe have to work out my brain." (Participant 3). Six participants described that the work-focused material made them reflect about their work: "It became clear to me what I was missing in my workplace; what I got too little of and what I had too much of." (Participant 9).
Discussion
The main purpose of the present study was to examine participants' experiences of having completed an iCBT and how the intervention could be improved to further enhance the efficacy and utilization in the target population. The results indicated that most of the participants experienced positive effects on their mental health and well-being in both life and at work. All participants emphasized the importance of having access to therapist support. In line with previous research, participants found the intervention to be extensive and demanding and expressed the need for extended treatment time and therapist support.
Combined CBT and work-focused interventions have previously been studied (Blonk et al., 2006;Lagerveld et al., 2012), but, as far as we know, never been evaluated in a qualitative study. The participants' experience of combined effects in their personal-and work life are promising, since the efficacy of iCBT and CBT usually extends to mental health and well-being (Andersson et al., 2019).
All participants found that the intervention was useful, relevant and user-friendly. Most useful were the sections regarding psychoeducation on stress, managing stress factors in both work and private life, and modules on sustainable working life. The task that the majority of the participants mentioned as least useful was the stress diary. The stress diary was described as too extensive and was often skipped. Perhaps the somewhat passive and repetitive nature of recording thoughts, feelings and behaviors linked to stressful situations, was less engaging than more active sections linked to behavioral change.
All participants emphasized the importance of having access to therapist support, which is in line with previous research suggesting better outcome when guidance is included (Baumeister et al., 2014). Some even stated that the contact was a prerequisite for completing the treatment. According to the participants, the main function of the therapist was to monitor progress and adherence and provide encouragement and support.
Aligned with previous research, participants found the intervention to be extensive and demanding, and session lengths and deadlines constituting a stress factor (Bendelin et al., 2011;Fleischmann et al., 2017;Halmetoja et al., 2014;Rozental et al., 2015a). Nevertheless, none of the participants wished to reduce the extent of the program. The present iCBT program contained both CBT-and work-focused material integrated and represented each module, covering between 5 and 16 regular pages of information. To our knowledge, 5 and 16 regular pages of information per module are comparable to other iCBT programs. However, given the challenge of developing iCBT interventions for stress and burnout that neither drain the participants' resources nor compromise the quality of the program, it could be feasible to test different iCBT formats for stressed participants. We have previous positive experiences of reducing content to 2/3 of the original format with sustained efficacy (Dagöö et al., 2014).
A majority expressed a need for extended treatment time, more frequent therapist support (when needed) and a follow-up or booster session. However, themes expressed in previous research regarding the need for more individualized feedback and support were not emphasized by the participants in the present study (Fleischmann et al., 2017;Gerhards et al., 2011;Ly et al., 2015;Rozental et al., 2015a). On the contrary, most of the participants perceived the treatment and the support as authentic and tailored to their needs, situation and preferences. In our previous studies, therapist has been Master's level students with limited past experience on stress-related disorders. In the present study, the therapists had long experience (20-30 years) as health care professionals and of treatment of stress and burnout. Perhaps their extensive experience enabled them to respond more authentically.
Limitations
The present study has several limitations in terms of selection, representativeness and interview procedure. First, participants were selected from the original study's treatment group using a criterion-based sampling approach, selecting those participants who had completed all treatment modules and the three-month follow-up assessment. The decision not to include dropouts potentially diminished the variety of perspectives, as they could have contributed with different experiences and suggestions for improvements of the intervention. Second, it is possible that the participants in the selected sample were more motivated (e.g. completed all treatment modules) and thereby reducing variation in experiences, as expressed by participants. Third, the participants in the present study cannot be seen as representative for the general working population given that all were women, eight had a university-level education, and eight suffered from severe stress disorder. Fourth, the interviews took place approximately three months after the end of treatment, which may have affected the results. Considering the effect of time, participants may have biased memories of the treatment and the treatment procedure. Fifth, telephone interviews were conducted. According to recommendations (Langemar, 2008), telephone interviews should only be used in cases where face-toface interviews are not feasible (e.g. geographical distance). This could have affected the allegiance of the participants to the interviewer and which experiences they chose to disclose. However, there are both advantages and disadvantages with telephone interviews. The disadvantages are that the non-verbal communication is missed and that the interview can be perceived as formal. On the other hand, interviews by phone can be perceived as more anonymous, which can appeal to some participants, facilitating disclosure of sensitive and important information (Carlbring et al., 2001). Due to the fact that the respondents were geographically dispersed, it was not feasible to conduct face-to-face interviews. Sixth, interviewer and coders where not blinded to the improvement of the participants, and therefore might have missed relevant themes. In addition, all therapists had long experience of treatment of stress and stress related disorders and this might have affected their perception and thereby missing important information that had been disclosed by a neutral interviewer. Finally, the therapists did the interviewing and it is likely that their positive view of the iCBT could influence the participants. Using therapists for both treatment and interviews was made due to limited resources. However, we assured that the therapists did not interview those participants they previously had treated.
Future recommendations
Considering the limitations of the study, future research could examine the feasibility of reducing session length, extending therapist support, fewer parallel home assignments, audio-and video files instead of written information and flexibility regarding deadlines, to further enhance the acceptance and utilization in the target population. Rozental et al. (2015a) stressed that some of the negative effects of internet-based interventions could be prevented by increasing flexibility of the treatment program. On the other hand, findings suggest negative effects of giving too much flexibility (e.g. extending treatment time with several weeks; Andersson, 2014). The challenge for future research targeting stressed individuals is to develop interventions that does not drain the participants' cognitive resources, nor compromises the quality of the program. | 2019-10-31T09:11:23.872Z | 2019-10-28T00:00:00.000 | {
"year": 2019,
"sha1": "bfda669ebb42db214f0ae696c17ddc2511f2e40c",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1016/j.invent.2019.100282",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "60748ec38a89d4ad895c142a3f712543b92e9b63",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
233404204 | pes2o/s2orc | v3-fos-license | Environmental determinants of minimum body temperature in mammals
Abstract. Physiological regulation of body temperature, set at a high level, is one of the key features of endothermic homeotherms, such as birds and mammals. However, many mammals and some birds have evolved the ability for temporal down-regulation of core body temperature. We investigated how variation in environment temperature and habitat primary productivity determine variation in daily body temperature down-regulation among mammalian species. Nearly half of the variation in minimum daily body temperature among species was explained by variation in both primary productivity and environmental temperature. Mammals expressing low minimum body temperature inhabited regions of low annual temperature with wide daily and seasonal temperature variation. Simultaneously, those regions were characterized by low productivity and low seasonality in productivity. Furthermore, regions characterized by a high level of among-year variation in environmental temperature, but not in primary productivity, were inhabited by species with low minimum body temperature, but only by those adapted to relatively humid conditions. Our results suggest that daily heterothermy can be selectively advantageous in the environmental circumstances when high energetic demands for maintaining endothermic homeothermy, physiological regulation of a high and stable body temperature, cannot be supported. The results corroborate the hypothesis that mammals that have evolved daily down-regulation of body temperature may have higher chances of surviving extinction events caused by climatic changes. Therefore, daily heterothermy adaptation in contemporary mammals represents a mechanism for surviving the ongoing global warming.
the circumstances when high energetic demands for maintaining endothermic homeothermy, a physiologically regulated high and stable body temperature, cannot be fulfilled (Hetem et al. 2016, Dammhahn et al. 2017, Lovegrove 2017. The high and stable core body temperature of endothermic homeotherms provides an optimal environment for enzymatic reactions, enables development of a complex nervous system, and allows a high reproductive rate, among other benefits (Nespolo et al. 2011, Preußner et al. 2017. However, homeothermic endothermy is energetically expensive, an endothermic mammal consumes five times more food than an ectothermic tetrapod of similar size (Pough 1980, McNab 2009). Temporal down-regulation of metabolic processes can reduce such costs, but it results in intraindividual body temperature variation observed, for instance, in hibernating and daily torpid mammals (Hetem et al. 2016, Dammhahn et al. 2017, Lovegrove 2017. The evolution of metabolic suppression, the physiological trait associated with heterothermy, is directly linked to energy conservation (Lovegrove 2000a). The adaptive benefits of temporal heterothermy are reduction of energy and water usage when those resources are unavailable or limited (Tieleman & Williams 1999, Vuarin & Henry 2014 and may have led to repeated evolution of heterothermy in mammals (Geiser 2008). It has been suggested that temporal heterothermy allowed mammalian lineages to survive the Cretaceous-Paleogene extinction event (Lovegrove 2017), with extinction risk potentially reduced in contemporary heterothermic mammals (Geiser & Turbill 2009). Hence, it was concluded that a heterothermic strategy can be beneficial if endothermic homeothermic animals are exposed to changing, fluctuating or unpredictable environmental conditions (Geiser & Brigham 2012, Dammhahn et al. 2017). Nevertheless, how environmental variation affects heterothermy is still debatable (Vuarin & Henry 2014, Levesque et al. 2016, Dammhahn et al. 2017).
Intra-individual temperature variation, or heterothermy, is a physiological strategy in which animals can adjust their body temperature depending on resource availability and ambient temperature (Bligh & Johnson 1973, Boyles et al. 2011, Geiser & Brigham 2012, Vuarin & Henry 2014. How multiple environmental variables simultaneously affect the level of heterothermy is still unclear (Vuarin & Henry 2014, Dammhahn et al. 2017. It was observed in hibernating mammals that metabolic rate reduces in response to decreased environmental temperature (Geiser 2004, Heldmaier et al. 2004. As a consequence, their body temperature also declines, as well as their overall energy expenditure (Geiser 1988). Surprisingly, little empirical evidence exists confirming that environmental temperature can predict body temperature down-regulation in daily heterotherms (Nespolo et al. 2010). Some evidence indicates that a decrease in ambient temperature can induce torpor intensity in some species of birds and mammals (Nowack et al. 2017(Nowack et al. , 2020. For example, in the lesser hedgehog tenrec (Echinops telfairi) daily torpor bounds were longer and minimum skin temperature was lower under a fluctuating versus constant temperature experimental condition (Dausmann et al. 2020). Daily heterothermy level, estimated from body temperature variation of free-ranging wood mice (Apodemus flavicolis) during a central European winter, was negatively related to minimum daily environmental temperatures ). However, experimental results suggest that environmental temperature can interact with habitat productivity in controlling expression of heterothermic physiology in mammals, complicating the temperature dependence pattern (Nespolo et al. 2010). Laboratory studies confirmed that food shortage alone can induce heterothermy. The euthermic Chilean mouseopossum (Thylamys elegans) showed daily fluctuations in body temperature, decreasing body temperature and entering torpor, only when food was absent (Bozinovic et al. 2005). Likewise, when pale kangaroo mice (Microdipodops pallidus) were provided with food in excess, animals did not enter torpor, but otherwise the torpor intensity increased with decreasing experimental temperature (Brown & Bartholomew 1969). The probability of entering torpor in sugar gliders (Petaurus breviceps) reflected differences in weather patterns, food availability and thermoregulatory costs in relation to ambient temperature (Körtner & Geiser 2000). These results from intraspecific experiments suggest that environmental temperature fluctuation, in combination with habitat productivity, can influence conditions in which evolution of the daily heterothermic strategy is promoted in otherwise homeothermic animals. However, the hypothesis that variation of daily heterothermy observed among mammalian species is predicted by both temperature and productivity has yet to be evaluated.
The main aim of this work was to estimate the relative contribution of thermal environment and energetic limitation caused by variation in habitat productivity on the expression of daily minimum body temperature among mammalian species. It was hypothesized that the level of daily heterothermy is determined by variation of ambient temperature (Brown & Bartholomew 1969, Körtner & Geiser 2000. Specifically, lower average and higher amplitude of daily and seasonal temperature fluctuations can stimulate temporal expression of a low body temperature in mammals, resulting in an increased heterothermy level. It was also hypothesized that the level of daily heterothermy is determined by variation in habitat productivity. Lower average and higher amplitude of seasonal variations in habitat productivity can stimulate expression of temporal body temperature down-regulation, when resources in the environment are too scarce to fulfil high energetic demands for homeothermic endothermy. High among-year fluctuations in ambient temperature and food availability could select for adaptive strategies to compensate for unpredictability in the environment, resulting in the evolution of daily heterothermic strategies. We tested the above hypotheses using comparative data for mammalian species known to express daily heterothermic physiology (Ruf & Geiser 2015, Nowack et al. 2020. In particular, we tested if and how variation in habitat productivity and environmental temperature predict variation in minimum body temperature. We predicted that species with daily heterothermy adaptation will inhabit regions with both low environmental temperature and productivity and high variation in environmental predictors. Due to food and water limitations, e.g. in deserts and semideserts, thermal and energetic processes can have differential consequences between arid and humid adapted species. As a consequence, and because of allometric scaling of multiple physiological traits, we evaluated how species body mass and aridity adaptation influence the covariation between minimum body temperature and environmental predictors.
Comparative data
Minimum core body temperature (BT min °C) and body mass (g) data for mammals expressing daily heterothermy was obtained from recent reviews (Ruf & Geiser 2015, Nowack et al. 2020). Environmental variables were derived for geographic distributions of terrestrial mammals obtained from Digital Distribution Maps (IUCN 2020). Environmental temperature variables (BIO1, BIO2 and BIO4) were derived from a historical climate data repository (Fick & Hijmans 2017) at a resolution of 2.5 minutes (downloaded on 30 October 2020). The range polygons for each species were used to clip the rasters of the variables and to estimate the median and standard deviations for each species. Environmental temperature variables in subsequent analyses included the annual median temperature (T M ) and its standard deviation (T SD ; among-year variation in temperature), median diurnal temperature range (T R ) and temperature seasonality (T S ). Normalized difference vegetation index (NDVI; the difference between near-infrared and red reflectance divided by their sum, with densely vegetated areas represented by higher values) was derived for the years 1982-2015 (National Center for Atmospheric Research Staff 2018, https://staff. ucar.edu/browse/orgs/NCAR, downloaded on 13 November 2020). Bimonthly measurements of NDVI were used to estimate the averages and standard deviations of primary productivity for each year. The mean and standard deviation of the average primary productivity across the years, and mean of within years standard deviations of primary productivity were estimated to summarize NDVI data (Fig. S1). Indices of annual primary productivity (P M ; median of annual productivity), among-year variation in productivity (P SD ; median of standard deviation of annual productivity) and seasonality of productivity (P S ; median of within years standard deviations) were calculated from summarized NDVI data and species ranges. Species were classified as inhabiting relatively arid or humid environments based on aridity index data (species with average aridity index < 0.5 were classified as arid and humid if average aridity index ≥ 0.5); Trabucco & Zomer 2019). Records deviating markedly from the distributions of the data on trait paired regression plots, five for BT min (Elephantulus rozeti, Microcebus berthae, Microcebus myoxinus, Tarsipes rostratus and Vespadelus vulturnus) and two for T M (Apodemus peninsulae and Phodopus sungorus), were excluded from analyses as outliers, resulting in complete records for 81 mammalian species (Supplementary online material: data). A trimmed mammalian phylogeny (topology with branch lengths: http://vertlife.org/ phylosubsets/; Upham et al. 2019) was used in comparative analyses.
Statistical analyses
Species and environmental variables were log 10 transformed, centred and scaled to improve normality of their residual variances and to place variation on the same scale. Because environmental variables were correlated (Table S1) principal component analysis was conducted to reduce multicollinearity between seven environmental predictors (T M , T SD , T R , T S , P M , P SD and P S ). Principal components (proportion of explained variable by component > 0.05, eigenvalue > 1) were extracted with varimax rotation (with "principal" command in the psych R package). Calculated principal components were used in the subsequent analysis. Differences between arid-and humid-adapted species ( Fig. 1) for studied variables were tested with ordinary (OLS) and phylogenetic (PGLS) generalized least squares regressions. Associations between BT min and principal components were tested with PGLS (phylolm R package; Rezende & Diniz-Filho 2012). The significance of the principal components on BT min was tested, with body mass as covariate and a binomial factor of aridity adaptation, distinguishing arid-from humid-adapted species. To test if size of species and their adaptation to aridity influenced response to environmental predictors, two-way interactions between environmental predictors with aridity factor and body mass were tested. The goodnessof-fit of the models were evaluated with the Akaike information criterion (AIC) and alternative phylogenetic regression R-squared values for PGLS analyses (Ives 2019). Finally, a backward step-wise selection method was applied on a full model, including all the above listed terms, and the model was simplified by removing the least significant interaction terms.
Minimum body
Annual temperature Diurnal temperature range Annual productivity Among-years temperature Among-years Temperature Productivity productivity Body mass temperature variation seasonality seasonality variation Humid species Arid species
Five rotated principal components explained 97% of the variation among the studied species in seven environmental predictors derived for the species distribution ranges ( Fig. 3; Table S3). The first component (RC1) was primarily loaded by diurnal temperature range, annual productivity and temperature seasonality. The second component (RC2) was loaded by annual temperature and temperature seasonality. Third (RC3) and fourth (RC4) components were loaded by among-year variation, either in productivity or in temperature, respectively. The fifth component (RC5) was loaded by annual productivity and productivity seasonality (Fig. 3).
BT min was
Annual temperature Among-years temperature variation Diurnal temperature range Temperature seasonality Annual productivity Productivity seasonality Among-years productivity variation (Table S3) are presented. (0) humid-(aridity index > 0.5) or (1) arid-climate species (aridity index < 0.5), and RC stands for rotated principal components (Fig. 3). Interpretation of standard regression R-squared are uncertain in PGLS analyses, thus alternative values are provided: R ls 2 = 0.52, R ce 2 = 0.55, R lr 2 = 0.42 (Ives 2019). Left panel presents results from an analysis including the non-significant terms of body mass and RC3, which are excluded (-) from the analysis presented in the right panel. negatively related to diurnal temperature range and seasonality, and positively related to annual productivity (Fig. 4). The positive effects of RC2 (β PGLS (SE) = 0.27 (0.11), t = 2.37, P = 0.020) and RC5 (β PGLS (SE) = 0.38 (0.12), t = 3.30, P = 0.0015) translate into positive associations with annual temperature, annual productivity and productivity seasonality, and to a negative association between temperature seasonality and BT min (Fig. 4).
Discussion
Physiological regulation of core body temperature is a key characteristic of endothermic animals (Ruben 1995, Lovegrove 2017. As a vital aspect of their biology, precise temperature regulation is sometimes postulated to have facilitated the ecological and evolutionary success of birds and mammals (Koteja 2000, Lovegrove 2017). However, there is growing recognition of deviations from tight homeothermic regulation in many lineages of mammals, resulting in the evolution of the Fig. 4. Bivariate plots for the relationships between traits from multivariate phylogenetically generalized least squares (PGLS) results (lines ± confidence intervals) plotted against the raw data (dots). Residuals of species minimum body temperature are plotted against body mass and principal components calculated from environmental predictors (Fig. 1). Residuals of minimum body temperature were calculated from PGLS regression of minimum body temperature on predictors in the multivariate PGLS analysis (Table 1), except the predictor plotted on the given panel. Presented are Pearson's product-moment correlations (r; **P < 0.01, *P < 0.05, + P < 0.08). X-axis logged and includes values of standardized loading of environmental variables to the principal components for variables with loading > 50% (Fig. 2). capacity to temporarily abandon homeothermy, such as daily heterothermy (Geiser 1988, Boyles et al. 2011, Geiser & Brigham 2012, Dammhahn et al. 2017. The phylogenetic analysis presented here showed that minimum body temperature can be determined by environmental predictors (Table 1). Variation in primary productivity and thermal environment together explained around half of the observed variation among studied species in minimum body temperature. In particular, low primary productivity and narrow seasonality in productivity predicted the low minimum body temperature observed in some species. At the same time, low environmental temperature and wide daily and seasonal variation in temperature predicted low minimum body temperature. Interestingly, among-year variation in primary productivity had a negligible influence on variation among species minimum body temperature. In contrast, among-year variation in environmental temperature had the opposite effect on minimum body temperature between mammals inhabiting relatively arid and relatively humid areas.
It was predicted that mammals exposed to warmer temperatures would express lower levels of heterothermy than those inhabiting colder areas (Nowack et al. 2017, Dausmann et al. 2020. Indeed, our analyses showed that higher annual environmental temperature (the main contributor to the second principal component: Fig. 3) resulted in higher minimum body temperature in the studied mammals (Fig. 4, Table 1). However, in addition to annual temperature we found that the parameters determining temperature variability substantially contributed to explained variation in minimum body temperature. A wide diurnal temperature range (main contributor to the first principal component: Fig. 3) was observed in areas inhabited by mammals with relatively low minimum body temperature (Fig. 4). Also, wide seasonal temperature variation (the main contributor to both the first and second principal components) was observed in areas inhabited by mammals with relatively low minimum body temperature (Fig. 4). Therefore, it can be postulated that lower annual environmental temperatures and wider amplitude of daily and seasonal temperature fluctuations are associated with the evolution of mechanisms for temporal down-regulation of body temperature in mammals. Interestingly, the effects of annual temperature level and daily temperature range contributed independently to variation in minimum body temperature. These environmental predictors were not correlated (Table S1) and contributed to two independent environmental principal components, both important predictors of mammalian daily heterothermy (Fig. 3). This statistical independence suggests that variation in the two predictors can lead to the origin of heterothermy independently, and when occurring simultaneously can reinforce evolutionary and physiological responses in exposed organisms. Consequently, areas of wide daily temperature variation and low annual environmental temperatures, such as relatively cold arid and semi-arid regions, and high-altitude regions around equator, are particularly anticipated to host mammals expressing daily heterothermy physiology (Geiser & Turbill 2009, McKechnie & Mzilikazi 2011, Okrouhlík et al. 2021).
Among-year variation in environmental temperature has a contrasting effect for mammals inhabiting relatively arid and relatively humid regions (Fig. 4, Table 1). This effect was strong in mammals from relatively humid areas (Tables S5, S6), predicting a lower level of minimum body temperature in species inhabiting regions with wide among-year variation in environmental temperature (Fig. 4). It has been previously hypothesized that an increased level of heterothermy, in otherwise homeothermic animals, could be a mechanism facilitating species to survive extreme climatic changes (Boyles et al. 2011, Geiser & Brigham 2012. The negative and strong covariation between minimum body temperature and among-year variation in environmental temperature is the first quantification of the relationship that supports the hypothesis. Our comparative results suggest that endothermic animals inhabiting areas with greater levels of temperature unpredictability have evolved temporal suspension of homeothermy, which could facilitate their survival during thermally harsh and unpredictable conditions. Such a mechanism seems not to operate in species inhabiting arid regions, where other environmental factors, such as the low level of primary productivity (Tables S5, S6), in addition to environmental temperature variation, are limiting to physiological and population responses (Lovegrove 2000b, Barros et al. 2018).
Daily heterothermy levels may be influenced by resource availability determined by variation in habitat productivity (Brown & Bartholomew 1969, Bozinovic et al. 2005, Nespolo et al. 2010). Low primary productivity may stimulate expression of temporal body temperature down-regulation, when resources in the environment are too scarce to fulfil the high energetic demands of homeothermic endothermy. Indeed, we established that low annual environmental productivity (a contributor to the second and fifth principal components: Fig. 3) predicted a low level of minimum body temperature in some mammals (Fig. 4, Table 1). Seasonality in primary productivity was positively correlated with its annual level (and also contributed to the fifth principal component: Fig. 3), suggesting that seasonal variation in productivity is higher in more productive areas. Together, the annual level and seasonality in productivity had the strongest positive association with minimum body temperature among mammals (Table 1). Annual productivity was negatively correlated with diurnal temperature range and temperature seasonality (and contributed to the same, first, principal component: Fig. 3). These environmental predictors, low annual productivity and wider temperature variation, set climatic conditions that promoted the occurrence of heterothermy in mammals (Fig. 4). In contrast, among-year variation in primary productivity was the poorest predictor of variation in minimum body temperature observed among the studied mammals. Thus, it can be concluded that a certain minimum level of primary productivity is needed to support energetically costly homeothermic endothermy and, whenever such condition is not fulfilled and species experience wide temperature fluctuations, heterothermic strategy is expected to evolve. Despite more than three orders of magnitude in body mass variation in our data the environmental effects on daily heterothermy were not associated with the body size of the studied mammals. We recognise that our results are derived only from 81 mammalian species, thus we do not provide overwhelming evidence for the absence of the effects of body mass on environmental constraints to daily heterothermy. However, if this result is representative it suggests that environmental predictors impose relatively similar selective forces on daily temperature regulation between large and small mammalian species.
Conclusions
Animals collect and transform energy from the environment and allocate it to physiological processes and ultimately to fitness (Boratyński & Koteja 2009, Boratyński et al. 2010, Dammhahn et al. 2017, Boratyński & Szafrańska 2018, Boratyński 2020, Desforges et al. 2020, Pettersen et al. 2020. We showed that the environmental predictors of primary productivity and environmental temperature, can determine the minimum body temperature observed in mammals. Minimum body temperature results from down regulation of physiological processes that can indicate the intensity of expressed heterothermy (Brown & Bartholomew 1969, Körtner & Geiser 2000. Thus, our results provide evidence that a physiological strategy to save energy, daily heterothermy, is expected to evolve in mammals (Fig. 2, Table 1). Among-year variation in environmental temperature, irrespectively of among-year variation in habitat productivity, strongly predicted low minimum body temperature observed among mammalian species. This result supports the hypothesis that mammals that evolved daily body temperature down-regulation may experience a greater chance of surviving extinction arising from environmental temperature changes (Lovegrove 2000b, Boyles et al. 2011. Our results suggest that such a mechanism does not apply in arid-adapted fauna, reinforcing our understanding of their sensitivity to habitat and climatic alterations (Brito et al. 2016). At the same time, an overall decreased level of primary productivity was a strong predictor of low minimum body temperature, confirming that heterothermy emerges when energetic requirements to support endothermic homeothermic regulation cannot be fulfilled (Rezende et al. 2020). To conclude, our results clearly show that low minimum body temperature is promoted by certain environmental conditions. In particular, organisms expressing low minimum body temperatures are predicted to inhabit regions of low but variable environmental temperature, and a low level and narrow seasonality in environmental productivity. Further research will verify whether results for minimum body temperature are confirmed by direct estimates of daily body temperature fluctuations, related to behavioural performance, locomotory activity, metabolic and food consumption rates (Boratyński et al. 2020, Nowack et al. 2020 Step 1: estimation of per pixel summary statistics (mean or SD) from 24 raster layers per year (two per month).
Supplementary online material
Step 2: estimation of summary statistics as in Step 1 for the years 1982-2015 resulting in 34 raster layers, one for each year.
Step 3: estimation of per pixel summary statistics (mean or SD) from the 34 raster layers produced in Step 2, resulting in one raster layer.
Step 4: extracting values from the single raster layer produced in Step 3 that overlap species distributions (shapefile polygons from IUCN terrestrial mammal dataset) to estimate median values within the range of the species. Summary statistics for NDVI estimated at each step for variables used in this study, medians of: averages of yearly means (PM: Step 1 mean, Step 3 mean, Step 4 median), averages of yearly SDs (PS: Step 1 SD, Step 3 mean, Step 4 median) and average SDs of yearly means (PSD: Step 1 mean, Step 3 SD, Step 4 median). Table S1. Correlation matrix of variables for 81 studied mammalian species. BT min and BM refer to minimum core body temperature and body mass, respectively. Bioclimatic variables include annual median (TM) and standard deviation (TSD) of environmental temperature (BIO1), median diurnal temperature range (TR; BIO2) and temperature seasonality (TS; BIO4). Indices of yearly levels of primary productivity (PM; median of yearly averages), among year variation in productivity (PSD; median of standard deviations of yearly averages) and seasonality of productivity (PS; median of within years standard deviations) were calculated from NDVI data and species ranges. *P ≤ 0.05, **P ≤ 0.001. Table S2. Differences in studied variables between arid (factor level = 1) and non-arid species (level = 0). Results for ordinary (OLS) and phylogenetic (PGLR) generalized least squares. Table S3. Results from rotated principal component analysis of environmental predictors calculated from species ranges of 81 mammals. Percentages of explained variation in environmental variables by the rotated principal components (RCs) and standardized loadings of environmental variables to the given components are presented. The main environmental predictor loadings to the components are indicated in bold. Table S4. Model selection procedure applied in the study. The full model included the factors aridity, body mass, five principal components and two-way interactions. The predictors "reduced predictor" refers to a predictor included in the above model (see "statistics of reduced predictor") but excluded in the given analysis. Table S5. Results from two rotated principal component analysis of environmental predictors calculated from species ranges of 46 arid and 35 non-arid mammals. Percentages of explained variation in environmental variables by the rotated principal components (RCs) and standardized loading of environmental variable to the given components are presented. Table S6. Parameter estimates for two phylogenetically generalized least squares (PGLS) analyses of environmental variables predicting minimum body temperature in arid (R ls 2 = 0.50, R ce 2 = 0.47, R lr 2 = 0.41) and non-arid mammals (R ls 2 = 0.44, R ce 2 = 0.44, R lr 2 = 0.31). | 2021-04-27T00:10:39.717Z | 2021-04-01T00:00:00.000 | {
"year": 2021,
"sha1": "b78965052f2086f024e616ba8ffe7cd14308992e",
"oa_license": "CCBY",
"oa_url": "https://bioone.org/journals/journal-of-vertebrate-biology/volume-70/issue-2/jvb.21004/Environmental-determinants-of-minimum-body-temperature-in-mammals/10.25225/jvb.21004.pdf",
"oa_status": "GOLD",
"pdf_src": "BioOne",
"pdf_hash": "b877b869c5fd442cab21ba69d4518570b35ab75c",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
251652840 | pes2o/s2orc | v3-fos-license | The Effect of Health Consciousness, Food Safety Concern, and Familiarity with Advertisements on Willingness to Pay to Purchase Healthy Food Products
. This study aims to determine the influence of health consciousness, food safety concern, and familiarity in advertising on the willingness to pay for the purchase of healthy food products, namely Lemonilo in West Nusa Tenggara. The type of research used is quantitative research with a causality association approach. This study's respondents are consumers who have never bought Lemonilo products in West Nusa Tenggara. The number of samples taken was 120 consumers aged 20-50 years. The sampling technique used was the purposive sampling technique. The data analysis used in this study was multiple linear regression analysis with SPSS 2.6 for windows applications. The results showed that health consciousness and food safety concern does not affect willingness to pay. Meanwhile, familiarity with the advertisement significantly affects willingness to pay.
INTRODUCTION
In the current era of globalisation, global competition is getting more demanding in all aspects. Factors that can support the quality of human resources are essential to consider competing. One of the factors that can help the performance of human resources in the era of globalisation is health. Today, more and more people are starting to realise and change their habits by paying attention to their lifestyles, such as healthy food consumption patterns to overcome health problems. At this time in Indonesia, the demand for organic products is quite significant. However, organic food marketing in Indonesia is still constrained by the perception of the price of organic food, which is considered expensive. For this reason, companies need to determine a suitable marketing strategy so that consumers are willing to pay for healthy food products. According to [7], willingness to pay is the highest price a consumer is willing to pay to get a benefit in the form of goods or services.
In general, some people consider organic food an expensive product, but for those who adopt a healthy lifestyle, this is not uncommon, and they are still willing to pay more for the food they want to consume. Nearly 64% of consumers in the world say that they are willing to pay more for food or drinks that contain the desired ingredients according to their health concerns [22].
The World Instant Noodles Association shows that Indonesia ranks second with the highest global demand for noodles. Seeing the high demand for instant noodles in Indonesia, producers that create instant noodle products that use organic ingredients have emerged. They make products that are safer to consume regularly, one of which is instant noodles launched by Lemonilo, which are made from organic ingredients and various selected spices. Since its launch in September 2017, Lemonilo has received quite a positive response. The advantages of Lemonilo and how Lemonilo consumers feel about the products have undoubtedly created a willingness to pay for their product.
Companies increase consumers' willingness to pay for the products offered by paying attention to the health consciousness factor, which is a concern to become better and motivated to im-prove, maintain, and maintain health and quality of life by implementing a healthy lifestyle. On the other hand, the increasingly fierce competition in the food industry can open up the possibility for companies to carry out unethical behaviour, for example, by using chemicals in producing these foods. Therefore, attention to food safety (food safety concern) by consuming organic food can be used as an alternative to reduce the risk of these health problems.
This research stems from the inconsistency of the results of previous studies. For example, research conducted by [1,18] shows that the health consciousness variable significantly affects willingness to pay in consumers' willingness to pay for organic products. But a different opinion was expressed by [2], where the results of his research show that women aged 17 years and over who routinely use cosmetics and are considering buying green cosmetics show that the health consciousness variable has no significant effect on willingness to pay.
In addition, research on the effect of a food safety concern on willingness to pay also shows inconsistent results. Research conducted by [3] shows that food safety concerns have a significant impact on willingness to pay for purchasing organic rice products. On the other hand, a different result was found by [4], who stated that food quality and food safety have the same impact on willingness to pay, indicating an inevitable confusion about their significance. In terms of familiarity, some studies showed varied results. For example, [5] research shows that most consumers are willing to pay more because they know the benefits or are familiar with these organic products. On the other hand, [6] stated in his research that familiarity with advertisements for purchasing food supplement products does not significantly affect willingness to pay.
With the inconsistency of previous studies' results, this research topic becomes interesting to study. In addition, there are still not many studies that specifically discuss willingness to pay for organic instant noodle products, so further research is sorely needed. In particular, this study will analyse the effect of health consciousness, food safety concern, and familiarity in advertisements on willingness to pay.
Literature review
Willingness to Pay. Willingness to pay is an economic value defined as a measurement of the maximum number of people's willingness to sacrifice goods and services to obtain other goods and services. The concept of a person's willingness to pay for goods or services produced by natural resources and the environment is formally referred to as willingness to pay [6]. Additionally, the author [7] explains the willingness to deliver the highest price given by the company to prospective consumers to get products that can provide a benefit and become a reference for how many potential consumers value the product.
The Effect of Health Consciousness on Willingness to Pay. Authors [8] stated that health awareness is oriented towards a healthy lifestyle. The author [9] re-conceptualises the dimensions that previous researchers have made and finds three dimensions in health consciousness, namely: (1) Self-health awareness, the tendency to focus attention on healthy behaviours such as activities, interests, and health-oriented opinions; (2) Personal responsibility, people who are aware of health feel more responsible for maintaining health and tend not only to engage in preventive behaviour and maintain health in daily life but also actively participate in the health community; (3) Health motivation indicates an individual's intrinsic motivation to maintain good health and its reflection or responsibility for health.
According to [36], the most frequent motive for buying organic food is consumers' perception that organic food is healthier for consumption. Consumers of organic foods may be more aware that the amount of food they consume can affect their health. They value health and pay more attention to choosing natural and healthy foods to maintain and improve their health. After the consumers know that organic food is healthy, the tendency to buy organic food is even greater. Therefore, this study proposes the first hypothesis as follows: H 1 : Health consciousness affects willingness to pay for the purchase of Lemonilo healthy food products.
The Effect of Food Safety Concerns on Willingness to Pay. According to [37], food safety is defined as the freedom of food from substances or materials that can harm the body without distinguishing whether the meaning is naturally present in the ingredients used or mixed intentionally or unin-tentionally into food ingredients finished food. On the other hand, [38] defines food safety as a condition and effort to maintain food quality to prevent contamination or foodborne disease.
Authors [39] define food safety as a consumer concern with food produced from chemical sprays, fertilisers, and preservatives. Consumers are generally willing to pay or buy a product that increases food safety concerns, so it is said that food safety concerns can influence consumers before making a purchase decision. Authors [40] support this statement by stating in their research that food safety concerns have become a motivating factor for consumers to seek safer food and guaranteed quality or attributes. Therefore, the second hypothesis proposed in this study is H 2 : Food safety concerns affect willingness to pay for purchases of Lemonilo healthy food products.
The Effect of Familiarity in Advertisement on Willingness to Pay. Authors [21] define familiarity as how much the brand relates directly and indirectly to the experiences that consumers have. Consumers can obtain an indirect brand experience by reading advertisements, the brand's exposure, knowing from other people, and directly buying products. Consumers tend to buy if they are already familiar with a brand. Authors [40] reveal the function of familiarity in identifying the quality of an item or service. Besides, the brand can increase the level of buyer efficiency. The brand makes it easier for consumers to search for goods or services.
Research by [22] states that forming brand awareness can be done instantly or in a relatively short time by using intense, unique and sensational advertising and attractive prizes in stores. A known brand can increase positive familiarity and the possibility of being included in the list of brands to buy. If consumers perceive them as a good quality brand, it will be easier for them to pay more for the brand. So it can be concluded that familiarity is the closeness between the consumer and the brand and product to be purchased. Based on this explanation, the third hypothesis proposed in this study is H 3 : Familiarity in Advertisement affects willingness to pay on purchases of Lemonilo healthy food products.
Conceptual Framework. The research concept framework is a concept in the form of a systematically arranged chart, used to provide a brief view of the object of research so that from this research model, the direction and purpose of the research to be carried out will be known. The conceptual framework of this research is as follows.
METHODS
Data collection is done by utilising the online Google form questionnaire application. Respondents come from various demographics (in terms of gender, origin, age, occupation and monthly income). The sampling technique used is purposive sampling, and then questionnaires are distributed to 120 respondents who are Lemonillo consumers aged between 20-50 years. Measurement of variable indicators adopted from previous research, namely health awareness from [12], food safety concern from [13], familiarity with an advertisement from [14] and willingness to pay from [15]. All items were evaluated using a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).
Validity and reliability tests were performed using IBM SPSS version 26 before examining the entire model. A variable is said to be valid and reliable if it gives Cronbach's alpha value ≥ 0.60 [10,11]. Data analysis in this study used multiple linear regression. Regression analysis is a statistical technique that analyses the linear relationship between two variables by estimating the coefficients for a straight line equation, one variable is expressed as the dependent variable, and the other variable is expressed as the independent variable [13].
This study collected responses from 120 respondents. Because it uses an online questionnaire form, all questions can be set to be mandatory so that respondents would answer all the questions asked. The characteristics of respondents are very varied. One hundred thirty-eight males and 82 females, most of whom (86.7%) were aged between 20 and 30 years. The most dominant occupation of respondents is a private employee (41.7%), while 63 of the respondents
RESULTS AND DISCUSSION
The Cronbach Alpha value was used; the results are shown in Table 2 below. According to [10,11], reliability testing is a tool used to measure a questionnaire which is an indicator of a variable or constructs. The variable is said to be valid and reliable if it gives Cronbach's alpha value ≥ 0.60. Based on the following table, the questionnaire in this study is said to be reliable. In addition, it can also be seen from the table that, on average, food safety concern is the priority for consumers. In terms of the scales' validity, it was evaluated by item-total correlation analyses. The result is displayed in Table 2, where all item-total correlation coefficients surpassed the critical value of r table 0.361 and were statistically significant at p ≤ 0.01. The scales were therefore reliable and valid. Moreover, the examination of the variance inflation factor (VIF) indicated that multicollinearity was not a problem in the data as they were within tolerable limit; i.e. tolerance values were between 1.076 to 2.319, which were <10 [13]. Therefore, the authors are confident in the regression results. For analysis and hypothesis testing, the authors used regression analysis using SPSS 26. As a result, a regression model was developed and displayed in table 4. Table 4 shows that Health consciousness and food safety concerns do not affect the willingness to pay. On the other hand, the data indicate that familiarity with advertisements significantly affects willingness to pay. Based on the information above, it can be seen that the value of the coefficient of determination or adjusted R 2 is 0.508. This means that the willingness to pay variable is simultaneously influenced by the Health Consciousness, Food Safety concerns, and Familiarity with Advertisement variables by 50.8%. The remaining 49.2% is influenced by other variables not discussed in this study. In addition, it is obtained that the f count is greater than the f table (f count 42.029 > f table 3.13 with a p significance value of 0.000, which is smaller than 0.05 so that it is said that the research model with variables X 1 (Health Consciousness), X 2 (Food Safety Concern) and X 3 (Food Safety Concern) can or is feasible to use to predict willingness to pay (Y).
The Effect of Health Consciousness on Willingness to
Pay. The research data shows that the t-count value of the health consciousness variable is 1.322 (smaller than t-table 1.980 with a significance of 0.189 (greater than 0.05). This result means that the health consciousness variable does not affect willingness to pay. This can also mean that in this case of Lemonilo healthy food products, although the health consciousness variable increase, it will have no significant effect on consumers' willingness to pay. The results of this study support the results of previous research conducted by [2], which showed that health consciousness did not affect purchasing green cos-metics. The dominant indicator of health consciousness is health alertness, namely the level of individual awareness of their physical health condition. However, this is not a factor that can affect consumers' willingness to pay more for healthy food products.
The Effect of Food Safety Concerns on Willingness to Pay. Based on the results of research conducted, food safety concerns do not affect willingness to pay. This is indicated by the t value of the food safety concern variable, which is 0.597 (smaller than t table 1,980 with a significance of 0.552 (greater than 0.05). This means that although the food safety concern variable increases, it does not affect the willingness of consumers to pay for healthy food products. The results of this study support the results of previous research conducted by [18], which shows that food safety concerns do not affect consumers' willingness to pay more. The most dominant indicator of food safety concern variables in this study is the knowledge indicator, where consumers assume it is essential to know how healthy the food is. But this is not a factor that can affect consumers' willingness to pay for healthy food products. This is explained by [19,20,21]. Consumers' willingness to pay is influenced by different factors such as age, number of family members, personal consumer concerns, family spending, and consumer trust in a product.
The Effect of Familiarity in Advertisement on Willingness to Pay. Based on the research results, it is known that the t value for the familiarity in the advertisement variable is 9.867 (larger than t table 1.980 with a significance of 0.000 (less than 0.05). Thus, it is concluded that familiarity with advertisements significantly affects willingness to pay. It is one of the many factors that can consider by consumers in buying or consuming a product. This means that familiarity in advertising is the closeness between consumers and the brand and the effect they will buy. If Lemonilo advertises heavily, it will increase the willingness of consumers to pay more for these healthy food products. Products that the consumers know can increase positive familiarity and the possibility of being included in the list of products to be purchased by consumers.
This study's results align with the respondents' answers to the five statements of familiarity in advertising. This indicates that in the willingness to pay more for healthy food products, consum-ers need to know or be familiar with the brand. In addition to consumers knowing that the ingredients used in nutritional food products are high quality, consumers will find it easier to pay for them. The results of this study support the results of previous research conducted by [22], showing that consumers' willingness to pay more for organic vegetables indicates that most consumers are willing to pay more because they know the benefits or are familiar with these organic products.
CONCLUSIONS
This study provides empirical evidence on whether there is an influence between health consciousness, food safety concern and familiarity in advertisements on willingness to pay. The results show that health consciousness does not affect willingness to pay. This is in line with the findings of [2] that health consciousness has no significant effect on purchasing green cosmetics. Food safety concern has no significant impact on willingness to pay either. This is in line with [23] research. On the other hand, familiarity with the advertisement significantly affects willingness to pay. This strengthens the research conducted by [24] that most consumers are willing to pay more because they know or are familiar with organic carrot vegetable products. This research has managerial implications for the Lemonilo company, namely that the company must pay attention to the reputation built to survive. Furthermore, companies can be directly involved in educating consumers to be more aware of their health and provide information on the importance of health awareness. In addition, Lemonilo, as a healthy lifestyle brand, can provide information and educate consumers about the ingredients in Lemonilo products so that consumers have more confidence in buying and consuming these Lemonilo products.
The limitations of this study are that the research object is only focused on healthy food products from the Lemonilo brand, which is only one of many healthy foods available and that the scope of the research is only in the West Nusa Tenggara area. Further research can be conducted with more comprehensive area coverage. Future researchers can also add other variables to the research model to enrich the findings. | 2022-08-19T15:12:56.350Z | 2022-07-31T00:00:00.000 | {
"year": 2022,
"sha1": "cf9545b690bf5d1acf7f04019aad3a0b1dff343f",
"oa_license": "CCBY",
"oa_url": "https://pathofscience.org/index.php/ps/article/download/1240/1044",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "c627a1766db2387067c6e87684c10325b16dd47b",
"s2fieldsofstudy": [
"Business",
"Agricultural And Food Sciences"
],
"extfieldsofstudy": []
} |
6991660 | pes2o/s2orc | v3-fos-license | Non-alcoholic Fatty Liver Disease Is a Risk Factor for Type 2 Diabetes in Middle-aged Japanese Men and Women
Objective Emerging studies have focused on the association between non-alcoholic fatty liver disease (NAFLD) and the risk of type 2 diabetes mellitus (T2DM). We aimed to investigate whether NAFLD diagnosed by ultrasonography could predict the risk of future T2DM in a Japanese middle-aged health check population. Methods We conducted a 10-year observational study in a health checkup population of middle-aged Japanese men and women at Hidaka Hospital from 2004 to 2013. We excluded cases with an alcohol intake exceeding 20 g/day and those with impaired glucose tolerance. The remaining 1,544 men and 864 women were classified into fatty liver and non-fatty liver groups based on the findings of abdominal ultrasonography. Both groups were followed for the development of diabetes. A multiple regression analysis was performed for each variable to predict the risk of future diabetes. Results The median age of the participants was 46.0 years at the entry, and the follow-up period was 10 years. The incidence of diabetes in the fatty liver group was 12.5% (29/232) in men and 26.3% (10/38) in women, whereas the incidence of diabetes in the non-fatty liver group was 2.5% (34/1,312) in men and 1.8% (15/826) in women. The relative risk of diabetes associated with fatty liver was 4.8 [95% confidence interval (CI) 3.0 -7.8, p<0.0001] in men and 14.5 (95% CI 7.0-30.1, p<0.0001) in women. Conclusion NAFLD was a significant predictor for future diabetes in a Japanese middle-aged health check population, especially in women.
Introduction
With the increasing prevalence of obesity, diabetes, and metabolic syndrome in the United States (US), non-alcoholic fatty liver disease (NAFLD) has become the most common disease of chronic liver damage in the US (1)(2)(3). The prevalence of NAFLD is also increasing in Japan due to the Westernization of lifestyles, such as high-fat and highcalorie diets with less physical activity (4,5). A high prevalence of fatty liver has been reported to be associated with type 2 diabetes mellitus (T2DM) (6)(7)(8).
NAFLD is characterized by significant lipid deposition in the hepatocytes in patients without a history of excessive alcohol intake and is often associated with obesity (9) and hypertension (10). Although these abnormalities are often categorized as insulin resistance syndrome or metabolic syndrome (11), they carry a risk of cardiovascular disease. In addition, diabetes, insulin resistance, and increased plasma fatty acids are considered to increase the risk of NAFLD (12, 13), and each of these metabolic factors is also characteristic of T2DM. It has been reported that NAFLD influences the severity of hepatic insulin resistance in T2DM (14). Furthermore, NAFLD has been correlated with hepatic insulin resistance independently of obesity and intraabdominal adiposity among non-obese men without T2DM (15).
The increased prevalence of NAFLD in relation to the development of diabetes has been reported in a cross-sectional study (16), and a close relationship between liver enzymes and diabetes has been reported in cohort studies (17)(18)(19)(20)(21)(22). However, the former does not prove a causal relationship, and the latter does not directly concern NAFLD. Shibata et al. first reported an observational cohort study in male workers over 40 years old from 1997 to 2005 as a health-check study of the direct relationship between NAFLD and the incidence of diabetes (23). The causal relationship between NAFLD and diabetes is not yet clear, but NAFLD might play an important role as a confounding factor.
The present study is a retrospective study clarifying the significance of NAFLD on the risk of diabetes in an observational population study in middle-aged men and women who continuously attended health checks from 2004 to 2013 at Hidaka Hospital, which is located in a typical, middlesized Japanese city near Tokyo, Japan.
Subjects
The study population comprised randomly selected individuals who had continuously received an annual health check at Hidaka Hospital, Takasaki, Japan, from April 2004 to October 2013. Subjects with the following were excluded: 1) alcohol intake >20 g/day (based on the frequency of drinking alcohol); 2) a recent history of acute illness; 3) taking medications known to modify lipid levels, blood pressure, or blood glucose; 4) taking medications known to induce a fatty liver; 5) being seropositive for hepatitis B or C; and 6) having any other health problem known to be associated with liver steatosis. All cases were aged 35-65 years old. Those with fatty liver in this study were confirmed to have NAFLD.
All of the patients were referred to the hepatology unit at Hidaka Hospital for ultrasound investigation of hepatic steatosis, performed as a part of routine clinical practice. Subjects were divided into those with and without a fatty liver according to the ultrasound results. The fatty liver-positive group included cases that presented with "bright liver" echo patterns according to standard criteria (i.e., evidence of a diffuse increase in the echogenicity of the liver compared with that of the kidneys). The non-fatty liver group was comprised of cases that did not exhibit any echogenic evidence of fatty liver.
Diabetes subjects were defined according to the Japan Diabetes Society (JDS) criteria: HbA1c (NGSP) ! 6.5% and fasting plasma glucose ! 126 mg/dL or the use of a prescribed anti-diabetic agent. In addition, pre-diabetes subjects [mainly those with impaired fasting glucose (IFG)] were also defined according to the JDS criteria: HbA1c (NGSP) <6.5%, fasting plasma glucose 110 to <126 mg/dL, and no prescriptions for anti-diabetic agents. "Suspected diabetes", as a type pre-diabetes defined by JDS criteria, was not included, since this could not be determined in a single examination.
The weight and height of each participant was measured, and the body mass index (BMI) was calculated to evaluate the presence of obesity. The retrospective study analysis in this health check population was approved by the Ethics Committee of Hidaka Hospital.
Biochemical analysis
After 12-h overnight fasting, blood samples were taken, and sera were isolated and stored at 4 . The following analytes were measured within 48 hours of drawing blood: liver enzymes, HbA1c, total cholesterol (TC), triglycerides (TG), and direct low-density lipoprotein cholesterol (LDL-C) and high-density lipoproteins (HDL-C) as previously reported (19). The reagent used to measure LDL-C and HDL-C by the direct method was obtained from Denka Seiken (Tokyo, Japan). Liver enzymes, including aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyltranspeptidase (GGT), and choline esterase (ChE), were used as markers of hepatic damages, and glucose, insulin, and hemoglobin A1c (HbA1c) were measured using commercial enzymatic kits on a Hitachi 917 autoanalyzer (Tokyo, Japan).
Statistical analysis
The data are presented as the median with the 25th and 75th percentile. Accordingly, the differences between groups were tested by Mann-Whitney U test, and multivariate analyses were performed after logarithmic transformation of the results for AST, ALT, GGT, TG, choline esterase, and HbA1c. A multiple regression analysis was performed for each variable with age and the presence of fatty liver (binary: 0 or 1) as the explanatory variables in order to examine the relative contribution to fatty liver. A multiple logistic regression analysis was also performed to evaluate the relative contribution of the variables in diagnosing the status of fatty liver. In addition, a factor analysis was performed to explore the commonality among the variables based on the principal component method with the varimax rotation of the factorial axes. A comparison of factor scores derived from the analysis between subjects with or without fatty liver was performed via a receiver operating characteristic (ROC) analysis. The area under the ROC curve was used as an index of accuracy in differentiating the two conditions. All of the statistical analyses were performed using the Stat-Flex Ver. 6 software program (Artech, Osaka, Japan). p values <0.05 were considered statistically significant. Table 1 shows the clinical and biochemical characteristics of the cases in this study group. The parameters of lipid metabolism, liver function, and glucose homeostasis were all significantly different between sexes (Table 1). BMI was also significantly different between sexes, as well as between the NAFLD and non-NAFLD cases. The most significant difference between sexes was in the prevalence of NAFLD, with values of 15.0% in men and 4.4% in women. These data indicated a 3.4-fold higher prevalence of NAFLD in men than in women. Table 2 compares the values between NAFLD and non-NAFLD cases for both sexes. All of the parameters were further divided into those with or without a NAFLD. We noted a stronger association of NAFLD with lipoproteins (elevated TG, low HDL-C) and markers of glucose homeostasis and liver function in those with NAFLD than in those without. HDL-C levels were significantly lower in those with NAFLD than in those without, and similar trends were noted for choline esterase (ChE) and uric acid (UA). In addition, NAFLD cases showed significantly higher levels of ALT, AST, and γ-GTP than those without, although most of these levels were still within the normal ranges.
Incidence of diabetes over 10 years by presence of NAFLD
The median age of participants was 46.0 years at the start of the study period, and the follow-up period was 10 years. Table 3A shows that the incidence of diabetes in the NAFLD group was 12.5% (29/232) in men and 26.3% (10/ 38) in women, whereas the incidence in the non-NAFLD group was 2.6% (34/1,312) in men and 1.8% (15/826) in women. The relative risk of diabetes associated with NAFLD was 4.8 [95% confidence interval (CI) 3.0-7.8, p< 0.0001] in men and 14.5 (95% CI 7.0-30.1, p<0.0001) in women. The incidence of diabetes over the 10-year period was significantly higher in women than in men, although the proportion of cases of diabetes in women with NAFLD was significantly lower than that in men with NAFLD. In addition, the incidence of diabetes in non-NAFLD cases was significantly lower than in NAFLD cases (p<0.0001). Table 3B shows that the relative risk of diabetes associated with the incidence of NAFLD during the 10-year observational period was 4.6 (95% CI 2.4-8.9, p<0.0001) in men and 4.0 (95% CI 1.5-10.7, p<0.01) in women. Although the relative risk of incidence of diabetes associated with NAFLD was significantly higher in women than in men (Table 3A), the relative risk of incidence of diabetes associated with the incidence of NAFLD was almost the same between sexes.
In addition, the samples were stratified into 3 groups by age: <41 years old, 41 to 50 years old, and ! 51 years old. Table 4 shows the results of the multiple regression analysis with age, gender, NAFLD, HbA1c, lipids, and lipoproteins set as the target variables. The values in Table 4 (A and B) represent the standardized regression coefficients that correspond to the partial (or adjusted) correlation coefficient between the target variables of diabetes and each explanatory variable.
Multiple logistic regression analysis of diagnostic markers suggestive of diabetes risk in 2013
To identify variables effective in predicting diabetes from the diagnostic markers (age, gender, ALT, AST, ChE, UA, HDL-C, LDL-C, and HbA1c), a multiple logistic regression analysis was performed by the backward elimination procedure. Table 4A shows the variables selected through the best regression model. The predictors of diabetes (from most to least important) were HbA1c, NAFLD, HDL-C, gender (fe-male), and age. The accuracy of prediction by the model was evaluated by an ROC analysis, with the area under the curve (AUC) at a high level of 0.865. After removing the strongest predictor HbA1c (Table 4B), NAFLD was effective in predicting the diabetes, with an accuracy of AUC=0.788.
To investigate the relationship between NAFLD and the diabetes incidence and cardiovascular risk markers in 2013, we carried out a multiple regression analysis using NAFLD as the target variable and diabetes incidence and cardiovascular risk markers in 2013 as the explanatory variables (Table 5). NAFLD was more strongly associated with diabetes incidence than cardiovascular risk markers in 2013.
Factor analysis for evaluation of commonality among the diagnostic parameters
The test results of 10 mutually related diagnostic parameters were subjected to an exploratory factor analysis to in- vestigate the commonality among them. Using the principal component method with varimax rotation of the factorial axes, we obtained five factors as shown in Table 6. The values in the table correspond to factor loadings, which represent the correlation coefficient between the identified common factor and each of the original variables. From the pattern of factor loadings, Factor 1 was highly positively correlated with gender, BMI, TG, and HDL-C. Therefore, the factor was interpreted as reflecting the serum lipid level and accumulation of visceral fat, and Factor 1 was referred to as "visceral fat". Factor 2 was interpreted as representing blood pressure. Factor 3 showed HbA1c and age. Therefore, Factor 3 was interpreted as representing the status of diabetes. Factor 4 was found to be highly positively correlated with HbA1c and NAFLD. Therefore, Factor 4 was interpreted as representing the incidence of diabetes. Factor 5 was highly correlated with BMI and LDL-C, representing obesity. The scores of each factor were computed for each individual.
The accuracy of each score in differentiating the status of fatty liver was evaluated by an ROC analysis. The area under the ROC curves for Factors 1, 2, 3, 4, and 5 were 0.726, 0.608, 0.825, 0.841, and 0.690, respectively, confirming a higher specificity of Factor 4 to the incidence of diabetes than other Factors.
Discussion
In the present health check study, we demonstrated that NAFLD is a strong risk factor for developing diabetes in middle-aged healthy Japanese men and especially in women. Specifically, the results indicated that women with NAFLD have a significantly higher incidence of diabetes than men with NAFLD, although the number of cases of NAFLD among women was significantly lower than among men. This study had several strengths, including the use of a population-based sample, sufficient ascertainment of a diabetes diagnosis, exclusion of individuals with prediabetes and diabetes at baseline, and adjustment for a number of potential confounding factors based on the criteria defined by the JDS. Table 3 shows that the relative risk of developing diabetes was 4.8 (95% CI: 3.0-7.8, p<0.0001) in men and 14.5 (95% CI 7.0-30.1, p<0.0001) in women among subjects with NAFLD, which was higher than in almost all previous cohort studies (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33). A systematic review and meta-analysis indicated that the pooled odds ratio for cohort studies was 2.42 (95% CI: 1.78-3.29). Most studies diagnosed T2DM only by fasting blood glucose level or by fasting blood glucose plus HbA1c level (with/without taking antidiabetic drugs). Only one nested case-control study in men made the diagnosis based on the 75 g OGTT with an odds ratio of 5.5 (95% CI 3.6-8.5, p<0.001) (23). Making a diagnosis not based on the 2-h postprandial glucose levels may ignore a proportion of potential diabetic patients. Since patients with NAFLD were more likely to have a risk of diabetes than those without NAFLD, incident diabetes would more likely be unnoticed in the NAFLD group, indicating that the risk of T2DM associated with NAFLD might be underestimated in these studies. Future large population studies are required to confirm that the risk of T2DM associated with NAFLD might be underestimated without performing OGTTs. Table 3A shows that, although the prevalence of NAFLD in men was higher than in women, women with NAFLD had a significantly higher incidence of diabetes than men. However, interestingly, Table 3B shows that the incidence of NAFLD and the relative risk of incidence of diabetes in women found during the observational period was almost the same as in men. In this study, it is possible that most female subjects entered menopause during the 10-year study period, as the median of age for female subjects was 46 in 2004, although we did not investigate the status of menstruation in female subjects. To clarify the effect of postmenopausal status on our findings, we stratified all of the subjects into three groups by age: <41 years old, 41-50 years old, and ! 51 years old, and calculated the agestratified relative risks. Table 3C shows that the age-adjusted relative risk of incidence in women <41 years old was the highest among women, and the relative risk in women ! 51 years old, where most were likely postmenopausal, was almost the same as that of men. Because estrogen, which has a protective effect against NAFLD (34) and diabetes (35) in women, decreases in postmenopausal women, the prevalence of NAFLD and relative risk of incidence of diabetes with NAFLD in women was higher than in men. The effects of estrogen may therefore have influenced the results obtained over the 10-year course of our study. Table 4 shows that HbA1c and NAFLD were stronger diagnostic markers of incidence of diabetes than others by multiple regression analyses (36). HDL-C, gender, and age were also strong markers of the incidence of diabetes. The accuracy of the prediction by the model was evaluated by an ROC analysis, and the AUC was extremely high (0.865). Af-ter removing the strongest predictor (HbA1c), NAFLD proved to still be effective in that order for predicting diabetes with an accuracy of AUC=0.788.
A multiple logistic regression analysis (Table 5) revealed that fatty liver and incident diabetes were significantly correlated. Although diabetes and visceral fat are known to be strongly correlated, fatty liver is independent from the visceral adipose tissue when predicted by plasma adiponectin levels as a specific fatty liver biomarker (36), as previously reported in the Framingham Study (37).
As an additional investigation, we obtained five factors by exploratory factor analysis, as shown in Table 6. These five factors were interpreted as reflecting or representing visceral fat (Factor 1), blood pressure (Factor 2), status of diabetes (Factor 3), incidence of diabetes (Factor 4), and obesity (Factor 5). Factor 4 showed moderate positive correlation with NAFLD and HbA1c. Therefore, Factor 4 was interpreted as representing the incidence of diabetes. Factors such as age, NAFLD, and HbA1c primarily represent the status and incidence of diabetes, while the others are primarily visceral fat, obesity, and blood pressure factors. Although HDL-C and gender (female) were shown to be more significantly associated with the incidence of diabetes than with age in Table 3A and B, only slight correlations were found with Factor 3 and 4 by the factor analysis. The individual scores of Factor 4 (DM score) were computed, and the diagnostic utility of the DM score in diagnosing the presence of incidence of diabetes was examined. The AUC representing the accuracy of the diagnosis was 0.841, while that of HbA1c was 0.802 (data not shown). The use of Factor 4 markers, which are most commonly ordered in routine clinical practice at hospital laboratories, could improve the accuracy of predicting the incidence of diabetes when a diagnosis of NAFLD is added.
Several limitations associated with the present study warrant mention. First, despite our best efforts, we had a relatively low follow-up rate over the 10-year study period, which was similar with most previous reports (30)(31)(32)38). It is challenging for a population-based cohort study (health check at the same hospital) to maintain an adequate rate of follow-up nowadays. The individuals that participated in the follow-up study, compared with those who did not, were elderly, had more fatty liver, and had higher BMI and metabolic parameters (blood pressure and lipids). However, no marked differences were observed in gender, lifestyle factors, and blood glucose levels between the followed-up and non-followed-up groups. Second, our study sample was not large enough to conduct subgroup analyses to explore the association between degree of NAFLD and the development of T2DM. Third, ultrasound was the preferred choice in our study because of its low cost, safety profile, and availability, although liver biopsy remains the current gold standard for the diagnosis of NAFLD (39). Nevertheless, ultrasound becomes less sensitive when the liver fat content is below 33% or with an increase in abdominal obesity (39)(40)(41), which might bias the results to some extent.
In conclusion, we found that NAFLD was a significant predictor for future diabetes in middle-aged Japanese men and especially in women. Although the prevalence of NAFLD was significantly lower in women than in men, the causal relationship between NAFLD and diabetes in women seemed to be stronger than in men in our 10-year observational study.
The authors state that they have no Conflict of Interest (COI). | 2018-04-03T03:34:46.803Z | 2017-04-01T00:00:00.000 | {
"year": 2017,
"sha1": "d4ccde022b32d5c6d82f8f5ae14eac61b777db68",
"oa_license": "CCBYNCND",
"oa_url": "https://www.jstage.jst.go.jp/article/internalmedicine/56/7/56_56.7115/_pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d4ccde022b32d5c6d82f8f5ae14eac61b777db68",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
153312915 | pes2o/s2orc | v3-fos-license | The $SU(2)$ Wess-Zumino-Witten spin chain sigma model
Classical strings propagating in $AdS_{3}\times S^{3}\times T^{4}$ supported with Neveu-Schwarz-Neveu-Schwarz flux are described by a Wess-Zumino-Witten model. In this note, we study the emergence of their semiclassical $SU(2)$ spectrally flowed sectors as the Landau-Lifshitz limit of the underlying quantum spin chain. We consider the propagator in the coherent state picture, and find that the time interval is discretized proportionally to the lattice spacing. In the Landau-Lifshitz limit, where both time and space become continuous, we derive a path integral representation of the propagator for each spectrally flowed sector. We prove that the arbitrariness of the global phase of coherent states is mapped to the gauge freedom of the $B$-field in the classical action. We show that higher order corrections in the Landau-Lifshitz limit are suppressed by inverse powers of the 't Hooft coupling.
The SL(2, R) Wess-Zumino-Witten (WZW) model constitutes a paradigm among exactly solvable realizations of string theory. It represents bosonic strings propagating in AdS 3 supported with Neveu-Schwarz-Neveu-Schwarz (NS-NS) three-form flux, and its spectrum is attainable through the study of the representations of the current algebra upon the worldsheet [1]. The use of integrability in the AdS/CFT correspondence has renewed the interest in this model, since it rises in type IIB superstring theory realized on the integrable AdS 3 × S 3 × T 4 background with mixed three-form fluxes [2]. Specifically, it corresponds to the pure NS-NS flux limit of the anti-de Sitter component of the bosonic truncation of the latter. In fact, a proposal for obtaining the spectrum of quantum strings on the whole background in that limit has been recently put forward on the basis of an integrable spin chain representation [3,4], both matching and generalizing the previously existing results [1,5]. This picture poses the problem of the emergence of the classical integrable structure, reached as the pure NS-NS flux limit of the mixed flux classical setting, arising from the worldsheet spin chain description. An appealing answer in this respect comes from the application of the Landau-Lifshitz limit to the spin chain. This method proved the equivalence between various sectors of four-dimensional N = 4 Yang-Mills theory and type IIB strings in the AdS 5 × S 5 background [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20]. As reviewed in [21], the procedure starts with the expectation value of the Hamiltonian of the spin chain in a general coherent state. A path integral over periodic coherent states is then derived by means of the usual discretization of the time interval. The matching with the effective action of the slow degrees of freedom coming from the sigma model is performed when a continuum limit for the spin chain sites is taken in the action in the path integral. Such a matching can be traced to the rearrangement of the perturbative expansions of both sides of the duality in an effective parameter which can be made small from the two points of view. The method has the advantage of showing the equivalence between integrable structures avoiding the reference to any particular solutions, since the identification of the two actions entails the correspondence between their respective conserved charges.
In this note we will follow this path to retrieve the classical WZW model for strings on R × S 3 with pure NS-NS flux from the the SU(2) sector of the underlying spin chain [22].
We will exploit the fact that the Landau-Lifshitz limit can be applied to any spin chain whose Hamiltonian is known to obtain an effective sigma model action for the worldsheet spin chain. In particular, we will derive a semiclassical path integral representation for the scalar product of two coherent states at different instants in every spectrally flowed sector of the theory. The partition of the time interval will reveal that time steps must be discretized in terms of the distance between spin chain sites. We will then apply the Landau-Lifshitz limit, which consists in simultaneous time and space continuum limits of the discrete expression. This feature is in sharp contrast to the way it is displayed in the AdS 5 /CFT 4 correspondence [6], where the time interval is not discretized but partitioned, and the Landau-Lifshitz limit is introduced as a space continuum approximation in an exact path integral representation. We will find both the path integral representation and the condition under which it belongs to a specific spectrally flowed sector. We will show that the continuous counterpart of global phases of coherent states realizes the classical gauge field of the B-field. We will further argue that subleading terms in the Landau-Lifshitz limit should emerge as quantum corrections.
We will start discussing briefly the worldsheet spin chain which realizes the SU (2) sector of the WZW model with pure NS-NS flux. The Hamiltonian under consideration is given by the shortening condition [23] where k = √ λ is the level of the WZW model, with λ the 't Hooft coupling, P is the momentum operator and M is an operator accounting for representation-dependent shifts of the dispersion relation. The level k bounds the number of sites L of the w-th spectrally flowed sector of the spin chain through [1,4] kw where w is a positive integer that parameterizes the spectral flow. Regarding the eigenvalues µ of M for states of the SU(2) spin chain, µ = 0 corresponds to the vacuum state, as a consequence of the BPS condition, while µ 2 = 1 for single-magnon states [4]. When µ = 1 the state transforms in a representation of the left-handed algebra of SU(2), whereas it transforms under the right-handed one if µ = −1. The action of M on composite-magnon states follows from these relations. We are thus able to construct a SU(2) spin chain of a given handedness by restricting the eigenvalue µ of the single-magnon states. The connection between µ and the angles in the Hopf fibration of the sphere, required for the construction below, can be elucidated as follows. Consider the metric of S 3 expressed as where ϕ 1 is the angle of the Cartan torus in S 3 with respect to which the light-cone gauge is imposed, and y i are the coordinates in the transverse directions. The ranges of these 2]. When the canonical light-cone gauge quantization scheme is applied, the left-handed and right-handed choices of µ correspond, respectively, to excitations in the Y = −y 1 − iy 2 andȲ = −y 1 + iy 2 complex directions [23]. The transformation to the coordinates in the Hopf fibering coordinates of the sphere, where the metric reads with θ ∈ [0, π/2], ϕ i ∈ [0, 2π), is given by Therefore, the orientation of ϕ 2 accounts for the handedness of the SU (2) representation.
In order to map the spin chain to a sigma model we need to introduce a continuum set of variables at each site of the chain, as reviewed for instance in [24]. An infinite set of states can be constructed by applying a SU(2) rotation to an eigenstate of the Cartan generators. If we parameterize such rotation in terms of Euler angles, the one-site coherent state reads [24] | n = e iχ e iϕ cos ϑ |1 + e −iϕ sin ϑ |2 , where |1 and |2 span the fundamental representation of SU (2), and the Euler angles are related with those in (4) as Coherent states form an overcomplete basis with the resolution of the identity where is the Haar measure. The extension to coherent states in a spin chain with L sites is obtained from the tensor product of L one-site coherent states and the associated measure is therefore equal to the product of one-site measures. It should be noted that the angles χ appear as global phases of one-site coherent states. If they were removed as in the case of N = 4 Yang-Mills, we would find the SU(2)/U(1) Landau-Lifshitz model rather than the SU(2) one in the subsequent derivation. Here we will keep them as they will be crucial in the comparison with the Landau-Lifshitz limit of the non-linear classical sigma model.
We also note that the operation ϕ 2 → −ϕ 2 that inverts the handedness of the SU (2) representation is realized by the exchange of ϕ and χ at each site.
The propagator of the theory to which the path integral analysis is applied merges an initial coherent state of the spin chain at time t 0 = 0, | n 0,a , and a final one at the instant where a = 0, ..., L − 1 labels the spin chain sites. We impose periodic boundary conditions n 0,L = n 0,0 and n N,L = n N,0 so that coherent states depict closed strings and limits thereof in the classical sigma model limit. Moreover, we will choose the spin chain frame rather than the string frame, as it will prove best suited to the setting considered here [4].
Now we must recall that the Hamiltonian above is taken to be realized in the proper representation of the algebra. However, the shortening condition (1) where µ denotes the eigenvalue of M for the state | n 0,a , we are able to write We will now slice [0, T ] in N subintervals [t α+1 , t α ] of equal length ∆t, with t α = αT /N and ∆t = T /N. If we introduce an insertion of the spectral decomposition of the identity between the endpoints of every pair of consecutive subintervals, we can express (12) as 1 We perform the derivation in Minkowskian time. The path integral expression that we would obtain in Euclidean time can be recovered from the one presented here by a Wick rotation.
where dµ α is the measure of the α-th insertion. At this point we must emphasize that the time dependence must be intrinsically discretized for the spin chain at issue. This follows from the expression of the operator appearing in (13), which involves the shift operator U = exp (iǫP ) raised to the power k∆t/2πǫ, where ǫ is the lattice spacing. In order for the action of the shift operator on the state space to be defined, the length of the time subinterval must then satisfy which implies the discretization of the time interval. Note that even if ∆t = 2πmǫ/k, with m a positive integer, is also admissible, the choice m = 1 is always achievable by diving the whole interval [0, T ] in enough subintervals. For concreteness we will consider the right-handed SU(2) spin chain. The operator in (13) is then realizable via 2 where σ a 3 is the diagonal Pauli matrix acting on the a-th site. If we choose this representation, we can write the propagator (13) as ϕα,a) cos ϑ α+1,a−1 cos ϑ α,a + e i(ϕ α+1,a−1 −ϕα,a+∆t) sin ϑ α+1,a−1 sin ϑ α,a .
(16)
The leading order action of the Landau-Lifshitz sigma model is obtained in the semiclassical limit of long wavelength, where the spin chain sites are taken to be parameterized continuously and the dependence of the angles on them is assumed to be analytic. In a conventional spin chain this limit consists in extracting the leading contribution inside the path integral after sending ǫ → 0 and L → ∞, with the length of the spin chain R = Lǫ fixed. Such path integral is constructed by applying to the propagator a prior continuum limit of the time coordinate defined by ∆t → 0 and N → ∞, while T = N∆t kept fixed.
However, space and time continuum limits are intertwined for the spin chain considered here. On the one hand, space and time step lengths are proportional as stated by (14), that implies that T = 2πNR/kL. Therefore, when both continuum limits are applied the condition N/kL ∼ O(1) is needed. On the other hand, the number of sites L of the chain 2 The choice of one or other vacuum determines classical actions which differ in a term without derivatives whose sign depends on the handedness of the SU (2) representation. This is however irrelevant in the Landau-Lifshitz limit. Besides, whatever the choice is, it is meant to represent the BPS vacuum of the R-R sector in the center of the Hodge diamond [4]. is bounded by the WZW level k in view of equation (2), and thus L → ∞ already assumes the semiclassical limit k → ∞, which implies that ∆t → 0. In fact, in order for the spin chain to remain within the w-th spectrally flowed sector, it must be satisfied that when both L, k → ∞. If we apply the Landau-Lifshitz limit to (16) taking these observations into account, the propagator becomes where the dot denotes the derivative with respect to t α , while the prime is the derivative with respect to x a = aǫ. The integrand in (18) may be regarded as the formal product of two Volterra continuous products when ∆t, ǫ → 0 once the short distance cut-off 1/∆t, or equivalently 1/ǫ, is introduced [24]. Therefore, in the continuum limit we can write where [dµ] denotes the path integral measure and The path integral extends over configurations subject to the periodicity condition n(t, x) = n(t, x + R) that satisfy the boundary conditions n(0, x) = n 0 (x) and n(T, x) = n N (x), where n 0 (x) and n N (x) are, respectively, the continuous counterparts of n 0,a and n N,a . It is worth to emphasize that we could have proceeded by expressing equation (12) as which leads to the same action as (20) once the expression of its associated Hamiltonian is considered. Nevertheless, if we had taken directly this path, some steps that are necessary in the attainment of equation (20) would have been obscured.
The Landau-Lifshitz limit can also be applied at the level of the classical string action.
It is obtained as the perturbative expansion with respect to √ λ/L of the Lagrangian density in the statically gauge fixed T-dual sigma model [10]. In particular, the Wess-Zumino term involving first order time derivatives appears at zeroth order, whereas the n-th order contribution consists in terms with n spatial derivatives. Such an expansion can be achieved in the action above by means of the change of variables t → t/k, which enables us to compare (20) with the result obtained from the classical non-linear sigma model. If we introduce this change of scale together with x → x/2π, the leading order action in the semiclassical limit reads If we choose now R = 1, we can replace ǫ by 1/L in front of the action, hence obtaining the NS-NS limit of the Landau-Lifshitz model found in [22]. The action (22) is the zeroth order in √ λ/L of the Landau-Lifshitz model. In principle, it should be corrected by subleading terms in the long wavelength limit [7,10]. However, these corrections are suppressed by additional powers of 1/ √ λ for the model at issue. We can show this by considering the most general possible term in (18), involving A time derivatives, B spatial derivatives and none derivatives in a factor φ C accounting for the contribution of M. The contribution to the action of this term, ∆t A+C ǫ BφA φ ′B φ C , after the change of variables t → t/k and x → x/2π is made (and R = 1 is set), reads Therefore terms with either A > 1, B > 1 or C > 0 carry additional powers of 1/ √ λ besides the factor L( √ λ/L) A+B+C−1 , and thus cannot be obtained from the Landau-Lifshitz expansion of the classical non-linear sigma model. 3 There are some natural extensions to the derivation in this letter. The most immediate question concerns the emergence of the classical Landau-Lifshitz model from other sectors of the worldsheet spin chain, such as the left-handed and right-handed SL(2, R) sectors, or those which include fermions. | 2019-05-15T14:28:54.101Z | 2019-05-14T00:00:00.000 | {
"year": 2019,
"sha1": "845098c328931552911442eb2b6b63f4efd461c8",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/JHEP06(2019)080.pdf",
"oa_status": "GOLD",
"pdf_src": "Arxiv",
"pdf_hash": "a4c22a770c827b1899ff5f1bda2f84e8b2057022",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
11843876 | pes2o/s2orc | v3-fos-license | Technique for Determining Bridge Displacement Response Using MEMS Accelerometers
In bridge maintenance, particularly with regard to fatigue damage in steel bridges, it is important to determine the displacement response of the entire bridge under a live load as well as that of each member. Knowing the displacement response enables the identification of dynamic deformations that can cause stresses and ultimately lead to damage and thus also allows the undertaking of appropriate countermeasures. In theory, the displacement response can be calculated from the double integration of the measured acceleration. However, data measured by an accelerometer include measurement errors caused by the limitations of the analog-to-digital conversion process and sensor noise. These errors distort the double integration results. Furthermore, as bridges in service are constantly vibrating because of passing vehicles, estimating the boundary conditions for the numerical integration is difficult. To address these problems, this paper proposes a method for determining the displacement of a bridge in service from its acceleration based on its free vibration. To verify the effectiveness of the proposed method, field measurements were conducted using nine different accelerometers. Based on the results of these measurements, the proposed method was found to be highly accurate in comparison with the reference displacement obtained using a contact displacement gauge.
Introduction
Fatigue damage in steel bridges is difficult to detect. Nevertheless, because there is a possibility that fatigue damage will lead to brittle fracture, appropriate inspection and countermeasures are necessary. To detect fatigue damage and take action before brittle fracture occurs, it is important to determine the displacement response of the entire bridge to external forces, in addition to that of each member, and to specify the displacement responses of members that can lead to stress concentrations [1,2]. In particular, the displacement response under live loads is a factor that dominates fatigue damage, and it is important that this can be accurately assessed to ensure steel bridges can be effectively maintained.
One method of directly measuring the displacement is to use fixed reference-based technologies, such as linear variable differential transformers (LVDTs), laser Doppler vibrometers (LDVs) [3], and vision-based systems [4][5][6]. Although these measurements are accurate, they are often impractical because they are costly and difficult to install and, more importantly, they require a fixed reference point, which is often unattainable in full-scale civil structures [7]. As an alternative indirect estimation approach, the use of a microelectromechanical systems (MEMS) accelerometer has recently been proposed for the measurement of the response to an external force [8,9]. Such accelerometers are easy to install on painted metal surfaces using magnets. Furthermore, because MEMS accelerometers are inexpensive and compact, their use in monitoring the health of structures is widespread [10][11][12][13][14][15][16][17][18]. Wireless MEMS accelerometers are easily attachable and detachable. Furthermore, unlike fixed reference-based approaches, the use of MEMS accelerometers does not require a fixed reference point, and thus improved workability is expected.
In theory, the displacement can be calculated from the double integration of the measured acceleration. However, data measured by accelerometers include measurement errors caused by the limitations of the analog-to-digital conversion process and sensor noise. These errors distort the results of the double integration. In addition to the effects of the measurement errors, boundary condition errors also have an adverse effect on the accuracy of the double integration results [19].
Park et al., [9] proposed an initial velocity estimation method in which the average velocity is assumed to be zero. Their method was found to provide highly accurate results when the initial displacement is zero and the times at which vehicles enter and exit the bridge can be determined. However, because a bridge in service is constantly vibrating due to the passing of vehicles, the initial displacement is not always zero. Furthermore, they proposed no method of determining when vehicles enter and exit the bridge. To measure the displacement of a bridge in service, a number of problems remain to be solved.
In theory, the displacement of a bridge can be obtained using an accelerometer or a strain gauge. However, in practice, it is difficult for an accelerometer to accurately measure low-frequency components. Additionally, accurately estimating the displacement using a strain gauge is difficult because the strain is easily affected by measurement noise, particularly in the high frequency range. To compensate for these drawbacks, the indirect displacement estimation using acceleration and strain (IDEAS) method was proposed in previous studies [20,21]. However, it is less time-consuming to attach an accelerometer than a strain gauge.
Nine different accelerometers were selected for use in the present study to compare and validate the accuracy of the displacement results obtained by different accelerometers with different specifications. First, the static characteristics of these nine accelerometers were analyzed. Next, to clarify the vibration characteristics of the test bridge, frequency analysis of the displacement of the test bridge was conducted. Then, based on the free vibration of a bridge in service, a method of calculating the displacement from the measured acceleration was proposed. Finally, to verify the effectiveness of the proposed method, measurements were performed on an actual bridge using the abovementioned nine different accelerometers. Based on the results of these measurements, the proposed method was found to be highly accurate in comparison with the reference displacement using contact displacement gauge.
Test Bridge and Installation of Accelerometers and Contact Displacement Gauge
Measurements of an in-service bridge were conducted using nine different accelerometers and a contact displacement gauge installed at the longitudinal center to determine the displacement response under a live load. The test bridge is shown in Figure 1. It is a 34.55 m-long single-span bridge that consists of six main girders with reinforced concrete (RC) deck plates. It is located in Tokyo and managed by the Metropolitan Expressway Co., Ltd. (Tokyo, Japan).
The experimental setup of the accelerometers and the contact displacement gauge (CDP-25, Tokyo Sokki) is shown in Figure 2. Each accelerometer was attached by screws to an L-shaped steel fixture that was fixed to the longitudinal center of the lower flange of the main girder (G3) with a C-clamp.
The accelerometers were aligned in the same direction because the performance of some accelerometers in the vertical direction differs from that in the horizontal direction. A contact displacement gauge was also fixed at the longitudinal center of the lower flange of the main girder (G3) to verify the accuracy of the displacement values obtained from the double integration of the measured accelerations. In this test bridge, the contact displacement gauge could be used on the ground because the bridge is located in an earthwork structure that connects to an elevated bridge. The specifications of the contact displacement gauge are listed in Table 1. ground because the bridge is located in an earthwork structure that connects to an elevated bridge. The specifications of the contact displacement gauge are listed in Table 1. The response of a bridge to a live load occurs at low frequencies [22]. In particular, the frequency range for forced displacement due to a live load is frequencies below 1.0 Hz. Assuming the bridge displacement is a superposition of sinusoidal waveforms, the acceleration can be calculated as ground because the bridge is located in an earthwork structure that connects to an elevated bridge. The specifications of the contact displacement gauge are listed in Table 1. The response of a bridge to a live load occurs at low frequencies [22]. In particular, the frequency range for forced displacement due to a live load is frequencies below 1.0 Hz. Assuming the bridge displacement is a superposition of sinusoidal waveforms, the acceleration can be calculated as
Specifications for Accelerometers Used to Determine Bridge Displacement
The response of a bridge to a live load occurs at low frequencies [22]. In particular, the frequency range for forced displacement due to a live load is frequencies below 1.0 Hz. Assuming the bridge displacement is a superposition of sinusoidal waveforms, the acceleration can be calculated as where A is the acceleration (m/s 2 ), D is the total amplitude of the bridge displacement (m), and f is the frequency (Hz). Therefore, because acceleration at low frequencies is small, an accelerometer with an inherently high resolution and little sensor self-noise at frequencies below 1.0 Hz must be selected to accurately measure the acceleration; for example, an amplitude of 1.0 mm and a frequency of 0.3 Hz correspond to an acceleration of 0.18 cm/s 2 . The specifications of the nine different commercial accelerometers, consisting of eight different MEMS accelerometers and a servo-type accelerometer, that were used to determine bridge displacement in the present study are listed in Table 2. However, the noise density, which is an important specification, is only listed for some of the accelerometers in Table 2. This is because in many cases, the noise density is not provided by the manufacturer. In addition, even in cases where the noise density is provided, different manufacturers may use different methods to calculate it. To allow the accelerometers to be compared directly, it is therefore necessary to evaluate the self-noise level for each sensor under the same measurement conditions. To perform such an evaluation, the static characteristics of the accelerometers were measured using the test setup shown in Figure 3. All accelerometers were attached by screws to the same L-shaped steel fixture to ensure the measurement conditions were the same. All measurements were carried out for a duration of 100 s. Although it is generally necessary to eliminate the effects of external factors such as microtremors and thermal sources when evaluating the sensor self-noise, this is very difficult to achieve in practice. Therefore, the noise level for each sensor was taken to be the self-noise for the entire sensor system, including external factors and the effects of analog-to-digital conversion. Table 2. Figure 4 shows the static characteristics of the different accelerometers obtained using this method. To present the data independent of the frequency resolution, the power spectral density (PSD) (µm/(s 2 ' Hz)) is plotted on the vertical axis. It can be seen that the PSD values at 1.0 Hz are approximately the same as the noise density specifications in Table 2, except for those of Accelerometers G and H. Figure 3. Test setup for analyzing static characteristics of accelerometers listed in Table 2.
Free Vibration Separation Method for Determination of Bridge Displacement
In theory, the displacement of a bridge can be calculated from the double integration of the acceleration as: where U is the displacement (mm), A is the acceleration (mm/s 2 ), V0 is the initial velocity (mm/s), U0 is the initial displacement (mm), t is the time (s), T0 is the initial time (s), and TT is the terminal time (s). However, data measured by accelerometers include measurement errors caused by the limitations of the analog-to-digital conversion process and sensor noise. Although these errors distort the results of the double integration, particularly at low frequencies, it is difficult to remove these errors at low frequencies because live loads on the bridge are random and the frequency range The higher values obtained for these two accelerometers are thought to be due to microtremors from the floor of the building. At frequencies below 1.0 Hz, all of the accelerometers exhibited very different PSDs, and the PSDs of Accelerometers G and H were found to be at least approximately 20 times smaller than those of the other MEMS accelerometers.
Free Vibration Separation Method for Determination of Bridge Displacement
In theory, the displacement of a bridge can be calculated from the double integration of the acceleration as: where U is the displacement (mm), A is the acceleration (mm/s 2 ), V 0 is the initial velocity (mm/s), U 0 is the initial displacement (mm), t is the time (s), T 0 is the initial time (s), and T T is the terminal time (s). However, data measured by accelerometers include measurement errors caused by the limitations of the analog-to-digital conversion process and sensor noise. Although these errors distort the results of the double integration, particularly at low frequencies, it is difficult to remove these errors at low frequencies because live loads on the bridge are random and the frequency range for forced displacement due to live loads is also low (below 1.0 Hz). In addition to the effects of the measurement errors, because a bridge in service is always vibrating as a result of the live load conditions, it is difficult to obtain the initial and terminal conditions for the numerical double integration. Therefore, accurately calculating the displacement is difficult.
Frequency Analysis for Determining Dominant Frequency of Displacement under Live Loads
To determine the dominant frequency of the bridge displacement under live loads, frequency analysis was conducted using the displacement record measured in the field with a contact displacement gauge. Figure 5 shows the displacement measured using the contact displacement gauge for a duration of 4 s at the longitudinal center of the lower flange of the main girder (G3) of the test bridge. Figure 5 shows the typical deflection due to the passage of a vehicle, and the vehicle passing time which depends on the span of the bridge and the vehicle speed, was approximately 2.2 s. Therefore, the frequency of the deflection due to the vehicle weight was approximately 0.5 Hz. In this case, as the length of bridge is 34.55 m, the speed of the vehicle was estimated to be 56.5 km/h. analysis was conducted using the displacement record measured in the field with a contact displacement gauge. Figure 5 shows the displacement measured using the contact displacement gauge for a duration of 4 s at the longitudinal center of the lower flange of the main girder (G3) of the test bridge. Figure 5 shows the typical deflection due to the passage of a vehicle, and the vehicle passing time which depends on the span of the bridge and the vehicle speed, was approximately 2.2 s. Therefore, the frequency of the deflection due to the vehicle weight was approximately 0.5 Hz. In this case, as the length of bridge is 34.55 m, the speed of the vehicle was estimated to be 56.5 km/h. Figure 6 shows the numbers of girder bridges within different ranges of span lengths. Most girder bridges have span lengths of more than 30 m. In addition, because the maximum allowed speed on expressways in Japan is 100 km/h, the response time should be above 1.2 s, and the frequency should be below 0.9 Hz. Figure 7 shows the dominant frequency range below 1.0 Hz due to the vehicle weight and that above 1.0 Hz due to the Figure 6 shows the numbers of girder bridges within different ranges of span lengths. Most girder bridges have span lengths of more than 30 m. In addition, because the maximum allowed speed on expressways in Japan is 100 km/h, the response time should be above 1.2 s, and the frequency should be below 0.9 Hz.
conditions, it is difficult to obtain the initial and terminal conditions for the numerical double integration. Therefore, accurately calculating the displacement is difficult.
Frequency Analysis for Determining Dominant Frequency of Displacement under Live Loads
To determine the dominant frequency of the bridge displacement under live loads, frequency analysis was conducted using the displacement record measured in the field with a contact displacement gauge. Figure 5 shows the displacement measured using the contact displacement gauge for a duration of 4 s at the longitudinal center of the lower flange of the main girder (G3) of the test bridge. Figure 5 shows the typical deflection due to the passage of a vehicle, and the vehicle passing time which depends on the span of the bridge and the vehicle speed, was approximately 2.2 s. Therefore, the frequency of the deflection due to the vehicle weight was approximately 0.5 Hz. In this case, as the length of bridge is 34.55 m, the speed of the vehicle was estimated to be 56.5 km/h. Figure 6 shows the numbers of girder bridges within different ranges of span lengths. Most girder bridges have span lengths of more than 30 m. In addition, because the maximum allowed speed on expressways in Japan is 100 km/h, the response time should be above 1.2 s, and the frequency should be below 0.9 Hz. Figure 7 shows the dominant frequency range below 1.0 Hz due to the vehicle weight and that above 1.0 Hz due to the Figure 7 shows the dominant frequency range below 1.0 Hz due to the vehicle weight and that above 1.0 Hz due to the free vibration. In this study, only the displacement response at frequencies between 0 and 20 Hz was considered because the power spectral density of the displacement response at frequencies above 20 Hz is negligibly small. Figure 8 shows the results of filtering the measured displacement shown in Figure 5 by separately applying a low-pass filter of 1.0 Hz and a bandpass filter between 1.0 and 20 Hz. Figure 8 also shows the original measured displacement. The displacement bandpass filtered between 1.0 and 20 Hz, which is the displacement of free vibration, vibrated about the zero-axis before vehicle entry and after vehicle exit. Conversely, the displacement low-pass filtered at 1.0 Hz, which is the forced displacement, free vibration. In this study, only the displacement response at frequencies between 0 and 20 Hz was considered because the power spectral density of the displacement response at frequencies above 20 Hz is negligibly small.
Free Vibration Separation Method
Based on the above considerations, we propose a new calculation method, hereafter referred to as the free vibration separation method, the basic concept of which is as follows. First, the initial and terminal conditions for the displacement are determined. Assuming that before vehicle entry and after vehicle exit the bridge is vibrating with sinusoidal oscillations about the zero-axis at its free vibration frequency, which includes the fundamental vibration frequency and higher-order-mode was considered because the power spectral density of the displacement response at frequencies above 20 Hz is negligibly small.
Free Vibration Separation Method
Based on the above considerations, we propose a new calculation method, hereafter referred to as the free vibration separation method, the basic concept of which is as follows. First, the initial and terminal conditions for the displacement are determined. Assuming that before vehicle entry and after vehicle exit the bridge is vibrating with sinusoidal oscillations about the zero-axis at its free vibration frequency, which includes the fundamental vibration frequency and higher-order-mode
Free Vibration Separation Method
Based on the above considerations, we propose a new calculation method, hereafter referred to as the free vibration separation method, the basic concept of which is as follows. First, the initial and terminal conditions for the displacement are determined. Assuming that before vehicle entry and after vehicle exit the bridge is vibrating with sinusoidal oscillations about the zero-axis at its free vibration frequency, which includes the fundamental vibration frequency and higher-order-mode vibration frequencies, the displacement of free vibration can be calculated. Second, the acceleration of the forced displacement component of the bridge displacement while a vehicle is passing over it is numerically integrated to obtain the forced displacement. Finally, the displacement, which includes both the free vibration and forced displacement components, is determined by summing the free vibration displacement and the forced displacement estimated above.
Field Measurements
As described above, measurement errors distort the results of double integration. To minimize such distortion, an accelerometer with an inherently high resolution and little sensor self-noise should be selected, and the free vibration separation method described above should be applied to the numerical integration process. To verify the effectiveness of this approach, measurements were conducted on an actual bridge using nine different accelerometers. Application of proposed free vibration separation method of determining bridge displacement.
The procedure of the free vibration separation method is shown in Figure 9. It consists of the following steps.
1.
The free and forced vibration regions are separated by detecting vehicle entry and exit.
2.
The acceleration is transformed from the time domain to the frequency domain by Fourier transform.
3.
By removing the frequencies below 1.0 Hz to eliminate the effect of forced displacement during vehicular passage, the acceleration of free vibration in the frequency domain is estimated. In the same manner, by removing the frequencies above 1.0 Hz to eliminate the effect of free vibration, the acceleration of forced displacement in the frequency domain is estimated.
4.
The accelerations of free vibration and forced displacement in the time domain are estimated by taking the inverse Fourier transform of the accelerations obtained in Step 3.
5.
Assuming the bridge is vibrating with sinusoidal oscillations about the zero-axis, the acceleration of free vibration should be zero when the displacement of free vibration is zero. Therefore, the acceleration of free vibration obtained in Step 4, which excludes the effects of vehicular forced displacement, is numerically integrated twice from the time when the acceleration of free vibration is zero before vehicle entered to the time when the acceleration of free vibration returns to zero after vehicle exited to obtain the displacement. 6.
Because of the measurement error in the measured acceleration, which is caused by sensor noise and the limitations of analog-to-digital conversion, integration error, and the constants of integration, the initial and terminal displacements of the integrated displacement are not equal to zero. To make the initial and terminal displacements equal to zero, the drift component is subtracted from the integrated displacement. 7.
The acceleration of forced displacement obtained in Step 4, which excludes the effects of free vibration, is numerically integrated twice to obtain the displacement. The lower and upper limits of the integration are the times when the first vehicle enters the bridge and the last vehicle exits the bridge, respectively. 8.
Because of the measurement error in the measured acceleration, which is caused by sensor noise and the limitations of analog-to-digital conversion, integration error, and the constants of integration, the initial and terminal displacements of the integrated displacement are not equal to zero. In the same manner as in Step 6, the drift component is subtracted from the integrated displacement to make the initial and terminal displacements equal to zero. 9.
By linearly summing the displacements of free and forced vibration, the total displacement, which includes free and forced vibration components, is estimated.
Field Measurements
As described above, measurement errors distort the results of double integration. To minimize such distortion, an accelerometer with an inherently high resolution and little sensor self-noise should be selected, and the free vibration separation method described above should be applied to the numerical integration process. To verify the effectiveness of this approach, measurements were conducted on an actual bridge using nine different accelerometers.
Detection of Vehicle Entry and Exit
To detect vehicle entry and exit, instead of applying the conventional method using a strain gauge [23], the acceleration was measured using accelerometers at the vertical stiffeners on both longitudinal edges of the main girder. The specifications of the accelerometers are listed in Table 3. The sampling frequency for vehicle detection was 500 Hz. The experimental setup for the accelerometer used to detect vehicle exit is shown in Figure 10. Figure 11 shows the acceleration record and the displacements measured using the contact displacement gauge. The acceleration record revealed the passage of one vehicle with two axles and another vehicle with five axles. Because the passing times of the two vehicles (1.72 and 1.74 s) were different, the results indicated that two vehicles were passing over the bridge. Furthermore, because the distances between the peaks in the acceleration record at the entry and exit sides were equal, it was concluded that the two vehicles passed over at constant speeds. Because the length of the test bridge is 34.55 m, the vehicle speeds, which were 72.3 and 71.5 km/h for the two-and five-axle vehicles, respectively, could be estimated. The entry time and the exit time for the two-axle vehicle (3581.11 s, 3583.04 s) and the entry time and the exit time for the five-axle vehicle (3581.37 s, 3583.73 s) are indicated in Figure 11. vehicle speeds, which were 72.3 and 71.5 km/h for the two-and five-axle vehicles, respectively, could be estimated. The entry time and the exit time for the two-axle vehicle (3581.11 s, 3583.04 s) and the entry time and the exit time for the five-axle vehicle (3581.37 s, 3583.73 s) are indicated in Figure 11. two vehicles passed over at constant speeds. Because the length of the test bridge is 34.55 m, the vehicle speeds, which were 72.3 and 71.5 km/h for the two-and five-axle vehicles, respectively, could be estimated. The entry time and the exit time for the two-axle vehicle (3581.11 s, 3583.04 s) and the entry time and the exit time for the five-axle vehicle (3581.37 s, 3583.73 s) are indicated in Figure 11.
Displacement Results Obtained Using Proposed Free Vibration Separation Method at Longitudinal Center
The displacement results obtained using the proposed free vibration separation method are shown in Figure 12.
The starting (3581.11 s) and ending (3583.73 s) times, which were obtained from the acceleration record at the vertical stiffener, were used as the limits of the numerical integration. By reducing the interval between the starting and ending times used in the integration of the measured acceleration, the integration error due to the measurement error of the measured acceleration can be diminished.
The results for Accelerometers E, F, and H and the servo-type accelerometer, all of which had resolutions of less than 9.8 µm/s 2 and sensor noise levels of less than 2.0 mm/(s 2 ' Hz), and the results obtained using the contact displacement gauge differed by 5.0% or less at the maximum displacement point. However, the results for Accelerometer G, the sensor noise level of which was approximately the same as that of Accelerometer H, were totally different from those obtained using the contact displacement gauge. The reason for this is thought to be that the measurement error due to the aliasing of the measured acceleration distorts the results. Furthermore, the results for Accelerometer D, the sensor noise level of which is approximately the same as those of Accelerometers E and F, were also markedly different from those obtained using the contact displacement gauge. The reason for this is thought to be that the resolution of Accelerometer D is too large to accurately measure the response at frequencies lower than 1.0 Hz. For example, the acceleration is 490 µm/s 2 for an amplitude of 0.1 mm and a frequency of 0.5 Hz. The results for Accelerometers A, B, and C and those obtained using the contact displacement gauge differed by more than 5.0% at the maximum displacement point. A low-frequency component was present in the results for these accelerometers because of the measurement error due to the limitations of the analog-to-digital conversion process and their sensor noise levels, which were more than 50 times that of Accelerometer H.
Displacement Results Obtained Using Proposed Free Vibration Separation Method at Longitudinal Center
The displacement results obtained using the proposed free vibration separation method are shown in Figure 12. The starting (3581.11 s) and ending (3583.73 s) times, which were obtained from the acceleration record at the vertical stiffener, were used as the limits of the numerical integration. By reducing the interval between the starting and ending times used in the integration of the measured acceleration, the integration error due to the measurement error of the measured acceleration can be diminished.
Validation of the Consistency of the Results Using Free Vibration Separation Method to Determine Displacement Response
In order to validate the consistency of the results using the free vibration separation method to determine displacement response, a displacement response for a time different from the time in Figure 12 is determined by the free vibration separation method using nine different accelerometers. The starting (1733.65 s) and ending (1735.69 s) times, which were obtained from the acceleration record at the vertical stiffener in the same manner as shown in Figure 11, were used as the limits of the numerical integration. The displacement results obtained using the proposed free vibration separation method are shown in Figure 13. As with the result in Figure 12, the results for Accelerometers E, F, and H and the servo-type accelerometer, all of which had resolutions of less than 9.8 µm/s 2 and sensor noise levels of less than 2.0 mm/(s 2 ' Hz), are very similar to the results obtained using the contact displacement gauge. On the other hand, the results of Accelerometers A, B, C, and G were markedly different from those obtained using the contact displacement gauge. Although the result of Accelerometer D give accurate peak displacement, the result of the displacement did not coincide, partially because of a low-frequency component due to the measurement error due to limitations of the analog-to-digital conversion process and its sensor noise levels.
Conclusions
In the present study, the static characteristics of nine different commercial accelerometers were first analyzed. Next, to clarify the vibration characteristics of the test bridge, frequency analysis of the displacement of the test bridge was conducted. Then, based on the free vibration of an in-service bridge, a method of calculating the displacement from the measured acceleration was proposed; this method is called the free vibration separation method. To determine the initial and terminal times for use as the lower and upper limits of the numerical integral, vehicle entry and exit times were detected using the response of the measured acceleration at the vertical stiffeners on both longitudinal edges of the main girder. Finally, measurements were performed on an actual bridge using nine different accelerometers to verify the effectiveness of the proposed method. The conclusions of the present study are as follows.
‚
Based on the free vibration of a bridge in service, the free vibration separation method for determining the bridge displacement was proposed.
‚
To apply the proposed method, the vehicle entry and exit times must be detected. In this study, this was achieved using the acceleration measured at a vertical stiffener of the main girder. Furthermore, using the distance between peaks in the acceleration record, the fact that multiple vehicles were passing over the bridge was able to be determined. In addition, using the distance between peaks in the acceleration record, the speeds of the vehicles were able to be estimated.
‚
To verify the effectiveness of the proposed method in determining the bridge displacement under a live load, field measurements were carried out using nine different accelerometers. Implementing the proposed method using an accelerometer with a resolution of less than 9.8 µm/s 2 and a sensor noise level of less than 2.0 mm/(s 2 ' Hz), the calculated displacement response was found to be in good agreement with the reference displacement measured by the contact displacement gauge.
‚
In order to validate the consistency of the results using free vibration separation method to determine displacement response, the displacement at different time is determined. As a result, by using accelerometer with a resolution of less than 9.8 µm/s 2 and a sensor noise level of less than 2.0 mm/(s 2 ' Hz), a result which is similar to the reference displacement measured by the contact displacement gauge could be determined at a different time. The consistency of the results could thus be demonstrated.
In the case of a short vehicle passing time, particularly for the passage of a single vehicle, the proposed method with high resolution and low sensor noise shows good agreement with the reference displacement measured by the contact displacement gauge. However, some traffic conditions may cause serious problems. For example, the adverse effect of the measurement error on the integration results become large when multiple vehicles continuously pass over the bridge because the integration time of the forced displacement is large. Such cases require careful attention. | 2017-09-02T08:55:31.643Z | 2016-02-01T00:00:00.000 | {
"year": 2016,
"sha1": "7db585dc060cb2a79fe0ab57fe418fc2347c51ba",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8220/16/2/257/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7db585dc060cb2a79fe0ab57fe418fc2347c51ba",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Engineering",
"Computer Science",
"Medicine"
]
} |
201837405 | pes2o/s2orc | v3-fos-license | Integrating thermodynamic and sequence contexts improves protein-RNA binding prediction
Predicting RNA-binding protein (RBP) specificity is important for understanding gene expression regulation and RNA-mediated enzymatic processes. It is widely believed that RBP binding specificity is determined by both the sequence and structural contexts of RNAs. Existing approaches, including traditional machine learning algorithms and more recently, deep learning models, have been extensively applied to integrate RNA sequence and its predicted or experimental RNA structural probabilities for improving the accuracy of RBP binding prediction. Such models were trained mostly on the large-scale in vitro datasets, such as the RNAcompete dataset. However, in RNAcompete, most synthetic RNAs are unstructured, which makes machine learning methods not effectively extract RBP-binding structural preferences. Furthermore, RNA structure may be variable or multi-modal according to both theoretical and experimental evidence. In this work, we propose ThermoNet, a thermodynamic prediction model by integrating a new sequence-embedding convolutional neural network model over a thermodynamic ensemble of RNA secondary structures. First, the sequence-embedding convolutional neural network generalizes the existing k-mer based methods by jointly learning convolutional filters and k-mer embeddings to represent RNA sequence contexts. Second, the thermodynamic average of deep-learning predictions is able to explore structural variability and improves the prediction, especially for the structured RNAs. Extensive experiments demonstrate that our method significantly outperforms existing approaches, including RCK, DeepBind and several other recent state-of-the-art methods for predictions on both in vitro and in vivo data. The implementation of ThermoNet is available at https://github.com/suyufeng/ThermoNet.
Introduction
RNA-binding proteins (RBPs) modulate the processing of cellular RNAs, including their production, transportation, splicing, stability, translation, and degradation [1,2]. There are more than 1,500 RBPs in the human genome which are identified with well-defined RNA-binding domains (RBDs) [3], including the RNA recognition motif (RRM) [4], the K-homology domain (KH) [4], and the C3H1 zinc-finger (ZF) domain [5]. Existing studies on different RNA binding domains indicate that their interaction specificities with RNAs are determined by various sequence-and/or structure-specific patterns. For example, sequence motifs on hairpins or loops are more accessible to many RBPs, while some RBPs, such as RBM22, RBM6 and PRR3, show a structural preference on bulged stems [6]. Understanding of the sequence/structure specificity of RNA-binding proteins is therefore critical for developing hypotheses and models of post-transcriptional gene regulation [7].
In vitro and in vivo methods have been developed for determining RBP binding specificities. RNAcompete is a high-throughput in vitro assay that quantifies the relative affinity of a specific RBP to a pre-defined set of more than 240,000 short RNAs. In a recent work, more than 200 human RBPs are analyzed by RNAcompete, generating the first large-scale dataset of protein-RNA interaction measurements [8]. A major limitation of RNAcompete is that the designed RNAs are only 41nt long and not structurally stable, therefore the motifs identified by this approach are predominantly in short unstructured contexts. SELEX (systematic evolution of ligands by exponential selection) [9] iteratively selects a set of high-affinity RNA sequences from a large pool, similar to the evolutionary procedure used in optimization, which is generally biased or suffers from undersampling. Different from in vitro methods, high-throughput in vivo techniques have been developed to measure genome-wide RBP-RNA interactions in their cellular contexts. For example, CLIP-seq, RIP-seq and their variants [10,11] provide high-resolution protein-RNA binding sites. However, it is usually not easy to clearly derive binding motifs from these experiments, because of the existence of protein cofactors, technical artifacts, RNA folding, and high levels of noise or non-specific background. Therefore, learning algorithms for computational prediction of protein-RNA binding from in vitro data may be less affected by the noise and provide insights that can be generalized to in vivo data.
Different computational methods have been introduced to protein-RNA binding prediction. Traditionally, the sequence specificities of an RBP is most commonly modeled by position weight matrix (PWM) [12] or hidden Markov models, which are solely based on the biases or enrichments of nucleotides on the binding sites. Such models can be learned from a collection of RNA sequences with high binding affinities. To take RNA structure into consideration, MEMERIS [13] applies MEME [14,15] to identify binding sites in unpaired loop regions. RNAcontext [16] learns a joint probabilistic model of both sequence and structure contexts. RCK [17], a recent improvement of RNAcontext, uses k-mer based contexts incorporating both sequence-and structure-based binding preferences. More recently, deep learning methods have been adapted to protein-RNA binding prediction. DeepBind [18] and DLPRB [19] utilize convolutional neural networks (CNN) to jointly extract binding preferences from both RNA sequence and structure and demonstrate substantial improvements, compared to previous approaches.
In both deep learning approaches, DeepBind [18] and DLPRB [20], RNA sequence contexts are modeled by a small number of convolutional filters, each resembling a k-mer or a binding site. RNA structural contexts are represented as a probability matrix for structure types (paired (P), hairpin loop (H), inner loop (I), multi-loop (M) or external region (E)) [16], each measuring the thermodynamic average of a structure type of the full ensemble of all possible structures. This matrix can be computed by an RNA folding algorithm with dynamic programming [21]. Models in both DeepBind and DLPRB were trained on the large-scale in vitro RNAcompete dataset. However, in RNAcompete, most synthetic RNAs are unstructured, which makes deep learning methods not effectively identifying RBP-binding structural preferences. Furthermore, RNA structure may be variable or multi-modal according to both theoretical and experimental evidence. In this work, we propose ThermoNet, a thermodynamic prediction model by integrating a new sequence-embedding convolutional neural network model over a thermodynamic ensemble of RNA secondary structures. First, our sequence-embedding convolutional neural network generalizes the existing k-mer based methods by jointly learning convolutional filters and k-mer embeddings to represent RNA sequence contexts. In this way, each k-mer is represented by a low-dimensional continuous vector, and convolutional filters combine a set of k-mer embeddings of consecutive positions and their corresponding structural contexts, thus providing more flexibility and higher expressiveness. Second, the thermodynamic average of structure-specific predictions explores structural variability and improves the prediction especially for the structured RNAs in in vivo datasets. Structural contexts in the high-probability structures are not simply averaged but used for providing structure-specific predictions. Extensive experiments demonstrate that our method significantly outperforms existing approaches, including RCK, DeepBind, DLPRB and several other recent state-of-the-art methods for predictions on both in vitro and in vivo data.
Materials and methods
We introduce a deep learning-based thermodynamic prediction model for protein-RNA binding prediction. The model integrates both sequence information and structural contexts to better capture RBP-binding preferences. As an overview (Fig 1), our method takes an RNA sequence as input and extracts informative features from the sequence and a thermodynamic ensemble of its corresponding secondary structures. A deep convolutional neural network is used to integrate the sequence and structure information and produce the prediction of binding affinity. We have two main contributions here: first, we use a sequence-embedding convolutional neural network (CNN) to represent each k-mer as a low-dimensional continuous vector, which is more flexible and expressive than previous methods that directly apply CNN on nucleotides [18,22] or solely use traditional k-mer based features [17]; second, we use a thermodynamic ensemble of RNA secondary structures to explore the structural variability and provide structure-specific predictions-this differentiates our method from previous Each k-mer is mapped to a low-dimensional continuous vector (called sequence embedding) through an embedding layer to construct a sequence representation of the input sequence. (c) An ensemble of possible secondary structures of the input RNA are sampled using the RNAsubopt function of Vienna RNA package. (d) High-probability structures from the structure ensemble are encoded using one-hot representation separately and the remaining structures are summarized by averaging their one-hot representations into an average structure profile. (e) Each of the high-probability structures is integrated with the sequence embedding by a convolutional neural network (CNN) to produce structure-specific predictions. The average structure profile is also combined with the sequence embedding to generate a prediction. (f) The final predicted protein-RNA binding intensity is obtained by computing the weighted average of individual predictions. (B) Prediction network structure. The convolutional neural network takes the sequence embedding and secondary structure representation as input. The network is composed of multiple 1D convolutional kernels, followed by a batch normalization layer, a ReLU activation layer and fully-connected (fc) layer. The output of the network is the predicted binding intensity for the input RNA. approaches that simply collapse the set of (predicted) secondary structures by taking the average, in which the structure-specific information of high-probability structures may be mixed. We describe the details of our method below.
Input sequence and structure representations
An RNA sequence of length ℓ is a string of ℓ nucleotides over the alphabet {A, C, G, U}. We represent the sequence using the one-hot encoding scheme, where A, C, G and U are represented by [1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], and [0, 0, 0, 1], respectively. The secondary structure of the RNA is a string of the same length ℓ over the alphabet {P, H, I, M, E}, representing the five types of structural contexts, namely, paired (P), hairpin loop (H), inner-loop (I), multi-loop (M) or external region (E). Existing approaches, including DeepBind [16,18], DLPRB [20] and RCK [17], represent the structural context of a position as a five-dimensional distribution vector corresponding to the probabilities of each type of structural contexts predicted by an RNA folding algorithm such as RNAplfold [21]. If only a specific structure is considered, the five-dimensional vector becomes one-hot.
Extracting sequence context features using k-mer embeddings
To better extract informative features from the RNA sequence, we propose a sequence-embedding convolutional neural network that first augments sequence features with k-mers and then uses a convolutional neural network (CNN) to extract higher-order features.
It has been demonstrated that incorporating k-mer based features, in addition to nucleotide based features, can encode larger sequence contexts, model dependencies between binding site positions and thus improve the prediction performance [17]. Here we generalize this approach to account for k-mers with various lengths (e.g., k = 1, . . ., 5). The straightforward way is to simply concatenate k-mers with various lengths together and encode it using onehot representation. However, the dimensionality of this representation increases exponentially (4 k ), which poses computational challenges in the prediction. To address this, we apply the embedding layer that is widely used in deep learning to do dimensionality reduction. The embedding layer can be thought of as a look-up table, which maps the 4 k -dimensional onehot vector to a d-dimensional continuous vector. The values of the d-dimensional vector are not pre-specified but rather learned from the data during the model training process. For an input RNA sequence of length ℓ and a fixed k-mer length k, the output of the embedding layer is a matrix with dimension ℓ × d, in which the i-th row is the low-dimensional representation of the k-mer starting at the position i (k-mers go beyond the RNA sequence length are zero-padded). To account for various k-mer lengths (i.e., k = 1, . . ., k m ), we stack the representations of each k-mer length to produce a sequence representation matrix E with dimension ℓ × k m d where k m d � 4 k . These low-dimensional representations of k-mers will be integrated by a CNN to further capture the higher-order dependencies in the RNA sequence (described below).
Compared with existing approaches, our method has more flexibility and higher expressiveness in extracting sequence features: Unlike RCK [17] that uses k-mer based features with only a specific length (k = 5), our method is more flexible in that it can handle various lengths of kmers through a dimensionality reduction process, capturing multi-resolution local sequence information. cDeepbind [22] directly applies CNN on nucleotide based input features, while our method incorporates k-mer based features and then uses a CNN on top of the k-mer embeddings to further extract higher-order features, providing richer information from the RNA sequence.
Incorporating thermodynamics contexts using an ensemble of structures
Previous studies have shown that RNA secondary structure information can provide additional prediction power in predicting protein-RNA binding [17,20,22]. These methods use the RNA secondary structure context profiles predicted by RNAplfold [16,21], which is a fivedimensional vector of the probabilities of five contexts (paired, hairpin loop, inner loop, multiloop or external region) for each position in the sequence. The structure context profile can be thought of as an 'average' of all possible secondary structures. However, informative structural features of high-probability structures may be buried in this uniform average. To enable structure-specific prediction, we propose to sample an ensemble of possible structures for a given RNA sequence. Each of the high-probability structures is then integrated with the sequence features to give structure-specific predictions separately, and the final prediction is obtained by combining the structure-specific predictions.
Specifically, for a given RNA sequence, we sample N possible secondary structures using the RNAsubopt function of the ViennaRNA package [21], which draws a specific structure with probability proportional to its Boltzmann energy. In our experiment, we choose N = 100 for computational consideration and we also found that an ensemble of this size generally reflects the distribution of the secondary structures of an RNA sequence. We then identify high-probability structures by counting the occurrence for each unique structures of the N sampled structures. Denote the U unique structures by R 1 , R 2 , . . ., R U and their associated frequency by m 1 , m 2 , . . ., m U , where R i (1 � i � U) is a ℓ × 5 one-hot matrix that represents the structural contexts of the i-th sampled structure. Here we assume the unique structures are sorted in descending order of their frequencies, i.e., m 1 � m 2 � . . ., �m U . The top T structures are considered as high-probability structures. The choice of the value of T is a joint consideration of structures diversity and computational cost: we want to include more high-probability structures, while also prevent T being too large otherwise the model training and prediction would be inefficient. Based on the analysis of the frequency histogram, we found T = 5 is a good balance between structures diversity and computation efficiency for both in vitro and in vivo datasets (S1 Fig and S1 Table). For example, in the in vitro dataset, the top 5 unique structures are highly frequent, contributing *50% frequencies in all sampled structures. To reduce noise, we thus consider the top T structures as high-probability structures, while for the remaining U − T structures, we summarize their structure information by computing an average structure profile Next, we build a series of rank-specific neural networks (described below) f y 1 , f y 2 , . . ., f y T , parameterized by θ 1 , . . ., θ T , respectively. Each neural network takes as input one of the highprobability structures R i as well as the sequence embedding E to produce a structure-specific prediction (a binding intensity). We also build an additional neural network f y Tþ1 that makes the prediction based on the average structure profile and the sequence embedding. Predictions of each individual neural network are combined as a weighted average to give the final prediction f(s) for an RNA sequence s, Note that previous methods like cDeepbind [22] make predictions using only the average structure profile as input, in which the structure-specific information may not be revealed. In contrast, our method, while accounting for the average structure profile, also explicitly teases high-probability structures apart from the uniform average profile to enable structure-specific predictions.
Convolutional neural network as base predictor
Deep learning continues to proliferate as a powerful set of tools to solve an increasingly diverse range of problems, including many in structural and systems biology [23][24][25][26][27][28][29]. We use the convolutional neural network (CNN) as the base model f θ to predict protein-RNA binding, where θ represents the network weights. CNNs are useful when local groups of data with high correlation and forming distinct patterns by combining lower level features and capturing more complex feature dependencies in the sequence input. In our model, we stack five convolutional layers. Every layer contains several one-dimensional convolutional kernels, each of which produces an output with the same size of the input. After applying the batch normalization technique [30] to the intermediate output from kernels, we fed it to the ReLU activation function [31] and give the output to the next convolutional layer. Following the last convolutional layer, we append a fully-connected layer to produce the predicted binding intensity of the protein-RNA binding. The overall structure is depicted in Fig 1B. Please refer to Results section for parameter-tuning details. In order to improve the training stability of ThermoNet, we train the T + 1 models individually. Each model f y i was trained using stochastic gradient descent to find the parameters of the embedding layer and CNNs that minimize the following loss function where y s is the (normalized) binding affinity of the RNA s to a given protein, f y i ðE; R i Þ is the predicted binding intensity by the model, and L Huber () is the Huber loss, which is defined by L Huber ðy s ; f ðsÞÞ ¼ We choose Huber loss as it is less sensitive to outlier data points than the squared error loss.
Implementation
We performed the grid search to find the optimal hyperparameters of our model and chose the one with the lowest validation cost using a two-fold cross-validation on the training data.
In particular, we grid-searched the initial learning rate in {0.001, 0.0001, 0.00001}, the filter lengths in {7, 12, 16}, the number of convolutional kernels in {16, 64}, the regularization coefficient of the L 2 norm in {0.001, 0.0001, 0.00001}, the size of sequence embedding vector in {10, 20, 30} and the maximum k-mer length k m in {2, 3, 4, 5}. A total of five convolutional layers were stacked in the CNN. All CNN parameters were initialized by Xavier initializer [32]. We chose Huber loss to define the loss function and Adam [33] as the optimizer for training the network. We reduced the learning rate to one-tenth of the previous one after the 5th epoch.
Evaluation of prediction performance on in vitro data
We first assessed the prediction ability of our method using in vitro protein-RNA binding dataset. We used the comprehensive RNAcompete dataset [8] which includes binding intensities for over 240,000 sequences across 244 experiments. The dataset consists of a training set of sequences (set A) and a test set (set B). For each experiment, we trained a model on sequences extracted from set A and then predicted the intensities for sequences from set B. To evaluate the performance, we used the Pearson correlation between predicted and measured intensities on set B as the metric. Similar to Deepbind [18], we removed outlier intensities in the dataset: we clamped all intensities greater than the 99.5 percentile to the 99.5 percentile. Furthermore, the scores were normalized to have a mean of zero and a variance of one. We compared our method against four existing methods that use both sequence and structure information for protein-RNA binding prediction, including two probabilistic model based method-RNAcontext [16] and RCK [17], and two deep learning algorithms-DLPRB with the CNN network (DLPRB-CNN) [19] and cDeepbind with the CNN network (cDeepbind-CNN) [22]. We found that our method consistently outperformed all other methods (Fig 2A) [17] and DLPRB-CNN [20], respectively. The p-values were calculated using one-sided Wilcoxon singed rank test.
respectively. The relative improvements achieved by our method over the for methods were 54.46%, 45.87%, 10.76% and 32.58%, respectively. In a pairwise comparison to the best performing probabilistic model RCK and the best performing deep learning model DLPRB-CNN (Fig 2B and 2C), we observed that our method achieves significantly higher Pearson correlation than the two baseline methods (Wilcoxon signed rank test p-value 4.50 × 10 −42 and 4.82 × 10 −38 , respectively). Moreover, we found that the improvements achieved by Thermo-Net are not biased toward certain predominant RNA binding domains (S2 and S3 Figs). For example, investigating the histogram of relative improvements achieved by our method (S4 Fig), we found that the subset of RBPs with the most noticeable improvements covered a wide range of RNA binding domains. Raw numbers of prediction performance of all methods in this experiment can be found in S2 Table.
Evaluation of prediction performance on in vivo data
We then assessed the performance of ThermoNet on in vivo protein-RNA binding dataset. We obtained a large-scale dataset of RBP binding sites from a previous work [34], including a compendium of 31 published CLIP-seq experiments on 19 RBPs. In this dataset, nucleotide positions with the highest cDNA counts were first identified as a pool of positive samples. A random sampling process preventing consecutive genomic positions was then applied to reduce the redundancy. Negative sites were randomly sampled from genes that were not detected as positive sites in any of the 31 experiments. Each experiment consists of 40,000 samples divided into a training set of 20,000 samples and a test set of 20,000 samples. The length of each RNA sequence is 101nt. Unlike the in vitro dataset that gives a real-valued binding intensity for every binding event, the in vivo dataset only gives a binary label (0 or 1) based on the cDNA counts to indicate whether a protein binds to an RNA or not. Therefore, we treated the prediction for in vivo data as a binary classification problem and used AUROC (Area Under the Receiver Operating Characteristics) as the evaluation metric.
We compared ThermoNet against two state-of-the-art deep learning algorithms, DeepBind and DLPRB-CNN (Fig 3), using the average of the AUROCs over the 31 experiments as the evaluation metric. We found that ThermoNet outperformed the other two methods, achieving an average AUROC of 0.864 against 0.835 for DLPRB-CNN and 0.836 for DeepBind. In a Integrating thermodynamic and sequence contexts improves protein-RNA binding prediction pairwise comparison of ThermoNet to DeepBind and DLPRB-CNN over the 31 experiments, we observed that the improvement gaps between ThermoNet and the two methods were also statistically significant (Wilcoxon signed rank test p-value 4.53 × 10 −6 and 2.11 × 10 −6 for DLPRB-CNN and DeepBind, respectively). Raw numbers of prediction performance of all methods in this experiment can be found in S3 Table.
Ablation analysis of ThermoNet
Having demonstrated the superior prediction ability of ThermoNet for both the in vitro and in vivo bindings, we proceeded to perform an ablation analysis to explain the sources of performance improvements achieved by ThermoNet. We built several variants of ThermoNet to investigate the importance of different novel designs in ThermoNet. The variants include i) "1-struc": using only 1-mer and the average structure profile as input; ii) "k-no-struc": using kmers of various lengths and without using structure information as input; iii) "k-struct": using k-mers of various lengths and the average structure profile as input; and iv) "k-struc-sampling": the full model, i.e., using k-mers of various lengths, individual secondary structures and average structure profile as input. All variants were trained and tested on the aforementioned in vitro dataset. We evaluated the performance of each variant using the average Pearson correlation over the 244 experiments in the in vitro dataset and the 31 experiments in the in vivo dataset (Fig 4; Detailed results in S4 Table).
We first observed that the k-struc model (green) improves the 1-struc model (blue) with a pronounced gap (average correlation 0.663 compared to 0.620 for the in vitro dataset and AUROC 0.861 compared to 0.856 for the in vivo dataset). Note that previous methods such as DeepBind, cDeepbind and DLPRB only used single nucleotide based sequences. The improvements of our method over these methods can thus be partially explained by the richer sequence information we have captured in the k-mer embedding process of our method.
In addition, the performance gap between the k-struc model (green) and the k-no-struc model (brown) was not significant: on the in vitro dataset, the k-struc model achieved an average correlation of 0.663 as compared to 0.661 for the k-no-struc model; on the in vivo dataset, both models achieved an AUROC of 0.861. The only difference between these two models is that k-struc used the average structure profile but k-no-struc did not. As the average structure profile is a uniform average of all sampled secondary structures, informative features of highprobability structures that are useful for protein-RNA binding prediction may not be revealed by the average profile, and thus the performance improvement was not significant. The small performance gap between the two variants may also be explained by the high expressiveness of the sequence embeddings, which already capture the information that the average structure profile can provide for the binding prediction.
Moreover, we found that adding individual high-probability structures as input (the kstruc-sampling model) further improved the prediction performance of the k-struc model (average correlation 0.671 compared to 0.663 for the in vitro dataset and AUROC 0.864 compared to 0.861 for the in vivo dataset). This result highlights the effectiveness of using the thermodynamic ensemble of secondary structures in ThermoNet. The unique features of structured RNAs revealed by the high-probability structures and the thermodynamic average of structure-specific predictions have resulted in a substantial improvement over existing methods.
Discussion
We have introduced ThermoNet, a deep learning-based thermodynamic model for protein-RNA binding prediction. ThermoNet incorporates the thermodynamic and sequence contexts by integrating a sequence-embedding convolutional neural network over a thermodynamic ensemble of RNA secondary structures. The model both explores structural variability and captures the higher-order dependencies in the RNA sequences, providing richer information and higher expressiveness for protein-RNA binding prediction. In addition, highprobability structures are utilized in ThermoNet to better extract informative structure features that enable structure-specific predictions. We compared ThermoNet to multiple stateof-the-art methods for protein-RNA binding prediction on both in vitro and in vivo binding datasets, and ThermoNet has achieved substantial improvements over other methods on both datasets. Ablation study performed on a series of ThermoNet also demonstrated the effectiveness of multiple novel designs in ThermoNet that lead to improved prediction performance.
ThermoNet is a flexible and scalable model that can be applied to a broad range of RNAbinding proteins. We expect ThermoNet to be an effective tool in practice for identifying novel binding sites for RBPs. In addition to its direct application in protein-RNA binding prediction, multiple novel model designs of ThermoNet can also be used as a stand-alone tool in other applications. For example, the sequence-embedding convolutional neural network can also be applied to understand signals in biological sequence data of other molecular events, including transcription factor (TF) binding, DNA accessibility, and histone modification. Interpreting the ThermoNet model is one of the directions worth pursuing in future work. A better understanding of what the deep learning model has learned, for example, what sequence and structure motifs contribute to protein-RNA binding, may reveal new biological insights. A great challenge of protein-RNA binding prediction is the in vitro-to-in vivo generalization. It was observed in previous works [17,20] that a model trained on in vitro data did not perform very well on in vivo data, possibly due to the biases of different experimental protocols [35]. We think the generalization requires not only a generalizable prediction model but also an in-depth characterization of the in vitro and in vivo datasets. Achieving a robust in vitro-to-in vivo generalization in protein-RNA binding prediction is also one of our future directions.
Supporting information S1 We compute the relative improvement over DLPRB-CNN achieved by our method for each experiment in the RNAcompete dataset. These relative improvements are sorted from the largest to the smallest and then discretized into bins of percentiles (0 th , 10 th ], (10 th , 25 th ], (25 th , 50 th ], (50 th , 100 th ]. The normalized counts of each RNA-binding domain within each bin are shown in the histogram. Note that the S1 binding domain has only one protein hence the relative count of the S1 domain in the (0 th , 10 th ] bin is 1.0. (TIF) | 2019-09-06T13:06:03.466Z | 2019-09-01T00:00:00.000 | {
"year": 2019,
"sha1": "6bed375adc563386608c45240d85fbbdefa0557d",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/ploscompbiol/article/file?id=10.1371/journal.pcbi.1007283&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "aea45e06b4ca1007d15318ebf31a16f28d35ba57",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
248841476 | pes2o/s2orc | v3-fos-license | Decidual stromal cell therapy for generalized lymphadenopathy as a special clinical manifestation of COVID‐19 infection: A case report
Abstract We are going through the greatest global health crisis of the last decades, the coronavirus disease 2019 (COVID‐19) pandemic. It may cause morbidity and mortality in some cases, and there is no therapeutic approach with reproducible and favorable outcomes. As clinical manifestations differ from patient to patient, any report regarding clinical symptoms has been beneficial for early detection and treatment. Due to the immunomodulatory effect of mesenchymal stem cells (MSCs), MSCs‐based therapy has been approved to be one of the therapeutic strategies for COVID‐19 management. For the first time in the literature, we reported generalized lymphadenopathy with fever and no sign of respiratory distress in a 16‐year‐old patient with confirmed COVID‐19 infection as the main clinical signs. We also introduce decidual stromal cells as a potential immunomodulatory treatment for COVID‐19–infected patients.
cause of the widespread transmission of COVID-19, resulting in its pandemic. In the absence of widely distributed vaccines, mobility and gathering restrictions were the primary strategy to diminish the COVID-19 transmission. 5 Treatment strategies for COVID-19 can be classified into three categories: (i) antiviral drugs, including Lopinavir/ritonavir, Darunavir/cobicistat, Remdesivir, Favipiravir, and Umifenovir; (ii) immunomodulatory drugs, like Chloroquine/hydroxychloroquine, Corticosteroids, Tocilizumab, Sarilumab, Anakinra, Emapalumab, Baricitinib, Mavrilimumab, Canakinumab, Colchicine, and interferons; and (iii) other drugs, such as neutralizing antibodies (intravenous immunoglobulins, hyperimmune plasma), low-molecular-weight heparin and unfractionated heparin, antibiotics (Azithromycin, Doxycycline), angiotensin receptor blockers and Angiotensin convertase enzyme inhibitors, and TMPRSS2 is a serine protease. The administration timing of these therapeutic agents is critical for their efficiency. For instance, although immunomodulatory drugs induce more favorable outcomes during the cytokine storm phase, antiviral agents should be ordered shortly after symptoms onset. 6,7 Mesenchymal stromal cells (MSCs) are wellknown immunomodulatory cells with promising results in controlling hyperinflammatory conditions. 8 Decidual stromal cells (DSCs) are MSCs derived from the placenta and fetal membrane. 9 Compared to the bone marrow-derived MSCs, DSCs have much stronger anti-inflammatory capacities that candidate them for cell therapy of inflammatory diseases. 9 Their safety and promising results in controlling inflammatory disorders, such as graft-versus-host disease, neural inflammation, and acute respiratory distress syndrome (ARDS) [9][10][11][12] have led us to apply them in the case of COVID-19-related hyperinflammation.
The clinical observations among COVID-19-infected patients may be surprisingly different, complicating the treatment process. Hence, numerous global reports about the diversity of COVID-19 clinical symptoms can be beneficial for early detection and on-time treatment. Here, we report a COVID-19-positive patient admitted to Taleghani hospital, Shahid Beheshti University of Medical Sciences, Tehran, Iran, with unusual generalized lymphadenopathy and no respiratory complications. The case is successfully treated with two intravenous doses of DSCs.
| CASE PRESENTATION
A 16-year-old boy who had suffered from fever (38.5 °C) for the past 24 h was admitted to the Taleghani Hospital, Tehran, Iran, on May 12, 2020. The patient showed no symptoms other than fever and generalized lymphadenopathy at the admission time. The CBC result showed a reduced WBC, Hb level, and MCV and MCH indexes; conversely, neutrophil count exhibited an abnormal increase (Table 1). However, the arterial blood gases were normal, and no sign of respiratory distress was observed at the admission time. Some inflammatory markers such as IL-6, CRP, ESR, CPK, and LDH, coagulation factors like troponin, D-dimer, and fibrinogen, and some biochemical parameters including ferritin, AST, ALT, and ALP were evaluated at the admission T A B L E 1 The results of CBC and oxygen saturation before DSC administration and after the first and second injection (Table 2). IL-6, CRP, ESR, CPK, LDH, and D-dimer parameters were higher than the reference range at the admission day, while troponin, fibrinogen, ferritin, AST, ALT, and ALP parameters were in the range. Initial medical examinations were performed on May 14, 2020, to assess lymphadenopathy by sonography of the neck, bilateral inguinal, abdomen, and pelvis. The results showed a few reactive lymph nodes at the jugular chain and bilateral submandibular; the most prominent lymph nodes with short-axis diameters (SAD) of 11 and 12 mm were found in the right and left submandibular areas, respectively. Besides, several reactive lymph nodes were observed at the axillary and inguinal regions with SAD of 10 and 12.5 mm, respectively.
T A B L E 2
The results of some inflammatory markers, coagulation assays, and biochemical tests before DSC administration and after the first and second injection On May 14, 2020, the spiral chest CT scan showed the bilateral multifocal ground-glass opacities with remarkable infiltrations in the lungs that strongly suggested COVID-19 infection as the main diagnosis. Furthermore, a mediastinal CT scan indicated hepatosplenomegaly, and bilateral axillary and bilateral lymphadenopathy with a maximum SAD of 12 and 15 mm, respectively ( Figure 1). However, COVID-19 real-time PCR tests on nasopharyngeal and oropharyngeal samples were negative. The presence or absence of viral infection must be confirmed regarding the patient's symptoms and lymphadenopathy. Eventually, SARS-CoV-2 RNA was detected in the bronchoalveolar lavage sample using real-time PCR, which confirmed SARS-CoV-2 infection.
As the generalized lymphadenopathy has not been reported as a clinical sign in COVID-19-positive patients, we considered other bacterial and viral infections or hematopoietic malignancies, including lymphoma, as the possible explanations for this phenomenon. In this light, the patient was confirmed to be negative for the varicella zoster virus, toxoplasmosis, cytomegalovirus (CMV), hepatitis B and C virus (HBV and HCV), human immunodeficiency virus (HIV), and Epstein-Barr virus. The transbronchial needle aspiration procedure was carried out for histological evaluations to find tumor cells within the lymphoid tissues. Unexpectedly, no sign of neoplasia was observed in the histological analysis.
The COVID-19 treatment with low-dose corticosteroids, Lopinavir/ritonavir (Kaletra) 200/50 mg, Hydroxychloroquine 400 mg, and azithromycin 250 mg was started following the COVID-19 positive result. However, high-dose dexamethasone was not prescribed, considering the possibility of neoplasia. Despite the COVID-19 treatment, the blood O 2 saturation decreased, lymphadenopathy and fever were not ameliorated, and the patient experienced a significant weight loss and disease progression.
On May 15, 2020, the patients underwent cell therapy with decidual stromal cells (DSCs). DSCs were isolated from the fetal membranes under GMP conditions based on a previously described protocol. 9 After providing informed consent, placenta tissues were obtained from healthy mothers who underwent elective cesarean section. The donors were screened for viral infections, including HIV, HBV, HCV, and CMV viruses. After transferring to the laboratory under sterile conditions, the fetal membranes were isolated, washed with PBS, dissected, and transferred into the 50 ml falcons containing trypsin/ EDTA for digestion and DSCs isolation. Then, the isolated cells were cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FSC). The fetal membrane residues were cultured in the T175 flasks containing DMEM to form the fibroblast-like cell colonies. Eventually, the cultured cells were harvested using trypsin/EDTA when they reached 90-95% confluence, then washed with complete DMEM, and cultured in new T175 flasks containing DMEM supplemented with 10% FSC. Isolated DSCs were expanded until passage five (P5). For DSCs characterization, the expression of CD105, CD90, CD73, CD44, CD45, and CD34 surface markers was examined using flow cytometry. The DSCs' viability and the specimens' microbial contaminations were checked prior to injection. 100 mg hydrocortisone was injected about half an hour before each DSC administration to prevent the side effects of the cell therapy. Two doses of 1-1.2 × 10 6 cell/kg, suspended in 0.9% NaCl injection solution, were injected into the central venous line within 5 min with 4-day interval. The vital signs were monitored every 3 h within the first 24 h following DSC injection and every 6 h within 48 h later. The patient presented no complications, even dizziness, and the most common adverse effect of cell therapy.
Evaluation of IL-6, as one of the main inflammatory cytokines, showed a significant decline following cell injection, where its concentration decreased from 12.38 pg/ ml before cell therapy to 6.2 pg/ml after cell therapy ( Figure 2). The CT scan results after the first and second doses of DSCs indicated the positive impact of this medical intervention.
On May 20, 2020, the patient was discharged from the hospital in good general condition. The PET/CT examination was performed on July 5, 2020, to rule out any malignancies, which showed no sign of lymphoproliferative disorders in the patient (Figure 3). The results of CBC analysis, oxygen saturation test, inflammatory markers (IL-6, CRP, ESR, CPK, and LDH), coagulation assays (troponin, D-dimer, and fibrinogen), and some biochemical tests (ferritin, AST, ALT, and ALP) before DSC administration and after the first and second DSC injection were presented in Tables 1 and 2. WBC count and the level of IL-6, CRP, ESR, CPK, LDH, and Ddimer were increased after DSC therapy.
| DISCUSSION
The clinical manifestations of COVID-19 include a wide range from mild to severe, including fever, tiredness, dry cough, and chest distress which may develop ARDS and even multiple organ failure. Lymphadenopathy is the less common symptom, along with headache, diarrhea, skin rashes, and lymphadenopathy. 2 Although CT scan has a leading role in COVID-19 diagnosis and management, 13 few studies have reported the mediastinal lymphadenopathy in COVID-19-positive patients as an atypical CT feature. Shi et al. detected lymphadenopathy in 6% of radiological analysis of 81 COVID-19-infected patients. 14 Valette et al. reported mediastinal lymphadenopathy in the six cases of nine COVID-19-positive patients. They suggested a relationship between mediastinal lymphadenopathy and COVID-19 infection severity. Accordingly, the patients with critical conditions exhibited larger mediastinal lymph nodes. 15 Lymphadenopathy could have several etiologies, including malignancies, infection, autoimmune disorders, and iatrogenic causes. 16 In this case, after ruling out the malignancy, COVID-19 real-time PCR and chest CT scan results suggested that SARS-CoV-2 infection was the reason for the generalized lymphadenopathy. The standard of practice in managing lymphadenopathy is to find and treat its underlying cause. We started low-dose corticosteroids, which did not improve the patient's condition and deteriorated the patients' symptoms. Although malignancy was ruled out, and since the generalized lymphadenopathy is not a typical symptom of COVID-19, we preferred not to prescribe high-dose corticosteroids due to the minimum possibility of malignancies. Therefore, the immunomodulatory approach was preferred over immunosuppressive treatments in hyperinflammatory conditions.
Regarding immunomodulatory and tissue-repairing effects of MSCs and trapping of a considerable part of infused MSCs in lungs, 16 they are considered promising choices for cell therapy of COVID-19, where hyperinflammation and lung tissue degeneration are the main causes of mortality and morbidity. 17 Numerous clinical trials are currently evaluating the efficacy of MSCs in controlling the cytokine storm and hyperinflammation of COVID-19, and 17-20 their safety is approved in several studies. 18,21,22 Several studies reported the effectiveness of DSCs, stromal cells with significantly stronger anti-inflammatory properties, 23 in controlling the hyperinflammatory disorders. First, Ringden et al. reported the safety and efficacy of DSCs and their significant privilege in alleviating inflammation. 11,12,23,24 This case report is the first report of using DSC infusion in a COVID-19-infected patient with a hyperinflammatory condition. Although a one-patient case report could not be conclusive, our results encourage using DSCs as an immunomodulatory element to attenuate hyperinflammation in COVID-19. The patients' lymphadenopathy was completely cured, and no abnormality was found after 1 year of monitoring. Since the patient was young, it can be considered that his improvement might be due to the clinical course of the disease rather than the DSC therapy. Since the conventional treatments for COVID-19 did not ameliorate the fever, lymphadenopathy, and weight loss and even worsened the O 2 saturation, it could be concluded that DSCs might, at least in part, have led to his improvement.
| CONCLUSION
Despite the ineffectiveness of the conventional therapies for COVID-19, DSCs showed a favorable effect on controlling the COVID-19 symptoms in the presented case. However, several controlled trials are required to prove the safety and efficacy of DSC therapy in severe COVID-19.
AUTHOR CONTRIBUTIONS
Elham Roshandel contributed to conceptualization; Abbas Hajifathali contributed to methodology; Sedigheh Sadat Akhlaghi contributed to formal analysis and investigation; Behrouz Farhadi Hosseinabadi contributed to writingoriginal draft preparation; Behnam Sadeghi contributed to writing-review and editing; Masoud Soleimani contributed to funding acquisition; Ali Pirsalehi contributed to supervision. All authors revised the manuscript and approved the final paper. | 2022-05-18T15:11:56.106Z | 2022-05-01T00:00:00.000 | {
"year": 2022,
"sha1": "5b3bfdddeebbc02c9546cbaac3dbb5d63befe728",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/ccr3.5851",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "aaad4b66b9f8c8a80f74ac33c4d51e7544f5d5f8",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
219563165 | pes2o/s2orc | v3-fos-license | Retinal Nerve Fiber Layer Thickness and Oxidative Stress Parameters in Migraine Patients without Aura: A Pilot Study
Background: Migraine is one of the most common disorders and its pathophysiological mechanisms are still under research, oxidative stress being emphasized as an important contributor. This study aimed to analyze the retinal nerve fiber layer (RNFL) thickness and oxidative/anti-oxidant balance in migraine patients. Methods: Two groups of subjects were evaluated: a group of patients with migraine and a control group of healthy volunteers. RNFL thickness was assessed for all subjects by the ocular coherence tomography spectral domain (OCT-SD). The oxidative stress parameter, namely nitric oxide (NOx), malondialdehyde (MDA), and total oxidative stress (TOS) were assessed. The antioxidant capacity of plasma was evaluated by assessing the level of catalase, and total anti-oxidative (TOS) capacity. Migraine severity was graded using the Migraine Disability Assessment Score (MIDAS) questionnaire. Results: All the oxidative stress parameters (NOx, MDA, and TOS) were significantly increased, and both parameters for anti-oxidative status were significantly decreased in the migraine group compared with the control group (p < 0.0001). Significant correlations with all the quadrants and different oxidative stress parameters were found, most involved being temporal quadrant. A significant positive correlation between catalase and macular RNFL thickness (inner ring, temporal quadrant) in migraine patients, for both eyes, was observed (p = 0.014 for the right eye and p = 0.12 for the left eye). Conclusion: The assessment of the oxidative stress/anti-oxidative balance together with RFLN thickness can constitute a promising method to evaluate the progression of the diseases. It can also contribute to the estimation of the efficiency of various therapies targeting oxidative stress and associated inflammation.
Introduction
Migraine is a common chronic disorder characterized by an episodic headache accompanied by various neurological, gastrointestinal and/or autonomic disturbances. A relevant percent of the population (up to 21% of women and 6% of men) suffer from migraine attacks associated with activity area [10]. The changes of RNFL thickness in migraine patients proved to be associated with the history of disease onset (number of years), the severity of disease (Migraine Disability Assessment Score (MIDAS)), the involvement of different retinal quadrants or chronic migraine onset [9,[31][32][33][34]. This study aimed to assess peripapillary RNFL thickness and macula thickness by OCT examination, oxidative stress parameters, and the relationship between oxidative stress and OCT results in migraine without aura patients and with no prophylactic medication.
Materials and Methods
A prospective, comparative, and analytical study was performed on a migraine group as compared to an age-matched group of healthy volunteers. The study protocol was approved by the Institutional Ethics Committee of the Rehabilitation Hospital, Cluj-Napoca, Romania (IECGH 4/2018). All participants included in the study signed prior to clinical examination, blood sampling, and OCT examination the written informed consent. The patients were recruited from the Neurology Clinic of Rehabilitation Hospital, Cluj-Napoca, Romania, from January to December 2018. The controls were recruited from the same hospital in the same period of time among the subjects that call for a visual check after the exclusion of those who recall migraine attacks. The Declaration of Helsinki specifications were followed. All the participants signed informed consent for participation in this study. The assessment was made in the first 24 h after the migraine attack onset.
The neurological and ophthalmological examination was done to all subjects included in the study. In the neurological examination, all the patients in the study group were selected following diagnostic criteria established by the International Headache Society (ICHD-2) [35]. Migraine severity was assessed by the neurologist using a Migraine Disability Assessment Score (MIDAS) questionnaire [36]. MIDAS score was divided into four grades: grade I (infrequent disability, score from 0 to 5), grade II (mild disability, score from 6 to 10), grade III (moderate disability, score from 11 to 20), and grade IV (severe disability, score >20). The duration of disease expressed in years and the number of attacks per month were established by anamnesis. After the neurological examination, patients with epilepsy, cerebrovascular or cardiovascular diseases, neurodegenerative diseases, hypertension, or diabetes mellitus were excluded. Patients have undergone a complete ophthalmological examination at the Ophthalmological Department of the Rehabilitation Hospital. The ophthalmologic examination included best-corrected visual acuity tested with Snellen chart, slit lamp biomicroscopy, dilatated fundus examination, intraocular pressure measurement, and visual field examination by automated perimeter (Humphrey Visual field Analyzer 750i, Zeiss, Germany). Ophthalmological examination excluded patients with glaucoma, optic nerve diseases, retinal vascular diseases, pre-proliferative or proliferative diabetic retinopathy, macular degeneration, hereditary retinal dystrophy. None of the patients with migraine had prophylactic medication prior to the presentation, and they only used triptans for migraine attacks. Other exclusion criteria included obesity, oral contraceptive medication, glucocorticoid medication, and anti-oxidant agents' treatments.
Retinal Fiber Layer Thickness Assessment by Optical Coherence Tomography
All patients and controls who met the neurological and ophthalmological criteria have been examined with ocular coherence tomography spectral domain (OCT-SD) (DRI Triton, Topcon, Tokyo, Japan) after pupillary dilatation, for both eyes. The software automatically recognizes eight retina layers. Peripapillary retinal fiber layer thickness (RNFL) and macular thickness (M) were assessed. Using macular thickness analysis mode and through a software algorithm, the device automatically constructed a topographic surface map (central field-1 mm centered on the fovea, inner ring bounded by 1 to 3 mm concentric circles, and outer ring bounded by 3 to 6 mm concentric circles). All fields were examined for both eyes: peripapillary RNFL thickness (superior, temporal, inferior and nasal quadrant), foveal thickness, inner macular thickness (superior, temporal, inferior, and nasal quadrant), and outer macular thickness (superior, temporal, inferior and nasal quadrant). The average macular thickness was manually determined [37].
Oxidative Stress Parameters
Blood samples were taken from all the subjects included in the study, in the morning, after fasting for 12 h, in the first 24 h after admission into the hospital for a migraine attack. The samples were assessed regarding the oxidative status and antioxidant capacity of plasma. The oxidative status evaluation consisted of the determination of total oxidative stress (TOS) using the method of Erel et al. [38], malonyl dialdehyde (MDA) using the method of Ohkawa et al. [39] and nitric oxide (NOx) using the method of Giustarini et al. [40]. Total NOx levels were used as a surrogate marker for serum nitric oxide levels [40]. The antioxidant capacity of plasma was assessed by measuring the level of catalase [41], and total anti-oxidative (TOS) capacity [42]. Spectroscopic measurements assessed the oxidative stress parameters. A Jasco V-350 UV-VIS spectrophotometer (Jasco International Co., Ltd., Tokyo, Japan) was used for all measurements. All the chemical substances were purchased from Sigma-Aldrich.
Statistical Analysis
Statistical analysis was done with Statistica program (v. 8, StatSoft, Tulsa, OK, USA), and p-values less than 0.05 were considered statistically significant. The numerical data were assessed for normal distribution and were summarized as mean (standard deviation) if normally distributed, respectively, and median (Q1 to Q3) (where Q is the quartile) otherwise. An independent t-test or Mann-Whitney test was used to evaluate differences between groups according to data distribution. The qualitative data were reported as a number, percentage and 95% confidence intervals (values provided in squared brackets) calculated with an exact method [43] and analyzed by the Chi-square test. For correlation analysis, Pearson's or Spearman's methods were used according to the distribution of data.
Twenty-nine (70.7%) women were included in the group with migraine and twenty-three (63.9%) in the control group, without significant differences between groups (χ 2 = 0.4, p = 0.5223).
The time in years from the migraine diagnostic varied from 3 to 8 years with a median of 5 years (IQR = (4-6) in the group with migraine. The frequency of migraine attacks per month varied from 2 to 8, with a median of 4 (3)(4)(5)(6).
A different pattern was observed according to genders regarding the MIDAS score, with significantly higher frequency of MIDAS I and II score among men, and MIDAS III and IV score among women (Fisher exact test; p = 0.0006).
A different pattern of oxidative stress parameters was observed between groups with significantly higher values of NOx, MDA, and TOS, and, respectively, significantly lower values of catalase and TAC among the migraine group as compared to the control group (Table 1). No significant differences regarding the investigated oxidative stress parameters were observed between genders, neither in the migraine group (Mann-Whitney test, p > 0.18) nor in the control group (Mann-Whitney test, p > 0.31).
The thickness of the peripapillary retinal nerve fibers proved to be, without any exceptions, significantly smaller in the migraine group as compared to the control group ( Table 2).
The averages of macular RNFL thickness for the outer and inner ring were significantly reduced in patients with migraine compared with the control group with a few exceptions: an inferior quadrant of the outer ring of the fovea for the left eye, an inferior quadrant of the inner ring and a superior quadrant of the outer ring of the fovea for the left eye ( Table 3).
The most frequent significant relationships were observed with catalase, and the most involved quadrant in these correlations was the temporal quadrant that showed a strong positive correlation with macular RNFL thickness, inner ring in both eyes ( Table 4). The strongest positive correlation was also found for catalase, and peripapillary RNFL thickness nasal quadrant left eye (p = 0.007) ( Table 4).
Discussion
Our study identified increased values of the oxidative stress parameters (NOx, MDA, and TOS), decreased values of both parameters for anti-oxidative status, decreased peripapillary retinal nerve fiber layer thickness as well as macular RNFL thickness among those with migraine as compared to those without migraine. Furthermore, several significant correlations were identified between oxidative stress parameters and, respectively, catalase on the one hand and peripapillary and macular thickness, on the other hand.
A higher percentage of women with migraine was observed in our study and this result is expected since female sex is considered a risk factor for migraine onset as well as for its transformation into a chronic disorder [44]. The prevalence of migraine in the general population is reported as three-times more in women than in men, especially during the reproductive years [45], our results being closed to this ratio. Furthermore, a more significant disability associated with female gender is reported [46], this also being observed in our study with a higher disability according to MIDAS score among women as compared to men. The role of estrogens in migraine risk and clinical characteristics, with decreasing levels in the premenstrual period associated with menstrual-related migraine, had already been demonstrated [45]. Gender is also associated with differences in oxidative stress, women appear to be less susceptible to oxidative stress due to the antioxidant properties of the estrogens [47,48].
Migraine and Oxidant/Antioxidant Balance Assessment
Oxidative stress is a plausible unifying principle behind the types of migraine triggers encountered in clinical practice [49]. All of the parameters that reflect the oxidant/anti-oxidant balance in our study (NOx, MDA, TOS, catalase, and TAC) were found to be significantly modified in migraine patients compared with the control group (Table 1). Endogenous reactive species such as reactive oxygen species (ROS) and reactive nitrogen species (RNS) are emerging as molecules that can mediate cell signaling responses with the activation of Ca 2+ -permeable membrane channels, encoded by the transient receptor potential (TRP) gene superfamily that are characterized by a wide variety of activation triggers that act from outside and inside on the cells [50]. Members of one class of TRP channels have emerged as sensors not only for reactive oxygen species (ROS), or reactive nitrogen species (RNS) but also for reactive carbonyl species (RCS), and gaseous messenger molecules including molecular oxygen (O 2 ), hydrogen sulfide (H 2 S), and carbon dioxide (CO 2 ) [51]. Transient receptor potential ankyrin-1 (TRPA1) ion channels, found on pain-sensitive nerve endings in the dura mater, play the role of nociceptor and reacts to environmental oxidative stress, transducing it into a neural signal, resulting in pain perception [50]. An associated neurogenic inflammation will also result, and will amplify the nociception process [52] reflected by pain-related disability (MIDAS score). Endogenous ROS can cause damage to DNA, mitochondria, membrane lipids, and proteins and, consequently, can contribute to neuronal cell loss in migraine patients [48,53,54] that were assessed in our study by OCT examination (Tables 2 and 3). There is a selective sensitivity of neurons in the central nervous system to oxidative stress [53], some of these regions being involved in the pathophysiology of a migraine attack (for example, functional connectivity between the hypothalamus and brainstem was found to be altered during a migraine attack, leading to the hypothesis that this network change might be the real driver of migraine attacks [55]). According to these data, we can also hypothesize that the vulnerability of specific neuronal networks to oxidative stress can initiate and maintain the migraine attacks. However, the oxidative stress during migraine attacks may constitute a physiologic environment conducive to stem cells [56]. Still, migraine attack is followed by the neuronal loss [57] and, as a consequence, the neuroplasticity phenomena can initiate an integrated mechanism for neural repair [49]. The balance between neuronal loss during migraine attack and neuronal repair stimulation after migraine attack has to be controlled in order to be used as a beneficial phenomenon for using it as a therapy for facilitating the survival, proliferation, migration, and differentiation of stem cells according to Borkum theory [49].
One of the most important molecules for reactive species is nitric oxide, the smallest signaling molecule known, that will enhance nitro-oxidative stress due to its increased production. Increasing NO production can be a result of the activation of three isoforms of NO synthase (NOS): neuronal NOS (nNOS), inducible NOS (iNOS), and endothelial NOS (eNOS) [57]. Their activation can exert various effects: nNOS activation contributes to synaptic plasticity in the central nervous system (CNS), central regulation of blood pressure, smooth muscle relaxation, and vasodilatation, iNOs activation contributes to the pathophysiology of inflammation, and eNOS activation has an important role for vasodilatation and vasoprotective effect [58]. Our results demonstrate a significantly increased NO molecule in migraine patients compared with control subjects (Table 1). Increased NO production can result from the activation of all of the three NOS isoforms. The NO excess reacts with superoxide, leading to the creation of peroxynitrite, a damaging product [58,59]. Several isoforms of O 2 •− -producing NADPH oxidase exists in the vascular wall [60]. Many types of vascular diseases appear as being associated with the up-regulation of NADPH oxidase [57] and evidence suggested that migraine pathophysiology is connected with vascular reactivity [61,62]. Endothelial NOS (eNOS) appears mainly involved in migraine pathophysiology. NO produced by eNOS that has been activated by serotonin (released by activated platelets), contributes to the vasodilation of migraines and associated neurogenic inflammation by stimulating the release of substance P that is an important molecule involved in neurogenic inflammation [63,64]. eNOS is also involved in endothelium, releasing VEGF and BDNF that are implicated in central sensitization during migraine attacks [65]. Among the constituents that contribute to increased oxidative stress in migraine patients, compared with the control group, malondialdehyde (MDA) is the constituent that features in our study (Table 1). A significantly increased MDA reflects an intense lipid peroxidation as an important contributor to oxidative stress and neuronal loss since neuronal membranes are rich in polyunsaturated fatty acids and are particularly susceptible to oxidative stress [66]. TOS was significantly increased and TAS was found to be significantly decreased in migraine patients compared to controls by our study, a result similar to the Yigit et al. [67] but opposite to the study conducted by Geyik et al. who reported no significant difference [68]. The age of the investigated patients can explain our different results regarding the oxidative stress changes as compared to Geyik et al. for our migraine group [68] aging itself being an essential contributor to increased oxidative stress [69]. Similar to our findings, Alp et al. reported significantly decreased TAC in migraine patients [70]. Several studies also reported that different types of headache are associated with decreased antioxidant defenses mechanisms [71][72][73]. As in our study (Table 1), the catalase level, as a contributor to the antioxidant system, was reported to be significantly decreased in migraine patients [74,75]. The persistence of oxidative stress and increasing the frequency of migraine attacks can lead to a chronic migraine [2]. Adding all of these pieces of evidence to our study, regarding the oxidative stress in migraine patients, we are considering that the disturbances of oxidant/antioxidant system balance has an important contribution in migraine pathogenesis.
Migraine and RNFL Thickness Assessment
The retinal nerve fiber layer (RNFL) contains the axons of the retinal ganglion cells. Therefore, measurement of the RNFL thickness by OCT is expected to provide information for monitoring the progressive loss of ganglion cells and axons in migraine patients [23].
Our study demonstrated that retinal fiber layers are significantly decreased in patients with migraine (peripapillary RNFL thickness for migraine group was decreased for all quadrants) ( Table 2). For the macular region, fovea RNFL thickness was significantly decreased for both eyes (Table 3). Our hypothesis is that oxidative stress contributes to RFLN thickness changes, and can be an effect of each recurrent attack since there are significant differences between the control group and migraine patients. Despite the inconsistent significant modification of macular thickness for various quadrants, macular volume significantly decreased in both eyes in migraine patients compared with the control group (Tables 2 and 3). Previous OCT studies had also demonstrated a significant decrease in RNFL thickness in various quadrants in patients with migraine. Assessing migraine patients by TD-OCT (time domain optical coherence tomography), different results were reported. Reduced RNFL thickness in the nasal sector was found by Sorkhabo et al. in a case-control study [34]. Martinez et al. found thinner TD-OCT RNFL in the temporal sector in migraine patients compared with healthy volunteers [76]. Colak et al. found on migraine patients using SD-OCT (spectral domain optical coherence tomography) no significant difference in the RNFL thicknesses of the temporal and nasal quadrants compared with control subjects but a significant difference of the superior and inferior quadrants [77]. In the same study, foveal, temporal, and nasal choroidal thickness measurements were significantly lower in the migraine group than in the control group [76]. Decreased thickness of RNFL in nasal sectors in patients with migraine (assessed by SD-OCT) was also found by different studies [78,79].
Correlation Between Oxidant/Antioxidant Balance and RNFL Thickness
Oxidative stress may be a final common pathway, signaling several unfavorable conditions in the brain, including neuronal degeneration and apoptosis [58]. Studies regarding retinal neuronal fiber layer involvement in migraine patients revealed a possible connection between the thickness of RNFL and migraine pathophysiology, even if different studies report discrepancies in the involved quadrants. Our results show a statistically significant correlation between the reduction in RNFL in the peripapillary region in different quadrants with different oxidative stress/anti-oxidative parameters with the strongest positive significant correlation (p < 0.007) between peripapillary thickness in nasal quadrants and catalase, in migraine patients (Table 4). Furthermore, a significant correlation for all quadrants with different stress oxidative parameters were also identified most often, the temporal quadrant being correlated for peripapillary and macular thickness with MDA, TOS, and catalase (Table 4). Our results suggest the absence of an identifiable rule regarding the involved quadrants of RNFL thickness and various oxidative parameters tested. However, the results obtained on this pilot study must be first validated on a larger sample, and if are reproducing, the mechanisms must be assessed. Reported results regarding the correlation of RNFL thickness with various parameters also showed associations that are not following a precise pattern. Abdellatif et al. [80] found that the duration of migraine is significantly correlated with the thickness of the ganglion cell layer, retinal nerve fiber layer, and all choroidal quadrants. The severity of migraine was only significantly correlated with the thickness of the ganglion cell layer and the retinal nerve fiber layer [79]. The duration of a migraine attack is the most important determinant factor for the decrease in the thickness of the superior retinal nerve fiber and in all the choroidal quadrants, while the severity of migraine is the most important determinant factor of inferior, nasal, and temporal retinal nerve fiber layer quadrants and the inferior ganglion cell layer [79]. Also using a spectral-domain optical coherence tomography assessment, some researchers found that inferior and superior quadrant RNFL thickness was significantly thinner in patients with migraine compared with the control [81]. Comparing both eyes in migraine patients with control subjects, the nasal RNFL thicknesses were significantly thinner on the right side in our study, while Gunes et al. reported a similar result but on both sides [82]. Other different significant correlations between RNFL thickness and migraine characteristics were also reported: the length of the history of a migraine [21], the migraine disability score [34], or visual disturbances amplitudes [83].
Consequently, we believe that the OCT-based evaluation of RNFL thickness could offer a non-invasive method for evaluating axonal loss in migraine patients. From our knowledge, this is the first study investigating the association between the oxidative stress parameters and RNFL thickness, a mechanism that could be an essential step for migraine treatment, targeting oxidative stress molecules. In light of this fact, personalized therapy could be applied according to which specific oxidative stress/antioxidant molecule is increased in each patient.
Study Limitation and Perspectives
Several limitations to our study could be listed. First, the presented study was designed as a pilot study on a small sample of patients having migraines without aura. The observed pattern of association between RNFL thickness in different quadrants and various oxidative stress parameters must be validated on larger sample sizes. Furthermore, the assessment of these associations among men and women on a larger sample size could be of scientific interest. Second, we evaluated patients with migraine with no prophylactic medication, but insufficient data were collected regarding the prophylaxis (such as were the medications never used or interrupted, were they interrupted recently or since a prolonged interruption, which medications were used, what doses were given and for how long, etc.). Third, the assessment of the group with migraine was made in the first 24 h after the migraine attack onset. Migraine attacks are known to be accompanied by oedema of the forearm, eyelids, and cheeks. The evaluation of the retinal parameters between attacks, out of painful periods, could bring more insights into how these parameters change during attacks and free of symptoms periods, allowing for a better assessment of the disease evolution. Fourth, we only evaluated the persons with migraines without aura. An evaluation also conducted on patients with migraine with aura could identify the same pattern or a different pattern of changes and associations between RNFL thickness and various oxidative stress parameters. Fifth, the results reported in correlation analysis must be interpreted with caution, because some statistically significant correlation could be observed, by chance, considering the number of the ocular coherence tomography spectral measurements.
Conclusions
Decreasing the RNFL thickness in migraine patients, in different quadrants, can be associated with an imbalance between oxidative stress and antioxidants by increasing oxidative stress or by reducing the antioxidant mechanisms, contributing to the axon degeneration in retinal layers. Because retina contains axons associated with ganglion cell neurons, and the RNFL is abundant in axonal tissue and has no myelin, the OCT assessment of RNFL thickness can be a potentially useful biomarker of axonal loss in the central nervous system. The association of RNFL thickness with various oxidative stress parameters in migraine patients can constitute an essential step for personalized therapy focused on targeting reducing oxidative stress molecules. However, the role of oxidative stress in migraine pathophysiology is not yet completely and worldwide established. This is a pilot study, and more exhaustive studies are needed to produce evidence for clinical practice.
Conflicts of Interest:
The authors declare no conflict of interest. | 2020-06-11T09:07:48.027Z | 2020-06-01T00:00:00.000 | {
"year": 2020,
"sha1": "464b01b945db6fb16f996fb07aedeefe1fd1516a",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2076-3921/9/6/494/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "571dd947fb465f9595d4fe7ca1332ef211a1fddd",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
72303330 | pes2o/s2orc | v3-fos-license | Influence of zinc on severity of common cold in children
Background Symptomatic treatment of common cold in children does not reduce the duration and severity of disease. Since zinc has been used to enhance cellular and humoral immunity, it has the potential to reduce the severity of the common cold. However, the effects of zinc on the common cold have been inconclusive. The use of zinc to treat cold symptoms deserves further studies. Objective To determine the effect of zinc supplementation on the severity of the common cold in children. Methods cold at primary health care centers in Gedongtengen, Umbulharjo consecutive sampling and their parents were interviewed. Severity of illness was categorized as mild, moderate or severe. Results One hundred fourteen patients with common cold were zinc supplementation while the other group received a placebo. of treatment, there were no significant differences in clinical Conclusion The severity of the common cold in children aged zinc supplementation compared to placebo. [Paediatr Indones. 2012;52:324-8].
Reprint requests to: Cahalafa Shinta Caesar, Department of Child cahalafa@gmail.comC ommon cold is an acute, viral respiratory tract infection.Cold symptoms are caused by the body's immune response against viral infection.In Indonesia, common cold has outpatient facilities in hospitals. 4The prevalence of 5 Ninety-four percent of the patients with common colds who visit a doctor get prescriptions for antibiotics, cough medicines, or symptomatic treatments.These medicines are of little benefit in treating the common cold and potentially cause side effects. 6There is still no standard therapy for the common cold.Good immune status is required to eliminate the virus in order to avoid complications.a role in the inflammatory process by inhibiting viral ing viral replication, stabilizing and protecting cell plasma membranes, as well as disrupting and inhibiting the release of proinflammatory mediators.
on the effects of zinc on the common cold, zinc had a significant effect in reducing the duration and severity of colds.However, Godfrey et al. reported inconsistent effects of zinc on cold severity.Since the evidence for the effects of zinc on the common cold have been inconclusive, a further study is needed.
Methods
common cold received either zinc supplementation or placebo therapy.A pharmacist from an outside primary health care center held the key, so subjects and researchers were blinded until the end of the study.Diagnosis of common cold were based on cough and runny nose with clear or mucopurulent secretions of good general condition and no thoracic abnormalities common cold for less than 2 days, had good general appearance, had parents willing to participate by signing the proxy consent and had an active phone number to maintain contact.Those suffering from a common cold with complications such as pneumonia, acute otitis media, sinusitis, exacerbation of asthma, as well as immunocompromised status such as HIV infection, malignancy or on steroid therapy for more than 4 weeks were excluded.eight was measured by a nurse using weight scales and nutritional status was determined based on weight-for-age according Adherence to therapy was considered to be good if > as exposure to cigarette smoke, kitchen smoke, or mosquito repellent smoke during the study.The study profile is shown in nurses on study forms.The outcome of the study was the severity of the common cold.The condition was considered to be improved if there was a reduction of one degree or more compared to the prior assessment of severity.
The independent variable in this study was the therapy type; zinc or placebo.The dependent variable was severity of the common cold.Confounding factors were age, nutritional states, and initial degree of severity.
This study was approved by the Commission on Medical Research Ethics and Health, Gadjah Mada University Medical School.All subjects provided proxy consent, signed by their parents.
of placebo therapy was analyzed by Chi-square test.Statistical significance was considered to be
Results
Subjects in both groups had similar characteristics, as shown in Table 1.
The decrease in severity of cold symptoms is shown in Table 2.There was no significant difference between the two groups in the decrease of cold
Discussion
Previous study showed that zinc administration reduced the severity and shortened the duration of common cold.The duration of common cold in subjects who were given zinc was less than 5 days.However, our results were in contrast to this previous study.The common cold is self limiting disease with In patients with a previous history of bronchial asthma, the common cold can be a trigger for asthma attacks.Thus reduction in the severity of common cold in children over time was expected.
The difference in our results may have been due to insufficient zinc dosage.Previous studies have used varying doses of zinc.According to Eby, only zinc ions were shown to have anti-rhinovirus activity by inhibiting the normal cleavages by which the viral polypeptides are processed, inhibiting ), inhibiting the release of histamine and leukotrienes from basophils and mast cells, protecting the plasma cell membrane, and being useful in allergy treatment.zinc dosage proved to be useful in other diseases.In addition, higher doses have been associated with adverse events. 22nother possible reason for a lack of difference in the effects of zinc and placebo may be due to the type of viruses causing colds in our subjects.Current evidence has shown zinc to have an anticold are caused by rhinovirus), but there has been no evidence that zinc inhibits other cold-causing that cause the common cold.gave zinc to patients with symptom duration of > 24 hours, whereas there is evidence that zinc provides benefits only if administered within the first 24 hours. 22 limitation of our study was that we did not perform intensive supervision after the treatment administration to both groups.Compliance to zinc administration was based on reporting by parents and families, hence, it is possible that the zinc was not taken regularly due to its unpleasant taste.
Co-intervention at home in the form of food and drink also were not explored.If the placebo group received more co-intervention than the zinc group, the results may have been influenced. 25,26The duration of the common cold was not defined as well because the data was compiled only up to the seventh day of observation.In addition, we found that side effects were not significantly different between the two years), they may not have been able to communicate their experience.supplementation once daily started in the first 48 hours of cold duration did not significantly reduce compared to a placebo.A further study needed to determine if a higher dosage of zinc and a different treatment duration can decrease the severity of the common cold.
Figure 1 .
calculated by unpaired categorical analysis with consecutive sampling.The treatment group received group received placebo powder.Any subject with fever received paracetamol.Subjects were assessed rd th days of treatment.Nurses assessed the cough, nasal symptoms, throat symptoms and systemic symptoms symptoms.After total scores were obtained, subjects effects of the therapy were recorded by parents or Allocated to placebo group Allocated to intervention Excluded Did not meet inclusion criteria Declined participation (n=6)
Table 1 .
Baseline characteristics of subjects
Table 2 .
Subjects with decreased cold severity in the zinc and the placebo groups | 2019-01-28T17:49:48.933Z | 2012-12-31T00:00:00.000 | {
"year": 2012,
"sha1": "3c5286a704b9315b8926a9072f4f2c483bd888c3",
"oa_license": "CCBYNCSA",
"oa_url": "https://paediatricaindonesiana.org/index.php/paediatrica-indonesiana/article/download/495/371",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "3c5286a704b9315b8926a9072f4f2c483bd888c3",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
248753512 | pes2o/s2orc | v3-fos-license | Age disparities in mental health during the COVID19 pandemic: The roles of resilience and coping
Background The COVID19 pandemic has caused a mental health crisis worldwide, which may have different age-specific impacts, partly due to age-related differences in resilience and coping. The purposes of this study were to 1) identify disparities in mental distress, perceived adversities, resilience, and coping during the COVID-19 pandemic among four age groups (18–34, 35–49, 50–64, and ≥65); 2) assess the age-moderated time effect on mental distress, and 3) estimate the effects of perceived adversities on mental distress as moderated by age, resilience and coping. Methods Data were drawn from a longitudinal survey of a nationally representative sample (n = 7830) administered during the pandemic. Weighted mean of mental distress and adversities (perceived loneliness, perceived stress, and perceived risk), resilience, and coping were compared among different age groups. Hierarchical random-effects models were used to assess the moderated effects of adversities on mental distress. Results The youngest age group (18–34) reported the highest mental distress at baseline with the mean (standard error) as 2.70 (0.12), which showed an incremental improvement with age (2.27 (0.10), 1.88 (0.08), 1.29 (0.07) for 35–49, 50–64, and ≥65 groups respectively). The older age groups reported lower levels of loneliness and perceived stress, higher perceived risk, greater resilience, and more relaxation coping (ps < .001). Model results showed that mental distress declined slightly over time, and the downward trend was moderated by age group. Perceived adversities, alcohol, and social coping were positively,whereas resilience and relaxation were negatively associated with mental distress. Resilience and age group moderated the slope of each adversity on mental distress. Conclusions The youngest age group appeared to be most vulnerable during the pandemic. Mental health interventions may provide resilience training to combat everyday adversities for the vulnerable individuals and empower them to achieve personal growth that challenges age boundaries.
Introduction
The COVID-19 pandemic has caused a global mental health crisis (e. g., Saladino et al., 2020). Older adults (age ≥65 years) have been found to be more vulnerable to the infection due to their increased comorbidities and weakened immunity compared to their younger counterparts (Centers for Disease Control and Prevention, 2021). However, it has been found that individuals at younger age are more vulnerable to stress, negative affect, depression, and anxiety in general or among those with certain health conditions such as cancer (Cohen et al., 2014;Hinz et al., 2009;Hopwood et al., 2010;McCleskey and Gruda, 2021;Varma et al., 2021). During the pandemic, younger adults (18-30 years of age) in multiple countries have reported the greatest increase in psychological distress compared to other age groups (Jung et al., 2020;McGinty et al., 2020;Pierce et al., 2020;Rossell et al., 2021;Varma et al., 2021). Some researchers have attributed the age disparities in mental health to younger adults' social roles, such as greater career demands, parenting duties, and economic burdens (Arndt et al., 2006;Kornblith et al., 2007) as well as older adults' increased experiences in coping with challenges and adversities with effective coping strategies (Brandtstädter, 1999). Despite the age disparities in social roles and coping resources, older adults may nevertheless experience an unprecedented degree of social isolation and loneliness due to COVID-19 (World Health Organization, 2020) that may adversely impact their mental health (Kowal et al., 2020;Robb et al., 2020).
Resilience is defined as positive adaptation in the context of significant risk of adversity (Cohen et al., 2014;Ong et al., 2009;Wood and Bhatnagar, 2015). It refers to the capacity to recover or bounce back from adverse life events through adjusting to changing situational demands (Cohen et al., 2014;Dyer and McGuinness, 1996;Ong et al., 2009;Richardson, 2002). Human beings cultivate resilient qualities to experience growth through adversities and to build resilient reintegration (Rutter, 1985). Resilience mitigates the negative effects of stressors on mental health (Wagnild, 2009;Wagnild and Young, 1993). For example, it has been found that a higher level of perceived personal control (an aspect of resilience) is related to a lower level of negative affect (Diehl and Hay, 2010), lower reactivity to daily life stressors (Hahn, 2000;Neupert et al., 2007;Ong et al., 2005), and reduced stress-anxiety association (Neupert et al., 2007). Resilience factors, such as meaningfulness and self-reliance, are inversely related to the level of depression, anxiety, and stress (Lenzo et al., 2020). During the COVID-19 pandemic, resilience is especially important in that individuals with lower resilience reported a greater increase in psychological distress (Killgore et al., 2020;Riehm et al., 2021).
Empirical evidence shows that resilience is a learned dynamic process or a skill set that grows with age (Leys et al., 2020;Nygren et al., 2005) and may be improved with education and therapeutic interventions (Loprinzi et al., 2011;Richardson and Waite, 2002). Resilience in older adults is especially demonstrated in emotional regulation and problem solving (Gooding et al., 2012), as they are more effective than young adults at regulating negative emotions, rationally assessing situations, applying prior knowledge, and predicting outcomes (Mather, 2006). Thus older adults typically experience lower anxiety than their younger counterparts (McCleskey and Gruda, 2021).
The concept of resilience is closely related to active coping, which is conductive to stress reduction and mental health amelioration. The resistance to social defeat stress may be considered as an adaptation that occurs with repeated exposures to stress (Wood and Bhatnagar, 2015). Active coping is characterized by behavioral responses using one's own resources to minimize the physical, psychological or social harm of a situation, such as creating a sense of coherence, exercising self-control, developing a sense of identity and purpose in life (Folkman and Lazarus, 1980). An active coping style was shown to be protective against psychotic symptomatology during the pandemic, independent of demographic factors (Song et al., 2020). Conversely, passive coping is characterized by feelings of helplessness, reliance on others for stress resolution, and behaviors such as avoidance, substance use, and blaming (Folkman and Lazarus, 1980;Billings and Moos, 1984;Yi et al., 2005).
Passive or maladaptive coping is associated with psychopathology. Previous research has revealed that psychosis is positively associated with avoidant coping, but inversely associated with adaptive coping (Mian et al., 2018;Ventura et al., 2004). However, whether a specific coping strategy is adaptive is a fluid concept: it varies with situations and types of stressors (Wood and Bhatnagar, 2015). For example, passive coping can be sometimes adaptive and increases chance of survival when time is limited or the stressor cannot be altered (such as past events).
Efficient coping during the unprecedented COVID-19 pandemic would be critical for mental health protection. The objectives of this study were: 1) to identify disparities in mental distress, perceived adversities, resilience, and coping behaviors by age group; 2) to assess the time effect on mental distress as moderated by age group, and 3) to examine whether and how age, resilience and coping behaviors moderated the detrimental effects of perceived adversities on mental health. Fig. 1 displays the conceptual model.
Data source
Data were drawn from Understanding America Study's (UAS) longitudinal online survey related to the COVID-19 pandemic (Understanding America Study, 2020). The UAS is a probability-based internet panel representative of the US population of adults excluding institutionalized individuals and military personnel (Angrisani et al., 2019). The UAS was first launched by the University of Southern California The present study used de-identified, publicly available data collected by UAS, and it was exempt from the ethics approval of the Institutional Review Board. Details on the UAS COVID data are available on their website (https://uasdata.usc.edu/index.php). The current study included 18 survey waves (Wave 4 to 21) over the period from April 2020 to January 2021, with 112,054 observations from 7830 unique participants. The first three waves were excluded due to the lack of key questions assessing resilience, loneliness, and certain coping behaviors.
Outcome variable
Mental distress was assessed by the 4-item Patient Health Questionnaire (PHQ4) (Kroenke et al., 2009), which asked the respondents how often they had been bothered by each of the four problems over the past 14 days: 1) Feeling nervous, anxious, or on edge; 2) Not being able to stop or control worrying; 3) Feeling down, depressed, or hopeless; 4) Table 3 Associations of mental distress with age groups, adversities, resilience, and coping from the hierarchical random effects models. Note. Var(SE) stands for variance (standard error). All models were adjusted for sociodemographics including gender, race/ethnicity, education, income, marital status, disability, and chronic conditions including asthma, high blood pressure, cancer, COPD, diabetes, heart disease, kidney disease, autoimmune disease, and obesity.
Little interest or pleasure in doing things. Responses were based on a 4-point Likert scale (0-3): not at all (0), several days (1), more than half of the days (2), and nearly every day (3). Cronbach's alpha for this construct was 0.91. The sum score was used to index mental distress (range 0-12).
Independent variables: perceived adversities
Perceived stress was measured with the 4-item Perceived Stress Scale (PSS4, Cohen et al., 1983), which asked how often the respondents felt stressed in the past 14 days: 1) unable to control the important things in your life; 2) confident about your ability to handle your personal problems; 3) that things were going their way; and 4) difficulties were piling up so high that they could not overcome them. Respondents rated each item on a 5-point Likert scale: never (1), almost never (2), sometimes (3), fairly often (4), and very often (5). Responses were reverse coded whenever appropriate so that a higher score would indicate greater stress. Cronbach's alpha for this construct was 0.65. Perceived stress was indexed by the sum score (ranged 4-20).
Loneliness was measured by a single item: "In the past 7 days, how often have you felt lonely?" Respondents rated loneliness on a 4-point Likert scale: not at all or less than 1 day (1), 1-2 days (2), 3-4 days (3), and 5-7 days (4). The exact score was used to index loneliness.
Perceived risk was assessed by perceived risk of death and hospitalization from COVID-19, contracting COVID-19, and running out of money in the next three months. Each item was rated based on a scale of 0%-100%. Cronbach's alpha for this construct was 0.72. The total risk was calculated as the average of these items and divided by 10, to keep its scale largely comparable to other independent variables (range 0-10).
and above.
Resilience was assessed with the six questions from the Brief Resilience Scale (BRS, Smith et al., 2008). Example items were "I tend to bounce back quickly after hard times"; "I have a hard time making it through stressful events"; "It does not take me long to recover from a stressful event." Respondents rated each question on a 5-point Likert scale: strongly disagree (1), disagree (2), neutral (3), agree (4), and strongly agree (5). Cronbach's alpha for this scale was 0.86. The Resilience was indexed by the sum score (range 6-30).
Coping behaviors included social coping, relaxation, and alcohol use. For social coping, respondents were asked to estimate the number of days that they did each of the following activities in the past seven days (range 0-7): connecting socially with friends or family (either online or in-person); posting or browsing social medias (e.g., Facebook, Twitter, Instagram, or Snapchat); having a phone or video call with a family member or a friend; messaging/emailing a family member or friend; interacting with a family member or friend in-person. Cronbach's alpha for this scale was 0.70. For relaxation, the respondents estimated the number of days they made time to relax out of the past seven days (range 0-7). Alcohol use was indexed by the product of the number of days of alcohol consumption in the past seven days and the average amount of consumption each day (range 0-210).
Covariates/confounders
Covariates included gender (male vs. female), education (less than high school, high school graduate, some college or 2-year college, and Bachelor's degree and above), household income (four quartiles), race/ ethnicity (Non-Hispanic White, Non-Hispanic Black, Hispanic, Native American, Asian and Pacific Islander, Non-Hispanic mixed race), marital status (married vs. unmarried), presence of a disability (yes vs. no), and chronic conditions (yes vs. no) including asthma, high blood pressure, cancer (other than skin cancer), chronic obstructive pulmonary disease (COPD), diabetes, heart disease, kidney disease, autoimmune disease, and obesity.
Statistical analysis
The weighted mean and standard error of the outcome variable mental distress, three independent adversity variables (perceived stress, loneliness, and perceived risk), resilience, and three coping behaviors were plotted over time (by survey wave) for each age group. Weighted distributions of all covariates (sociodemographics, chronic conditions, and disability), independent variables, and the outcome variable were derived from the initial wave and compared across age groups with the Chi-Square test for categorical variables and ANOVA test for continuous variables. Weighted correlations among all independent variables and the outcome were derived from the initial wave. Hierarchical randomeffects models (random intercept) were employed to assess the theoretical effects of covariates on mental distress. Of all the variables included in the models, age group, gender, race/ethnicity were timeinvariant and assessed in the initial wave (wave 4) or the first wave for those newly added individuals, and the other variables were timevariant. To assess the linear effect of time, survey wave was entered into all models as a continuous variable. We implemented the sandwich estimator to generate robust standard errors in all the random-effects models. Error covariance matrix was specified as unstructured. Model 1 included the age group as the independent variable, adjusted for survey wave, and covariates including sociodemographics, chronic conditions, and disability; Model 2 added each adversity variable, resilience, and three coping variables (alcohol, social coping, and relaxation) as independent variables, adjusted for survey wave and other covariates; Model 3 further added two-way interactions between 1) age group and survey wave, 2) age group and each adversity variable, 3) age group and each coping strategy, and three-way interactions between 1) age group, resilience, and each adversity, 2) age group, each coping strategy, and each adversity, and 3) age group, resilience and each coping strategy. To keep Model 3 parsimonious, only the significant interactions remained in the model. With a focus on linear relationships, no quadratic terms were added. Additionally, a sensitivity analysis including age as a continuous variable in all three models was conducted to assess the generalizability of our findings. All above analyses incorporated survey weights and were conducted in SAS 9.4.
Our random-effects model results were derived from complete observations from any single wave. The complete data on all covariates in any wave included 109,424 observations, accounting for 97.65% of all available observations from 7755 unique respondents. About 46.8% of respondents participated in all 18 waves. The number of actual waves completed by each respondent ranged from 1 to 18, with a mean of 14.1 waves per person. The proportion of missing observations out of available observations due to the inclusion of only complete data was 4.05%, 2.52%, 1.66%, and 1.39% for age group 18-34, 35-49, 50-64, and ≥65 respectively. With the assumption of missing at random (MAR), we implemented a multiple imputation procedure (n = 10) within the Bayesian framework in Blimp 2 as a sensitivity analysis for the randomeffects models. The model-based multiple imputation procedure was demonstrated in simulations to accurately estimate the interaction effects for multilevel (repeated measures) models (Enders et al., 2020). The Markov chain Monte Carlo (MCMC) method was employed to obtain the parameters of the models for the multiple imputation.
Descriptive statistics
The first set of results (Fig. 2) shows the mean level of mental distress, perceived adversities, resilience and each coping strategy over time by age group. Note that the Y-axes in Fig. 2 were truncated to better contrast the four age groups. Table 1 shows the covariate and outcome distribution by age group in the initial wave (April 2020). Age disparities were consistent both in the initial wave and over time: the youngest age group (18-34) showed the greatest mental distress, the highest levels of loneliness and perceived stress, and the lowest level of resilience. For instance, in the initial wave, the youngest age group showed the greatest mental distress compared to the other age groups, with mean (standard error) as 2.70 (0.12), 2.27 (0.10), 1.88 (0.08), and 1.29 (0.07) for the youngest to the oldest age groups, respectively. The oldest age group perceived the highest COVID risk. In terms of coping, the younger groups (18-34, 35-49) used more social and less relaxation coping. There was no significant overall difference in alcohol coping across all age groups in the initial wave (Table 1). Table 2 reports the weighted correlations among mental distress, perceived adversities, and resilience by age group in the initial wave. All correlations were significant (p < .0001). The correlations between these variables and coping behaviors were small (r < 0.2) or not significant and thus not displayed in the table. As shown in Table 2, mental distress had moderate and positive correlations with loneliness and perceived stress, similar across age groups. Mental distress had a weak and positive correlation with perceived risk, and its magnitude was weakest in the youngest group. Mental distress had a moderate and negative correlation with resilience, with similar magnitude across age groups. Resilience was also negatively correlated with each adversity perception, with strongest correlations in the youngest group. Table 3 reports the fixed and random effects and model fit from three hierarchical random-effects models to assess the adjusted associations of Table 5 Slope estimates (B) and standard errors (SE) of loneliness, perceived stress, and perceived risk on mental distress by resilience levels, corresponding to Fig. 4 Note. n.s. Refers to not significant at p = .05 as compared to slope coefficient = 0.
Random-effects model results
independent variables with the outcome variable mental distress. These hierarchical models showed incremental model improvement with model complexity. Model 3 indicated the best fit. The random effect of the intercept was significant in each model. For fixed effects, Model 1 regression coefficients decreased with older age groups, suggesting mitigated mental distress with increasing age bracket. This was consistent with the mental distress graph in Fig. 2. Survey wave had a negative association with mental distress, suggesting improved mental health over time. For a complete set of covariate estimates from Model 1, see Appendix Table 1. Being male, married, Non-Hispanic Black or Native American race, making higher income, and not having certain chronic conditions such as high blood pressure, asthma, autoimmune disease, obesity, or disability, were associated with lower mental distress. Our finding on Non-Hispanic Blacks faring better in mental health during the pandemic was consistent with a previous finding using UAS data (Owens and Saw, 2021). The age group effect and wave effect remained significant in Model 2 after adding perceived adversities, resilience, and coping strategies. Adversities (loneliness, stress, and risk perception) and maladaptive coping (e.g., alcohol use) were positively associated with mental distress, whereas resilience and relaxation were negatively associated with mental distress. Model 3 further included two-way and three-way interaction effects, but only the significant (p < .05) interactions were kept. Moderation effects derived from Model 3 are also displayed in Fig. 4. Fixed effects of loneliness (4a), perceived stress (4 b), and perceived risk (4c) on mental distress, as moderated by age groups and resilience levels, corresponding to Table 5. Table 4. Fig. 3 series shows age-moderated associations of mental distress with time / wave (3a), loneliness (3 b), perceived stress (3c), perceived risk (3d), resilience (3e), social coping (3f), alcohol coping (3g), relaxation coping (3h). The slope estimates and standard errors are shown in Table 4. Over time, mental distress did not change among the youngest group and age group 50-64, but showed a slight decrease among age groups 35-49 and ≥ 65. Across all age groups, loneliness, stress, perceived risk, and both social and alcohol coping were associated with higher mental distress, whereas resilience was associated with lower mental distress. The youngest group had a slightly elevated slope of loneliness [B(SE) = 1.13 (0.06), p < .0001] and perceived stress [B(SE) = 0.25 (0.01), p < .0001] and the slope gradually decreased with older age, suggesting the effects of loneliness and stress were most detrimental on the youngest age group. The youngest group had a flattened slope of perceived risk [B(SE) = 0.06 (0.03), p = .0260] but an elevated baseline fixed (average) intercept on mental distress. This suggests that when perceived risk was zero and other variables were held constant, the youngest group showed highest mental distress, but their distress level tended to be least affected by perceived risk. The protective effect of resilience was smallest in the youngest group [B(SE) = − 0.05 (0.01), p < .0001] and became stronger with older age [B(SE) = − 0.08 (0.01), − 0.10 (0.01) and − 0.11 (0.01), all ps < .0001, for age groups 35-49, 50-64, and ≥65 respectively]. For social coping, the youngest age group had a steeper slope [B(SE) = 0.08 (0.02), p < .0001] compared to other age groups. For alcohol coping, the slope was the same across age groups Table 6.
[B(SE) = 0.003 (0.001), p = .0105]. Relaxation coping had no effect for the youngest age group and the age group of 50-64, but it decreased mental distress for the other two age groups.
To correctly interpret the modification effects of resilience on three adversity variablesloneliness, perceived stress and perceived risk, please refer to Table 5 (i.e., significant slope estimates by increasing resilience) together with Fig. 4 (i.e., the combined effects of the fixed intercepts and slopes). Table 5 shows the predicted slope estimates of loneliness, perceived stress and perceived risk on mental distress across different levels of resilience (rows) and age groups (columns). Most slopes were significantly different from 0. The slope of loneliness on mental distress consistently decreased with resilience level for all age groups, suggesting a buffering effect of higher resilience. This is also shown in Fig. 4a. The youngest age group had the highest slope of loneliness at each fixed resilience level, indicating the greatest effect of loneliness on mental distress for this group. For perceived stress, greater resilience unanimously and consistently reduced its effect on mental distress across all age groups, indicating a consistent buffering effect of resilience. This is also shown in the gradually flattened slopes with higher resilience in Fig. 4b. Similarly, the slopes of perceived risk decreased with higher resilience across age groups, and Fig. 4c confirmed this finding as indicated in the combined effects of slopes and their fixed intercepts. Taken together, resilience was an effective buffer for the three adversity variables on mental health for all age groups.
In terms of the modifying effect of coping on perceived adversities, there was only one significant interaction, which was between loneliness, relaxation, and age. With greater relaxation, the slope of loneliness on mental distress become smaller across all age groups, as shown in Fig. 5 and Table 6. Thus, relaxation was an effective buffer for the detrimental effect of loneliness on mental distress during the pandemic. Social coping and alcohol coping did not moderate the effect of perceived adversity on mental distress.
In the multiply-imputed data, the parameter results from 3 models were very close to those from the complete data analysis (see Appendix Table 2). In the sensitivity analysis where age was entered into the models as a continuous variable, age was also negatively associated with mental distress (see Appendix Table 3). Although the interaction effects did not completely overlap with Model 3 in which age was entered as a categorical variable, some main results remained unchanged, such as the effects of loneliness and perceived stress on mental distress as moderated by age and resilience.
Discussion
This study aimed to identify age disparities in mental health, perceived adversities, resilience and coping during the COVID-19 pandemic, and to assess the moderation of resilience and coping behaviors on the detrimental effects of perceived adversities on mental health among different age groups. Taken together, the results largely identified the youngest age group (aged 18-34) as the most vulnerable group for mental distress, with the highest perceived adversities and lowest resilience relative to other age groups. Resilience had an agemoderated negative association with mental distress and it also effectively buffered the relationship between perceived adversities and mental health. Maladaptive coping such as alcohol use increased with mental distress, and the relationship was not moderated by age group. Similar to resilience, adaptive coping such as relaxation had both a direct negative relationship with mental distress and mitigated the agemoderated association of loneliness with mental distress. Greater social coping, usually framed as adaptive, however had an age-moderated positive association with mental distress, and it did not moderate the link between adversity and mental distress.
Age disparities in mental health, perceived adversities, resilience, and coping
The results of the current study consistently identified the youngest age group (18-34) as the most vulnerable age group, with the highest mental distress, loneliness, and stress. All of these were reported to be associated with poor psychological well-being (Beutel et al., 2017;Kim et al., 2020;Kwag et al., 2011;Mushtaq et al., 2014;Wiegner et al., 2015). Additionally, the slope analysis suggested that loneliness is psychologically most detrimental for the youngest adults. These are consistent with previous findings of a differentially stronger inverse relationship between negative affect and perceived personal control for younger adults (Diehl and Hay, 2010) and poorer psychological outcomes in younger age groups than their older counterparts (Cohen et al., 2014;Hinz et al., 2009;Hopwood et al., 2010;Varma et al., 2021). Due to the lockdown measures implemented during the pandemic, it is not surprising that there is an elevated level of loneliness due to disconnections from real-life social networks. According to Statistics Canada, the number of people living alone has drastically increased, mostly among adults aged 35 to 64 years (Tang et al., 2019). Unlike some middle-aged and older adults who are accustomed to living alone, younger adults may experience greater impacts of social isolationduring the pandemic. The poorer mental health outcomes in younger adults could be driven by their higher career demands, parental duties, and economic needs (Arndt et al., 2006;Kornblith et al., 2007). The pandemic has led to a high unemployment rate, surpassing that of the Great Recession in the United States, according to the data reported by the Congressional Research Service (Falk et al., 2021). Thus, social disconnection, the toll of COVID-19, and financial and employment insecurity combined may explain why younger adults were most psychologically distressed and perceived the highest adversities in the present study.
We found that older adults perceived highest COVID risk, which may be due to their elevated susceptibility to the virus, such as infection, complications, and death. Nevertheless, older adults showed a generally more positive mental health profile out of all age groups, in line with earlier findings that older adults fare better during the pandemic in terms of depressive symptoms and affective experiences (Carstensen et al., 2020;Zach et al., 2021).
Consistent with earlier research (MacLeod et al., 2016;Zach et al., 2021), our study demonstrated a gradual increase in resilience with older age, which supports the conceptualization of resilience as a process instead of a personal trait, and thus could be modified and gained through experience with aging (Zach et al., 2021). In our study, resilience was negatively correlated with all three perceived adversities, suggesting that highly resilient individuals (such as some older adults) also perceive low adversities during the pandemic. Thus, it is expected Note. Relax=Relaxation.
that intervention programs aiming at building resilience in vulnerable individuals (such as young adults) may also reduce their perceived severity of adversities. The conjunction of increased resilience and reduced perceived adversities may provide an optimal protection for mental health outcomes. Evidence-based interventions, such as attention training, interpretation exercises directed away from fixed prejudices, and cultivation of skills such as gratitude, compassion, acceptance, forgiveness, and purpose in life, have been reported to increase resilience and decrease stress and anxiety (Loprinzi et al., 2011). In terms of coping strategies, older age groups were more likely to adopt relaxation as a coping strategy and the youngest group used more social coping relative to other age groups. This age disparity in different coping strategies might be linked to the age-related differences in perception of the pandemic and in the availability and ease of use of different coping resources. Younger individuals perceived COVID-19 infection as less of a threat so they were still differentially more socially engaged as measured by the social coping scale, whereas older adults perceived the infection as more severe and may reduced social interactions to minimize infection risk. In addition, compared to older adults, younger adults may have a greater need for socialization and greater access to social resources (such as social media platforms) and advanced technology which became critical during the pandemic. Older adults, mostly retired, tend to engage in more relaxation, potentially due to their flexible schedule and reduced workload. Although alcohol use had no overall significant difference across all age groups in the initial wave, with further analysis, we found a downward trend of alcohol use over time and a significant interaction between age group and time. The decline was significantly slower in older age groups (50-64, ≥65) compared to the youngest group. Multiple studies worldwide found that alcohol use decreased from the pre-to post-COVID-19 period among young adults (Vera et al., 2021;Evans et al., 2021;Ryerson et al., 2021). Specific to the US, the reasons for alcohol consumption reduction among college students above legal drinking age were stipulated as loss of access to establishments (e.g., bars and restaurants) that sold alcohol and relocation from the campus to the family home setting during the pandemic (Ryerson et al., 2021).
The moderation effects of resilience and coping
The current study largely supports the protective effect of resilience and adaptive coping such as relaxation. In contrast, perceived adversities, social coping, and alcohol use showed detrimental effect on mental health. These results are consistent with past research that denotes maladaptive or passive coping as linked to psychopathology (Folkman and Lazarus, 1980), whereas adaptive coping could be beneficial to mental health (Song et al., 2020).
In the current study, we quantified and visualized the buffering effects of resilience and relaxation on mental distress. Resilience greatly mitigates the effect of perceived adversities on mental distress for all age groups, consistent with the earlier findings (Killgore et al., 2020;Wagnild, 2009). Relaxation unanimously buffers the detrimental effects of loneliness on mental distress. Consistent with literature (Gao et al., 2019), this suggests that an adaptive coping strategy could reduce the negative impact of stressors on psychological health.
However, in a time when social isolation is required and beneficial to health, social coping may not be as protective for mental health as expected from pre-pandemic literature (Mian et al., 2018;Ventura et al., 2004). Our results concur with a U.K. study (Fluharty and Fancourt, 2021) that found worries related to finances, basic needs, and the coronavirus during the pandemic were positively associated with endorsement of social coping. The lack of consistently beneficial effects of social coping may be intrinsic to the infectious nature of the pandemic and the prevention measures that discouraged socialization. In other words, social coping may not be perceived as adaptive and could even be perceived as discomforting or risky during the pandemic compared to other situations. Nevertheless, we should note that this study did not intend to assess the causal relationship between social coping and mental distress, and the result should be interpreted as correlational, with potential residual confounding.
Limitations
This study has several limitations. The study results should be interpreted as associations, not causal relationships. The results were derived from a nationally representative online sample, and thus this study has inherent limitations of drawing data from an established database, such as unmodifiable data collection strategies and measurements. Although administrative procedures were taken to minimize the sampling bias, such as providing tablets and Internet access if needed, the sample may still show selection bias especially for the older adults. As shown in Table 1, the oldest age group had relatively low disability prevalence as compared to age group 50-64, which was at odds with literature. A report based on the data from the Behavioral Risk Factor Surveillance System found that the prevalence of any type of disability (vision, cognition, mobility, self-care, and independent living) was higher among those aged 65 and above (35.5%) than those aged between 45 and 64 years (26.2%) and 18-44 years (15.7%, Courtney-Long et al., 2015). Thus, the prevalence of disability may be underestimated for all age groups, especially for the oldest one, presumably due to sample selection bias and lack of clear definition of disability in the survey question. Generalizability of the study can be improved by including individuals who are not regularly online, but this may only represent a small segment of the population, given the prevalence of online communication. According to Pew Research, the percentage of Internet use was 96% among US adults aged 50-64 and 75% among those aged 65 and above in 2021 (Pew Research Center, 2021). Given that all responses were self-reported and collected online, there could be some generational differences (e.g., comfort with self-disclosure and technology and mental health stigma) confounded with age differences. The social coping scale can be further refined to reduce redundancy and increase reliability. Future studies may consider including other coping behaviors such as religious coping and substance use.
Conclusions and implications
Despite of the limitations of the current study, the results nevertheless revealed pronounced age disparities in mental distress, perceived adversities, resilience and coping, as well as highlighted how resilience and coping could adaptively mitigate mental distress in this unprecedented pandemic. The results provide important insights into the mental health status and associated factors in different age groups at the early stage of the COVID-19 pandemic in the United States. This study used timely reported longitudinal data from a nationally representative sample. The validity of the results was confirmed by further analysis with multiple imputation of missing data points. Our study is informative for mental health practitioners and policy makers to design interventions to mitigate mental distress and ameliorate psychological risk factors during a public health crisis, especially for the most vulnerable individuals. It will be important to continuously monitor and evaluate resilience and coping as the effects of COVID-19 continue to evolve and hopefully dissipate.
Author contribution
Ling Na has contributed to the conceptualization, methods, data acquisition, data analysis and interpretation, and writing of the manuscript. Lixia Yang and Peter Mezo have contributed to methods, results interpretation, and writing of the manuscript. Rong Liu has contributed to multiple imputation and writing of the manuscript. All authors approved the final revision. | 2022-05-14T13:07:44.374Z | 2022-05-01T00:00:00.000 | {
"year": 2022,
"sha1": "f5fe81548d9bb656bf251c428993af01e57fb8a5",
"oa_license": null,
"oa_url": "https://doi.org/10.1016/j.socscimed.2022.115031",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "9f8416829c4af091190a151c65f35762b38d7c80",
"s2fieldsofstudy": [
"Psychology",
"Medicine",
"Sociology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
247101106 | pes2o/s2orc | v3-fos-license | Consumers’ Perception of Food and Agriculture Education in Farmers’ Markets in Taiwan
Following the trend of green product consumption, farmers’ markets sell green products and have gradually developed the promotion of environmentally friendly knowledge of food and agricultural education. However, past research on farmers’ markets has rarely been combined with the function of food and agricultural education. Therefore, this study explores the relationship between product knowledge, green perceived value, and purchase intention from the perspective of farmers’ market food and agricultural education. Furthermore, research and investigation were conducted with trust and local attachment as multiplicative adjustment variables. This study adopted structural equation modeling (SEM) to examine the structural relationship among local attachment, product knowledge, green perceived value, purchase intention, and model fit. We distributed 396 surveys at three large farmers’ markets in Taiwan, of which 88.6% were used in this study. The research results showed that product knowledge and green perceived value had a significant positive impact on purchase intention; trust positively affected product knowledge and purchase intention; the multiplication of local attachment and trust also positively impacted product knowledge and purchase intention. Based on the research results of this study, it is suggested that while transferring knowledge, the farmers’ markets can incorporate trust and emotional relationships, which will serve as stimulating factors that can increase consumers’ purchase intentions. Although the findings are local to Taiwan, its characteristics are typical of food and agricultural education, as well as farmers’ markets worldwide, especially in Asia.
Introduction
In the modern age, when health awareness is rapidly increasing, the issues of environmentally friendly land, healthy eating, and food safety are getting more and more attention than ever. However, in the pursuit of technological advancement and modern development, the public's understanding of and exposure to agriculture, food, environment, and land culture have been declining nowadays [1,2]. Based on the above problems, the quality and safety of agricultural products have gradually become the public's primary concern [3], which highlights the importance of food and agricultural education [4].
The rise of planting methods that are more environmentally friendly, with the increase in food materials, have led the public to pursue production methods that are more environmentally friendly, and food materials [3], and the green agricultural products they produce, have indirectly promoted the development of the farmer's market [1]. Specifically, local farmers deliver fresh, local food to a growing number of shoppers who demand food that is not only healthy, but also environmentally friendly. However, farmers' markets take agricultural education. Therefore, the farmers' market can reflect the spirit of food and agricultural education more directly and convincingly [17].
In the past, many related studies in the farmers' market, and the majority of the relevant literature, conducted analyses from the perspective of consumers or producers, such as research on the determinants of consumer attitudes and behavior-with the farmers' market as an example [18], knowledge of food safety and sociodemographic factors of farmers' market consumers [19], hidden benefits of the short-term food supply chaintaking Italian farmers' markets as an example [20], consumers' purchase intention of agricultural products from the farmers' market [21], the purchase intentions of farmers' markets using mobile payment [22], and even the willingness of consumers to purchase and loyalty to farmers' market products [23]. However, Taiwan's farmers' markets that promote food and agricultural education are now emerging gradually, but there are few relevant studies to discuss them from the perspective of food and agricultural education. In addition, according to the previous literature and research, there is quite a lot of research on local attachment and trust, but few research studies of the three-order interaction of the two multiplying each other. Therefore, to fill the gaps in the research on the farmers' markets, this study took food and agricultural education as the entry point to explore the relationship between consumer knowledge of the product, green perceived value and purchase intention, and used trust and local attachment as a three-way interaction for research and investigation.
Based on possible food and agricultural education and concepts of the farmers' market that sells green agricultural products, this study attempted to explore the relationship between product knowledge, green perceived value, and purchase intention from the perspective of food and agriculture education in farmers' markets, and conducted research and investigations with trust and local attachment as multiplicative moderator variables. For these reasons, the study reviews the literature on farmers' markets and food and agricultural education, product knowledge and purchase intention, perceived value and green purchase intention, trust in product knowledge and purchase intention, and local attachment to trust, product knowledge, and purchase intention. Additionally, then conducting a questionnaire distribution from three large farmers' markets in Taiwan. After analyzing the results, this study is expected to determine the perception of consumers about food and agriculture education in farmers' markets in Taiwan. The results of this study should help promote Taiwan's food and agricultural education and the achievements of green products, and contribute to the concrete implementation of sustainable environmental development and the transmission of knowledge about food safety.
Farmers' Market and Food and Agricultural Education
"Green products" are based on the principle of not wasting resources [24]. In the process of product production from manufacturing to the finished products, the risk of environmental damage must be minimized, waste and hazardous materials must be reduced, and safe and environmentally friendly resources and energy must be used to achieve the production method of environmental friendliness and environmental protection [25]. Compared with traditional products that do not pay too much attention to regulations, green products focus on saving resources and energy, reducing waste discharge, and having the efficiency of recycling, which can significantly reduce the harm to the environment. It also has advantages for environmental protection and sustainable development [26]. Green agricultural products are rice, fruits and vegetables, meat, fresh aquatic products, eggs and milk, etc., as well as processed agricultural products that meet the conditions and standards of green product specifications. Nowadays, when energy and resources are being wasted and the ecological environment is gradually deteriorating, the emergence and widespread use of green agricultural products can improve the harm of harmful substances to the environment [27]. For this reason, Fraccascia, Giannoccaro, and Albino [28] suggested that any product that can cut down and reduce damage to the environment in the life cycle from its growth to recycling can be called a green product. The green products in this study refer to green agricultural products. It is defined as using environmentally friendly resources and production methods for agricultural products to reduce damage to the environment and achieve the functions of a friendly environment and sustainable development.
The emergence of the farmers' market is to help farmers have a platform for selling agricultural products. It enables farmers to increase their income and drives the development of agriculture, local communities, as well as economic growth. As it takes into account the environment to achieve sustainable development, it has gradually become a better choice for promoting green products [29]. In addition, the market encourages the public to break away from the previous consumption mode and switch to another way of purchase that is friendly to the environment and land and can make food a safe choice for farmers and consumers as well [1].
Food and agricultural education can be defined as an experiential education activity that cultivates the public's understanding of the basic agricultural production process and achieves the connection between agriculture, environment, and diet [30]. Via such an experiential education activity, the public's respect for the land, the environment, the producers, and the food itself is cultivated to achieve the sustainable development of health and land agriculture [31][32][33]. In other words, food and agricultural education not only attaches importance to experiencing agricultural education activities, but also the need to understand environmental education. The public can better experience the production of nature, cultivate cherishment toward the environment, and transform the learned experience and knowledge into their own experience accumulation through their personal experience of agricultural activities and understanding of environmental resources, ecology, and pollution [34].
Promoting food and agricultural education allows people to contact agriculture. Its related industries have shifted their educational background to agricultural education and health, nutrition, and natural resources [35]. In order to implement the spirit of "food and agricultural education" in daily life, the localization of the product and focusing on the locally produced crops must be put into consideration [36]. People should learn about the source of the diet and its nutritional value via getting to know the food materials, learn to cook the food materials, and know the place of production and the producer. They can establish the knowledge of food and agricultural education in the most basic direction and maximize the benefits of food and agricultural education. Accordingly, the deepened relationship between food and the environment, and the closer connection between food and local emotions through food and agricultural education can better promote food safety and the culture of green consumption [37].
In summary, nowadays, when energy and resources are being wasted and the ecological environment is deteriorating, the emergence and widespread use of green products are one of the ways to eliminate the environmental hazards due to harmful substances [38]. To extend the definition of green products, it should be considered as the use of environmentally friendly resources and production methods for agricultural products to reduce damage to the environment, achieve a friendly environment and sustainable development. That is, under the above premises, how to promote the emerging and widespread use of green products is a cause for concern. In the process, there should be more discussions on the functions of farmers' markets for the promotion and knowledge education of green products [1,2].
Research on Correlation between Product Knowledge and Purchase Intention
Knowledge is one of the key factors that affect consumers during the purchase process [39]. Consumers' perceived value of a product and its price, quality, product attributes, and social culture are all important factors affecting purchase behavior and intention [40]. As suggested in the research of Chung et al. [41], product knowledge is an important resource for a certain place or area, and the amount of product information will affect the purchase behaviors of individuals and groups, and can be used as the information for purchase decision making [42].
Shih's [43] research mentioned that consumers will use product-related knowledge for the evaluation of purchasing a certain product. When the externally transmitted and provided product information is more complete and detailed, the consumers' product knowledge will increase. However, when consumers lack product information and related knowledge, they must seek external professional knowledge to support their purchase decisions to evade the risk of uncertainty [44]. Therefore, the research of Chao [45] further shows that consumers will obtain product-related information and knowledge from their own experience and external information, resulting in a series of intentions and behaviors to purchase products.
Based on the familiarity and understanding of the product, it is easy to generate value and increase confidence and attitude, which in turn affects the purchase intention [46]; on the contrary, those consumers who lack product information will change their acceptance of product information due to the change in their environment [47]. Therefore, the existence of external expertise has a significant impact on consumers with high product knowledge [48]. The research of Kong and Chang [49] also pointed out that when consumers understand product information and have a preference for and valuable perception of the product, the product-related knowledge obtained in the purchase of products will indirectly affect consumers' purchase intentions.
Based on the aforementioned research, product knowledge has an impact on consumers' purchase intentions. The more product information and knowledge are provided by external expertise; the more consumers' product purchase intentions will be stimulated [50,51]. Consumer behaviors and preferences will be affected by the cognition and acquisition of product-related knowledge. Consumers with low product knowledge are more likely to be influenced in their purchasing behaviors due to external recommendations [52]. The level of product knowledge has significant impacts on purchase intention [53]. The research results also show the real understanding of the product knowledge by consumers can promote the purchase intention, that is, product knowledge will have a positive impact on the purchase intention [54]. Therefore, if consumers have sufficient product knowledge, it is more likely to encourage consumers to have a higher purchase intention of products and services [55].
In summary, according to previous scholars' research and research results, product knowledge usually positively affects consumers' purchase intentions and behaviors. Therefore, this study inferred that the product knowledge conveyed through the food and agricultural education viewpoints of farmers in the farmers' market will affect consumers' purchase intentions. The following hypothesis is inferred from this study: Hypothesis 1 (H1). Product knowledge positively and significantly affects consumers' purchase intention.
Research on Correlation between Green Perceived Value and Purchase Intention
The research of Petrick [56] considers that the quality and feelings of products and services will have an impact on consumers' purchase behavior and intention [57]. In other words, when the value of products and services is high, the perceived value will increase, making the purchase intention relatively increased. The perceived value is derived from the judgment of consumers on the product or service, and the cost will become one of the factors that affect the consumers' perceived value.
Continuing the above, Jamal and Sharifuddin [58] pointed out that if the cost paid is higher, the perceived value will decrease; conversely, if the cost paid is lower, the perceived value will increase. The study by Lai [59] also showed that when the values and feelings perceived by consumers are higher, indicating that the benefits are higher than the cost, it means that the positive value is obtained, thereby affecting the consumer's purchase intention. Perceived value is one of the most significant factors affecting consumers' purchase intentions and has a positive impact on consumers' behaviors and intentions to purchase products [60]. Gan and Wang [61] mentioned that the perceived value has a positive impact on consumers' purchase behaviors. Patterson and Spreng [62] indicated that consumers' views explain the generation of perceptual value, and the concept of "green" is added for discussion. The definition of the green perception value is that consumers feel the concept and expectation of environmental sustainability from the products and services they receive or pay for. This study defines the green perception value as the consumer's subjective feeling that when purchasing green products, they believe that the product does not use pesticides and contributes to the sustainable development of the environment.
Regarding green perceived value, Yu [63] pointed out that the development of an environmental perspective from a green perspective will affect consumers' green perceived value and their attitude and willingness to purchase green and environmentally-friendly products. However, Chen's [64] research mentioned that consumers' perceived value of green and environmentally-friendly products affects their purchase intentions and helps inspire producers to produce more high-quality products. Therefore, for consumers, the green perceived value is quite important. In addition, Chia's [65] research results indicated that the green perceived value obtained from green products positively and significantly affects consumers' purchase intention. It was even revealed by the research results of Han et al. [66] that if the product's green perceived value can reach consumers' expectations, it will have a positive impact on consumers' purchase intentions.
In summary, according to previous studies and research results of scholars, the green perceived value usually positively affects consumers' purchase intention. Therefore, this study inferred that the green perceived value of consumers in the farmers' market will affect their purchase intention. Below is the hypothesis inferred by this study: Hypothesis 2 (H2). Green perceived value positively and significantly affects consumers' purchase intention.
Analysis of Impact of Trust on Product Knowledge and Purchase Intention
Trust is an important factor bridge between the person who is trusted and the person who trusts, while the source of establishing trust is the person who trusts receives the professional knowledge and capacity provided by the person who is trusted [67]. In this regard, the research of Koohang et al. [68] showed that the amount of knowledge helps to increase consumers' trust in products. In other words, when consumers have confidence in the product brand and the provider of product knowledge, they have a certain degree of understanding of product-related information [69].
The degree of consumers' trust in a product will affect their choice and intention to purchase a product [70]. In other words, since trust is a consumer's response of dependence and affirmation to the product and its brand, different products will reflect different relevant information and knowledge, which will also affect consumers' choices of products. If the understanding and knowledge of the product are enhanced, trust will improve consumers' uncertainty about the product and increase their degree of trust [71]. Therefore, consumers' increased trust in the product can have more influence on consumers' behavior in purchasing the product [72]. Furthermore, past experience and word-of-mouth marketing promote trust in the product, product-related knowledge, and personnel, which will enhance consumers' purchase intention of the product and brand.
In continuation of the above, the research of Ridings et al. [73] proposed that the existence of trust significantly affects the product knowledge acquired by consumers, and the previous product-related knowledge also affects their own attitudes, thus, affects the trust in the product. For an in-depth discussion, the research results of Deng [74] illustrated that consumers' product knowledge has a positive impact on their trust. In addition, as shown in the research results of [75][76][77][78] Chen and Barnes, Mainardes and Cardoso, Guo, Yan, and Feng [75][76][77][78], trust has a positive and significant impact on consumers' willingness to buy products. Furthermore, Chou [79] pointed out that if the damage to consumers' interests is reduced, the product and product-related knowledge gained can make consumers have a trust relationship, which in turn affects their behavior and willingness to buy products. Finally, according to the research results of Lin [80], the higher the degree of trust that consumers have, the stronger the influence on the relationship between their product knowledge and purchase intention.
In summary, according to previous studies and research results of scholars, trust not only positively affects product knowledge, but also positively affects consumers' purchase intentions. Therefore, this study inferred that consumers' trust in the farmers and the agricultural products they have sold will strengthen the relationship between the product knowledge acquired by consumers and the purchase intention. The following hypothesis was inferred by this study: Hypothesis 3 (H3). When consumers have a higher sense of trust in farmers of the farmers' market, it will increase the effect of product knowledge and purchase intention.
Impact of Local Attachment on Trust, Product Knowledge, and Purchase Intention
According to the research of Kyle and Mowen, and Scannell and Gifford [81,82], local attachment is an emotional connection between people and the environment and place. When an individual has a special emotional relationship and identification with the place and environment, it makes the individual generate more interest and investment in cognition and behavior. Wang [83] pointed out that emotional relationships between individuals and places, specific places, such as product knowledge and cognitive preferences, as well as product beliefs and behaviors will influence each other. In other words, the individual's emotional connection to a place and a specific place will produce cognitive and emotional behavioral input to the product [81]. Then, Chiu [84] pointed out that the higher degree of consumers' understanding of the knowledge about the place and products, the more it will affect the emotional connection that it produces.
As seen from the above discussion, local attachment is the cognitive, emotional, and functional connection between consumers and a place, and when local attachment is combined with behavior, emotion will affect the behavior of individuals [85]. However, the degree of local attachment is affected by contextual preferences. At the same time, the amount of previous experience also significantly influences local attachment [86]. In this regard, the level of willingness to consume depends on the perception of the local environment and place. Therefore, the degree of local attachment increases or decreases consumers' purchase behavior and intention [87].
In addition, the amount of trust also affects the degree of local attachment. When consumers have a higher level of trust in cognition, their emotions and attitudes towards local attachment will also increase [88]. The results of Yen [89] showed that consumers' perceived involvement affects their emotional relationships. Similarly, the results of Kyle et al. [90] also supported that consumers' involvement has a significant impact on local attachment. Furthermore, it was found in the results of Chen [91] that the greater the emotional degree of local attachment, the more it strengthens consumers' purchase behavior and intention. At last, according to Chiu's [84] research results, knowledge has a positive relationship with local attachment.
In summary, according to previous studies and research results of the scholars, local attachment has a positive impact on product knowledge. Moreover, the amount of emotional relationship strengthens consumers' purchase intentions and behaviors. In addition, the level of trust can cause interference in local attachment. Therefore, this study inferred whether local attachment will strengthen the relationship between trust and consumer product knowledge and purchase intention. The following hypothesis was inferred from this study: Hypothesis 4 (H4). When consumers have more local attachment to the farmers' market, it will increase the relationship between their trust in the farmers' market and the consumer's product knowledge and purchase intention.
Sample and Data Collection
The subjects of the questionnaire (please refer to Appendix A for details) for this study were set to be consumers who go to farmers' markets to promote food and agriculture education. Therefore, the researcher personally went to three large farmers' markets in Taiwan to distribute the questionnaires from July to August 2020, and a total of 396 questionnaires were collected. After 45 invalid questionnaires with incomplete responses were removed, there were 351 valid questionnaires and the response rate of valid questionnaires was 88.6%. The demographic description of the survey samples are presented in Table 1, the majority were women (N = 280, 79.8%), with a total of 280 women (counting 79.8% of the total samples); in the age distribution, the majority were 41-50 years old (N = 104, 29.6%), followed by 51 years old (inclusive) and older (N = 88, 25.1%), and 21-30 years old (N = 88, 25.1%); in terms of education level, the majority were colleges/junior college (N = 216, 61.5%); in the part of marital status, the majority were married and have children (N = 176, 50.1%), followed by unmarried people (N = 145, 41.3%). In terms of habitual place of shopping (multiple choice), the traditional market was the majority (N = 258, 40.7%), followed by the farmers' markets (N = 139, 21.9%); in terms of monthly purchase frequency, the majority was 2-3 times (N = 129, 36.8%), followed by 4 times (inclusive) and above (N = 96, 26.8%); finally, in terms of the monthly average disposable amount for food, the majority was NTD 5001-6000 (N = 87, 24.8%), with a total of 87 people (approximately 24.8% of total samples), followed by NTD 5000 (inclusive) or less (N = 75, 21.4%).
Statistical Analysis
This study used structural equation modeling (SEM) to investigate the structural relationships between local attachment, product knowledge, green perception value, purchasing intentions, and model adaptation. SEM is a method of testing and improving practical models that enables the testing of theoretical models and can explain the causal relationship between variables in a model based on statistical dependence hypotheses. SPSS 22 (IBM, Armonk, NY, USA) analysis is used for descriptive statistics, reliability analysis, and hierarchical regression, and Amos 22 is used to analyze test factors and to model structural equations. Figure 1 presents a research framework to understand the relationships between product knowledge, green perceived value, purchase intention, local attachment, and trust to promote food and agricultural education on consumer purchase intention in Taiwanese farmers' markets based on literature reviews. In this framework, the independent variables are product knowledge and green values, purchase intentions are dependent variables, and local engagement and trust are modulator variables. The questionnaire elements were designed based on the review of relevant literature and included the use of a Likert scale of five points, and the five represent 'strongly agreed' and the one represents "strongly disagree" (for details, see Appendix A). In the product knowledge section, three items were set, including items such as 'food and agricultural education promoted by farmers who are familiar with their agricultural products help me increase my knowledge of agricultural products', and was adapted from Lee with a Cronbach's α of 0.841 [92]. The three-item scale of the survey's perception value section includes items such as 'I think buying green products can contribute to the sustainable development of the environment', and can be adapted from Chia with a Cronbach's α of 0.902 [65]. The four-item scale in the purchasing intentions included items such as 'If there are agricultural products sold on the farmers' market of food and agricultural education, I would like to buy them', and was adapted from Huang with a Cronbach's α of 0.894 [47]. The survey's four-item trust section includes items such as 'I think that the agricultural products sold by farmers on the farmer' market are trustworthy', and was adapted from Hsiung with a Cronbach's α of 0.924 [93]. In the local attachment section, the four-item scale includes items such as 'I think the green products sold in the local farmers' market are attractive to me', and was adapted from Pan with a Cronbach's α of 0.882 [94]. The Cronbach's α coefficient of these structures exceeded the suggested 0.70 levels. [95], furthermore, each scale is listed in Table 2. Table 2 shows the results of the reliability and convergence validity analysis of each construction. Stewart and Volt [96] stated that Cronbach's α coefficient is high at 0.7, while the coefficient is low at 0.30. Cronbach's alpha for each dimension is greater than 0.80, indicating high reliability [95]. Fornell and Larcker [97] argue that the composite reliability of persistent variables is greater than 0.60. The high CR of the latent variables indicates that the test variable is valid to measure the latent variable. The coefficient of the CR of variables studied ranges from 0.842 to 0.922, showing that the model is well unified internally. The average difference for each factor (AVE) is 0.640-0.763, higher than the recommended 0.5 standards [98]. However, the factor load value (0.57-0.915) is higher than the recommended 0.5 [99]. In Table 3, the average, standard deviation and correlation between the constructions are presented. There has been found to be a significant positive correlation between knowledge about products and intention to purchase. (r = 0.640, p < 0.01), green perceived value and in green of purchase intention (r = 0.472, p < 0.01), trust and purchase intention (r = 0.503, p < 0.01), and local attachment and purchase intention Table 2 shows the results of the reliability and convergence validity analysis of each construction. Stewart and Volt [96] stated that Cronbach's α coefficient is high at 0.7, while the coefficient is low at 0.30. Cronbach's alpha for each dimension is greater than 0.80, indicating high reliability [95]. Fornell and Larcker [97] argue that the composite reliability of persistent variables is greater than 0.60. The high CR of the latent variables indicates that the test variable is valid to measure the latent variable. The coefficient of the CR of variables studied ranges from 0.842 to 0.922, showing that the model is well unified internally. The average difference for each factor (AVE) is 0.640-0.763, higher than the recommended 0.5 standards [98]. However, the factor load value (0.57-0.915) is higher than the recommended 0.5 [99]. In Table 3, the average, standard deviation and correlation between the constructions are presented. There has been found to be a significant positive correlation between knowledge about products and intention to purchase. (r = 0.640, p < 0.01), green perceived value and in green of purchase intention (r = 0.472, p < 0.01), trust and purchase intention (r = 0.503, p < 0.01), and local attachment and purchase intention (r = 0.440, p < 0.01). The results show that consumers understand products more, perceive green values, and trust and support local communities. They are more likely to buy green products because they have more green products.
Structural Equation Modeling and Empirical Analysis
We used Moment Structure (AMOS) 22 analysis to model structural equations (SEM) and assess path relationships between structures [70]. The results show that the measurement model is good for data (χ 2 /df = 1.318, GFI = 0.976, AGFI = 0.959, CFI = 0.995, NFI = 0.981, IFI = 0.995, RMR = 0.012, RMSEA = 0.030). The ratio χ 2 /df is less than 2 [100], the GFI, the AGFI, the NFI of the CFI and the IFI exceed the recommended threshold of 0.90 [101], and the RMSEA and SRMR values are below the 0.08 separated [102]. This indicates that the data methods used to model the studies are appropriate. Figure 2 and Table 4 show the hypotheses tested from the model data.
(r = 0.440, p < 0.01). The results show that consumers understand products more, perceiv green values, and trust and support local communities. They are more likely to buy green products because they have more green products.
Structural Equation Modeling and Empirical Analysis
We used Moment Structure (AMOS) 22 analysis to model structural equations (SEM and assess path relationships between structures [70]. The results show that the measure ment model is good for data (χ 2 /df = 1.318, GFI = 0.976, AGFI = 0.959, CFI = 0.995, NFI 0.981, IFI = 0.995, RMR = 0.012, RMSEA = 0.030). The ratio χ 2 /df is less than 2 [100], the GFI the AGFI, the NFI of the CFI and the IFI exceed the recommended threshold of 0.90 [101] and the RMSEA and SRMR values are below the 0.08 separated [102]. This indicates tha the data methods used to model the studies are appropriate. Figure 2 and Table 4 show the hypotheses tested from the model data. Hypothesis H1 was supported. As a result, knowledge of products is strongly and positively related to the intentions of customers to purchase. In the case of the pathway coefficient, 11 is 0.680 (p < 0.001). In other words, consumer product knowledge is positiv and important with respect to purchase intention; that is, the greater the knowledge o products that consumers acquire, the greater their intention to buy them. Hypothesis H was supported. In other words, the perceived value of green is highly positive and related to customers' purchase intention. The path coefficient 12 is 0.148 (p < 0.01). This mean that consumers have a positive and important relationship between the value of thei green products perceived and purchase intention; the higher the value perceived by con sumers in green, the greater their desire to buy. Hypothesis H1 was supported. As a result, knowledge of products is strongly and positively related to the intentions of customers to purchase. In the case of the pathway coefficient, 11 is 0.680 (p < 0.001). In other words, consumer product knowledge is positive and important with respect to purchase intention; that is, the greater the knowledge of products that consumers acquire, the greater their intention to buy them. Hypothesis H2 was supported. In other words, the perceived value of green is highly positive and related to customers' purchase intention. The path coefficient 12 is 0.148 (p < 0.01). This means that consumers have a positive and important relationship between the value of their green products perceived and purchase intention; the higher the value perceived by consumers in green, the greater their desire to buy.
Testing Moderating Effects
Hierarchical regression was used to analyze the moderating effect of trust on the relationship between product knowledge and purchase intention. The analysis results are Table 5. The regression model M1 in this table shows that the value for the effect of product knowledge on purchase intention in the first stage was β = 0.640, p < 0.001, the coefficient reached a significant level, and it showed a positively significant correlation. The regression model M2 shows that the value of the impact of trust on purchase intention in the second stage was β = 0.344, p < 0.001, the coefficient reached a significant level, and it showed a positively significant correlation. The regression model M3 shows that after adding the interaction value between product knowledge and trust in the third stage, its value was β = 0.093, p < 0.05, the coefficient reached a significant level, it showed a positively significant correlation, and the explanatory variation (R 2 ) increased to 0.497, the change value of variance explained (∆R 2 ) was 0.009, which showed that trust had a moderating effect on the relationship between product knowledge and purchase intention. In addition, based on the results of the hierarchical regression analysis, the values were divided into two parts, low-trust groups and high-trust groups based on the average value of trust, and the effects of these two groups were observed to plot the diagram of the moderating effect of trust on product knowledge and purchase intention, as shown in Figure 3. When consumers had a higher sense of trust in the farmers of farmers' market, the more the effect between product knowledge and purchase intention was increased; in addition, when consumers had low trust in the farmers of farmers' market, it also increased the effect between product knowledge and purchase intention. Although both high trust and trust can produce an increased effect compared to low trust, high trust was more effective in predicting the effect between product knowledge and purchase intention. Therefore, Hypothesis H3 was supported. It means that the more trust that consumers have in the farmers of farmers' markets, the more the effect between product knowledge and purchase intention will be increased. Table 5. Results of hierarchical regression analysis.
Variables
Purchase Intention
Testing Moderating Effects
Hierarchical regression was used to analyze the moderating effect of trust on the relationship between product knowledge and purchase intention. The analysis results are presented in Table 5. The regression model M1 in this table shows that the value for the effect of product knowledge on purchase intention in the first stage was β = 0.640, p < 0.001, the coefficient reached a significant level, and it showed a positively significant correlation. The regression model M2 shows that the value of the impact of trust on purchase intention in the second stage was β = 0.344, p < 0.001, the coefficient reached a significant level, and it showed a positively significant correlation. The regression model M3 shows that after adding the interaction value between product knowledge and trust in the third stage, its value was β = 0.093, p < 0.05, the coefficient reached a significant level, it showed a positively significant correlation, and the explanatory variation (R 2 ) increased to 0.497, the change value of variance explained (∆R 2 ) was 0.009, which showed that trust had a moderating effect on the relationship between product knowledge and purchase intention. In addition, based on the results of the hierarchical regression analysis, the values were divided into two parts, low-trust groups and high-trust groups based on the average value of trust, and the effects of these two groups were observed to plot the diagram of the moderating effect of trust on product knowledge and purchase intention, as shown in Figure 3. When consumers had a higher sense of trust in the farmers of farmers' market, the more the effect between product knowledge and purchase intention was increased; in addition, when consumers had low trust in the farmers of farmers' market, it also increased the effect between product knowledge and purchase intention. Although both high trust and trust can produce an increased effect compared to low trust, high trust was more effective in predicting the effect between product knowledge and purchase intention. Therefore, Hypothesis H3 was supported. It means that the more trust that consumers have in the farmers of farmers' markets, the more the effect between product knowledge and purchase intention will be increased.
Testing Three-Way Interaction Effects
This study used product knowledge as the independent variable, purchase intention as the dependent variable, and local attachment and trust as the moderator variables. After the multiplication of the moderator variables, the three-way interaction was adopted to examine whether the multiplication of local attachment and trust have a moderating effect between product knowledge and purchase intention. At the same time, several evaluation criteria were used as the basis for determining the existence of the moderating effect. First, if the coefficient value is positive, it means that the moderating effect is positive moderation; otherwise, it is negative moderation. Furthermore, if the value obtained reaches a significant level and is within the confidence interval of the low limit confidence interval (LLCI) and the upper limit confidence interval (ULCI), and the value 0 is not included, there is a moderating effect; otherwise, there is no moderating effect.
In this study, the results of the analysis of the effect of the multiplication of local attachment and trust on product knowledge and purchase intention are summarized in Tables 6 and 7. According to the results of the three-way interaction research, the coefficient value of the multiplication of local attachment and trust on product knowledge and purchase intention was positive, that is, it was a positive moderation. Furthermore, its value was p = 0.033, and the change value of variance explained ( R 2 ) was 0.006, indicating that the coefficient reached a significant level. In addition, the values of the range in the confidence interval did not include the value 0, indicating that the multiplication of local attachment and trust had a moderating effect on the relationship between product knowledge and purchase intention. However, in order to clarify the type of three-way interaction, it is necessary to verify whether the slope of the hypothetical interaction is significant. As shown in Table 8, regardless of high trust and low trust, the simple slopes are all significant for different levels of local attachment (b = 0.32-0.75, p < 0.01). In addition, according to the standard of Dawson and Richter [103], a three-way interaction diagram was drawn by this study, as shown in Figure 4. When consumers have high trust in and high local attachment to the farmers' market, it can better improve the relationship between product knowledge and purchase intention of consumers than low trust and high local attachment. Therefore, as mentioned above, Hypothesis 4 proposed in this study was established. intention of consumers than low trust and high local attachment. Therefore, as mentioned above, Hypothesis 4 proposed in this study was established.
Relationship between Product Knowledge, Green Perceived Value, and Purchase Intention
According to the analysis of this study, the results showed that product knowledge positively and significantly affected consumers' purchasing intentions. When farmers in the farmers' market convey their knowledge of agricultural products from the perspective of food and agricultural education, consumers can deepen their knowledge of agricultural products, thus increasing their purchase intention of the agricultural products. Therefore, the more knowledge of agricultural products transmitted by farmers through food and agricultural education viewpoints that was acquired by consumers, the more likely it was to produce a higher purchase intention. The results of this study are consistent with the research results of Horn and Salvendy, and Chocarro et al. [50,51], that is, it is confirmed that in Taiwan, the knowledge of the product will affect consumers' purchase intentions, and the higher the knowledge of the product, the more it will stimulate consumers' purchase intentions for products.
In addition to the above, the results of this study also showed that the perceived value of green positively and significantly affected the intention of consumers to purchase. This result is in line with the research results of Chia [65], who mentioned that the green perceived value that consumers feel from green products will affect the intention to purchase the product of consumers. When consumers feel that the agricultural products sold on the farmer's market contain the value and benefits that contribute to the sustainable development of the environment, it will increase their green perceived value and increase their
Relationship between Product Knowledge, Green Perceived Value, and Purchase Intention
According to the analysis of this study, the results showed that product knowledge positively and significantly affected consumers' purchasing intentions. When farmers in the farmers' market convey their knowledge of agricultural products from the perspective of food and agricultural education, consumers can deepen their knowledge of agricultural products, thus increasing their purchase intention of the agricultural products. Therefore, the more knowledge of agricultural products transmitted by farmers through food and agricultural education viewpoints that was acquired by consumers, the more likely it was to produce a higher purchase intention. The results of this study are consistent with the research results of Horn and Salvendy, and Chocarro et al. [50,51], that is, it is confirmed that in Taiwan, the knowledge of the product will affect consumers' purchase intentions, and the higher the knowledge of the product, the more it will stimulate consumers' purchase intentions for products.
In addition to the above, the results of this study also showed that the perceived value of green positively and significantly affected the intention of consumers to purchase. This result is in line with the research results of Chia [65], who mentioned that the green perceived value that consumers feel from green products will affect the intention to purchase the product of consumers. When consumers feel that the agricultural products sold on the farmer's market contain the value and benefits that contribute to the sustainable development of the environment, it will increase their green perceived value and increase their willingness to buy agricultural products. Therefore, when consumers' green value is higher, their purchase intention is higher.
Moderating Effect of Trust on Product Knowledge and Purchase Intention
Based on the research results of Ridings et al. [73], the existence of trust will significantly affect the knowledge about the product acquired by consumers. The higher the trust, the higher the acquired knowledge of the product. According to the analysis of this study, the results also showed that trust had a moderating effect between product knowledge and purchase intention, that is, when consumers had a higher sense of trust in farmers at the farmers' market, it would strengthen the relationship between product knowledge and purchase intention. In further discussion, the research results of Chen and Barnes, and Mainardes et al. [75,76] all show that trust has a positive effect on consumers' willingness to buy products. In addition, the research results of Lin [80] mentioned that the higher the degree of trust that consumers have, the more likely it will strongly affect the relationship between product knowledge and purchase intention. It can be seen that the results of this study are similar to the research results of Ridings et al. [73], Chen and Barnes [75]), Mainardes et al. [76], and Lin [80]. That is, if farmers on the farmers' market give consumers a sense of trust so that consumers feel trust in the farmers and the agricultural products they sell, it can increase consumers' knowledge of the agricultural products sold on the farmers' market, and promote their intentions to purchase agricultural products.
In addition to the above, according to the research results of Chiu [84], the higher the consumer's knowledge of places and products, the more positive emotional connections they will have to place attachment. The results of this study further illustrated that the multiplication of place attachment and trust had a moderating effect between product knowledge and purchase intention, that is, when consumers had a more local attachment to the farmers' market, it would strengthen the relationship between trust with the knowledge of the product acquired by consumers and their purchase intention.
Furthermore, according to the research results of Kozak and Rimmington [87], the degree of local attachment increases or decreases the purchase behavior and intention of consumers. Furthermore, Lee and Yun [88] noted that the degree of trust accumulates through the relationship generated by emotions. When consumers have different trust in cognition, their local attachment emotions and attitudes will vary. Therefore, the results of this study also confirmed the results of the research of Kozak and Rimmington [87] and Lee and Yun [88]. That is, when consumers are more attached to the farmer's market, the more it will affect consumers' trust in farmers and the agricultural products they sell. The degree to which consumers obtain knowledge about agricultural products will affect the possibility of consumers' purchase intention of agricultural products.
Implications and Contributions
Farmers' markets are the ultimate green sector of the economy. They are major successes and spurring sustainable economic development. From this study, what additional 'value' can 'green add' to the functionality of green farmers' markets would be illustrated in multiple dimensions of functionality. Though there exist 'functionality' differences between sales of 'fresh products' through traditional shops and farmers' markets, intangible functionality (leisure food safety, ease of purchase, etc.) and emotional functionality (personal contact, trust, localization, etc.) would play an important role in addition to physical functionality and economic functionality in consumers' purchase intention in Taiwan. Although the findings are local to Taiwan, its characteristics are typical of food and agricultural education, as well as farmers' markets worldwide, especially in Asia.
Research Limitations and Recommendations
First, in terms of gender in the samples of this study, there were a larger number of female samples than male samples. However, the previous research and literature on farmers' markets or purchase intentions for agricultural products show that the proportion of women is relatively high, such as the research of Chung [10] on the impact of product knowledge and quality of relationship on the intention of purchasing in the organic farmer market which shows that the proportion of women is 76.4%. The study by Wang [104] of the impact of market orientation and community customer relationship management on organic small farmers with intention of purchasing as an example shows that the proportion of women is 83%. It is estimated that one of the limitations of this research is that women were the main consumer group in farmers' markets and for the purchase of agricultural products.
In addition, this study aimed to understand the relationship between the purchase intentions of farmers' markets consumers and the promotion of food and agricultural education. Only the five items of product knowledge, perceived value of green, purchase intention, trust, and local attachment were adopted as variables to discuss the impact of the promotion of food and agricultural education on the purchase intention of consumers. Of course, there are other more important factors that affect the purchase intention of the consumers of farmers' markets that promote food and agricultural education, such as experience value. According to the previous research and literature, experience is the important basis for food and agricultural education. So, this is the second limitation of this study. Finally, in terms of research field, this study only conducted research and investigations in farmers' markets that promote food and agricultural education. However, farmers' markets are not only classified by promotion of food and agricultural education or not, but also divided into farmers' markets with a fixed time and place and farmers' markets of the type of holidays. This is the third limitation of this study.
In view of the limitations of the above-mentioned research, it is recommended that more different factors be included in subsequent research to explore consumers' purchasing intentions in the farmers' market after food and agricultural education promotion and their impacts. This is because it cannot only facilitate a better understanding of the factors influencing consumer purchase intentions, but also provide the corresponding suggestions and policies to farmers in the farmers' market based on the results of the discussion. Finally, and also the most importantly, since the main purpose of the study focuses on consumers' purchase intention with respect to farmers' markets, it would likely fail to bridge the intention-behavior gaps. When it comes down to 'green', there would be a big difference between what people say (often politically correct) and what people do (more selfish oriented). To address the gaps, research about what actually happens in the buying processes in the markets is strongly urged. First, the results of this study showed that the knowledge about the product acquired by consumers in farmers' markets had a positive impact on the purchasing intentions of consumers. Consumers pay more attention to the knowledge of the product and related information with improved health awareness. Changes in consumer demand make them more willing to choose green products that are environmentally friendly and the ingredients themselves. Furthermore, all green products sold on the farmers' markets are grown and sold by farmers themselves, so they can provide consumers with valuable information and knowledge about green products. In other words, farmers on the farmer's market play an important role in acquiring knowledge about green products for consumers. According to the investigation by Mercier [35], the promotion of food and agricultural education can allow the public to be personally exposed to agriculture and its related industries, make the education background not limited to agricultural education, but also cover health, nutrition, and natural environmental resources. Therefore, the farmers' market should be able to cooperate actively with the central government and local government units to jointly organize special exhibitions of agricultural products to make the public better understand green products and other related food and agricultural experiences. When the farmers' market provides more information and knowledge about the agricultural products sold, consumers can clearly and fully understand green products, which will further promote their intention to purchase agricultural products.
Secondly, according to the results of this study, the green perceived value had a positive impact on consumers' purchase intentions. Based on the fact that green products are important for saving resources and energy, as well as reducing waste emissions, and have recycling efficiency, which can greatly reduce the harm to the environment, they have more advantages for environmental protection and sustainable development [37]. Therefore, when the farmers' market is more widely established and the sale of green agricultural products is expanded, the perceived value of green by consumers should be further improved. As a result, consumers can think that the agricultural products sold in the farmers' market are environmentally friendly, and it should increase their willingness to buy agricultural products, which indirectly helps reduce environmental damage and achieve environmental sustainability.
Strengthening Trust to Enhance Agricultural Product Knowledge and Green Product Purchase Intention
Furthermore, more discussions should be held on the functions of green product promotion and knowledge education in farmers' markets [1,2]. The results of this study showed that trust had a moderating effect between product knowledge and purchase intention. The higher the consumer trust in farmers in the farmers' market and the agricultural products they sell, the more they will recognize the knowledge and information about the product they propose, and this increases the information related to the product that consumers can obtain and increases the intention of purchasing at the same time. Therefore, farmers in the farmers' markets are recommended to use social networking sites (such as Facebook and LINE) to establish a friendship and trust relationship with consumers through the dissemination of information, exchanges, and communication. When farmers in the farmers' market frequently communicate and interact with consumers, consumers will feel a sense of trust. They are also more willing to believe in information about green products delivered by farmers, and thereby increase their purchase intention of agricultural products.
Finally, the results of this study showed that the multiplication of local attachment and trust had a moderating effect between product knowledge and purchase intention. The higher the consumer's local attachment to the farmers' market, the more it will affect the level of trust in the farmers of the farmers' market and the agricultural products sold on the product-related information that consumers can obtain, and thereby increase the purchase intention. Therefore, it is recommended that the farmers' markets cooperate with local government units to promote food and agricultural education, and promote the farmers' markets through newspapers, magazines, advertisements, and even through the YouTube platform. In this way, when farmers in a farmers' market can accumulate consumers' attachment to the place through more local cohesion, it should also be able to increase consumers' trust in the product knowledge they have obtained and increase their purchase intention of agricultural products. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement:
The datasets generated for this study are available on request to the corresponding author.
Conflicts of Interest:
The author declares no conflict of interest. | 2022-02-25T23:59:39.859Z | 2022-02-22T00:00:00.000 | {
"year": 2022,
"sha1": "db95a1bbe0a57945999187a50ce9be348cb8bccd",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2304-8158/11/5/630/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "842a30ef2848188b7ba9940bdcc5cba3073bfe13",
"s2fieldsofstudy": [
"Agricultural and Food Sciences",
"Education",
"Economics"
],
"extfieldsofstudy": [
"Medicine"
]
} |
51910699 | pes2o/s2orc | v3-fos-license | Polymorphisms of Catechol-O-Methyl Transferase (COMT) Gene in Vulnerability to Levodopa-Induced Dyskinesia
Purpose. Parkinson's disease (PD), a common neurodegenerative disorder, is usually treated with Levodopa (L-DOPA). The use of this drug, however, is severely limited by the development of side effects of the motor system: Levodopa-induced dyskinesia (LID). The aim of this study is to investigate the association between seven COMT gene single-nucleotide polymorphisms (SNPs) and the development of LID in patients with PD. Methods. 232 Caucasian patients with PD were investigated. 212 patients with PD received Levodopa therapy. Dyskinesia was assessed with the use of the Abnormal Involuntary Movement Scale (AIMS). Genotyping was carried out on seven SNPs of the COMT gene (rs4680, rs6269, rs4633, rs4818, rs769224, rs165774, rs174696) using a real-time PCR method, and blind to the clinical status of the subjects. Results. We found association between four SNPs, rs165774, rs4818, rs4633, rs4680, and LID. When the duration of disease was added as a covariate in regression analysis, however, the results did not reach statistical significance. Only the additive model for rs165774 was found to be close to be statistical significance (OR = 1.627 [0.976–2.741], permutation p = 0.057). Conclusions. The results failed to clearly support a contribution of the studied polymorphisms; this may be related to a dominant relationship with the disease duration confounding the effect on the prevalence of LID.
INTRODUCTION
Over 55 years after the discovery of its beneficial effects, 1,2 Levodopa remains the mainstay of treatment for Parkinson's disease (PD). The use of Levodopa for the treatment of motor symptoms at all stages of PD is supported by strong evidence. 3 Longterm treatment with Levodopa for PD, however, is frequently complicated by motor fluctuations and dyskinesias. [3][4][5][6] Monotherapy with dopamine agonists in the early phases of the disease does reduce the risk for dyskinesias compared with the use of Levodopa, but dopamine agonists are unable to prevent dyskinesias once Levodopa is added, and which is always required once disease severity progresses. 7 Chronic Levodopa treatment has been reported to result in LID in up to 45% of Levodopa users within five years. 8 Clinical heterogeneity of LID suggests a significant role of endogenous factors in determining their prevalence, although no effective drug treatment of levodopa-induced dyskinesia (LID) has yet been developed. 9 Several theories have attempted to explain the pathophysiology of this treatment complication with an ultimate goal to develop such treatments [9][10][11] , but the exact pathological mechanism has also not yet been elucidated. Recently, Ivanova and Loonen developed the hypothesis that LID is related to an increased vulnerability to excitotoxicity of indirect pathway striatal medium spiny neurons potentiated due to the contribution of increased intracellular oxidative stress. 12 This last phenomenon is expected to be caused by the same genetic composition which causes the degeneration of dopaminergic nigrostriatal neurons in patients with PD. Dopamine can be considered to be a particularly neurotoxic endogenous substance: dopamine metabolism results in the release of hydrogen peroxide, which, in turn results in the production of free radicals, which then cause cell damage. 13 This damage is limited by the influence of certain enzymes such as manganese superoxidedismutase (MnSOD), which scavenge these free radicals. 14 This neurotoxicity is expected to be potentiated by biological factors which increase the intracellular levels of dopamine.
Catechol-O-methyltransferase (COMT; EC 2.1.1.6) is one of the most important enzymes in levodopa metabolism. It was identified by Julius Axelrod at the National Institute of Mental Health (USA) in the second half of the 1950s. 15,16 COMT catalyzes methyl group transfer from S-adenosyl-Lmethionine to one of the hydroxyl groups of catechol in the presence of Mg2+ ions. COMT substrates include a wide variety of catechols (catecholamines, their hydroxylated metabolites, catecholestrogens, ascorbic acid, dietary phytochemicals, and medicinal compounds). 15,16 The major physiological role of COMT is the elimination of biologically active or toxic catechols. Of special importance is the methylation of Levodopa to 3-O-methyldopa in Levodopa/aromatic amino acid decarboxylase inhibitor-treated Parkinson's disease (PD) patients, because this metabolism markedly limits the availability of Levodopa to the brain. 15 It has also been suggested that COMT plays a relevant role in modulating prefrontal dopamine neurotransmission. 17 Polymorphic variants of the COMT gene determine the activity level of this enzyme. [17][18][19] COMT exists in two forms: the soluble form (S-COMT) which is located in cytosol, and the membrane-bound form (MB-COMT) which is anchored to the rough endoplasmic reticulum. The two forms differ only by a 50 residue long extension in the MB-form, which is the signal sequence for membrane anchoring. 20,21 In humans, the COMT gene is located on chromosome 22 band q11.21, and is composed of six exons. 15 The first two exons are noncoding, while the translation initiation codons for the membrane bound and soluble isoforms are located on the third exon. Two separate promoters direct the synthesis of two partially overlapped transcripts: one of 1.5 kb that is constitutively expressed, and another of 1.3 kb which is subject to tissue-specific transcription regulation. The short transcript translates S-COMT and the longer transcript translates MB-COMT, and also the soluble form by the leaky scanning mechanism of translational initiation. 15 The long transcript has been found in all tissues analyzed. The short transcript, on the other hand, is only found in very small amounts within the human brain. There is, however, no direct correlation between transcript and protein levels. In the human brain S-COMT represents about 30% of the total COMT. 15 Although several polymorphisms have been reported for the COMT gene, most of them appear not to have any physiological significance. However, the genetic variant Val158Met (rs4680) with low thermal stability and low COMT activity is a wellestablished polymorphism that could contribute to various neuropsychiatric manifestations. 18,19 Low COMT activity appears to have predominantly clinical effects within the prefrontal cortex (PFC) as in the PFC DA-transporters are expressed in low abundance within synapses, contrary to, for example, the striatum. 19,22 Furthermore, the nonsynonymous Ala72Ser (rs6267) polymorphism has been associated with reduced COMT enzyme activity and with a risk for schizophrenia 23 Apart from the above mentioned Val158Met (rs4680), three singlenucleotide polymorphisms (SNPs) were found: one in the S-COMT promoter region (rs6296) and two in the S-and MB-COMT coding region (c.186C>T, p.62His (rs4633), c.408C>T, p.136Lys (rs4818), next to rs4680). These were associated with differences in pain sensitivity and the likelihood of developing a chronic musculoskeletal pain condition. 24 It was shown that these haplotypes modulate protein expression by altering mRNA secondary structure, thus stressing the functional significance of synonymous variations and importance of haplotypes. 18 Another relevant synonymous variation includes c.597G>A, p.199Pro (rs769224). 18 Xu et al demonstrated an association of the G-G-G vs. A-G-G haplotype of rs4633, rs4818 and rs769224 concerning response to non-SSRIs in major depressive disorder. 25 Recently published results revealed that homozygosity for rs4633 (TT), rs4680 (AA) and of the two linked rs4633-rs4680 (TT/AA) was significantly associated to Levodopa effects in PD patients, while no significant differences were observed in patients who carried individual rs6269 and rs4818, the two linked rs6269-rs4818 and the four combined COMT SNPs. 26 These last observations contrast to the findings in patients with Attention-Deficit/Hyperactivity Disorder (ADHD) (64% using medication for it) in who the c.-98A>G (rs6269) was associated with the largest influence on the hyperactivity/impulsivity scores. In PD patients, any existing significant differences were found for individual SNPs (rs6269; rs4633; rs4818; and rs4680) in the allele and genotype frequencies between PD cases and controls, but their division into haplotypes with low, medium and high COMT activity revealed that the high activity haplotypes needed a significantly higher levodopa dosage. 27 In their study, the COMT haplotype seemed to have little influence on the development of Levodopa-induced dyskinesias. Meloto et al identified a functional marker situated in the 39 untranslated region of a newfound splice variant (c.615+739G>A; rs165774) which displays unique substrate specificity, exhibiting enzymatic activity with dopamine, but not with epinephrine. 28 They established that the pain-protective A allele of rs165774 coincides with lower COMT activity due to increased dopaminergic tone. Associations of rs165774 were also found by Higashiyama et al in patients with schizophrenia vs. healthy controls and by Seib et al in older women with depression. 29,30 Ittiwut et al found an association between rs4680 and c.615+1354C>T (rs174696) in European American patients with cocaine-induced paranoia. 31 In our current study we have investigated a possible existence of an association between rs4680, rs6269, rs4633, rs4818, rs769224, rs165774, as well as rs174696 and levodopa-induced dyskinesia in 232 Caucasian PD patients from the neurological department of the Siberian State University Hospital of Tomsk, Russian Federation. The rationale for our study was that these polymorphisms have been demonstrated to induce certain functional changes. It was hypothesized that at least some of the biochemical alterations causing these functional changes might also affect the intracellular dopamine concentration, which could result in changed vulnerability for developing levodopa-induced dyskinesia.
Patients
The work described in this article was carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki 1975, revised in Fortaleza, Brazil, 2013) for experiments involving humans. Patients were recruited from the neurology department of the Siberian State Medical University Hospital in Tomsk, Russian Federation. Written informed consent was obtained from each patient after obtaining approval for the study (protocol N 3604 10.02.2014) from the Local Bioethics Committee of the Siberian State Medical University. None of the participants had a compromised capacity/ability to consent; therefore consent from the next-of-kin was not necessary, and was not recommended by the local ethics committee. The inclusion criteria were a clinical diagnosis of Parkinson's disease according to ICD-10 (G-20) and also prior long-term use of Levodopa. Exclusion criteria were non-Caucasian physical appearance (e.g., Mongoloid, Buryats or Khakassians), relevant psychiatric symptoms or any organic brain disorders other than Parkinson's disease. These patients had not been treated with dopamine antagonists (antipsychotic or antiemetic drugs, including clozapine) for at least three years. Patients were assessed for the presence or absence of dyskinesia during 'on'-phase according to the abnormal involuntary movement scale (AIMS). [32][33][34] The AIMS scores were transformed into a binary form of LID (presence or absence of dyskinesia) with Schooler and Kane's criteria. 35 DNA analysis DNA was isolated from the leukocytes in whole peripheral blood using the standard phenolchloroform micro method. 36 Genotyping was carried out on the MassARRAY Analyzer 4 (Agena Bioscience) using the set SEQUENOM Consumables iPLEX Gold 384. DNA sample preparation for SEQUENOM MassARRAY Analyzer 4 includes several steps: a standard PCR reaction to obtain the amplification products, a shrimp alkaline phosphatase reaction to neutralize the unincorporated dNTPs in the amplification products, the PCR iPLEX Gold extension reaction, and then placing the samples on a special chip (SpectroCHIP array) using Nanodispenser RS1000 prior to loading them into the analyser. Genotyping was carried out on 7 SNPs of COMT genes (rs4680, rs6269, rs4633, rs4818, rs769224, rs165774, rs174696), which were selected after reviewing the literature for possible relevance.
Statistical analysis
Statistical analyses were performed using SPSS software for Windows, release 17. The Mann-Whitney test (MWT) was used to compare qualitative traits and χ2 test for categorical traits. The Hardy-Weinberg equilibrium (HWE) of genotypic frequencies was tested using the χ2 test. The χ2 test and the Fisher's exact test, if necessary, were used for between-group comparisons of genotype or allele frequencies.
The analysis of association between the SNPs and the phenotype was carried out using logistic regression with adjustment for covariates. Additive, dominant and recessive models were tested, and the optimum model has been chosen using Akaike Information Criterion (AIC). Experiment-wise permutations were used to address the multiple testing issue, hence Bonferroni correction was not applicable. Models with permutation p-value < 0.05 were considered statistically significant. Odds ratio (OR) and 95% confidence intervals were estimated to assess the strength of genetic effect. The following filters were applied to the SNPs: minor allele frequency, 5%; deviation from Hardy-Weinberg equilibrium, p=0.007 (assuming α=0.05 for 7 SNPs).
RESULTS
A total of 232 Caucasian patients with PD were included, consisting of 149 females and 83 males (age range from 40 to 86 years, average age 68.7±7.6 years). These patients demonstrated typical PD demographics, with a mean age of onset of 59.0 ± 9.5 years, and a mean disease duration of 9.8 ± 5.6 years. The distribution of genotypes of studied genes corresponds to the Hardy-Weinberg equilibrium (data not shown). According to the predefined criteria, 58 patients suffered from dyskinesia. We found that four out of seven tested SNPs passed the significance thresholds for association with LID. Since associations can in part be explained by the effects of additional factors, such as sex and age, we carried out a regression analysis using LID as the dependent variable, the genotypes as the fixed factor and sex and age as covariates. Using linear regression models, we found association between four SNPs, rs165774, rs4818, rs4633, and rs4680, and LID; the best model (according to AIC values) was additive for rs165774 and rs4818 and dominant for rs4633 and rs4680 (Table 1). In case of rs165774, the rare allele was associated with an increased risk of LID (OR = 1.75 [95% CI 1.14-2.72]), while in case of rs4818, rs4633 and rs4680, the rare allele or homozygote genotype for the rare allele were protective against LID (OR = 0.57 [0.34-0.92], 0.45 [0.23-0.89] and 0.46 [0.23-0.91], respectively). However, when the duration of disease was added as a covariate in regression analysis, the results did not reach statistical significance. Only the additive model for rs165774 was found to be close to statistical significance (OR = 1.627 [0.976-2.741], permutation p = 0.057).
DISCUSSION
We have failed in this study of 232 white patients with Parkinson's Disease (PD) to establish a clear relationship between any of the seven polymorphisms of the COMT gene and the prevalence of Levodopa-induced dyskinesia (LID).
The associations that were found lost their statistical significance after correcting for disease duration. This last observation may correspond with the previously described association between some COMT gene polymorphisms and the onset of disease in 143 patients with PD. 26 Watanabe et al found a slightly higher prevalence of PD in 30/118 patients with homozygosity for the low-activity allele of Val158Met (rs4680), 37 but these findings were not reproduced by Contin et al in 104 patients, and by Cheshire et al in 285 patients. 38,39 We had observed previously a significant association with PD for the c.615+739G>A (rs165774), but not Val158Met (rs4680), polymorphism and PD of the same patient population vs. 127 healthy controls. 40 De Lau et al described the existence of a prospectively assessed correlation between the A-allele of the COMT Val158Met polymorphism and an increased risk of developing dyskinesias in 219 patients with PD. 41 A comprehensive meta-analysis in which a total of 363 datasets were included, consisting 56,998 cases and 74,668 healthy controls from case control studies, as well as 2,547 trios from family based studies, showed a definite relationship between several psychiatric disorders, including, for example, attention-deficiency hyperactivity disorder (ADHD) and panic disorder for Caucasian samples. 42 We conclude that the tested polymorphisms, and in theory at least the COMT Val158Met polymorphism as described above, may have an influence on the prevalence of LID in PD, but this influence is small, and in our study confounded by a possible increased likelihood for PD to manifest itself, thus causing an increased disease duration in our patient population. In previous studies, we have observed that in both tardive dyskinesia as well as in LID, 14,36,43 limbtruncal dyskinesia may have a genetic background other than orofacial dyskinesia. Unfortunately, the number of patients in this study with LID is relatively small (N = 58), and too many suffered from both orofacial and limb-truncal dyskinesia to make further statistical analysis feasible.
The strengths of our study include the relatively large number of patients (N = 232); the thorough assessment of the severity of dyskinesia by direct clinical assessment with AIMS; the application of Schooler and Kane's criteria to conclude to the presence of LID, and the assessment of seven putatively relevant polymorphisms of the COMT gene after carefully reviewing previous findings. The weaker points are the relatively low number of patients suffering from LID (N = 58, 25%), and the co-occurrence of limb-truncal and orofacial dyskinesia in many of them. The results of this study illustrate that genetic variations of the activity of the COMT enzyme have only a modest influence on the clinical presentation of PD, and its treatment complications such as LID. The vulnerability in PD apparently differs from that in other neuropsychiatric disorders, such as ADHD. 42 It can been suggested that this may be related to low expression of DA-transporters within the prefrontal cortex and not within synapses, contrary to, for example, the striatum. 19,22 It is possible that the abundance of DA-transporters within the striatum limits the influence of variations in COMT activity, because after re-uptake within dopaminergic terminals dopamine is generally processed differently (e.g., through mitochondrial monoamine oxidase). Moreover, the highly adaptive capacity of the functional nigrostriatal extrapyramidal connectivity may limit the influence of genetic variations in enzyme activity. This influence on the activity of the cortico-striatothalamo-cortical circuitry may be less than that within the prefrontal cortex or corticoid amygdala, which may play a role in causing other neuropsychiatric disorders. 12 However, other PD manifestations, including cognitive symptoms, may be more dependent on the functioning of the prefrontal cortex. Recently, Zhang et al have demonstrated that COMT Val158Met polymorphism is probably not associated with increased risk of PD, but has an effect on the prefrontal executive function interacting with gender and dopaminergic medication. 44 Specifically addressing manifestations related to dysfunction of the PFC (vigilance), corticoid amygdala (delusions) and hippocampal complex (hallucinations) may, in future, be more awarding when studying COMT gene polymorphisms.
CONCLUSIONS
The possible association between seven putatively relevant genetic variants of COMT gene and prevalence of LID in 232 Caucasian patients with PD was studied. Although the association of four polymorphisms reached statistical significance, the relationship was masked after discounting the time frame of the proportion of patients in the sample suffering from PD.
ACKNOWLEDGMENTS
This work resulted from a collaboration between the Mental Health Research Institute in Tomsk and the Groningen Research Institute of Pharmacy (GRIP) of the University of Groningen. The Russian part is carried out within the framework of Tomsk Polytechnic University Competitiveness Enhancement Program. The authors have no conflicts of interest to report. | 2018-08-14T14:12:29.265Z | 2018-08-03T00:00:00.000 | {
"year": 2018,
"sha1": "410056b26e442cad04a8c1129bd02173a3a2282f",
"oa_license": "CCBYSA",
"oa_url": "https://journals.library.ualberta.ca/jpps/index.php/JPPS/article/download/29903/21450",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "9da6164792db2efbe5cf10b8da44218aa75be5de",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
235784923 | pes2o/s2orc | v3-fos-license | Epitranscriptomics: A New Layer of microRNA Regulation in Cancer
Simple Summary MicroRNAs are small non-coding RNAs, acting as post-transcriptional regulators of gene expression. In the last two decades, their role in cancer as oncogenes (oncomir), as well as tumor suppressors, has been extensively demonstrated. Recently, epitranscriptomics, namely the study of RNA modifications, has emerged as a new field of great interest, being an additional layer in the regulation of gene expression. Almost all classes of eukaryotic RNAs, including miRNAs, undergo epitranscriptomic modifications. Alterations of RNA modification pathways have been described for many diseases—in particular, in the context of malignancies. Here, we reviewed the current knowledge on the potential link between epitranscriptomic modifications of miRNAs and cancer. Abstract MicroRNAs are pervasive regulators of gene expression at the post-transcriptional level in metazoan, playing key roles in several physiological and pathological processes. Accordingly, these small non-coding RNAs are also involved in cancer development and progression. Furthermore, miRNAs represent valuable diagnostic and prognostic biomarkers in malignancies. In the last twenty years, the role of RNA modifications in fine-tuning gene expressions at several levels has been unraveled. All RNA species may undergo post-transcriptional modifications, collectively referred to as epitranscriptomic modifications, which, in many instances, affect RNA molecule properties. miRNAs are not an exception, in this respect, and they have been shown to undergo several post-transcriptional modifications. In this review, we will summarize the recent findings concerning miRNA epitranscriptomic modifications, focusing on their potential role in cancer development and progression.
Introduction
MicroRNAs (miRNAs) are a class of short, non-coding RNAs that control gene expression at the post-transcriptional level via either translational repression or mRNA degradation.
Since miRNAs act as pervasive regulators of gene expression, it is not surprising that they were involved in normal animal development and in a variety of biological processes [1,2]. The aberrant expression of miRNAs is also associated with many human diseases [3,4].
A large amount of literature documents the wide involvement of miRNAs in cancer as key players in the development and progression of different malignancies (reviewed in reference [5]), as diagnostic/prognostic biomarkers (reviewed in reference [6]) and as potential therapeutic targets [7].
One hundred and seventy-two post-transcriptional modifications of RNAs have been reported thus far [8], collectively known as the "epitranscriptome" [9]. Some of these pseudogene-encoded RNAs or circRNAs) sequester miRNAs, thereby preventing their interaction with other targets, whose repression is therefore relieved.
MiRNAs participate in gene regulatory networks that control diverse biological processes in multicellular organisms, such as animal development (reviewed in reference [1]), cell fate specification and differentiation [58], the immune response [59] and inflammation [60]. Changes in the miRNA expression levels have been associated with a wide range of human diseases, including diabetes, cardiovascular and kidney disease and cancer [3,4]. A huge number of miRNAs are downregulated or upregulated in human cancers, where they exert oncogenic or tumor suppressor functions, depending on the cellular context. Alterations of miRNAs in different malignancies have been linked to genetic deletion or amplification, as well as to DNA methylation of the miRNA genomic loci, to the modulation of the pri-mRNA transcription level by transcription factors or to the dysregulation of one or more steps in miRNA biogenesis (reviewed in reference [5]). Recently, epitranscriptomics is emerging as an additional layer of the regulation of the miRNA function in cancer.
Several other "reader" proteins have been shown to bind m6A-modified RNAs thanks to a so-called "m6A switch" [68]. Indeed, m6A installation may trigger a conformational switch that allows the binding of these "reader" proteins, which, in fact, do not directly bind to the m6A residue itself [13]. This mechanism is exploited by several members of the hnRNP (heterogeneous nuclear ribonucleoprotein) family. Finally, insulin-like growth factor 2 mRNA-binding proteins (IGF2BP) were also reported to bind m6A-modified RNAs, promoting their stability [69].
Although two enzymes able to "erase" m6A from mammalian RNAs have been reported, i.e., FTO Alpha-Ketoglutarate-Dependent Dioxygenase (FTO) and AlkB Homolog 5, RNA Demethylase (ALKBH5) [29,30], the specificity and the relevance of these enzymes in physiological conditions are still a matter of debate [70].
In 2015, two pivotal contributions by Alarcón and colleagues [78,79] provided evidence that a m6A modification by METTL3 globally enhances miRNA processing in mammalian cells. Mechanistically, a novel nuclear m6A "reader", namely hnRNPA2B1, binds to m6A-modified pri-miRNAs and interacts with the Microprocessor complex, promoting miRNA processing in a METTL3-dependent manner. Interestingly in Arabidopsis thaliana, an ortholog of METTL3 has been recently shown to mediate m6A installation on pri-miRNAs, thus promoting an interaction with the Microprocessor complex, suggesting that the regulation of miRNA processing by m6A is a widely conserved mechanism [80].
Neither FTO nor ALKBH5 were proven yet to be able to catalyze the demethylation of miRNAs. However, it is worth mentioning that DEAD-box RNA helicase 3 (DDX3) controls the methylation status of microRNAs and interacts with both the AGO2 and the ALKBH5 enzymes [81], suggesting a possible role of ALKBH5 in miRNA demethylation.
In cancer, the relevancy of m6A in miRNA maturation was first unveiled for miR-126 in hepatocellular carcinoma (HCC) [82]. The authors showed that a METTL14-dependent m6A modification on pri-miR-126 is required for its processing by the Microprocessor complex. Indeed, in hepatocellular carcinoma, METTL14 is downregulated, thus decreasing m6A on pri-miR-126 and reducing the miR-126 abundance. Furthermore, their findings support the hypothesis that the miR-126 modification by METTL-14 is instrumental to prevent cell invasion, as assessed by in vitro assays.
From that moment on, increasing evidence has disclosed the relevance of the m6A modification of miRNA in cancer progression. Most of the literature confirms that m6A mainly promotes pri-miRNA processing and that the deregulation of the enzymes involved in writing or reading m6A is correlated with tumor onset. Notably, alteration of the m6A deposition on miRNAs is not only a common feature of different tumors but also participates in tumorigenesis processes ( Figure 1 and Table 1).
Cancers 2021, 13, x 4 of 20 miRNA processing in a METTL3-dependent manner. Interestingly in Arabidopsis thaliana, an ortholog of METTL3 has been recently shown to mediate m6A installation on pri-miR-NAs, thus promoting an interaction with the Microprocessor complex, suggesting that the regulation of miRNA processing by m6A is a widely conserved mechanism [80]. Neither FTO nor ALKBH5 were proven yet to be able to catalyze the demethylation of miRNAs. However, it is worth mentioning that DEAD-box RNA helicase 3 (DDX3) controls the methylation status of microRNAs and interacts with both the AGO2 and the ALKBH5 enzymes [81], suggesting a possible role of ALKBH5 in miRNA demethylation.
In cancer, the relevancy of m6A in miRNA maturation was first unveiled for miR-126 in hepatocellular carcinoma (HCC) [82]. The authors showed that a METTL14-dependent m6A modification on pri-miR-126 is required for its processing by the Microprocessor complex. Indeed, in hepatocellular carcinoma, METTL14 is downregulated, thus decreasing m6A on pri-miR-126 and reducing the miR-126 abundance. Furthermore, their findings support the hypothesis that the miR-126 modification by METTL-14 is instrumental to prevent cell invasion, as assessed by in vitro assays.
From that moment on, increasing evidence has disclosed the relevance of the m6A modification of miRNA in cancer progression. Most of the literature confirms that m6A mainly promotes pri-miRNA processing and that the deregulation of the enzymes involved in writing or reading m6A is correlated with tumor onset. Notably, alteration of the m6A deposition on miRNAs is not only a common feature of different tumors but also participates in tumorigenesis processes ( Figure 1 and Table 1). Colorectal cancer (CRC) displays a low expression of METTL14, and this is associated with an impaired m6A modification and processing of pri-miR-375. Low levels of miR-375 result in the overexpression of its targets YAP1 (Yes-associated protein 1) and SP1, thus increasing cell proliferation, migration and invasion [83]. Colorectal cancer (CRC) displays a low expression of METTL14, and this is associated with an impaired m6A modification and processing of pri-miR-375. Low levels of miR-375 result in the overexpression of its targets YAP1 (Yes-associated protein 1) and SP1, thus increasing cell proliferation, migration and invasion [83].
A recent report identified a further putative m6A "reader" protein, RALY (also known as hnRNPCL2), which interacts with miR-483, miR-676 and miR-877 in CRC and is required for their m6A-dependent processing. Mechanistically, RALY interacts with DROSHA and DGCR8 to enhance pri-miRNA processing [90].
Recently, it has been proposed that a m6A modification reduces the ability of miRNAs to suppress target mRNA translation. Indeed, Konno and colleagues [91] showed that the m6A modification of let-7a-5p and miR-17-5p caused a large structural change in the RISC complex, which affected the target RNA recognition. In pancreatic and CRC tissues, the m6A levels on let-7a-5p and miR-17-5p increased without affecting the miRNA expression level.
A-to-I Editing
A-to-I editing is catalyzed by enzymes highly conserved in vertebrates, called Adenosine Deaminases Acting on RNA (ADAR) [92]. Mammalian genomes encode for three members of the ADAR family: ADAR1, ADAR2 and ADAR3 [93].
ADAR enzymes bind double-stranded (ds) regions of coding and noncoding RNAs [94]; in RNAs forming imperfect dsRNA structures, A-to-I editing involves only one or two adenosines (site selective editing), while, in the case of long perfect dsRNA regions, the random modification of several A residues is observed (hyper-editing) [95][96][97].
Inosine is recognized by the cellular machinery as guanosine, causing a change in the RNA sequence. As a consequence, depending on the modification site, this type of RNA editing can influence the RNA stability [98][99][100], splicing [101][102][103], localization and translation, as well as redefine its interactions with specific factors [104,105]. In mRNAs, the modification of A-to-I can lead to a codon change, thus affecting the primary structure of the encoded protein [106,107].
A-to-I editing mainly targets noncoding regions of RNA, such as introns and UTRs, containing repetitive Alu elements and Long Interspersed Elements (LINEs) that fold into dsRNA structures recognized by ADARs [108].
In most types of cancer, the activity of ADAR enzymes is significantly decreased, as witnessed by the extensive hypoediting of Alu RNAs, as well as by the reduced expression of ADAR enzymes [109].
The first evidence of the editing of a miRNA was shown in 2004 by Luciano and colleagues [33], who reported A-to-I conversion within the miR-22 precursor in Homo sapiens and Mus musculus. Soon after, it was shown that the A-to-I editing of pri-miR-142 prevents processing by DROSHA [110]. ADAR enzymes have a degree of specificity for different miRNA precursors, depending on their secondary structure [111]. The ADAR1 interaction with DICER was associated with enhanced miRNA processing in oral squamous cells carcinoma [112] and in melanoma [113], although, in both cases, the authors did not assess the editing of the miRNA precursors. Furthermore, ADAR editing has been shown to affect the DICER-dependent processing of viral miRNAs [114]. Of note, ADARs can also alter miRNA metabolism independently from their editing activity [115][116][117].
Several examples showed that the A-to-I editing of miRNA precursors inhibits the biogenesis of mature miRNAs ( Figure 1 and Table 2). The deregulation of ADAR1 and/or ADAR2 in glioblastoma and in chordoma affects the expression levels of miR-21, miR-221 and miR-222 [118] and of miR-10a and miR-125a [119], respectively. Furthermore, the impairment of let-7 biogenesis by means of ADAR1-mediated A-to-I editing drives leukemia stem cells renewal [120]. A further mechanism by which A-to-I editing alters miRNA functions is the remodulation of potential targets of mature miRNAs [121]. Indeed, A-to-I editing in the "seed" sequence causes a loss-of-function when no more targets are recognized [122] or a gainof-function when a new target is recognized by the edited miRNA [123,124]. Therefore, changes in the relative abundance of the edited and unedited forms of the miRNA lead, in turn, to altered gene expression profiles.
In this context, in 2012, it has been shown that the loss of mir-376a-5p editing results in the increased invasiveness of glioblastoma multiforme (GBM). Mechanistically, unedited miR-376a-5p promotes aggressive glioma growth by its ability to target Ras-Related Protein Rap-2a (RAP2A), a member of the RAS oncogene family, and the concomitant inability to target Autocrine Motility Factor Receptor (AMFR) [125]. These findings were further corroborated by the discovery that miRNA hypoediting is widespread in GBM [126].
Similarly, the impairment of ADAR-mediated editing of miR-455-5p enhances the progression in melanoma. Indeed, unedited miR-455-5p targets the tumor suppressor Cytoplasmic Polyadenylation Element-Binding Protein 1 (CPEB1), thus promoting metastasis, while edited miR-455-5p exerts the opposite effect [127]. In a follow-up of their work, the authors showed that, in melanoma, the editing of miR-378a-3p allows the targeting of the oncogene Parvin Alpha (PARVA). Hence, the loss of miR-378a-3p editing promotes melanoma progression [128].
In the brain, ADAR2 edits the seed sequence of miR-589-3p. In glioblastoma, the editing of miR-589-3p decreases. Higher levels of the unedited version of miR-589-3p promote proliferation and invasion by targeting the tumor suppressor Protocadherin 9 (PCDH9). On the contrary, editing within miR-589-3p retargets the miRNA to the ADAM Metallopeptidase Domain 12 (ADAM12) to contrast the progression of the tumor [129].
In different contexts, miRNA A-to-I editing stimulates the progression of the tumor by altering the selection of miRNA targets. In thyroid cancer, the slight overexpression of ADAR1 corresponds to a higher expression of ZEB1, a master regulator of Epithelial-Mesenchymal Transition (EMT). It has been demonstrated that editing of the seed sequence of miR-200b by ADAR1 impairs its ability to inhibit ZEB1 expression, favoring the progression of the cancer [130,131].
Another target of A-to-I editing is miR-381, a microRNA involved in stemness and chemoresistance [132,133] that is overedited in non-small cell lung carcinoma (NSCLC) cell lines harboring the genomic amplification of ADAR1. Edited miR-381 promotes cell viability [134].
5-Methylcytosine (m5C)
m5C is one of the most representative post-transcriptional RNA modifications [140], and it has long been known to be present in all three kingdoms of life [141,142]. m5C was originally reported in tRNAs, rRNAs [62] and coding RNAs [143]; later, it was identified in other noncoding RNAs, thanks to technologies such as bisulfite treatment and Next-Generation Sequencing (NGS) [144][145][146].
In vitro and in vivo studies have demonstrated that aly/REF nuclear factor (ALYREF) is a putative "reader" of m5C sites on mRNAs and that, following the knockdown of NSUN2, ALYREF loses its RNA-binding ability and is retained in the nucleus, suggesting a role for m5C in mRNA exports from the nucleus [26]. A further m5C "reader" is Y-Box-Binding Protein 1 (YBX1) that recognizes and binds m5C-modified mRNAs and stabilizes their target mRNAs by recruiting ELAV-like Protein 1 (ELAVL1) [161,162]. m5C "writers" and "readers" are primarily implicated in fundamental cancer-related processes such as cell differentiation, motility [163,164], proliferation [165,166], cell cycle progression [167] and senescence [155].
In particular, NSUN2 is aberrantly expressed and plays important roles in the development and pathogenesis of different types of tumors, such as breast, colorectal, lung, skin, ovarian and bladder cancers [168].
m5C has been only recently characterized in miRNAs. Interestingly, methylation, but not an abundance of miR-200c-3p and miR-21-3p, was increased in pancreatic and colorectal cancer tissues, as well as in serum samples from pancreatic and colorectal cancer patients [91] (Figure 1 and Table 3). Lung cancer let-7 family (m7G) n.d. processing m7G methylation within miRNAs regulates cell migration [173] Colon cancer let-7e (m7G) ↓ processing Down-regulation of mature let-7e results in the activation of its targets HMGA2 thus stimulating colon cancer cell viability and mobility [174] 1 Increase (↑) or decrease (↓) of epitranscriptomic modifications (n.d., not detected; HMGA2 High Mobility Group AT-hook 2).
Cheray and colleagues [170] proposed that the DNMT3A/AGO4 complex promotes the methylation of cytosine residues of miRNAs at CG dinucleotides. In glioblastomaderived cell lines and glioblastoma tumor samples, the methylation of mature miR-181a-5p by the DNMT3A/AGO4 complex inhibits the recognition of its target mRNA BIM, a proapoptotic gene, also known as B-cell chronic lymphocytic leukemia/lymphoma (Bcl-2)like 11 (BCL2L11) [170]. This preliminary evidence highlights that the profiling in tumor samples of m5C in miRNAs deserves further investigation.
Recently, we described that m5C is widely spread in human miRNAs in various sequence contexts by taking advantage of a novel NGS analysis of bisulfite-treated small RNAs (BS-miRNA-seq) [171].
Finally, 5mC is oxidized by the Ten-eleven translocation (TET) enzymes both in DNA and RNA. TET enzymes are Fe(II)-and 2-oxoglutarate-dependent dioxygenases that mediate the conversion of 5mC to 5hmC, then to 5-formylcytosine (5-fC) [175] and, finally, to 5-carboxylcytosine (5-caC) [18,176,177]. In DNA, these subsequent conversions have the purpose of demethylating 5mC [178], but it is still not clear if this mechanism is conserved in RNA.
hm5C has been detected in RNA isolated from different mouse and human tissues, including the brain, heart, pancreas and spleen [179]. Transcriptome-wide analyses of hm5C in mouse and in Drosophila RNAs have revealed the presence of 5hmC on hundreds of messenger RNAs, mainly in UC-rich motifs [180,181]. The deposition of hm5C in mRNAs has been associated with the differentiation of murine embryonic stem cells and brain development in Drosophila via controlling the mRNA stability or translation, respectively.
To date, decreased levels of hm5C in RNA have been shown in tumor tissues, such as CRC and hepatocellular carcinoma [176].
Interestingly, we recently unraveled not only the presence of m5C but, also, of hm5C on several miRNAs in human cancer cell lines [171]. However, no evidence of the role of hm5C modification in miRNAs in tumors has yet been reported.
N7-Methylguanosine (m7G)
m7G is a positively charged modification installed cotranscriptionally at the 5' Caps of eukaryotic mRNAs [182]. This modification protects and stabilizes transcripts from exonucleolytic degradation [183] and influences all the events responsible for the processing of the mRNA molecules, from transcript elongation to translation [184,185].
Notably, the presence of internal m7G sites was found not only in tRNA and rRNA molecules [186][187][188] but also in mammalian mRNAs [188]. Internal m7G could affect mRNA translation, and this modification typically occurs near the start and stop codons in a GA-enriched motif [189].
The enzyme responsible for this internal m7G modification is METTL1, which cooperates with the cofactor WD Repeat Domain 4 (WDR4) [189,190]. Interestingly, METTL1 has been linked to tumor vascular invasion and poor prognosis in hepatocellular carcinoma [173,191].
Recently, by high-throughput screening, several miRNAs were identified as harboring internal m7G sites [192]. In particular, METTL1-dependent m7G was discovered in a subset of tumor-suppressor miRNAs involved in the inhibition of cell migration, including the let-7 family. METTL1-mediated m7G occurs on pri-miRNA within G-rich regions that display the propensity to form G-quadruplexes, i.e., structures known to be inhibitory to miRNA processing [174,193,194] (Figure 1 and Table 3).
Indeed, m7G in the let-7 family affects G-quadruplex formations, thus facilitating the formation of a canonical stem-loop structure and miRNA processing [192]. In line with this study, Liu and colleagues showed that, in colon cancer, the downregulation of METTL1 leads to a decrease in the let-7e levels. The alteration of let-7e expression affects its downstream target High Mobility Group AT-hook 2 (HMGA2), thus promoting cell proliferation, invasion and EMT [195].
m6A in miRNA Targets
The installation of m6A in miRNA targets may affect the pairing with miRNAs, thus affecting miRNA functions. Such a mechanism was first suggested in 2015, when Ke and colleagues [72] reported that the majority of m6A sites on mRNAs were in the last exon. The authors also reported a significant overlap between the m6A residues and AGO-binding sites. This finding was further corroborated by a later study [196].
Importantly, m6A modification may positively or negatively affect miRNA-mRNA pairing by several distinct mechanisms. On the one hand, m6A modifications within miRNA-binding sites may directly affect the miRNA:mRNA duplex stability. On the other hand, m6A modifications nearby AGO-binding sites may alter the mRNA secondary structure and/or recruit other RNA-binding proteins, thus modifying the mRNA accessibility to miRNA. Thus far, a few examples have been reported.
In the liver, YAP is a target of miR-582-3p, which binds at residues 313-321 of YAP 3 UTR. Such binding is enhanced by m6A modifications of YAP 3 UTR at residue 355. In HCC, the m6A modification at residue 355 of YAP 3 UTR is impaired, resulting in the loss of YAP repression by miR-582-3p [197].
In cancer cell lines, m6A modification promotes the recruitment of IGF2BP1 onto Serum Response Factor (SRF) mRNA 3 UTR. Interestingly, IGF2BP1 acts as a m6A "reader", and its recruitment is instrumental to reduce miRNA-mediated AGO binding to SRF. Therefore, m6A deposition on the 3 UTR of SRF mRNA relieves the negative post-transcriptional regulation by miRNAs [198]. Furthermore, a genome-wide analysis of m6A in glioma stem cells highlighted that m6A deposition on the 3 UTR of several mRNAs may affect their targeting by different miRNAs [199]. On the other hand, Cheng and colleagues [200] showed an effect at odds with this. Indeed, in neuroblastoma, m6A modification in N-Myc 3 UTR near a miR-98-binding site is necessary to promote the miR-98-mediated posttranscriptional repression of N-Myc. Overall, these examples highlight that the effect of m6A within mRNA 3 UTRs on miRNA targeting is ambiguous.
A-to-I Editing in miRNA Targets
A-to-I editing can influence microRNA targeting. Indeed, when A-to-I editing occurs within the miRNA-binding site in the 3 UTR of a mRNA, this can remodulate the interactions between miRNA and mRNA in different ways.
The editing of the 3 UTR of Rho GTPase Activating Protein 26 (ARHGAP26) by ADAR1 blocks the interaction with miR-30b-3p and miR-573 to favor the expression of the protein [204]. On the contrary, in HCC cells, for instance, the RNA editing catalyzed by ADAR1 of 3 UTR of Aryl hydrocarbon Receptor (AhR) creates a miR-378-binding site to negatively regulate the expression of this protein [205]. Through a similar mechanism, editing of the 3 UTR of the mRNA encoding for the tumor suppressor Phosphatase and Actin Regulator 4 (PHACTR4) mediated by ADAR1 prevents the binding of miR-196a-3p. Accordingly, the decreased activity of ADAR1 in gastric cancer results in the repression of PHACTR4 by miR-196-3p [206].
Interestingly, a pan-cancer RNA editing study highlighted that the 3 UTR of the Mouse double minute 2 homolog (MDM2) oncogene underwent A-to I editing in 11 out of the 14 cancer types investigated within a region of the 3 UTR complementary to the miR-200 "seed" region. This editing impaired MDM2 repression by miR-200 [207].
The above-mentioned examples suggest that A-to-I editing within regions of 3 UTR pairing with miRNA "seed" affects miRNA binding. However, a genome-wide report pinpointed that the editing of A residues lying outside of the miRNA:mRNA pairing region affects the mRNA structure, thus modulating the accessibility to AGO2-miRNA complexes [208].
m5C in miRNA Targets
To date, little is known about the mechanisms through which m5C can regulate the function of miRNAs, but, interestingly, a possible role for m5C in miRNA targeting was suggested by the overlap between the AGO2-binding sites and m5C positions reported in the 3 UTR of human mRNAs [209].
Methodological Challenges and the Potential Limits of Current Knowledge
Our understanding of the mechanisms underlying the epitranscriptomic regulation of miRNAs is still potentially biased by the relatively small number of modifications which have been widely investigated in miRNAs, with most reports focused on m6A and A-to-I editing. Indeed, further mechanisms through which epitranscriptomic modifications may affect miRNA function are conceivable. For example, the recent report by Konno and colleagues suggested that m5C modification in position 9 of miR-200-3p might affect the interaction with AGO proteins [91]. However, further investigation will be required to assess whether this may represent a general paradigm. Epitranscriptomic modifications might also modulate miRNA half-life, or control miRNA subcellular localization (either in membrane enclosed organelles, or by liquid-liquid phase separation) or secretion of modified miRNAs. Secreted miRNAs have been widely reported as potential biomarkers for a variety of diseases and they also serve as signaling molecules to mediate cell-cell communications [210,211].
Our current knowledge of the modifications installed on miRNAs is likely incomplete because of technical limits which still prevent application of several high-throughput NGS methods for the quantification and characterization of epitranscriptomic modification to miRNAs. Indeed, most epitranscriptomic modifications are investigated through NGS methods which exploit the block of reverse transcription at the modified nucleobase of interest by means of different biochemical treatments (i.e., antibody cross-linking or chemical modifications). These protocols result in reads truncated at the modified positions, thus allowing the genome-wide mapping of the epitranscriptomic modification at singlenucleotide resolution [212,213]. Those methods cannot be applied to mature miRNAs, as premature reverse trascriptase termination would yield reads too short to be effectively mapped on the genome. Currently, single-nucleotide resolution analysis in miRNAs is only possible for those modifications assessed through methods based on mismatched nucleotides introduced during the reverse transcription step. Furthermore, techniques relying on unmodified nucleobases conversion (e.g., bisulfite treatment) require dedicated pipelines for data analysis as alignment of short reads is severely affected by complexity reduction associated with those techniques [171].
Finally, emerging methods to assess epitranscriptomic modifications through third generation sequencing are best-suited for long RNA molecules and cannot be easily adapted to the study of miRNAs [214].
Conclusions
In this review, we summarized the current knowledge on the epitranscriptomic modifications of miRNAs that play a role in cancer development and/or progression. In most cases, epitranscriptomic modifications exert their effect on miRNAs by affecting either their biogenesis or their binding to target mRNAs.
The current lack of mature, high-throughput technologies to profile epitranscriptomic modifications of miRNAs, with the notable exception of A-to-I editing, is likely one of the main reasons why the prognostic/diagnostic use of the "miRNA epitranscriptome" has been poorly explored thus far. Nevertheless, a plethora of studies focused on the mechanistic role of specific modifications in cancer development and/or progression point out how promising the miRNA epitranscriptome is.
Adaptation of the existing NGS methods to yield techniques allowing cheap, highthroughput and quantitative assessments of epitranscriptomic modifications in miRNAs could potentially propel research in this field by allowing mechanistic investigations of further modifications. Furthermore, such methodologies would be fundamental to investigate the potential prognostic role of miRNAs, including miRNA secreted into extracellular fluids.
Conflicts of Interest:
The authors declare no conflict of interest. | 2021-07-11T05:30:05.086Z | 2021-07-01T00:00:00.000 | {
"year": 2021,
"sha1": "6f2983f0a6db730d33fe92e7c7b86d6943f9c67a",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2072-6694/13/13/3372/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "6f2983f0a6db730d33fe92e7c7b86d6943f9c67a",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
258069299 | pes2o/s2orc | v3-fos-license | EFFNet-CA: An Efficient Driver Distraction Detection Based on Multiscale Features Extractions and Channel Attention Mechanism
Driver distraction is considered a main cause of road accidents, every year, thousands of people obtain serious injuries, and most of them lose their lives. In addition, a continuous increase can be found in road accidents due to driver’s distractions, such as talking, drinking, and using electronic devices, among others. Similarly, several researchers have developed different traditional deep learning techniques for the efficient detection of driver activity. However, the current studies need further improvement due to the higher number of false predictions in real time. To cope with these issues, it is significant to develop an effective technique which detects driver’s behavior in real time to prevent human lives and their property from being damaged. In this work, we develop a convolutional neural network (CNN)-based technique with the integration of a channel attention (CA) mechanism for efficient and effective detection of driver behavior. Moreover, we compared the proposed model with solo and integration flavors of various backbone models and CA such as VGG16, VGG16+CA, ResNet50, ResNet50+CA, Xception, Xception+CA, InceptionV3, InceptionV3+CA, and EfficientNetB0. Additionally, the proposed model obtained optimal performance in terms of evaluation metrics, for instance, accuracy, precision, recall, and F1-score using two well-known datasets such as AUC Distracted Driver (AUCD2) and State Farm Distracted Driver Detection (SFD3). The proposed model achieved 99.58% result in terms of accuracy using SFD3 while 98.97% accuracy on AUCD2 datasets.
Introduction
In the past few decades, the rapid increase in road accidents due to the lack of driver attentiveness, has gained researchers' attention [1]. For instance, in 2016, the World Health Organization (WHO) reported "1.4 million humans lost their lives due to road accidents globally". In addition, road accident is the eightieth major cause of death [1]. A study by the government of India in 2017, reported that approximately half a million road accidents occurred in India, in which several people lost their lives and many of them obtained serious injuries [2]. In another article reported in 2018 by the Ministry of Road Transport and Highway (MRTH), almost half a million road accidents have been recorded in different states in India, in which roughly 0.15 million people lost lives and almost 0.48 million people obtained serious injuries [3]. Similarly, the report of National Highway Traffic Safety Administration (NHTSA) in the USA concluded that around 64.4% of people lose life due to diversion of attention from driving [3]. Moreover, their report also declared that 94% of car accidents are caused by driver's inactiveness [3], while a large number of road accidents are due to the usage of electronic devices such as Bluetooth devices, mobile phones, and so on.
Prior studies have demonstrated that drivers' attention is changed by engaging in other activities when they are driving, which can lead to road accidents. These activities include engaging with electronic devices while driving such as calling, talking, texting, and so on. Researchers are thus motivated to find out the easiest way to reduce the number of road accidents. Therefore, several researchers have presented different computer visionbased methods to alert the driver in case of engaging in other activities while driving. These methods are broadly categories into two major fields such as traditional Machine Learning (ML) and Deep Learning (DL)-based methods [4]. For instance Vural, et al. [5], used a traditional ML approach such as Adaboost and multinomial ridge regression to determine the drivers' drowsiness based on the 30 facial actions from the Facial Action Coding system. In addition, their resultant technique obtained 90% accuracy across subjects based on two datasets such as, Cohn-Kaneda DFAT-504 and spontaneous expressions dataset. As a follow-up research Babaeian et al. [6], proposed a method by the use of advanced logistic regression using a ML algorithm that can detect driver's drowsiness based on computing heart rate. Chen et al. [7], used AdaBoost algorithms to fabricate a driving behavior classification model to analyze the behavior of a driver and analyze whether it is safe. In another article, Kumar et al. [8] proposed a method of real-time driver's drowsiness detection system. The researchers recorded a video through a webcam (Sony CMU-BR300) and detected the driver's faces using image processing techniques. The researchers used a Support Vector Machine (SVM)-based classification. However, the limited performance, high false alarm rate, and time complexity of traditional ML models are the major factors of failure. Furthermore, in the traditional ML-based models, the handcrafted features extraction and classification are very tedious, error prone, and time-consuming processes. These factors motivated the researchers to explore the DL-based model for driver distraction detection.
For instance, Hssayeni, et al. [9], proposed deep learning models for the detection of drivers' attentions, although their resultant works require more improvement in terms of accuracy. Kapoor, et al. [10], proposed a light-weight pretrained technique with some fine-tuning strategies for real-time detection of driver distraction. However, their approach generated a false alarm rate due to the rapid movements of the body based on low performance. A DL-based model for drowsiness detection is presented in [11], to determine the driver attentiveness based on facial landmark key point detection. The researchers used the NTHU-DDD dataset and achieved 80% in terms of accuracy. However, the accuracy of their proposed method needs further improvement.
Driver distraction detection is a problem to be solved, the aforementioned techniques based on traditional ML and DL models are time-consuming and required further enhancement in terms of accuracy and time complexity. In addition, such techniques generate false alarms due to the low performance. Moreover, it is a challenging task to detect driver behavior to overcome road accidents. To deal with the problem in a satisfactory way, we proposed an EfficientNetB0 with CA for the real-time efficient detection of driver distraction. The major contributions of the proposed work are as follows: • Inspired by the transfer learning technique, we trained different types of pretrained models without dense layers and applied CA mechanism for obtaining optimal performance. In addition, we compared the performance of our proposed model with other architectures including VGG16, VGG16+CA, ResNet50, ResNet50+CA, Xception, Xception+CA, InceptionV3, InceptionV3+CA, and EfficientNetB0.
•
The results of a detailed ablation study showed that the EfficientNetB0 with channel attention (CA) achieved the highest performance compared with all other methods. Based on these findings, we selected EfficientNetB0 with CA as the model of choice for driver distraction detection. In addition to its superior performance, the proposed model is also lightweight, enabling fast processing times compared with other architectures. The faster processing time of the EfficientNetB0 with CA mechanism can reduce the risk of accidents and improve the overall safety of drivers and passengers. Furthermore, the lightweight and fast processing nature of the proposed model makes it highly applicable for real-world scenarios that require real-time detection, such as medical diagnosis, video surveillance, and robotics.
•
We evaluated the performance of the proposed model on the SFD3 and AUCD2 datasets. Our results showed that the proposed model achieved higher accuracy and faster processing times compared with other baselines. This highlights the potential of the proposed model as a more efficient and effective solution for driver distraction detection in real-world scenarios.
The rest of the article is formatted as follow: in Section 2 we highlight related works with previous literatures and their approaches, Section 3 presents the methodology of our work, discussion and result are available in Section 4, and finally, in Section 5 we provide the conclusion and future work.
Related Work
Drivers' distraction is a major cause of accidents that affects human lives and their resources. To cope with these issues, several researchers have proposed different techniques to notify the driver of their distraction based on alarm or messages using a Traditional Deep Learning (TDF) approach. For instance, Alzubia et al. [12], presented a CNN-based method which alerts the drivers by their distraction while driving. In this study, the researcher utilized an ensemble technique to detect driver distraction using their custom dataset. Their method is not only limited to determining drivers' distractions but also can work in real time using resource constraint edge devices. However, their technique needs further improvement in evaluation matrices. As a follow-up study, Leekha et al. [13] proposed a CNN method and trained the existing method on two publicly available datasets, such as the State Farm Distracted Driver Detection (SFD3) and the AUC Distracted Driver dataset (AUCD2), additionally their proposed method achieved 98.48% and 95.64% performance, respectively. Despite that, their technique is time-consuming as they trained the complex model on datasets. In another research, Varaich et al. [14] used two competing DCNN architectures named InceptionV3, and Xception. In addition, the authors compared the results of both architectures and applied them to recognize ten unique actions of the drivers in the SFD3 dataset. The resultant technique was complicated compared with state-of-the-art techniques. The next method, devised by Jamsheed et al. [3], is a technique for alerting distracted drivers and reducing the ratio of the road accidents based on deep learning. Their technique consists of three models, namely, vanilla CNN, vanilla CNN based on data augmentation, and CNN with transfer learning. Differently, false classification of distraction can happen based on performance. Similarly, Moslemi et al. [1] derived a benefit from temporal information by using a 3D CNN and optical flow to improve the driver monitoring system. Their resultant model achieved 90% performance based on the Kinetics and the SFD3 datasets, but their method is computationally inefficient, in addition, their technique requires further improvement. The next article proposed by Qin et al. [15], introduced a new D-HCNN model based on a declining filter size with only 0.76M parameters, a much smaller number of parameters compared with SOTA based on two available datasets such as AUCD2 and SFD3, through which their model obtained 95.59% and 99.87% performance in terms of accuracy, respectively.
Another study, presented by Dua et al. [16], was focused on enhancing the performance of four deep learning models: AlexNet, VGG Face, Flow ImageNet, and ResNet. The models detect four types of different features such as hand gestures, facial expressions, behavioral features, and head movements. The authors used NTH Drowsy Driver Detection (NTHU-DDD) video dataset in this article. They passed the RGB videos as input and the goal of that input is detecting the driver drowsiness. Their resultant model achieved 85% accuracy; However, the resultant model is limited in drivers' behavior classes.
Alotaibi et al. [17] used a TDF approach and tried to enhance the performance of the proposed model. Moreover, their research is focused on the three popular pretrained CNNs architectures, such as Inception, ResNet, and Hierarchical Multiscale Recurrent Neural Network [18]. Based on Inception, ResNet, and Hierarchical Multiscale Recurrent Neural Network, they obtained promising performance. Additionally, Dhakate et al. [2] implemented four pretrained DL architectures, i.e., VGG16, ResNet50, Xception, and InceptionV3 for the efficient classification of drivers' distraction, whereas their proposed architecture obtained 97% performance using well-known datasets SFD3 and AUCD2. However, their experiments were performed based on computationally large models such as, VGG16, ResNet50, and so on. The next approach devised by Jabbara et al. [11] proposed a real-time drowsiness detection technique based on Deep Neural Network (DNN). The researchers designed a method using facial landmark key points detection to show whether the driver is active or not. Their work is based on the (NTHU-DDD) dataset and their proposed model obtained 80% accuracy; however, their proposed method requires a proper setup for real-time detection to save the driver privacy.
The approach presented by research Hssayeni et al. [9] utilized a computer vision and ML technique to detect drivers' behavior based on a dashboard camera. Their experimental results depended on three transfer learning architectures, such as AlexNet, VGG16, and ResNet50 and their proposed model obtained 85% accuracy. However, their proposed architecture creates false detection due to the rapid movement of a body and low accuracy. The other research introduced by Streiffer et al. [19] proposed a convolutional and recurrent neural network that can analyze driving image and IMU sensor data to detect up to six classes of driving behaviors with high performance.
In another study, Valeriano et al. [20] compared different deep learning methods for the classification of driver behavior. However, their proposed method achieved high accuracy of 96.6% based on three rounds of 5-fold cross validation; however, their proposed model needs to deploy edge devices. Masood et al. [21] proposed a CNN-based model that not only detects distraction but also analyzes the images that are captured inside of the vehicle. In addition, their proposed method achieved 99% accuracy using the SFD3 dataset. Furthermore, the VGG16 and VGG19 methods were utilized for the identification of driver distraction in this article. However, their experiments are computationally expensive based on large models. In another approach, Majdi et al. [22] presented an automated supervised learning method called DriveNet for driver distraction detection based on two other popular machine-learning approaches: an Recurrent Neural Network (RNN) and Multi-Layer Perceptron (MLP). Moreover, their presented method reached 95% accuracy, but their experimental setup is complex.
Wöllmer et al. [23] proposed a Long Short-Term Memory (LSTM) technique that figures out real-time distractions of drivers and their resultant technique achieved 96.6% in terms of accuracy; however, the privacy of the driver is a critical issue in real-time distractions. Xing et al. [24] presented a driver behavior recognition system based on DCNN based on a low-cost camera (use for image acquisition). Their work related to three different pretrained CNN architectures, for instance, AlexNet, GoogLeNet, and ResNet50, and their CNN-based models obtained 81.6%, 78.6%, and 74.9%, respectively. These models are also trained for binary classification problems whether the driver is distracted or not. The binary classification rate achieved 91.4% accuracy. The summary of the literature is tabulated in Table 1; however, their models need further enhancement for multiclass classification.
Reference
Description Method [12] Proposed an ensemble-based technique for the classification of driver distraction. DL Ensemble Technique [13] Utilized a deep learning architecture based on CNN for driver distraction detection using two well-known datasets. DL [14] Implemented two DCNN pretrained networks named (InceptionV3 and Xception) for the recognition of driver action using publicly available SFD3 dataset.
Reference
Description Method [3] The authors implemented several architectures namely, vanilla CNN with and without augmentation technique, and pretrained CNN model for driver distraction detection.
Vanilla CNN [1] The authors implemented a 3D CNN technique for driver behavior monitoring. 3D CNN [15] Utilized a novel D-HCNN algorithm, which detects driver action in early stages while driving using AUC2 and SFD3 datasets.
D-HCNN [16] Proposed an ensemble technique which contains four DL pretrained architectures using video data. DL Ensemble Technique [17] Trained a DL pipeline named inception using some fine-tunning strategies for accurate classification of driver behaviors.
DL [2]
Used a stacking technique for obtaining optimal results. Initially, they stacked all the feature vectors and feed to the CNN for training purposes.
Stacking Ensemble Technique [19] The authors presented a deep learning framework called DarNet which classifies driver behavior using input sensor data.
DL [20] The researchers utilized the deep convolutional neural network for efficient and effective classification of driver distraction using the SFD3 dataset. In addition, their experimental results are focused on three rounds of 5-fold cross validation.
DCNN [21] The authors used forward machine learning based on convolution neural network which not only classifies the driver's distraction but also finds the reason of their distraction.
ML and CNN [22] Presented a method named Drive-Net based on supervised learning for the accurate detection of driver behavior while driving using the well-known publicly available SFD3 dataset.
DL [23] The authors introduced a novel framework called LSTM to detect online driver activity. DL LSTM [24] Evaluated three CNN-based transfer learning techniques using some fine-tuning strategies for the recognition of seven common driver distractions using low-cost camera collected images.
CNN
Ye et al. [6] implemented a pretrained Xception network as a backbone for features extraction and incorporated channel attention for selection of more optimal features for detecting driver distraction behavior. Their proposed network (SE-Xception) obtained 92.60% performance in terms of accuracy. Another article presented by Liu et al. [7], utilized channel expansion and attention mechanism to improve YOLOv7 (namely CEAM-YOLOv7) for driver distraction detection using an in-vehicle camera. Additionally, their proposed architecture achieved promising performance among SOTA techniques. As a follow-up research, Zhang et al. [8] introduced a novel attention mechanism-based architecture for driver distraction behavior detection in real time. In this paper, the authors evaluated their proposed method using two datasets such as publicly available dataset and their custom dataset. Lin et al. [9] proposed a novel lightweight architecture known as LWANet. In other words, to decrease the computation cost and number of parameters that can be trained, the classic VGG16 architecture is optimized by reducing its trainable parameters by 98.16% through replacing standard convolution layers with depth-wise separable convolutions. Moreover, the proposed LWANet achieved 99.37% accuracy on SFD3 dataset and 98.45% accuracy using AUC dataset. Another study presented by Wei et al. [10] presented a technique named ENet-CBAM which is based on EfficientNet and Convolutional Block Attention Module for effective detection of driver distraction. Overall, their proposed ENet-CBAM is capable of detecting effectively driver distraction in a real-time scene with few parameters. Similarly, Hu et al. [11] proposed a deep learning-based technique to learn dominant features from the input data. In addition, their proposed technique is improved by two aspects: firstly, use of a multi-scale convolutional block with various kernel sizes to generate a hierarchical feature vector. They also adopted a maximum selection unit that concatenates multi-scale information in an adaptive manner. Secondly, the researchers added an attention mechanism to learn pixel and channel saliency between convolutional features. Furthermore, their experimental results demonstrated that the proposed technique (MSA-CNN) achieved higher performance for driver distraction behavior recognition.
As evident from the literature, numerous researchers have proposed several methods for driver distraction detection. It is worth noting that these techniques suffer from substantial shortcomings including limited performance and required huge computational hardware. In addition, such techniques generate false alarms due to the rapid movement of the body owing to their low performance. Furthermore, the selection of a suitable DL model to deploy over a resource constraint device in real time is a challenging task. To cope with this, in the upcoming section, we briefly explained the proposed model that can be easily deployed over resource constraint devices and can improve the performance over SOTA methods.
Proposed Method
We provided a brief discussion about the proposed model to solve the aforementioned problems in a satisfactory way. The proposed model is composed of two main steps such as (1) preprocessing, to prepare data for training and testing, and (2) training the traditional DL model for accurate driver distraction detection. Furthermore, we fine-tuned a pretrained DL model to enhance the driver distraction detection performance and minimize the false alarm rate. In the proposed work, we employed a CA module with several DL models and validated their performance against SOTA over the benchmark datasets. The proposed framework employing the CA module as is presented in Figure 1. The following subsequent sections explain the details of step 1 and step 2. Convolutional Block Attention Module for effective detection of driver distraction. Overall, their proposed ENet-CBAM is capable of detecting effectively driver distraction in a real-time scene with few parameters. Similarly, Hu et al. [11] proposed a deep learningbased technique to learn dominant features from the input data. In addition, their proposed technique is improved by two aspects: firstly, use of a multi-scale convolutional block with various kernel sizes to generate a hierarchical feature vector. They also adopted a maximum selection unit that concatenates multi-scale information in an adaptive manner. Secondly, the researchers added an attention mechanism to learn pixel and channel saliency between convolutional features. Furthermore, their experimental results demonstrated that the proposed technique (MSA-CNN) achieved higher performance for driver distraction behavior recognition. As evident from the literature, numerous researchers have proposed several methods for driver distraction detection. It is worth noting that these techniques suffer from substantial shortcomings including limited performance and required huge computational hardware. In addition, such techniques generate false alarms due to the rapid movement of the body owing to their low performance. Furthermore, the selection of a suitable DL model to deploy over a resource constraint device in real time is a challenging task. To cope with this, in the upcoming section, we briefly explained the proposed model that can be easily deployed over resource constraint devices and can improve the performance over SOTA methods.
Proposed Method
We provided a brief discussion about the proposed model to solve the aforementioned problems in a satisfactory way. The proposed model is composed of two main steps such as (1) preprocessing, to prepare data for training and testing, and (2) training the traditional DL model for accurate driver distraction detection. Furthermore, we fine-tuned a pretrained DL model to enhance the driver distraction detection performance and minimize the false alarm rate. In the proposed work, we employed a CA module with several DL models and validated their performance against SOTA over the benchmark datasets. The proposed framework employing the CA module as is presented in Figure 1. The following subsequent sections explain the details of step 1 and step 2.
Preprocessing
Data preprocessing has a vital role in the ML and DL models and is considered as a fuel for their training [25][26][27][28]. Data preprocessing is a technique for cleaning and or-ganizing unusual data to make them well-known information. In simple words, data preprocessing is a task of data mining that prepares the raw data into an understandable form for model training [29]. Furthermore, the useful and error-free data provide optimal results at the time of evaluation. Additionally, there are several techniques of preprocessing, for instance, augmentation, enhancement, data transformation, and data reduction, among others.
Data Augmentation
Data augmentation is a technique that prevents ML and DL models from acquiring irrelevant information. In addition, ML and DL models require a huge amount of data (which are not available easily) for predicting accurate results. In some cases, the available datasets are expanded artificially by applying augmentation techniques [30]. After applying the augmentation technique, the network learns the same object located in the picture with a different view, which enhances the performance of the model [31].
Furthermore, there are different steps available in geometric augmentation, for instance, resizing, cropping, rotating, flipping, scaling, and so on. These transformations expand the available dataset and bring the network toward optimal results.
Resizing plays a major role to train any ML or DL models. In addition, our traditional DL models train very quickly and accurately on small images. Moreover, all the DL models need the images to be the same size. The mathematical formulation of resizing is provided below: Normalization is a scaling technique of translating the low and high intensity pixel into the range of 0 and 1 called Min-Max scaling. It mainly keeps the numerical data in a specific range without changing its shape.
Horizontal flip means "flip" or "mirror" look. Horizontal flipping means transforming all the layers of images horizontally, from left to right or right to left. It only changes the position of the pixel on x-axis without losing any information.
Rotation is a method which is applicable to rotate the object around the center, which simply means, rotate the images in a clockwise or counterclockwise direction. However, we rotated the images 10 • in a clockwise position to generate new images.
Image enhancement is a method used to process the image adjustment, so the resultant image looks more suitable. This method is implemented on input images to avoid noise from an image. The equation is formulated below: In the above equation, g(x, y) is the output of the image, while f (x, y) is the input pixel data; where a 1 , a 2 and a 3 are scaling factors for many grayscale areas and s 1 ,s 2 , r 1 and r 2 are the adjustable parameters.
The Proposed Model
We utilized EfficientNetB0 as a backbone architecture followed by a CA module to increase the performance over the state-of-the-art models. The EfficientNet was proposed latterly as a series of eight networks named, such as, B1, B2, B3, up to B7. The top network version B7 on the ImageNet dataset revealed state-of-the-art results in terms of accuracy by achieving 84.4% top-1 accuracy while using 66 million parameters. In addition, Swish f (x) = x.sigmoid (β.x) = x 1+e βx is an activation function introduced with the version of EfficientNet architecture [32]. Moreover, the Swish Activation (SA) function has better performance than the ReLU activation. It obtains better performance on deeper networks throughout of challenging datasets [33].
EfficientNet was introduced by the researcher of Google Tan et al. [34], which is based on the inverted bottleneck residual block (MBConv), which was originally proposed with MobileNetV2. The major goal of the EfficientNet block is to enlarge the channels and then squeeze them; this technique diminishes the number of channels for the upcoming layers [28]. Moreover, this network also brings down the computational weight; hence, it works in in-depth separable convolutions.
Furthermore, we used EfficientNetB0 as a proposed model, which focuses on detecting the driver's distraction in the early stage based on the optimal performance. On the other hand, EfficientNetB0 is a lightweight architecture where it can easily deploy on edge devices. This model works in block-wise separable convolution neural networks, moreover, it has 237 layers. The proposed model is capable of scaling up or down and it exhibited enormous performance compared with previous state-of-the-art ConvNets [35] on CIFAR-100. The architecture of the proposed model is presented in Figure 2. In the implementation, we used EfficientNetB0 without classification layers, where the features vectors are 7 × 7 with 1280 number of channels (F) and integrated CA mechanism for further strengthening of model performance as is discussed in Section 3.3.
( . ) = is an activation function introduced with the version of EfficientNet architecture [32]. Moreover, the Swish Activation (SA) function has better performance than the ReLU activation. It obtains better performance on deeper networks throughout of challenging datasets [33].
EfficientNet was introduced by the researcher of Google Tan et al., [34], which is based on the inverted bottleneck residual block (MBConv), which was originally proposed with MobileNetV2. The major goal of the EfficientNet block is to enlarge the channels and then squeeze them; this technique diminishes the number of channels for the upcoming layers [28]. Moreover, this network also brings down the computational weight; hence, it works in in-depth separable convolutions.
Furthermore, we used EfficientNetB0 as a proposed model, which focuses on detecting the driver's distraction in the early stage based on the optimal performance. On the other hand, EfficientNetB0 is a lightweight architecture where it can easily deploy on edge devices. This model works in block-wise separable convolution neural networks, moreover, it has 237 layers. The proposed model is capable of scaling up or down and it exhibited enormous performance compared with previous state-of-the-art ConvNets [35] on CIFAR-100. The architecture of the proposed model is presented in Figure 2. In the implementation, we used EfficientNetB0 without classification layers, where the features vectors are 7 × 7 with 1280 number of channels (F) and integrated CA mechanism for further strengthening of model performance as is discussed in Section 3.3.
Channel Attention Mechanism
To optimally select the features in images that contribute to achieving the targeted task effectively, our experiments were performed by using the CA module between two basic layers to acquire the features. In this article, the CA technique contains global average pooling layer, max-pooling layers, three fully connected layers, and a multiplication operation [36]. However, the main objective of channel attention is to show the relationship between each channel of the feature map and to acquire a 1-D weight ∈ and then multiply it to a specific channel. For that reason, it can provide more attention to the important information of images in the target task. To learn optimal weights, we used two parallel connections of pooling operation after (F), which is average pooling and max pooling to make two descriptors for each channel. Then we concatenated the output of both channels and fed into shared multilayer perceptron with 3 fully connected layers to create more effective feature vectors. Lastly, we obtained CA by using the SoftMax function as mentioned in Figure 3. The formula is presented as follows:
Channel Attention Mechanism
To optimally select the features in images that contribute to achieving the targeted task effectively, our experiments were performed by using the CA module between two basic layers to acquire the features. In this article, the CA technique contains global average pooling layer, max-pooling layers, three fully connected layers, and a multiplication operation [36]. However, the main objective of channel attention is to show the relationship between each channel of the feature map and to acquire a 1-D weight W c ∈ R C×1×1 and then multiply it to a specific channel. For that reason, it can provide more attention to the important information of images in the target task. To learn optimal weights, we used two parallel connections of pooling operation after (F), which is average pooling and max pooling to make two descriptors for each channel. Then we concatenated the output of both channels and fed into shared multilayer perceptron with 3 fully connected layers to create more effective feature vectors. Lastly, we obtained CA by using the SoftMax function as mentioned in Figure 3. The formula is presented as follows: x FOR PEER REVIEW 9 of 19
Results and Discussions
In this section, the results are conducted on two benchmark datasets and the performance of the proposed model is evaluated. First, we provided a detailed explanation about the experimental setup, followed by performance parameters, datasets, and finally presented the results of both datasets in terms of quantitative and qualitative analysis.
Experimental Setup
Our experimental results were conducted in TensorFlow 2.3.0 with Nvidia CUDA support. All the experiments were performed on Ubuntu 20.04.3 LTS operating system, equipped with a Core i7-9700KF CPU, 62 GB Memory, and NVIDIA Corporation TU104 [GeForce RTX 2070 Super GPU] with 8 GB of VRAM.
Performance Parameters
Many frozen CNNs with CA mechanism were used in this study. All the models obtained optimal performance based on a variety of metrics such as testing accuracy, testing loss, F1-score, precision, and recall. True positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN) are the confusion matrix instances, through which we determined the performance of a specific network. Accuracy is a confusion matrix term that indicates the performance of the model for all classes. In simple words, it can measure the number of accurate samples to the total number of samples. The recall is also called sensitivity or True Positive Rate (TPR). This instance evaluates the model to detect driver's distraction in positive image samples. The specificity of a confusion matrix is determined by the ability to correctly classify negative samples in all true negative cases. Confusion matrix can manage the model that keeps away the model from misidentifying the driver's distraction. F1-score manages the stability between recall and precision. These matrices are briefly explained in [37][38][39] and the mathematical formulation of these matrices are provided below: 1 − ( ) = 2 * + (8)
Results and Discussions
In this section, the results are conducted on two benchmark datasets and the performance of the proposed model is evaluated. First, we provided a detailed explanation about the experimental setup, followed by performance parameters, datasets, and finally presented the results of both datasets in terms of quantitative and qualitative analysis.
Experimental Setup
Our experimental results were conducted in TensorFlow 2.3.0 with Nvidia CUDA support. All the experiments were performed on Ubuntu 20.04.3 LTS operating system, equipped with a Core i7-9700KF CPU, 62 GB Memory, and NVIDIA Corporation TU104 [GeForce RTX 2070 Super GPU] with 8 GB of VRAM.
Performance Parameters
Many frozen CNNs with CA mechanism were used in this study. All the models obtained optimal performance based on a variety of metrics such as testing accuracy, testing loss, F1-score, precision, and recall. True positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN) are the confusion matrix instances, through which we determined the performance of a specific network. Accuracy is a confusion matrix term that indicates the performance of the model for all classes. In simple words, it can measure the number of accurate samples to the total number of samples. The recall is also called sensitivity or True Positive Rate (TPR). This instance evaluates the model to detect driver's distraction in positive image samples. The specificity of a confusion matrix is determined by the ability to correctly classify negative samples in all true negative cases. Confusion matrix can manage the model that keeps away the model from misidentifying the driver's distraction. F1-score manages the stability between recall and precision. These matrices are briefly explained in [37][38][39] and the mathematical formulation of these matrices are provided below: F1 − Score(k) = 2 precision * recall precision + recall (8)
Dataset Description
In this manuscript, the experiments of driver distraction detection were conducted on the two well-known benchmark datasets: the State Farm Distracted Driver Detection [40] (SFD3), which is publicly available; and the AUC Distracted Driver [40] dataset is a private dataset.
State Farm Distracted Driver Detection (SFD3) Dataset
The State Farm Insurance (SFI) company published a challenging dataset of distracted drivers, which is publicly available on the Kaggle competition. The SFD3 contains around 102,150 images of different driver behaviors that are separated into 10 categories as provided in Figure 4, which are labeled as in Table 2. , x FOR PEER REVIEW 10 of 19 [40] (SFD3), which is publicly available; and the AUC Distracted Driver [40] dataset is a private dataset.
State Farm Distracted Driver Detection (SFD3) Dataset
The State Farm Insurance (SFI) company published a challenging dataset of distracted drivers, which is publicly available on the Kaggle competition. The SFD3 contains around 102,150 images of different driver behaviors that are separated into 10 categories as provided in Figure 4, which are labeled as in Table 2. Table 2.
Class
Class Name Number of Images c0 Safe driving 2489 c1 Texting-right 2267 c2 Calling on the phone-right 2317 c3 Texting-left 2346 c4 Calling on the phone-left 2326 c5 Operating the Radio 2312 c6 Drinking 2325 c7 Reaching behind 2002 c8 Makeup 1911 c9 Talking to the passenger 2129 Total ---22,424
AUC Distracted Driver (AUCD2) Dataset
The AUCD2 is a challenging dataset that was created by Abouelnaga et al. [40], there are thirty-one drivers from different nations, who participated in this dataset. The dataset contains 11,678 images of the different drivers with different postures as tabulated in Table 3; moreover, the images are separated into 10 different folders, where Figure 5 is the visual representation of the AUCD2 dataset. We split both datasets into three sub-sets such as training, testing, and validation. In the training set we have a total of 60% of data, testing 20% of data, and validation 20% of data. Table 2.
AUC Distracted Driver (AUCD2) Dataset
The AUCD2 is a challenging dataset that was created by Abouelnaga et al. [40], there are thirty-one drivers from different nations, who participated in this dataset. The dataset contains 11,678 images of the different drivers with different postures as tabulated in Table 3; moreover, the images are separated into 10 different folders, where Figure 5 is the visual representation of the AUCD2 dataset. We split both datasets into three sub-sets such as training, testing, and validation. In the training set we have a total of 60% of data, testing 20% of data, and validation 20% of data.
Results Evaluation Using SFD3 Dataset
We evaluated and compared the performance of different pretrained traditional DLs with the proposed model using SFD3. To evaluate the performance of our model, we used the Stochastic Gradient Descent (SGD) optimizer with 50 epochs. The training and validation accuracy are illustrated in Figure 6a, while Figure 6b illustrates the training and validation loss, where the confusion matrix of our experimental results is provided in Figure 7. It is clearly shown in the graph that training and validation accuracy of the proposed model are significantly increasing with each epoch, while our proposed model converged above 90% approximately within a few numbers of epochs. After reaching 30 epochs, the model accuracy or loss line graph did not change further and continues as a straight line, until the training process ends. Table 3. Brief detail of AUCD2 dataset.
Class
Class Name Number of Images c0 Safe driving 2706 c1 Texting-right 1438 c2 Calling on the phone-right 976 c3 Texting-left 844 c4 Calling on the phone-left 1040 c5 Operating the Radio 843 c6 Drinking 796 c7 Reaching behind 754 c8 Makeup 764 c9 Talking to the passenger 1517 Total -11,678 1 c0 c1 c2 c3 c4 c5 c6 c7 c8 c9 Figure 5. Visual representation of AUCD2 dataset where the details of c0~c9 are given in Table 3. Figure 5. Visual representation of AUCD2 dataset where the details of c0~c9 are given in Table 3. Table 3. Brief detail of AUCD2 dataset.
Class Class Name Number of Images c0
Safe driving 2706 c1 Texting-right 1438 c2 Calling on the phone-right 976 c3 Texting-left 844 c4 Calling on the phone-left 1040 c5 Operating the Radio 843 c6 Drinking 796 c7 Reaching behind 754 c8 Makeup 764 c9 Talking to the passenger 1517 Total ---11,678
Results Evaluation Using SFD3 Dataset
We evaluated and compared the performance of different pretrained traditional DLs with the proposed model using SFD3. To evaluate the performance of our model, we used the Stochastic Gradient Descent (SGD) optimizer with 50 epochs. The training and validation accuracy are illustrated in Figure 6a, while Figure 6b illustrates the training and validation loss, where the confusion matrix of our experimental results is provided in Figure 7. It is clearly shown in the graph that training and validation accuracy of the proposed model are significantly increasing with each epoch, while our proposed model converged above 90% approximately within a few numbers of epochs. After reaching 30 epochs, the model accuracy or loss line graph did not change further and continues as a straight line, until the training process ends. Furthermore, the proposed model is compared in terms of evaluation metrics with Stacking Ensemble [2], ConvoNet [13], HRRN [17], VGG19 [20] without pretrained weights, and Drive-Net [21]. We notice that Stacking Ensemble obtained 97.00% accuracy using SFD3 dataset as provided in Table 4.
In addition, ConvoNet achieved 98.48% performance in terms of accuracy, while the HRRN method has 96.23% accuracy based on the SFD3 dataset. Comparably, VGG19 and Drive-Net obtained 99.39% and 95% performance in terms of accuracy, respectively, the details are listed in Table 4. Our proposed model surpasses these methods by achieving higher accuracy, which is 99.58% accuracy using SFD3 dataset. In addition, the visual results of our proposed model using SFD dataset are shown in Figure 8. Table 4. Comparison between proposed model with different SOTA model using SFD3 dataset.
Proposed Model 99.58%
In addition, ConvoNet achieved 98.48% performance in terms of accuracy, while the HRRN method has 96.23% accuracy based on the SFD3 dataset. Comparably, VGG19 and Drive-Net obtained 99.39% and 95% performance in terms of accuracy, respectively, the details are listed in Table 4. Our proposed model surpasses these methods by achieving higher accuracy, which is 99.58% accuracy using SFD3 dataset. In addition, the visual results of our proposed model using SFD dataset are shown in Figure 8.
Results Evaluation Using AUCD2 Dataset
Detailed reports of each model across test data using AUCD2 are presented in Table 5. We trained several baseline models for 50 epochs where the proposed model achieved optimal results compared with other models in terms of testing accuracy and testing loss as we observe in Table 5. The training and validation graphs of the proposed method using AUCD2 dataset are shown in Figure 9. Furthermore, the classification reports of the proposed model can be retrieved from the confusion matrix as presented in Figure 10.
Results Evaluation Using AUCD2 Dataset
Detailed reports of each model across test data using AUCD2 are presented in Table 5. We trained several baseline models for 50 epochs where the proposed model achieved optimal results compared with other models in terms of testing accuracy and testing loss as we observe in Table 5. The training and validation graphs of the proposed method using AUCD2 dataset are shown in Figure 9. Furthermore, the classification reports of the proposed model can be retrieved from the confusion matrix as presented in Figure 10. Table 5. Comparison between proposed model with different SOTA models using AUCD2 dataset.
Results Evaluation Using AUCD2 Dataset
Detailed reports of each model across test data using AUCD2 are presented in Table 5. We trained several baseline models for 50 epochs where the proposed model achieved optimal results compared with other models in terms of testing accuracy and testing loss as we observe in Table 5. The training and validation graphs of the proposed method using AUCD2 dataset are shown in Figure 9. Furthermore, the classification reports of the proposed model can be retrieved from the confusion matrix as presented in Figure 10. Table 5. Comparison between proposed model with different SOTA models using AUCD2 dataset.
Ablation Study
This section provides the discussion and results over several DL-based models with and without CA mechanism. The comparison of proposed model with other DL-based models using evaluation metrics such as, F1-score, precision, recall, testing accuracy, and We compared the proposed model using AUCD2 dataset with HRRN [17], C-SLSTM [22], D-HCNN [15], and ConvNet [13], where we examine that HRRN [17] obtained 92.36% accuracy using AUCD2 dataset. In addition, C-SLSTM [22] achieved 92.70% performance in terms of accuracy. Similarly, the D-HCNN [15] and ConvNet [13] methods have 95.59% and 95.64% accuracy, respectively. The proposed model obtained the highest accuracy, 98.97%, using the AUCD2 dataset as mentioned in Table 5. Moreover, Figure 11 is the visual results of the proposed model.
Ablation Study
This section provides the discussion and results over several DL-based models with and without CA mechanism. The comparison of proposed model with other DL-based models using evaluation metrics such as, F1-score, precision, recall, testing accuracy, and
Ablation Study
This section provides the discussion and results over several DL-based models with and without CA mechanism. The comparison of proposed model with other DL-based models using evaluation metrics such as, F1-score, precision, recall, testing accuracy, and testing loss over SFD3 and AUCD2 datasets are briefly explained in the following subsequent sections.
The proposed model and other baselines were trained for 50 epochs with 32 batch size using a low learning rate of 0.001. Further, we set Stochastic Gradient Descent (SGD) with a momentum of 0.9 to ensure that the network retains most of the previously learned information. In these experiments, the proposed model was used to update the learning parameters moderately, which resulted in optimal performance on the target dataset. Additionally, we used the default input size (224 × 224) for each network.
In the experimental results, we conducted extensive experiments to evaluate the effectiveness of the proposed model and other baselines with and without CA for driver distraction detection using SFD3 and AUC2. We compared these models using several evaluation metrics including F1-score, precision, recall, testing accuracy, and testing loss. Our experimental results indicate that the models with CA outperforms among the models without CA across all metrics, indicating that the inclusion of CA enhances the proposed model's effectiveness.
Significantly, the proposed model with CA achieved an F1-score of 1.00, precision of 1.00, recall of 1.00, testing accuracy of 0.9958, and testing loss of 0.0202 for the SFD3 dataset as provided in Table 6. Furthermore, the proposed model with CA also obtained promising performance using the AUC2 dataset based on F1-score, precision, recall, testing accuracy, and testing loss, which were 0.99, 0.99, 0.99, 0.9897, and 0.0425, respectively as tabulated in Table 7. These results justify that the CA can help the model better attend to important features in the input data, which lead to enhancing the model performance. The classification reports of all the models across test data using SFD3 are presented in Table 6 where the proposed model achieved 0.9958 testing accuracy and 0.0202 testing loss. We observe that the proposed model is comparatively better, which exhibits the efficiency of our model. Table 7 shows the results over AUCD2 dataset, where the VGG16 and VGG16+CA obtained the worst results in the experiments. Similarly, ResNet50 and ResNet50+CA also achieved the lowest results comparatively, which is approximately the same as shown in the Table 7. To compare with Xception and Xception+CA, these models achieved optimal performance in terms of testing accuracy. However, it is not suitable to deploy on resource constraint devices. Furthermore, InceptoinV3 and InceptionV3+CA achieved better results; however, the proposed model achieved the highest performance based on testing accuracy. In addition, we proposed this model due to the highest performance and lightweight model capabilities. These two reasons prove that it can be easily deployed on resource constraint devices.
Time Complexity
In the visual domain, achieving lower time complexity is a more challenging task than obtaining promising performance and achieving the smallest error rate in real time. Therefore, we compared the proposed model with four different baseline methods in terms of inference time. In addition, numerous experiments are conducted based on two different hardware such as CPU and GPU as tabulated in Table 8. In these experiments, the ResNet50 and ResNet50+CA have lower inference speed than the InceptioV3 and InceptioV3-CA. The proposed model achieved higher frame per second (FPS) rates for both CPU and GPU than other baseline models, which is 21.73, and 83.75, respectively. In addition, the inference time of the proposed EFFNet-CA can be further enhanced based on hardware improvement. Hence, the inference speed justifies that the proposed model can be easily deployed over resource constraints for real-time decision-making.
Conclusions
Driver distraction leads drivers toward accidents that affect lives, i.e., driver death or major injuries and causes of economic losses, globally. In the literature, several techniques have been introduced to detect driver distraction in an efficient way. However, their techniques are time-consuming, have a high false alarm rate and are difficult to deploy on edge devices due to the high number of parameters. To solve a certain problem, we proposed a novel framework for an efficient and effective driver distraction based on a CNNs with the integration of CAmechanism. Moreover, the proposed model contains three steps, such as training, testing and evaluation. Additionally, our proposed model is compared with various baseline CNNs where only the classification layers were fine-tuned while the rest of the models' layers were frozen. Moreover, the proposed model achieved optimal results in terms of testing accuracy and testing loss using two well-known datasets. The proposed model indicated 99.58% testing accuracy using the SFD3 dataset and 98.97% testing accuracy on the AUCD2 dataset. In other words, the proposed model can easily be deployed on resource constraints devices due to its size and less computational complexity. Further, due to the rapid increase in the developing technologies, the metaverse provides us a great opportunity for better contributions such as the implementation of our proposed work in metaverse-based 3D modeling.
In the future, our goal is to make the proposed model more effective, reduce the false alarm rate, and try to reduce the number of parameters of using model compression techniques such as pruning and quantization. Furthermore, we also aim to deploy our proposed architecture on resource constraints such as Raspberry Pi and Jetson Nano. | 2023-04-12T15:34:28.484Z | 2023-04-01T00:00:00.000 | {
"year": 2023,
"sha1": "5e5b01f77a21f059e08c08d05ce5d84aed2d8f4e",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8220/23/8/3835/pdf?version=1681464159",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d8b58b1aee2f297b6c485d9a2771f9d2a154e105",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
67853998 | pes2o/s2orc | v3-fos-license | Efficacy of Metarhizium anisopliae in the control of the horn fly , Haematobia irritans ( Diptera : Muscidae ) , under natural infestation conditions
Distributed under Creative Commons CC-BY 4.0 Abstract The aim of this study was to determine the efficacy of the entomopathogenic fungus, Metarhizium anisopliae strain 134 (Ma134), for dairy cattle naturally infested with Haematobia irritans (L.) An aqueous formulation of 1 x 108 conidia/ml was applied by aspersion on a group of eight cows every seven days for a total of four applications, while a control group was sprayed with the same solution without conidial content. The degree of horn fly infestation was estimated daily in both groups by the direct count of adult flies found resting or feeding on the animals, and Abbott’s formula was used to calculate the effectiveness of the formulation. The Ma134 formulation resulted in a 68.6 % reduction of the fly infestation during the study period, and a significant reduction was observed from the first week post-application (p < 0.05). The results show that the Ma134 formulation is effective in reducing infestations of horn fly, H. irritans, on dairy cattle under natural conditions. Thus, the fungus, Ma134, can be considered a candidate for biological control for the integrated management of H. irritans.
Introduction
The horn fly, Haematobia irritans (L.), is common and primarily affects cattle on grass pastures in Mexico.The horn fly was introduced from Europe to North America in the late 19 th century.Currently, it is widely distributed across the American continent and is considered the major pest of livestock. 1,2H. irritans is an obligate bloodsucking ectoparasite, always remains on cattle and feeds 24 to 38 times a Original Research day, consuming an average of 1.71 mg of blood per meal. 2 The feeding habits of horn flies on cattle cause irritation and discomfort.In an effort to repel the flies, the animals develop defensive behaviours such as tail flicks, leg stomps, skin twitches and head throws.4][5] Hence, treating horn fly infestations are of sanitary and economic importance.
It has been estimated that weight loss can be up to 220 g/day and milk production can be between 10 and 20 % of regular production when the infestation is high (> 300 flies/animal).Furthermore, H. irritans is an efficient transmitter of several viral, bacterial, protozoal, and rickettsial diseases and helminth infections. 2,6,7he bites of the horn fly reduce the quality of the hides, and insecticide application is costly.][13][14] The use of insecticides has a considerable impact on public health due to the contamination of meat and milk with toxic residues and through environmental pollution.An alternative for reducing the intensive use of insecticides is to use non-chemical measures such as biological control.For example, the entomopathogenic fungi is a safe and sustainable alternative with minimum risk for vertebrates, humans and the environment.This fungus infects the insect through contact, invading it through the cuticle and killing it.After the death of the insect, the fungus emerges from the cadaver to produce new conidia. 15urrently, the efficacy of the fungus Metarhizium anisopliae as a biological control of H. irritans has been demonstrated under laboratory conditions in both immature and adult stages, [16][17][18][19] as well as in adults under controlled infestation conditions. 20However, the efficiency of this entomopathogenic fungus has been scarcely studied under field conditions. 21To advance the development of this biological control, the objective of this study was to determine the effectiveness in reducing the infestation of H irritans on dairy cattle under natural conditions, by applying the strain Ma134 of M. anisopliae by aspersion.
Study site
This work was performed at the livestock unit of El Llano Technological Institute, located in a municipality of the same name in Aguascalientes, Mexico.The place is 2 020 masl, with a semiarid climate, summer rains, and an average temperature of 15.5 °C.In the region, natural infestation of H. irritans is seasonal, being the highest from summer to autumn. 22,23
Entomopathogenic fungi
The strain M. anisopliae sensu lato 134 (Ma134) was isolated from soil in the dairy unit at El Llano Technological Institute using the larvae of Galleria mellonella L. (Lepidoptera: Pyralidae) as a reservoir host.Taxonomic identification was based Original Research on morphological criteria of the reproductive structures; the strain forms part of the entomopathogenic fungi collection at the same Institution. 24The strain Ma134 was evaluated in Musca domestica and Stomoxys calcitrans adults under in vitro conditions with good results 24 and high efficacy with S. calcitrans under field conditions. 25The strain was cultured in the laboratory according to the protocol previously described by Cruz-Vázquez et al. 25 The isolate was cultured on Sabouraud's Dextrose Agar enriched with 1 % yeast extract 26 containing 500 ppm chloramphenicol and incubated at 25 ± 1 ºC for 21 days in a 12:12 hr light/dark regime.Conidia were harvested by scraping, suspended in sterile distilled water containing 0.1 % (v:v) Tween 80, and homogenized on a vortex mixer.Spore viability, which exceeded 98 %, was determined by seeding 100 µl of conidial suspension on Sabouraud's Dextrose Agar and counting colonies 48 h later.Mass reproduction was performed on grains of rice according to that described by Ángel-Sahagún et al. 27
Field trial
Two experimental groups were formed with eight Holstein cows, each group was maintained in an area of 5 000 m2, separated from each other by 500 m.Both areas had ryegrass pasture (Lolium perenne), water tanks for drinking and natural shade provided by various native trees.The cows did not receive any treatment to control fly infestation during the year prior to this trial.The experiment began the first week of September, before the second fly population peak of the season according to information reported in the area. 22,23The first group received an application of the aqueous formulation of Ma134 fungi (1 x 10 8 conidia/ml concentration), water, Polyoxyethylene sorbitan monooleate (Tween 80®, Sigma-Aldrich Co., St. Louis, MO, USA) solution (0.01 %) and an agricultural adjuvant at 0.1 % (Inex-A, Cosmocel, Mexico).Five litres of the formulation were applied per animal by aspersion using a back sprayer with a cone-type nozzle at a pressure of approximately 40 lb/in2, on four occasions with seven-day intervals between treatments.The formulation was prepared 30 min prior to administration.The control group received a water solution with Tween 80 (0.01 %) plus an agricultural adjuvant at 0.1 %.Sprayings were between 7 and 8 pm to avoid sunlight.The number and interval of applications were similar to that used in a previous study with the stable fly. 25The Committee on Use and Care of Animals of the Instituto Tecnológico El Llano Aguascalientes approved this project, and adequate veterinary care was provided to all of the animals under study.
Estimation of H. irritans infestation
An Infestation index (average number of flies per animal) was estimated daily by the direct count of adult flies found resting or feeding on the animals; each cow was considered an experimental unit.This estimation consisted of photographing only one side of the body of each animal (including regions of the head, neck, back, sides and limbs).using a digital camera with a 5x optical zoom function.This activity was done daily between 8 and 9 am, always by the same person, who was not aware of the status of the cows.Counting the number of flies was performed by computer image analysis.The obtained number was multiplied by two to obtain Original Research an estimate of the total number of flies. 20,28Climatic factors were registered as reported by the El Llano weather station, the measurements included the average daily temperature (T °C), the average daily relative humidity (RH %) and the rainfall (Rf) during the study period.
Data Analysis
The information recorded in the groups under study (treated and control) was analysed by analysis of variance and Student's t-test (p < 0.05), with the objective to demonstrate the differences between the groups in each study week using the corresponding week's average index of infestation values.Then, Abbott's formula was used to determine the efficacy of the formulation 29 by identifying the percent reduction in the average infestation index for each evaluated week. 20
Results and Discussion
The population curve of the horn fly, H. irritans, in the groups under study is shown in figure 1.On the second day post-treatment, the daily index infestation began to decrease compared to the control group, maintaining this trend until the end of the study.Infestation index at day 0 in the treated group was 530 and in the control group, 532.The aqueous formulation reduced the index infestation of 538, which was determined at the start of work (day 1), to 169 on day 28 post-treatment, representing 68.6 % efficiency.
In Table 1, the average infestation index is shown in each of the study weeks.The infestation index was reduced from week to week in the treated group.In the control group, it was maintained between 530 and 569 in the first three weeks; however, in the fourth week, it was reduced to 479.The number of flies in the treated group was always less than in the control group.Statistical differences were detected each week between treated and control groups: first week (t 14 = -3.14, p = 0.006), second week (t 14 = -10 -19 , p = 7.3 x 10 -8 ), third week (t 14 = -25.88,p = 3.17 x 10 -13 ), and fourth week (t 14 = -9.08,p = 3.04 x 10 -7 ).The cattle remained healthy during the study period.Climatic data during the study period indicated that the average temperature was 20.2 °C with a range from 18 to 21.6 °C, while the RH was on average 59.7 % with a range from 38.4 to 80.3 %, and no rainfall was registered (figure 2).
The results obtained in this study show that the formulation of the Ma134 strain was able to control horn fly infestation under field conditions.Compared to the control group, an infestation reduction of the fly in the treatment group was consistent during the study, and the weekly applications helped reinforce the pathogenic action of the fungus.Thus, it can be considered that the treatment had an additive effect.The aqueous formulation of Ma134 was added to an agricultural adjuvant to promote adhesion of conidia to the flies and animal hair and provide assistance in protecting the conidia from exposure to sunlight.In the literature, it is mentioned that conidia are sensitive to UV rays and to high environmental temperatures, and when applied to the body of animals conidia are sensitive to body temperature. 30The design of a formulation for topical application should consider at a minimum these factors; hence, the incorporation of an agricultural adjuvant or another compound with a heat/UV protecting ability can improve the performance of the conidia. 21,27However, Angel-Sahagún et al. 27 did not find that environmental conditions had any effect on the performance of entomopathogenic fungi used in experimental grasslands artificially infested with larvae of Rhipicephalus microplus.The tolerance of entomopathogenic fungi to environmental conditions appears to be related to the site where the fungus was isolated, 31 and it can be assumed that the Ma134 strain is adapted to a semiarid climate, the same as the study site, which probably contributed to its good performance in the study.
M. anisopliae has been the most studied entomopathogenic fungus because of its cosmopolitan and highly pathogenic nature.In addition to the fungus being specifically pathogenic for common insects affecting domesticated animals, such as ticks and flies, it is an organism that has minimum risk for plants, animals, humans and the environment. 15,30,32In this study, no adverse effects of the fungus were observed in the cattle.Other authors reported similar findings in cattle infested with ticks or stable flies and treated with M. anisopliae formulations under field conditions. 25,33,34The homogenous weather conditions in the study period and an ad-
Study week Treated group Control group Efficacy (%)
Original Research equate combination of humidity and temperature were ideal for the development and maintenance of the population curve of H. irritans.However, at the end of the study, the change in these environmental conditions, especially the low average daily temperature, favoured a decline in the infestation of flies.Cruz-Vázquez et al. 22,23 describe similar population dynamics in the region, characterized by a close relationship with temperature and relative humidity.In these studies, it is mainly suggested that temperature induces a decline in population growth and in the late fall temperature can cause diapause.Recently, in Tecoman, Colima, Mexico, a region with a tropical dry climate, the effect of several strains of M. anisopliae on steers with an experimental infestation of H. irritans was reported: mortality was observed from day 1 post-treatment (20 %), on day 4 it was 61 %, by day 8 it was 84 %, and from days 12 to 13, it was between 94 % and 100 %.However, the absence of reinfestation limited the possibilities of estimating the longevity of the effect of the fungus. 20In São Paulo state, Brazil, a region with a sub-tropical climate, a study was performed to evaluate the susceptibility of adult H. irritans flies to a water-based formulation of a strain of M. anisopliae under natural infestation conditions.For this purpose, the formulation had a concentration of 3 x 10 10 conidia /ml and was applied to cattle every five days on four occasions.This study found that the level of infestation reduced in the treated group from the second spraying of the formulation, while in the untreated group infestation always increased.At the end of the study, the treated group had an average of 28.5 flies/head while the untreated group had 65.4 flies/head. 21n our study, an additive effect was also found.This phenomenon was observed both in cattle naturally infested with S. calcitrans 25 and in cattle infested with R. microplus ticks under field conditions. 34The use of multiple applications of an
Table 1 .
Infestation index average and efficacy of the Ma134 aqueous formulation applied on Holstein cows naturally infested with H. irritans.Different subscripts (a,b) on the same row indicate significant differences (p < 0.05).Efficacy: reduction percentage of the average infestation index calculated with Abbott's formula. | 2018-12-21T02:47:37.850Z | 2017-04-06T00:00:00.000 | {
"year": 2017,
"sha1": "11647bc20efbf75fc760759c883bc227da1a3c69",
"oa_license": "CCBY",
"oa_url": "http://veterinariamexico.unam.mx/index.php/vet/article/download/384/496",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "11647bc20efbf75fc760759c883bc227da1a3c69",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
52949558 | pes2o/s2orc | v3-fos-license | Factors Affecting Pricing in Patent Licensing Contracts in the Biopharmaceutical Industry
: This paper analyzes factors affecting pricing in patent licensing contracts in the biopharmaceutical industry based on a dataset that includes royalty-related data such as running royalty rate, up-front payment, milestones, and deal value. Data on drug candidates for 11 drug classes is obtained for regression analysis between royalty-related data and multiple input descriptors such as market factors, licensor factors, and licensee factor in order to derive the formula for predicting royalty-related estimates such as royalty rate, up-front payment, milestones, and deal value. Data is gathered from multiple sources including MedTrack and is processed through merging and cleaning. We found that the three most important factors in pricing in patent licensing in the biopharmaceutical industry are CAGR (Compound Annual Growth Rate), PDELR (Previous Deal Experience of Licensor), and AR (Attrition Rate). We found that factors in the formula used to estimate the license fee are totally different by drug class. We found that the three most important factors in the frequency in the formula used to estimate the license fee are PDELR, RnDLR (R&D Costs of Licensor), and PDELE (Previous Deal Experience of Licensee). This study suggests a method of estimating the proper royalty rate, up-front payment, milestones, and deal value of the drug candidates of 11 drug classes by using easily obtained input data.
Introduction
Drug development is expensive and risky because it requires a lot of time and money from an initial concept to approval [1,2].Licensing is recognized as a good exit strategy to overcome the economic crisis in biotech companies and royalties are recognized as one of the major revenue sources in the biotech industry [3,4].Therefore, the number of licensing transfers in the biotech industry is at least 10 times that of other industries [5,6].
Valuation for the drug under development is required for various purposes including merger and acquisition (M&A) transaction and two major quantitative valuation approaches are applied in the biotech industry, that is the discounted cash flow (DCF) method and the real options method [7,8].The risk-adjusted Net Present Value (rNPV) method is an NPV method that uses only the attrition rate at development stage as a discount rate to consider the risk at each development step and was developed to overcome the disadvantages of DCF and real options methodology.The rNPV method, which considers the drug type as well as the development phase, recently emerged [9], but it is not enough to reflect the characteristics of the biotech industry and to provide accurate valuation for early-stage drugs.Paik's study confirmed that technology providers (licensor) and technology adopters (licensee) differ in factors that are important to technology transfer, while past studies tend to focus on only one type of licensors/licensees [10,11].
The purpose of this study is (1) to identify the factors that affect each license fee through previous studies, (2) to analyze the effect of each factor on the license fee by multiple regression, and (3) to make a formula through the values of the influencing factors by drug class by using backward elimination method.
The biotech industry is facing a challenging situation with decreasing financial performance and declining R&D productivity.Licensing has become an important element of the business model in the biotech industry and a key activity for biotech companies and determining the right royalty rate in licensing deal is very important to pharmaceutical and biotechnology success.The valuation is vital to valuate the technology in finding licensing partner, fundraising for further development, portfolio management in biotech industry [12].The value of technology depends on a number of factors in licensing deal between technology provider and technology consumer [6].Main technology providers (licensors) are biotech, academia including national research institutes.Main technology consumers (licensees) are pharma company and biotech [12].This study is meaningful to provide royalty-related estimates such as royalty rate, up-front payment, milestones, and deal value by drug class based on historical deal data in more simple way.The existing valuation methods like DCF and the real options require long time more than one month and complex steps to derive these royalty-related estimates.Licensing deals between pharmaceutical companies and biotech/academia are active in the biotechnology area [8].
The types of factors in view of β (Non-standardized regression coefficients) that influence all aspects of the license fee including deal value, royalty rate, upfront payment, milestones is as follows: 1 CAGR (Compound annual growth rate) 2 PDELR (Prior Deal Experience of Licensor) 3 AR (Attrition Rate) 4 RNDLR (Licensor R&D Costs) 5 RNDLE (Licensee R&D Costs) 6 SALESLR (Licensor Sales) 7 SALESLE (Licensee Sales).In fact, these factors can be later determined in order (the rank of influencing weights) since we investigate the correlation between the predictor and the dependent variables, leading to the size of regression coefficients.We found that the three most important factors in pricing in patent licensing in the biotech industry are CAGR, PDELR, and AR.We found that factors in the formula used to estimate the license fee are totally different by drug class.Total frequency of factors in the formula to estimate all the technical fee including deal value, royalty rate, upfront payment, milestones by drug class is as follows: 1 PDELR (22 times) 2 RNDLR (16) 3 PDELE (15) 4 SALESLE, CAS (14) 5 AR (9).We found that the three most important factors in the frequency in the formula used to estimate the license fee are PDELR, RNDLR, and PDELE.This rest of this paper is organized as follows.The second section reviews licensing in the biotech industry and the development status of the valuation method.The third section describes the proposed research framework.The fourth section presents the data, empirical model, and results.The final section contains the discussion and implications.
Licensing in the Biopharmaceutical Industry
The discovery and development of new drugs is a highly risky investment because it is a lengthy and expensive process.Few companies have the capability to develop products from the discovery of a new lead compound to delivering the approved drug to patients.Even those that do have the capability may not have the capacity to have pipelines that are broad enough to provide a smooth flow of new drugs [6].Small pharmaceutical biotech companies with good technology generally have limited R&D budgets and drug development periods are so long that they are often close to budgetary crisis, which makes licensing a good strategy and business model to overcome monetary crisis [3,13].
Licensing between competing firms is popular in many countries.Several papers have focused on the horizontal product differentiation approach.Using this approach, Erkal (2005) finds that royalty licensing is always profitable for the licensor regardless of its technological lead over the competitor.Furthermore, when the technology gap between competing firms is small, it is socially optimal to discourage licensing.Sinha (2010) shows that when licensing takes place between a multinational and a domestic firm, if FDI (Foreign Direct Investment) is the multinational's mode of entry into the domestic market, royalties are more profitable compared to fixed fee licensing [14].
Development of Valuation Method
Value is established through trade and is dependent on market supply and demand as well as other factors.Essentially, value is the amount that both the buyer is willing to pay and the seller is willing to accept.Both sides will be interested in achieving the best deal and will want to know what other buyers might pay or sellers accept.Comparable products or technologies that have recently been traded may provide an indication of the expected value.The value of technology depends on many factors.There will be more than a dozen variables in even the simplest eNPV (Expected Net Present Value) model for a product in early clinical development.These include the size of the target market for the final therapeutic product, the anticipated clinical qualities of the drug, and the extent of competition for the drug and these will include the phase-specific success probabilities, development costs and timelines, the expected market size and market share, the costs of goods, and marketing and administration [6].
Ruckman and McCarthy examined the determinants of licensing using 93 licensing agreements in the global biopharmaceutical sector in a panel dataset spanning 1993-2007 and found that licensors with strength in technological prestige, licensing experience, and combined technological depth and breadth are more likely to license-out their patents and have a greater chance of being chosen by licensees [15].
Research on technology transfer has been carried out mainly through analysis of policy factors, factors for promoting technology transfer, and ways to promote the success of technology transfer.According to a recent study by Hwhang (2015), it was found that technology transfer through exploration of potential demand is more effective than 33 government-sponsored research institute technologies [16].
Understanding how fees or compensations are determined in patent licensing contacts is a critical issue for firms trying to profit from their inventions as well as for firms that source technology.Patent licensing has become an increasingly widespread practice [17,18].
Patent licensing is particularly important for companies that have limited resources to commercialize their own inventions.Companies may choose licensing if the invention falls into areas that are not core to their business.The licensing of patents has also become an important concern for universities, which have been encouraged to profit from their own inventions since the Bayh-Dole Act of 1980 gave universities the right to own and license inventions resulting from federally-funded academic research.Several explanations for the phenomenon of patent under exploitation have been suggested, including managerial myopia, inertia, management incompetence [18], the marginality and low value of patents, and transaction costs or other impediments in the technology market [19].
The valuation of patents has caught widespread attention in innovation research because patents are one of the few readily available measures of research output.Mariko (2010) added the proximity of a licensed patent to a licensor's core technology (the number of a licensor's patents in the same IPC as the licensed patent compared to the total patent stock of the licensee), licensor patent stock (the number of patent applications by a licensor), and licensee patent stock as the determinants of the price of patent licensing.Sakakibara's research added the sales per capital of licensor and licensee as a size indicator among the determinants of the price of patent licensing [11].However, due to the limited availability of data, past studies typically used proxies for the value of patents, such as citations [20], Tobin's Q [20,21], patent renewal [20], and patent litigation costs [22].Scherer and Harhoff [23] and Gambardella et al. [24] use a survey-based measure.Research based on actual patent transaction data such as survey-based research [25], examination of those contract features which do not include pricing [17], and studies of university patents [26,27] is very limited.Patents without established market values (e.g., no negotiated royalty rates) are often valued by comparing the number of citations the patent has received to those received by other patents whose market values are established.Caviggioli and Ughetto (2013) investigated the main drivers of companies' decisions to engage in patent licensing and sales transactions and then examined how improvements in the marketplace for patents and more intense involvement of patent brokers might impact the factors hindering the development of patent transactions [28].For recently issued patents, which have not had time to accumulate citations, this [29].
Determining a reasonable royalty rate for a licensing deal is very challenging, and proper valuation is essential to make a reasonable determination.There are two major valuation approaches used in the field of life sciences: discounted cash flow (DCF) and real options [1].The valuation methodology for new drug pipelines includes rNPV or eNPV (expected NPV) or the probability-adjusted NPV method in which technical risk is reflected on the cash flow, the scenario-implemented decision tree method, which calculates the weighted average NPV according to the probability of specific scenarios, and real options, which is often used by financial professionals but is not favored by drug experts.The rNPV method is most frequently used in practical work in the field of life sciences [1].Attrition rates proposed by DiMasi [30] are most often used for valuation among technology traders in biotechnology and pharmaceutical companies.A recent study by Thomas, D.W. et al. derived the attrition rate at each development stage for four drug types [9].
Lee et al. suggested a method of estimating the proper royalty rate and up-front payment using the formula derived from the regression of a dataset of historical licensing data [1].However, further in-depth research is necessary to investigate the relationship between royalty-related data and more input variables that can be converted to numerical value and can be used for the input for prediction [1].Lee's study group suggested a method of estimating the proper royalty rate and up-front payment using multiple data descriptors.In the regression analysis study on Anticancer activity candidates, factors that influence the royalty rate are ranked as follows: (1) Licensee Revenue (+) (2) Market Size (−) (3) TCT median value (−) (4) CAGR (−) (5) Attrition Rate (+).In the regression model to predict the royalty rate, Royalty Rate is directly proportional to Licensee Revenue and Attrition Rate and Royalty Rate is inversely proportional to Market Size, TCT median value, and CAGR [8].
Byeon (2013) studied the influence of factors on the outcome of the technology trading contract.The positive order of influence on the success of the technology trading contract was found to be the technical transaction environment factor, ability, and the [31].Jong-ilBaek and Byung-hwan Hyun (2017) analyzed the priority factors of technological, marketability, internal factors, and technical supply agencies, which act as complex variables in license fee negotiation, using AHP analysis from the viewpoint of licensee [32].Another study by Jong-ilBaek and Byung-hwan Hyun study (2017) shows that both parties including the government-funded research center and private company acknowledge "Technical considerations for determining the profitability of the technologies" and "The interest and willingness of the management group" as critical factors for the determinants of royalties.The difference between the parties is that private companies acknowledge "Available budget plan" as a critical factor while government-funded research centers value "Market competitiveness" [33].
Research Framework
Paik's study separated licensor and licensee when considering the factor influencing technology transfer, while past studies tend to focus on only one type of licensors or licensees [10,11].The rNPV method became the major valuation approach since DiMasi suggested the attrition rate at development stage as a factor for valuation in the biotech industry [1,3,8,10,[34][35][36][37].According to Baek's study, Market Size and CAGR were the major factors to estimate the running royalty in the biotech industry [8,32,33].
Market
The attrition rate at development stage was proposed by DiMasi's study [30] and is used most often for the valuation in the biotechnology industry.According to Baek's study, Market Size and CAGR were the major factors to estimate the running royalty in the biopharmacuetical industry [33] and several studies followed [8,32].
Licensor
The licensor seeks to extract value in the form of licensing revenues and may be motivated to license-out their technologies because they lack the financial, physical, or intellectual resources to commercialize it [15].Past studies tend to focus on only one type of licensor or licensee [11].However, Paik's research confirmed that licensor and licensee differ in factors that are important to technology transfer, such as previous deal experience [10], and there have been several subsequent studies on licensor factors [38,39].Svačina and several groups added R&D licensor costs as factors affecting the value of intangibles [18,40].Sakakibara added the sales per capital of the licensor and licensee as a size indicator among the determinants of the price of patent licensing [11].
Licensee
Paik's research confirmed that licensor and licensee differ in factors that are important to technology transfer, such as previous deal experience [10], and several studies for licensee factor followed [31,38,40].Pavel Svačinaand several groups added the R&D costs of licensee to as the factors affecting the value of intangibles [18,41].Sakakibara added the sales per capital of the licensor and licensee as a size indicator among the determinants of the price of patent licensing [11].
Empirical Model
We employed linear regression model like the followings.Technical Fee is a dependent variable and PDELR, PDELE, MS, CAGR, AR, RNDLR, SALESLR, RNDLE, SALESLE are independent variables.
Model I.
Data
Data is gathered from multiple sources such as MedTrack and Harvard WRDS (Wharton Research Data Services) for analysis and is processed through merging and cleaning.
Market
The attrition rate at development stage was proposed by DiMasi's study [30] and is used most often for the valuation in the biotechnology industry.According to Baek's study, Market Size and CAGR were the major factors to estimate the running royalty in the biopharmacuetical industry [33] and several studies followed [8,32].
Licensor
The licensor seeks to extract value in the form of licensing revenues and may be motivated to license-out their technologies because they lack the financial, physical, or intellectual resources to commercialize it [15].Past studies tend to focus on only one type of licensor or licensee [11].However, Paik's research confirmed that licensor and licensee differ in factors that are important to technology transfer, such as previous deal experience [10], and there have been several subsequent studies on licensor factors [38,39].Svačina and several groups added R&D licensor costs as factors affecting the value of intangibles [18,40].Sakakibara added the sales per capital of the licensor and licensee as a size indicator among the determinants of the price of patent licensing [11].
Licensee
Paik's research confirmed that licensor and licensee differ in factors that are important to technology transfer, such as previous deal experience [10], and several studies for licensee factor followed [31,38,40].Pavel Svačinaand several groups added the R&D costs of licensee to as the factors affecting the value of intangibles [18,41].Sakakibara added the sales per capital of the licensor and licensee as a size indicator among the determinants of the price of patent licensing [11].
Empirical Model
We employed linear regression model like the followings.Technical Fee is a dependent variable and PDELR, PDELE, MS, CAGR, AR, RNDLR, SALESLR, RNDLE, SALESLE are independent variables.Model I.
where u i is the error term.
Data
Data is gathered from multiple sources such as MedTrack and Harvard WRDS (Wharton Research Data Services) for analysis and is processed through merging and cleaning.
Data collection is based on the following resources:
Variables
In general, we should consider whether control variables influence the relation between independent and dependent variables.To investigate various factors affecting the deal values of the licensing contract, at first, we introduced potential candidates for control variables such as whether the licensee is a large firm or small-and-medium size firm (which is categorized to nominal variables).But in this work, we put all potential control variables to the subset representing independent variables, because we found no influence effect in that there exists no statistical significance without showing no cases less than significance level of 0.05 (5%) even if we fix one for control variable and remain the rest for independent variables.
In case of PDELR, PDELE, we calculated the number of deals prior to the deal date.In case of RNDLR, SALESLR, RNDLE, SALESLE, we used the value of the previous year based on the deal date.DiMasi's attrition rate [30] is applied for the development phase and the average royalty rate is used after data cleaning.
Table 1 presents the category, criteria, and related studies of each variable and Table 2 describes the statistical characteristics of independent variables Sales (Revenue) of Licensee (M$) [11,15,37,59] Table 2 presents the statistics of independent variables.
Regression Analysis of Total Dataset
As one of the most popular regression methods in use, we hereby employed the regression analysis of "Enter method" in Table 3 but found that most of the variables returned beyond the significance level, except for specific metrics such as "PDELR, CAGR, RNDLR, SALESLR, RNDLE" (for DV) and "PDELR, CAGR, RNDLR, SALESLR" (for MP).Thus, we applied to an alternative of "Backward elimination method" in Table 4, which might be well-known as to provide high fidelity to model fitness and "metric significance" and much more appropriate for our work.From the results of multiple regression analysis, we did not find the occurrence of multicollinearity issues in that tolerances for all four cases (DV, Milestone, Royalty and Upfront) are no less than 0.1, and variable inflation factors (VIF) are no greater than 10.0.Thus, our models are justified without no collinearity problems.
Factors that influence the deal value are ranked as follows: 1 CAGR (+) 2 Prior Licensor Deal experience (+) 3 Attrition Rate (−) 4 Licensor R&D Costs (+) 5 Licensee R&D Costs (+) 6 Licensor Revenue (−).In the regression model to predict the deal value, deal value is directly proportional to CAGR, Previous Licensor Deal experience, Licensee R&D Costs, and deal value is inversely proportional to Attrition Rate and Licensor.Factors that influence the Total Upfront Payments are ranked as follows: 1 Previous Licensor Deal experience (+) 2 Licensor R&D Costs (+) 3 Licensor Revenue (−).In the regression model to predict the total upfront payments, Total Upfront Payments is directly proportional to Previous Licensor Deal experience, Licensor R&D Costs and Total Upfront is inversely proportional to Licensee Revenue.
Regression Analysis of Deal Value by Drug Class
Based on the total dataset, we performed the regression analysis on deal value (dependent variable) by 11 drug class in Tables 5 and 6 and obtained the algebraic formulae sensitive to each drug class (independent variable) as below.
The formula to predict deal value by drug class is as follows. (
Regression Analysis of Royalty Rate by Drug Class
In a similar way with the afore-mentioned cases for "influencing factors-deal value" relationships, we performed the regression analysis on royalty rate (dependent variable) by drug class in Tables 7 and 8 and obtained the algebraic relationships between each independent variable and royalty rate.
Regression Analysis of Total Upfront Payment by Drug Class
Here, we got to the regression analysis on Total Upfront (TU) payment (dependent variable) by drug class in Tables 9 and 10, and obtained the algebraic relationships between each independent variable and TU.The formula to predict total upfront payment by drug class is as follows.
(1) CV (3) Derma There were five samples in the Dermatology dataset, which was too small to perform the regression analysis.
There were four samples in the Haematology dataset, which was too small to perform the regression analysis.
There were five samples in the Musculoskeletal dataset, which was too small to perform the regression analysis.(
Regression Analysis of Milestone Payment by Drug Class
Lastly, we performed the regression analysis on Milestone Payment (dependent variable) by drug class in Tables 11 and 12 and obtained the algebraic models for MP.
Discussion and Conclusions
In this study, we have investigated influencing factors directly coupled with pricing in patent licensing contracts in the biopharmaceutical industry based on a dataset that includes royalty-related data such as running royalty rate, up-front payment, milestones, and deal value.We have also considered multiple input descriptors such as market factors, licensor factors, and licensee factor, assuming that royalty-related data could be statistically treated or analyzed by drug class in order to derive the formula for predicting various types of royalty-related deal value estimates.
Depending on the drug class, most of the factors related to the amount of technology transactions such as the costs for the clinical phases, the attrition rate, the drug development time, the average sales amount, and the sales curve, and control over pricing by country-specific regulators change [12,59].Cardiovascular disease drugs have a higher probability of entering the market than oncology drugs, and the clinical success rate of protein therapeutics is higher than that of chemical drugs [59].Real price statistics show that the real price for eye therapeutic class is the lowest (USD 90.87) and the real price for viral infections is highest (USD 637.69) [60].However, existing valuation methodologies, such as r-NPV, do not take into account factors that vary depending on the drug class.The afore-mentioned valuation methodologies would only consider successful probability and investment cost at each pre-clinical/clinical stage as well as typical principal variables such as technology economic life, cash flow estimation, discount rates, technology factor (or contribution rate).Here, successful probability and investment cost are not related nor sensitive to drug class yet.For this reason, a regression study that can reveal these variables that can be changed according to the drug class is meaningful.If the expected development cost of drugs is relatively low and the expected operating profit is relatively high, the drug is of course profitable.An example is the orphan drug, a drug for rare diseases.The market size of orphan drug is lower than that of general medicine, but the cost of medicine per patient is much higher than the average cost of general medicine.Orphan drugs are eligible for special tax deductions and other benefits [59].For this reason, we have reached the possible hypothesis that some of the variables used in the regression analysis according to the drug class may work properly, and some variables may not, leading to the fact that we might find significant variables with corresponding coefficients from the regression analysis.In previous similar studies, we have derived a formula that predicts the royalties and upfront payment through regression analysis for anticancer and cardiovascular drug classes [1].In this study, the formula type of the regression model for the oncology drug was "Rational Model, 3 Parameter Type 2 Regression" and the formula type of the regression model for the cardiovascular drug was the "Polynomial Model, Inverse Second Order-type Regression".Curve types for oncology drugs and cardiovascular drugs were also different [1].
While taking the integration of regression analysis results in preceding sub-sections in Section 5, we find that most regression models involve two or more independent variables in either (+) or (−) proportionality relations, except for several cases such as Royalty Rate for Derma, InI, Onco and Milestone Payment for EMGO, InI.
The proportional relationship between dependent and independent variables in the regression analysis on factors affecting pricing in patent licensing contracts in the biotech industry are illustrated in detail in Table 13.Technical fees such as DV, MP are in directly proportional to PDELR, CAGR, RNDLR, and SALESLE while DV is inversely proportional to AR. DV and MP is inversely proportional to SALESLR, but TU is directly proportional to SALESLR.Factors that influence all technical fees including deal value, royalty rate, upfront payment and milestones are ranked as follows: 1 CAGR, 2 PDELR, 3 AR, 4 RNDLR, 5 RNDLE, 6 SALESLR, 7 SALESLE.MS and PDELE are excluded in the factors affecting pricing in patent licensing contracts, because p-value for MS and PDELE was statistically not meaningful in the regression result.Although the market size for the sub-classical drug class can be used for regression to obtain more accurate results, it is difficult to obtain market size information corresponding to the subclassification.The market size used for the actual regression is 11 drug classes corresponding to the major category.For this reason, we assume the p-value for MS is not statistically significant.We need further study why MS and PDELE are excluded in the factors affecting pricing in patent licensing contracts.We found that three most important variables to consider in the licensing deal in biotech industry are CAGR, PDELR, and AR.Knowing the factors to consider in the licensing deal is very important to prepare for the licensing deal negotiation and it is advantageous to have a better positioning in the negotiation.
In Table 13, we find that statistically CAGR has a positive influence on DV and MP, but it has no relations with RRA and TU.This implies that RRA and TU can be influenced by licensee's financial status and company circumstances.In the meanwhile, DV and MP might be enhanced in accordance with economic growth and relevant factors such as a firm's technical necessity and willingness-to-pay, etc.The proportional relationship between dependent and independent variables by drug class in the regression analysis on factors affecting pricing in patent licensing contracts in biotech industry are illustrated in detail in Tables 14-16.The total frequency of factors in the formula to estimate all technical fees including deal value, royalty rate, upfront payment, and milestones by drug class is as follows: 1 PDELR (22), 2 RNDLR (16), 3 PDELE (15), 4 SALESLE, SALESLR (14), 5 AR (9).We found that the three most important factors in the frequency in the formula to estimate the license fee are PDELR, RNDLR and PDELE.Factors that influence all technical fees including deal value, royalty rate, upfront payment and milestones are ranked from the viewpoint of B as follows: 1 PDELE, 2 PDELR, 3 We can see from the above that factors in the formula to estimate the license fee are totally different by drug class.This study is meaningful in that it is the first one to suggest a method of estimating four types of technical fee for most of drug classes.
Usually the DCF method often results in a negative value when applied to early-stage drugs [1].In fact, when evaluating the economic value of a technology or business project including biotechnology fields, we need to consider the period and cost for commercialization due to pre-clinical/clinical demonstration.Since the discounted cash flow (DCF) method has limitations in that it cannot consider consecutive investments or does not reflect the probabilistic property of commercialization cost, we often take it desirable to apply the concept of 'Real Option' with key metrics of underlying asset value, commercialization cost, and volatility, while regarding the value of technology and investment as the opportunity value [61].More elaborated real options model, which reflects the uncertainty in the option pricing model (OPM), can provide a feasible solution to overcome the challenges of rNPV, widely used in biotechnology or pharmaceutical industries.Therefore, the Real Option method often has advantages over the DCF method in that it has the flexibility to cope with changes in the external environment but has often been criticized for offering too high a valuation price.Estimating future product sales by the Real Option method is very difficult due to market changes due to fierce competition, the emergence of new science and technology, and possible additional indication [59].rNPV is an effective tool to approximate value in late stage products with well-defined markets, but the predictive power of rNPV method is heavily dependent on accurate estimates of multiple factors like market size, market share, project costs, project timing, probabilities of success at each major milestone and discount rate.However, it is not easy and one to require a lot of time to estimate multiple variables accurately.Overall, valuation of drug candidates for licensing remains more art than science.Objective valuation methods exist but require a lot of subjective inputs to generate estimates of value [62].On the other hand, the method presented in this study can be used to estimate the license fee as an input of objective data that can be easily obtained, and it is also applicable to early-stage drugs as well as late stage drugs.The valuation is essential in finding licensing partner, fundraising for further development, portfolio management in biotech industry [12] and the result of regression for 11 drug classes can be a good starting point to be referred to by a manager in case of negotiations in biotech licensing deals (for finding licensing partner and fundraising) for a specific drug classification.The result can be a good starting point to manage the portfolio within biotech company.The regression models of this study could be used for the following purposes: (1) Derive the base amount that will be the starting point of the negotiations for drug licensing tasks in Biotech and in-License/Out-License departments of biotech and pharmaceutical companies (2) Assist decision-making of go/no-go of a specific project among multiple drug pipelines.(3) Pre-economic evaluation of national research project related to medicine.Technology licensing, M&A, and partnership are main open innovation strategies on the closed innovation system [59].So, the results of this study are expected to be helpful tools for implementing open innovation strategies.Since the Fourth Industrial Revolution affects business models, it could reshape the organizational forms to deliver values also [63].So, the results of this study are expected to be helpful to reshape the organizational forms to deliver better value in the Fourth Industrial Revolution age.Further in-depth research is necessary for the following topics in the future: (1) Regression analysis using more independent variables like technical life cycle, number of patented countries, quality of technology and so on.(2) Case study comparing the regression model in this study with DCF and rNPV method.
Figure 1 20 Figure 1 .
Figure 1 presents the framework to research factors affecting pricing in patent licensing contracts in the biopharmaceutical industry.Sustainability 2018, 10, x FOR PEER REVIEW 5 of 20
Figure 1 .
Figure 1.Factors affecting pricing in patent licensing contracts in the biopharmaceutical industry.
Table 1 .
Summary of variables.
Table 2 .
Statistical characteristics of independent variables.
Table 3 .
Descriptive statistics for technical fee factors (Enter Method).
Table 4 .
Descriptive statistics for technical fee factors (Backward method).
Deal Value = 70.773093+5.91506×PreviousLicensorDeal experience + 19.758378 × CAGR − 1.963919 × Attrition Rate + 0.176483 × Licensor R&D Costs − 0.028014 ×Factors that influence the royalty rate are ranked as follows:1Licensee R&D Costs (−)2Licensee Revenue (+).In the regression model to predict the royalty rate, Royalty Rate is directly proportional to Licensee Revenue and Royalty Rate is inversely proportional to Licensee R&D Costs.
Table 5 .
Results of regression analysis of deal value by drug class (1).
Table 6 .
Results of regression analysis of deal value by drug class (2).
Table 7 .
Results of regression analysis of royalty rate average by drug class (1).
Table 9 .
Results of regression analysis of total upfront payment by drug class (1).
Table 10 .
Results of regression analysis of total upfront payment by drug class (2).
Table 12 .
Results of regression analysis of milestone payment by drug class (2).
Table 13 .
Proportional relationship in the regression analysis.
+: Dependent variables are directly proportional to independent variables, −: Dependent variables are inversely proportional to independent variables.
AR,4RNDLR,5RNDLE,6SALESLR, 7 SALESLE.In most cases technical fees are directly proportional to PDELE, PDELR, RNDLR, and SALESLE while inversely proportional to SALESLR.It is interesting that factors in the formula to estimate the technical fee are totally different by drug class.
Table 15 .
Proportional relationship in the regression analysis by drug class (2).
Table 16 .
Proportional relationship in the regression analysis by drug class (3). | 2018-09-27T17:04:53.538Z | 2018-09-03T00:00:00.000 | {
"year": 2018,
"sha1": "d3dfb5eef21fe9158aabbc398f3f054e81aaaccb",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2071-1050/10/9/3143/pdf?version=1535976899",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "d3dfb5eef21fe9158aabbc398f3f054e81aaaccb",
"s2fieldsofstudy": [
"Business",
"Medicine"
],
"extfieldsofstudy": [
"Economics"
]
} |
229114473 | pes2o/s2orc | v3-fos-license | Regulation Improvement of Right to Education Policy for Street Children in Padang City
Corresponding Author: Roni Ekha Putera, email: roniekhaputera@soc.unand.ac.id, Kampus Unand Limau Manis, Pauh, Padang, Sumatera Barat 25175 Abstract: The fulfillment policy of the right to education for street children has been included in some policies in Padang City. However, it still hasn't been able to solve the problem. Hence, policy improvements are as impressive as research topics. The propose of this research is to describe and analyze the implementation process of education rights fulfillment for street children in Padang city. This research was conducted using qualitative research methods and descriptive design by employing library research and field research. The data were collected through interviews and documentation. The results of this research show that the policy needs improvement because there have been obstacles during the implementation of the policy, which make the policy not function optimally.
A. Introduction
The fulfillment of the right to children's education is one of the indicators of whether a country falls into the category of a developed, or developing country. Education is also interpreted as a part of both public goods and not public goods. In this context, education can be a publicly-owned service item (public), where every community has the same right to get education and teaching as outlined in the mandate of law. 1 The state has guaranteed the fulfillment of children's right to education through a series of the legal constitution that has been previously issued by the government, central government, and regional governments. In the amendment to the 1945 Constitution in 2000 article 28c verse (1) it is explained that everyone has the right to develop themselves through fulfilling their basic needs, right to education, and to benefit from science and technology, arts and culture, to improve their quality of life and for the welfare as human".
Besides, law No. 23 of 2002 concerning Child Protection Article 19 verse (1) affirms that every child has the right to education and teaching in the context of their personal development and level of intelligence according to their interests and talents. 2 Besides, the fulfillment of children's right to education is also strengthened through a ministerial regulation, namely Minister of Women's Empowerment and Child Protection Regulation No. 5 of 2011 concerning Policies on Fulfilling Children's Right to Education. 3 However, it seems that those policies have not been able to guarantee the fulfillment of children's right to education. The increasing dropout rates can be an indicator of the difficulty of fulfilling the right to education. This is evidenced by an increase in dropout rates that occur in many areas, one of which is the City of Padang. According to the Document Center for Education and Culture Data and Statistics In the 2017/2018 Academic Year, Padang City was ranked first in West Sumatra Province with the highest number of __________ dropouts, with a total of 622 students. This phenomenon can also be affected by other factors, as Jamaluddin et al. assert that the lack of knowledge about the significance of education in their life is also one of the reasons for the lack of awareness to attain a decent education. One way to resolve this problem is to approach them personally. 4 The phenomenon of dropping out of school will have an impact on the environment. Tamara's research showed the same thing, that another factor often the one which is the cause of the problems faced by children, is the environmental factor. 5 As written by Gunawan in Muamalah's research which said that the problem of dropping out of school, especially at the lower level of education, then not working and earning a steady income, can be a burden of the community and even often disturbs the peace of the community, the problem of dropping out of school can cause disturbances in society in the form of delinquency which is contrary to positive social norms. 6 One form of public peace can be in the form of the presence of street children. 7 Street children are children who, most of the time, are on the streets or public places with a minimum of four hours a day within one month. 8 These two things are closely related to where children who do not have activities (studying/schooling) will tend to spend their time outside the home or on the streets. In addition to dropping out of school, the current urbanization that is happening in Indonesia also causes an increase in the number of street children that occur every year, so this requires comprehensive treatment. 9 __________ 4 Jamiludin Jamiludin et al., "Street Children's Problem in Getting Education: Economic and Parental Factors," Mediterranean Journal of Social Sciences 9, no. 1 (2018) Street children are the most vulnerable groups in the effort to fulfill the right education. Similarly, IEC and Thompson in Kisirkoi & Mse say that it is difficult, for street children to enroll in learning centers and schools. 10 Moreover, Alam and Wajidi found that they (street children) Complained that they were denied admission in public and private schools. 11 The children protested that they were treated in the society as rejected a lot. Consequently, a specific approach is needed to fulfill their right to education so that it runs well and minimizes the discriminatory actions and the violation of the law.
Deawinadry and Hashim, in their research, mention that the state, local governments, communities, families, and parents are obliged to provide protection and ensure the fulfillment of child rights based on the duties and responsibilities. 12 The problem of education and street children has been a concern of Padang City government for a long time, and this is indicated by the policies issued by the local government, namely the Padang City Regional Regulation No. 2 of 2012 concerning Child Development and Protection. The body of this policy states that the local government is obliged to organize social welfare in the form of educational services for street children? This education policy is made to solve problems faced by the people. To solve the problems needs the support and participation of people. 13 To reinforce the policy of fulfilling the right to education of street children in Padang City, the government issued Padang Mayor Regulation No. 41 of 2017 concerning Procedures for the Coaching of Street Children. However, the existing policies have not been able to solve the problem of street children's education. This is indicated by the continued increase in the number of street children who have not received education in the last few years. Here is a diagram of the increase in the number of street children who do not get an education ( Figure 1). It appears that there is an increase in the percentage of street children who do not get an education. It means the policy of fulfilling the right to education in Padang City has not run optimally. Aulia, in his research, explains that so far the attention given to street children has not been sufficient since the government has not empowered street children. Many children still wander along the streets. 14 Therefore, a study on the implementation of policies on fulfilling the right to education of street children in Padang is needed, so that the findings obtained can be used as a reference for the improvement of regulations on the fulfillment of the right to education for street children in Padang in the future.
B. The Implementation of Right to Education Fulfillment Policies for Street Children in Padang City
Implementation becomes an essential stage in the policy process to achieve policy objectives. Policy implementation can be seen by comparing the policy objectives issued by the government (goals and benefits) to the beneficiaries of the policy, which in this case is the community. This is to say that if the contents of the policy issued can provide excellent benefits for the community, the policy is considered auspicious. Conversely, if the community considers the program issued by the government is not sufficient, the policy is considered a failure. On the other hand, Keban states that policy implementation is a stage where policies that have been adopted are implemented by specific administrative units by mobilizing existing funds and resources. 15 To analyze the implementation of the fulfillment policy of street children's right to education in Padang City, the theory used as a reference is the theory of policy implementation stated by Mazmanian and Sabatier. Mazmanian and Sabatier argue that the crucial role of implementation analysis is to identify the factors which affect the achievement of the objectives throughout this entire process 16 . Mazmanian and Sabatier describe the factors that influence the implementation of the policy into three major parts, namely tractability of the problem being addressed by the statute (the ability to control easily), the ability of the problem to favorably structure the implementation process (the ability of the policy to structure the implementation process) and the net effect of a variety of "political" variables on the balance of support for statutory objectives (variables outside the policy that affect the implementation process) and the stages (or dependent variables) in the implementation process (referred to as independent variables, and the implementation stages are referred to as dependent variables). 17 This research will be analyzed using the theory because the theoretical model fits the characteristics of the fulfillment policy of the street children's right to education, which is a top-down policy. The influence of the four factors on the implementation of the street children's right to education policy fulfillment will be described as follows:
The Feasibility of Controlling a Problem
In implementing the street children's right to education policy fulfillment, the City Government of Padang, together with the Padang City Social Service, the Padang City Education Office, and the Padang City SatPol PP, are the primary activators. Each SKPD involved has its primary duty and function. The technical implementation of this policy is carried out based on each primary function of each SKPD involved and refers to the guidelines for the handling of street children issued by the Padang City Social Service based on the results of the agreement (MoU) in the context of handling street children through an integrated coaching pattern promoted by the Social Service Kota Padang as the leading sector.
The technical implementation of this policy will be carried out first by the PP Padang, Padang Police Unit as the authorized party to screen street children. The street children who have been saved will be handed over to the Padang City Social Service Agency. After the street children are handed over to the Social Service Agency, the Agency will conduct an assessment and data collection on the street children. Then the street children will be validated their feelings or desires whether they have the desire to be able to get their education back or not. If the street child expresses his desire to be re-educated, the Social Service Agency will coordinate with the Padang City Education Office to be further facilitated by the education office. This is following their duties, as stated in the guidelines for handling street children. However, getting to this stage is not easy.
This research finds that the problem of controlling the street children are also influenced by the diversity of the behavior of the target group. This diversity is influenced by several factors, i.e. internal and external factors. Internal factors can be interpreted in their inclination, while external factors can be interpreted as environmental influences and knowledge that make them respond differently to policy. In the implementation of the policy of fulfilling the right to education to street children in Padang, it can be seen that the tendency that occurs among the behavior of these target groups is the refusal of them to be facilitated their education.
Apart from the diversity of the target group's behavior towards this policy, documenting the data well also has an impact on whether or not this problem is easy to control. This relates to the percentage of the target group to the population, where it will be a difficult problem to control when the implementor does not know and cannot map which target group and which population. Based on the findings, it is known that the Padang City Social Service Agency itself does not have well-documented data, so they also do not know how to map the target groups for this policy, Besides, the behavioral change targeted by the policy will also affect the feasibility of controlling the problem. In the context of implementing the policy of fulfilling the right to education to street children, the expected behavior change is related to their awareness to be aware of their education. But as stated by Mazmanian and Sabatier in Subarsono that a program that is aimed at providing knowledge or is cognitive will be relatively easier to implement compared with programs aimed at changing people's behavior 18 . in term of the policy of fulfilling the right to education to street children, it can be concluded that the implementation of the policy of fulfilling the right to education to street children is a difficult problem to control.
The Capability of a Policy to Structure the Implementation Process
Every policy must be able to structure the implementation process, which can be seen through several points that are considered essential and able to explain the purpose of this variable, namely by setting out the formal goals that will be achieved clearly (clarity of policy content), the adequacy of the causal theory, accurate allocation of finance or financial resources for policies, the integrated hierarchy within and between implementing agencies, clarity, and consistency of rules that exist in implementing agencies, the level of commitment of the apparatus to policy objectives, and by providing access or opportunities to participate for external or private parties or non-governmental organizations in the implementation process.
Regarding the ability of this policy in describing the clarity of the contents of the policy objectives, it is found that the clarity of the contents of the policy of fulfilling the right to education to street children in Padang is still not optimal. This is because the policy of fulfilling the right is not a policy promulgated with a separate nomenclature, namely 'the policy of fulfilling the right to education to street children,' but contained in the body of other policies and with other nomenclature. Based on the results of the interviews with informants, it is known that several implementors do not understand this policy and it influences the implementation of the policy of fulfilling the right to education for street children. In general, the policy of fulfilling the right to education of street children in Padang City has had a sufficiently causal theory. It can be seen from Padang Mayor Regulation No. 41 of 2017 concerning the Procedures for the Coaching of Street Children. However, when viewed further, the regulations which are owned by Padang City dealing with the fulfillment of the right to education for street children are sill minimal, because they lack clarity of the fulfillment policy.
From the budget side, the Padang City government has recorded a reasonably good budget. The fulfillment policy is budgeted in Padang City Regional Budget in the field of education. According to Padang City Social Service Agency's Work Plan of 2018, the budgeted funds increased in 2018 to Two hundred thirty billion Rupiah, which previously amounted to 209 Billion Rupiah in 2017. This fund will be allocated for all types of educational needs, including for education funds that can be accessed and enjoyed by street children, namely equality/package of education. Padang City Education Office is one of the SKPDs that receives funding from the Padang City Regional Budget. The equality of education program is also budgeted based on the Padang City Education Office Work Plan of 2016 until 2018. Nevertheless, the budget allocated by the education office is still insufficient; this is evidenced by the finding that the amount of the budget in the implementation of equality of education during the last three years tends to be less than the number of funds budgeted in the budget plan. Furthermore, PKBM itself, as an operational institution that runs equality of education program, also feels the impact and is experiencing financial difficulties due to the uncertainty of assistance to be channeled through the education office.
Meanwhile, Padang City Social Service Agency itself does not have a budget specifically for the fulfillment of the rights to education of street children. However, the budget for this activity is combined with several other activities related to street children, the budget of the overall street children handling program. To conclude, in terms of the allocation of finance/financial resources, it is not running optimally.
Concerning the ability of policies to structure the implementation process, a policy should be able to explain and describe the integration of hierarchies within and between the implementation agencies. Hierarchical cohesiveness can be seen through the quality of coordination carried out by the implementing agencies. Regarding the implementation of the policy of fulfilling the right to education of street children in Padang City, it can be concluded that the integrated hierarchy of coordination that exists is still not optimal. In general, this is caused by the weakness of the policy in outlining and explaining the contents of the policy, so that it will have an impact not only on one aspect. Like the integration of this hierarchy, weak coordination is caused by unclear rules, Another factor also influencing the policy of the implementation process to which extent the opportunity of participation is opened to private and nongovernmental organizations. It appears that the Padang City government has opened the opportunity for private sectors and non-governmental organizations. This is evidenced by the involvement of PKBM (Community of Learning Centers). However, access to participation from the PKBM as an outsider does not have legal certainty. So, PKBM can reject street children who want to attend education arbitrarily. In this case, PKBM cannot be blamed ultimately.
Policy Environment
In implementing a policy, the environment outside of the policy itself also has an effect that is no less influential on the achievement of the goals set by the policy. The environment referred to in this variable is related to the socioeconomic and technological conditions, public support, attitudes of the voting group, support from superior officials, as well as the level of commitment and skills of the apparatus and implementor.
Referring to socio-economic and technological conditions, in terms of implementing the policy of fulfilling the right to education of street children in Padang City, it is found that street children tend to reject this policy due to weak economic conditions. They are forced to remain on the road than following education fulfillment programs. Meanwhile, when viewed from social conditions, it is known that social problems related to street children have not become a concern of the government itself. This is reflected in the actions taken by Padang City Social Service and its Education Office.
Based on the results of the interview conducted with several informants at the social and education offices, it shows that there are still many problems that are more urgent than the problems of street children's education. So that the trend that occurs is an attitude that "half-measures" in solving this problem, this is also felt by the public, and it appears that public support in this policy is still somewhat "between there and nothing" because there are still many people who do not care about this problem, including the voter group.
Voter groups are defined as groups that can provide intervention to decisions made by the implementing body. In this study, Padang City DPRD was made part of the voter group because it was relevant to the meaning of the voter group itself. Based on the finding, Padang City DPRD, as the party that has the authority to make draft policies or regional regulations, does not yet have an intention to formulate more specific policies related to fulfilling the right to education of street children. It can be said that the role of the voting group has not been able to have a significant impact on the successful implementation of the policy of fulfilling the right to education of street children in Padang City.
The implementation of the policy of fulfilling the right to education of street children also needs the support of officials. The size of the support can have a significant impact on the policy implementation process. This form of support can be done in several ways. In Padang City, the support has been provided support materially, in the form of a budget and non-materially, in the form of motivation. However, the support of the supervisor of the Padang City Education Office is considered not optimal, specifically on material support. This is evidenced by the large amount of the budget which was budgeted by his subordinates for the implementation of equality/package of education is not necessarily approved as the amount that has been designed. What happens is that the approved funds are always below the targeted budget prepared by his subordinates.
The things that have been explained before will ultimately lead to the commitment of the policy implementers. When viewed and considered further, every point of the problem in this policy environment variable, which starts from the socio-economic and technological conditions, public support, the attitude of the voting group, to the support of superior officials will be closely related to the commitment of the policy implementers. However, it cannot be denied that other things become obstacles in the process of implementing the policy of fulfilling the right to education of street children in Padang City, as explained in the previous sub-sections.
Steps in the Implementation Process
In this variable, there are some stages, which can be called an endpoint or a dependent variable. Each of these stages is also an input for success in the other stages. The stages of the implementation process in question are the policy output from the implementing agency, the target group's compliance with the policy output, the real impact of the policy, the perceived impact of the policy output, and ultimately lead to the improvement of regulations. Policy outputs from the implementing agency, namely Padang City Social Service Agency, and Padang City Education Office make some efforts so that the fulfillment of the right to education of street children can be carried out properly. Based on interviews with some informants, it can be concluded that the output of the policy of fulfilling the right of education to street children is the availability of street children to pursue education without being discriminated against or made difficult to obtain by cooperating with the education office. Nevertheless, what happens is that the target group's compliance with policy output is not in line with the policy output itself. This is indicated by the refusal of street children as the target group to be facilitated their education.
The rejection occurs due to inherent factors in the target group itself. Besides whether or not there is a sanction in policy will also have an impact on the target group's compliance with the policy. In the policy of fulfilling the right to education of street children in Padang City, there are no sanctions for violations committed by the target groups if they are not cooperative with the implementation apparatus. Such things like this will impact the alignment of the real impact of a policy with the impact of perceived policies.
After all the stages of the implementation process have been studied more deeply, then the end of this long process is to improve the regulations, as stated by Mazmanian and Sabatier. In implementing the policy of fulfilling the right to education of street children, it can be seen that several aspects need to be addressed in the future.
Several informants suggest the improvement may be done by enacting the policy of fulfilling the right to education of street children to become a unified product of a stand-alone policy, not a policy that is within the body of another policy. This step needs to be taken due to the lack of understanding of the implementors of this policy. By formatting it into a stand-alone policy, it is expected to be able to provide explicit content and consistency of strong objectives for the implementor and the target group.
C. Conclusion
In general, the implementation of the policy to fulfill the right to education of street children in Padang City had not yet run optimally. This is indicated by several obstacles that were discovered when this policy was implemented-the technical difficulties faced by the implementors such as the diversity of the behavior of the target group that was difficult to capture and control, unorganized and inadequate data, and the difficulty in changing the behavior of the target group.
Besides, the clarity of the contents of the policy which was still blurred because this policy was in the body of other policies so that it made it difficult for the implementors to understand the contents of the policy itself. The lack of understanding of the implementors in translating the policy also had an impact on the lack of commitment and coordination between institutions, which should have been done much more optimally. Therefore, the policy to fulfill the right to education of street children requires the improvement of regulations so that detailed policies are created so that they do not cause multiple interpretations.
The improvement of the regulation may be in the form of a whole separate regulation with the nomenclature of the policy 'fulfillment of the right to education of street children,' with the hope that when this policy has become independent, it can become a more detailed policy and easily understood by all circles, both from the perspective of the implementors, target groups, and ordinary people. In addition to drafting and issuing separate regulations, improvement steps can be in the form of the issuance of Regional Action Plans related to the fulfillment of the street children's right to education, so that everything related to the implementation of this policy can run optimally. [s] | 2020-11-19T09:15:52.660Z | 2020-04-21T00:00:00.000 | {
"year": 2020,
"sha1": "0f2e9c6a0963fef69bdac4948236a21101c94b05",
"oa_license": "CCBYSA",
"oa_url": "https://journal.walisongo.ac.id/index.php/sawwa/article/download/4822/2772",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "0246df18cbadec2e54d09dfba0363b2c8f9d922c",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Sociology"
]
} |
245859904 | pes2o/s2orc | v3-fos-license | A simple predictor of interface orientation of mesogenic (cid:29)uids and its implications for organic semiconductors
From classical molecular dynamics simulations, we identify a simple and general predictor of molecular orientation at solid and vapour interfaces of isotropic (cid:29)uids of anisotropic particles based on their shape and interaction anisotropy. For a wide variety of inter-particle interactions, temperatures, and substrate types within the range of typical organic semiconductors and their processing conditions, we (cid:28)nd remarkable universal scaling of the orientation at the interface with the free energy calculated from pair interactions between close-packed nearest neighbours and an empirically derived universal relationship between the entropy and the shape anisotropy and bulk volume fraction of the (cid:29)uid particles. The face-on orientation of (cid:29)uid particles at the solid interface is generally predicted to be the equilibrium structure, although the alignment can be controlled by tuning the particle shape and substrate type, while changing the strength of (cid:29)uid(cid:21)(cid:29)uid interactions is likely to play a less e(cid:27)ective role. At the vapour interface, only the side-on structure is predicted, and conditions for which the face-on structure may be preferred, such as low temperature, low interaction anisotropy, or low shape anisotropy, are likely to result in little orientation preference (due to the low anisotropy) or be associated with a phase transition to an anisotropic bulk phase for systems with interactions in the range of typical organic semiconductors. Based on these results, we propose a set of guidelines for the rational design and processing of organic semiconductors to achieve a target orientation at a solid or vapour interface.
Introduction
Anisotropic molecules, whose shape deviates significantly from spherical, are hugely important in many technologies. Liquid crystal molecules, for example, form ordered phases that can be controlled by temperature or concentration, and are the basis of a multi-billion dollar display industry, with an extensive history dating back to the late nineteenth century. 1 Additionally, organic molecules with extended π-conjugation, many of which can also be classified as liquid crystals, have been shown to display interesting opto-electronic properties, and devices based on these organic semiconductors (OSCs) are growing in importance as their performance improves. 2,3 As important electronic processes often happen in the vicinity of interfaces, [4][5][6][7] controlling the alignment of molecules at these interfaces is important for optimising the performances of OSC-based devices and liquid-crystal displays (LCDs). Due to the anisotropic shape of these molecules, symmetry breaking at both solid and vapour interfaces often leads to a preferred orientation, 8 even when the bulk phase is isotropic. This interface alignment has broad implications for device performance.
For a uniaxial molecule (in which two of the principal axes are equivalent), alignment can vary between the extremes of the non-equivalent molecular axis aligned parallel (planar anchoring; side-on orientation for an oblate ellipsoid) or perpendicular (homeotropic anchoring; face-on orientation for an oblate ellipsoid) to the interface (Fig. 1). For biaxial molecules (in which all three principal axes are different), three alignment extremes are possible: face-on, end-on, and side-on.
In OSCs in particular, orientation at the solid and vapour interface is important for improving the performance of a variety of devices, with different functionalities requiring different alignments. In organic field-effect transistors (OFETs), for example, charge mobility is greater when molecules are aligned side-on at the solid dielectric interface, as charge transport can occur in the π-stacking direction rather than having to proceed through the often insulating side-chains. [9][10][11][12][13][14] Conversely, in organic photovoltaics (OPVs), the face-on orientation at the electron donoracceptor bulk heterojunction (BHJ) interface is favoured as it reduces charge recombination by allowing efficient charge transport away from the interface, while also ensuring better charge generation by increasing donor-acceptor orbital overlap. [15][16][17][18][19] Organic light-emitting diode (OLED) performance also depends on the orientation of the emitter's transition dipole moment with respect to the substrate, with planar alignment preferred. [20][21][22][23] The design of molecules for these devices should therefore focus not only on the electronic properties of the individual molecule, but also on how the molecules align at important interfaces in the device in order to maximise performance.
Although the importance of controlling interface orientation in OSC devices is generally well understood, and there is no shortage of experimental [16][17][18]24 or computational [25][26][27][28][29][30] examples of preferential alignment of anisotropic molecules at interfaces, there remain few general rules for predicting alignment at either the solid or vapour interface from the chemical structure of OSC molecules. We have recently extensively reviewed the subject in the context of OSCs, 31 and despite a number of factors being implicated in controlling molecular orientation, in many cases an understanding of the general physical principles that dictate this behaviour was lacking, especially at the solid interface.
There is an extensive history of simulations of liquid crystals using simple computational models (see ref 32 for a review), which have been used to provide a basic understanding of the behaviour of these anisotropic molecules. A smaller subset of these studies have examined the interplay of repulsive and attractive interactions in controlling the interfacial orientation. At a free interface (such as that with a vapour phase), the orientation of anisotropic particles has been studied at various levels, ranging from mean field theories, 33 to particle-based molecular dynamics (MD), 34 and Monte-Carlo (MC) 35,36 simulations. It has been found that switching between the face-on and side-on orientation at this interface can be achieved through tuning the anisotropy of the attractive interactions. In the absence of attractive interactions, the long axis of an anisotropic molecule has been predicted to align perpendicular to the interface due to excluded-volume entropic effects. 33 MD simulations of the Gay-Berne (GB) fluid have indicated a dependence of the orientation at the vapour interface of nematic fluids on the ratio of the shape anisotropy to the interaction anisotropy of the particles. 34,36 Fluids that are isotropic in the bulk have also been shown to display preferential alignment at a vapour interface. 35 In contrast, at the solid interface, free volume is maximised in the face-on orientation, for which the long molecular axis is parallel to the interface. 37 Density functional theory calculations of hard platelets, 38,39 and theoretical calculations of hard spherocylinders 40 and spheroplatelets 41 at hard impenetrable walls show a preference for alignment with the short axis of the molecule perpendicular to the wall due to entropic effects and minimisation of the surface tension. The inclusion of attractive interactions between fluid particles has been predicted to introduce a temperature dependence to interface orientation, with systems at high temperatures showing the same alignment as the purely repulsive systems due to the dominance of entropy, and a transition to the opposite alignment at lower temperatures as energetic effects become more significant. 42,43 Many computational studies of specific systems have shown similar dependences on temperature and interaction anisotropy. [44][45][46][47][48][49] In general, the orientation at the solid interface depends on both the strength and anisotropy of the fluid-substrate interactions, 43 and the strength of the fluid-fluid interactions. 50 A number of experimental studies report control of orientation at this interface for specific small molecules, 19,44,46,[51][52][53][54][55] and the strength of the interactions with the substrate has been shown to directly influence the orientation, again for a specific set of interaction parameters. 45 An interesting illustration of the contrasting behaviour of hard particles at solid and vapour interfaces is the case of a penetrable wall, from which a particle's centre-of-mass is repelled but which the rest of the molecule can penetrate to varying degrees. 56,57 As the wall interacts isotropically with the centre-of-mass of the anisotropic particle, a highly penetrable wall is qualitatively similar to a free interface, whereas a impenetrable wall represents a solid interface. MC and density functional theory calculations of these interfaces show a transition between face-on alignment at an impenetrable surface and side-on alignment at a fully penetrable wall, 56,57 consistent with previous theoretical predictions at solid and vapour interfaces, respectively, due to excluded-volume entropic effects. 33,42 Although a number of general rules that appear to be relatively predictive for determining the orientation of anisotropic particles at both solid and vapour interfaces exist, and many specific examples of orientation control can be found in the literature, a more detailed understanding of how molecular shape and interactions control the orientation in general, and the ability to predict orientation based on combined energetic and entropic contributions using a simple analytical expression, is lacking. In this work, we investigate a series of OSC-like systems using equilibrium coarse-grained MD simulations of GB particles between solid and vapour-like interfaces. These simulations allow for systematic tuning of the strength of the interactions, shape of the anisotropic molecule, and its interactions with the substrate, to gain an understanding of how the interplay of these features influences the orientation at both the solid and vapour interfaces. The use of GB particles, which have a simple ellipsoidal shape, means that these results should be generally applicable to a wide range of molecules, whether OSCs or not, that share similar shapes and interactions, rather than specific to a single type or class of chemical structure.
We focus our work here on systems that are isotropic in the bulk in order to isolate the effects of the interfaces, to remove the additional complications associated with alignment of additional layers, and to broadly understand what happens under experimental conditions in which a thin film of anisotropic molecules is annealed from the melt, deposited from solution, or subjected to high operating temperatures. Generalisation to interfaces with bulk fluids that are anisotropic is possible. We also focus on simulations of equilibrium fluid structure. While thin-films in OSC devices are commonly deposited under non-equilibrium conditions, systems will evolve towards equilibrium over time. As device performance depends strongly on interface orientation, a shift away from a desired non-equilibrium morphology towards the thermodynamic minimum may result in a reduction in performance over time. It is therefore desirable for devices to have an as-deposited morphology that is close to equilibrium, as this will confer higher thermal stability on device performance. It is important then to understand the equilibrium behaviour of such molecules in order to develop materials that continue to perform well over the entire device lifetime.
The paper is organised as follows. We first describe the simulation and analysis methods in section 2. We then show that molecular orientation of anisotropic fluid particles with both repulsive and attractive interactions at the solid and vapour interface of an isotropic fluid cannot generally be predicted using a previously proposed metric, and propose a semi-empirical mean-field estimate of the free-energy difference between face-on and side-on orientations as an alternative. We use simulations of purely repulsive particles confined between a solid and vapour interface to obtain quasi-universal relationships for the scaling of the entropic component of the free energy with the shape anisotropy and bulk density of the fluid particles (section 3.1). Combining this estimate of the entropy with an estimate of the energetic component of the free energy from nearest-neighbour pair interactions (section 3.2), we show that this free energy parameter accurately predicts molecular orientation at the solid (section 3.3.1) and vapour (section 3.3.2) interfaces. In section 3.4 we generalise these results into a number of practical design principles for OSCs, with the goal of providing guidelines towards rational design of OSC interfaces.
Methods
The orientation of a fluid of anisotropic particles at both solid (impenetrable) and vapour-like (penetrable) interfaces was studied using classical molecular dynamics (MD) simulations. A range of uniaxial oblate Gay-Berne (GB) ellipsoids, representative of typical small OSC molecules, were studied to elucidate the general effects of shape and interaction anisotropy. These simple models greatly increase computational efficiency compared with all-atom simulations, allowing a wide variety of parameters and conditions to be examined. Simulations were carried out using LAMMPS (version 3 March 2020), 58,59 and visualisation and analysis of simulation trajectories were conducted using OVITO. 60 The GB potential 61 is an anisotropic form of the widely used Lennard-Jones (LJ) pair potential, and captures the short-ranged excluded-volume repulsion and longer ranged van der Waals attraction between uncharged anisotropic molecules. The potential is characterised by parameters σ and ε that define the length and energy scales of the potential, respectively, parameters ν and µ that tune the shape of the potential, and, for uniaxial particles, the short and long principal diameters, σ F and σ S , and relative well depths, ε F and ε S , along the corresponding axes (the subscripts "F" and "S" denote "face" and "side", respectively). The anisotropy of the uniaxial particles is described by the shape anisotropy parameter κ = σ F /σ S and the interaction anisotropy parameter κ = ε S /ε F . A full description of the GB potential and parameters is given in the ESI, section S1.
In all simulations, we have used ν = 1 and µ = 2, as is common for the GB potential, in line with its original parameterisation 61 and previous parameterisations of OSC systems, 62 and and have set ε S = 1 and σ F = 1.03σ . A non-bonded interaction cutoff of 3σ S was used for all systems. In order to match experimentally relevant systems, published GB parameters for biaxial models of several simple OSCs 62 were used as a starting point to define the interactions in the systems studied here. To obtain a simpler uniaxial representation of each molecule, for which the end-end and side-side interactions and dimensions are equivalent, the published biaxial principal-diameter and well-depth parameters for end-end and side-side interactions were averaged. This typically resulted in changes to these parameters of less than Table 1 Shape anisotropy parameter κ and interaction anisotropy parameter κ for uniaxial models of some typical organic semiconductors (OSCs) and benzene based on published parameters for biaxial models. 62 The sideside parameters used to calculate the anisotropies in each uniaxial model were obtained as the average of the sideside and endend parameters in the biaxial model, as described in the main text. The resulting shape and interaction anisotropies of several small organic molecules are given in Table 1 and inform the values of κ and κ explored. We have used LJ reduced units throughout this work, with energies, lengths, temperatures, pressures, and times given in units of ε, σ , ε/k B , ε/σ 3 , and τ ≡ m 0 σ 2 /ε, where m 0 is the unit of mass and k B is the Boltzmann constant. We note that setting ε = 1.8 kcal/mol and σ = 3.2 Å gives behaviour consistent with the experimental behaviour of perylene, i.e. a density of 1.1 g/cm 3 at 560 K and 1 atm for a system with κ and κ approximately those of perylene in Table 1, 62 and so the parameter range studied should be representative of the behaviour of a variety of OSCs given this choice of ε and σ .
To mimic the scaling of molecular mass with molecular size typical of OSCs, the particle mass m was scaled in proportion to the particle volume v f = πσ F σ 2 S /6 (calculated as the volume for an ellipsoid with principal diameters σ F , σ S , and σ S ) to give a constant mass density within the particle. Although approximate, this behaviour is reasonably representative of real systems, for which the mass density within the particle generally varies between 1.4 and 1.7 g/mol/Å 3 when the particle volume is calculated from published biaxial GB parameters 62 (ESI Table S3). To facilitate comparison between systems with different sized particles, we have analysed results in terms of the volume fraction, φ , rather than the number density, ρ b , in the bulk fluid. The two are related as To encompass the region of parameter space in Table 1 κ > κ, as interfacial behaviour has previously been predicted to exhibit a transition around κ/κ = 1 in some circumstances. 36 Several systems with κ = 1.2 were also studied to examine the behaviour when side-side interactions are stronger than faceface ones. While side-side interactions that are stronger than the face-face ones are not typical of OSCs-like molecules, they can potentially be accessed through, for example, functionalisation of the aromatic core and may present an interesting means of tuning interfacial orientation. The systems with κ > 1 in this work had weaker face-face interactions than the systems with κ < 1 in order to maintain an isotropic bulk fluid as the side-side interactions were increased, meaning face-face interactions were weaker than would be the case for a typical OSC. As the alignment at both solid and vapour interfaces is expected to depend on temperature, due to competition between excluded-volume entropic effects and attractive intermolecular interactions, 42 several temperatures between 0.56 and 0.84ε/k B were studied, corresponding to a temperature range of 507-761 K for the representative OSC systems described above, encompassing a range of temperatures where the OSCs-type molecules that are represented by these GB parameters are liquid. A full list of the specific systems studied is given in the ESI, Table S6.
Simulations were conducted with the fluid confined in the z direction between two fixed interfaces parallel to the (x, y)-plane, as illustrated in Fig. 2a: a particle-based solid surface at the bottom and a perfectly flat wall at the top that interacts with the center of each fluid particle with a repulsive harmonic potential, where z w is the wall position, ε w describes the strength of the wall-particle interaction, and z is the vertical position of the centre of the fluid particle. This repulsive wall was implemented at the top surface of the fluid to maintain an approximately constant average fluid density between systems with particles of the same shape and to prevent evaporation. Similar to a vapour interface, this wall constrained only the vertical position of the centre of a fluid particle and placed no constraints on the position of the particle surface defined by the repulsive part of the GB potential.
We will refer to this vapour-like interface as the "vapour interface" in what follows, and note that it is comparable to the penetrable walls used in previous studies of interface ordering of anisotropic particles. 56,57 In contrast, no part of a fluid particle could penetrate the solid surface.
The solid substrate was modelled as a single layer of either atomistic graphene or silicon, positioned with atoms centred at z = 0. These substrates have been shown experimentally to interact strongly (graphene) or weakly (silicon) with OSCs and to give rise to face-on and side-on orientations, respectively, of the common OSC polymer poly(3-hexylthiophene) (P3HT). 51 Parameters (non-bonded interactions and bond length) for graphene were taken from the OPLS-AA force field 63 for aromatic carbons (ε LJ = 0.039ε, σ LJ = 1.11σ , bond length = 0.44σ , for ε = 1.8 kcal/mol and σ = 3.2 Å) and atoms positioned in the hexagonal lattice of graphene with separation taken from the OPLS-AA force field as the bond length for aromatic carbons. 63 OPLS parameters were also used for silicon 64 (ε LJ = 0.056ε, σ LJ = 1.70σ ) and atoms positioned to give the fcc(001) plane of silicon's diamond cubic lattice (lattice spacing of 1.69σ ). 65 Although ε LJ is larger for silicon, the larger lattice spacing means that the overall attraction of a single GB fluid particle to the substrate is weaker for silicon than graphene. An additional substrate with the same structure as graphene but with interactions that were twice as strong (ε LJ = 0.078ε, σ LJ = 1.11σ ) was also examined, which we will call the "strong" substrate. The structures of these substrates are shown in the ESI, Fig. S1. The mixing rules for combining GB parameters of two dissimilar particles, 62 defined by eqns (S3)-(S8) of the ESI, were used to model the fluid-substrate interactions so 2 (a) Example of the system setup, with GayBerne (GB) uid particles conned between two surfaces: the impenetrable solid surface consisted of a single layer of Lennard-Jones (LJ) particles, centered at z = 0 (black dash-dot line), while a repulsive harmonic potential acted on the centre of each uid particle at the penetrable vapour-like interface (dashed black line) to prevent evaporation. The plot at the top right is representative of the form of this potential. The bulk region was dened as being within ±3σ S of the z coordinate of the centre of the box (z mid , black solid line). z max is the maximum z coordinate of any particle during the simulation, σ sub the diameter of a substrate particle, and dotted black lines indicate the bounds of the bulk region. This region was suciently far from both interfaces that, except for a couple of the systems that formed anisotropic phases, the uid had uniform density here. The position of the vapour interface, z v (green line), was dened as the z-coordinate where the average density was equal to half the average bulk value. The average molecular orientation at the vapour interface, s vap , was calculated as the average of s(z) from eqn (2) over particles whose centres were at z coordinates within z v ± σ S /2 (region indicated in gure). The average molecular orientation at the solid interface, s sub , was calculated by averaging s(z) over uid particles whose centres were within σ S of the surface of the substrate (i.e. at z < σ sub /2 + σ S , cyan line). These denitions selected, to a good approximation, the particles within the rst uid layer at either interface. Particles are coloured by the orientational order parameter s(z) (which is −0.5 and +1, respectively, for fully side-on and face-on orientations), calculated for layers of width z = σ S for a snapshot of the sample system (κ = 0.3, κ = 0.3, T * ≡ k B T /ε = 0.73, substrate = graphene). (b) Fluid density ρ(z) and (c) orientational order parameter s(z) proles calculated over the entire simulation of this system. The shaded areas correspond to the interfacial uid regions at the solid or vapour interface. Dashed and dotted horizontal lines in (c) correspond to the values of s sub and s vap respectively. Additional ρ(z) and φ (z) proles can be found in the ESI, section S5. as to maintain the correct shape dependence of the interactions. The interaction range σ in eqn (S2) of the ESI was replaced by the arithmetic mean, (σ + σ LJ )/2, of the fluid and substrate interaction ranges for the fluid-substrate interactions. The substrate parameters, and the influence of the type of substrate on the orientation at both interfaces are given in full in the ESI (Table S4, Fig. S2).
To give comparable steepness for the repulsive potential at the vapour interface to the interaction with the solid substrate, the value of ε w was determined by fitting eqn (1) to the repulsive part of the potential energy versus z of a single fluid particle representative of perylene (κ = 0.36, κ = 0.19) interacting with the solid substrate in the face-on orientation for the region where the potential energy was < 2k B T at T = 0.62ε/k B . This gave a value of ε w = 17.8ε for the graphene substrate and ε w = 8.9ε for the silicon substrate. The same values for each substrate were used for both the attractive and repulsive surfaces, and the graphene value also used for the "strong" substrate, as the strength of the wall potential did not significantly affect the behaviour at the vapour interface (ESI Fig. S2).
All simulations were conducted at constant volume and temperature with the positions of the substrate particles fixed. Each system consisted of 6000 fluid particles and 4680 (graphene) or 800 (silicon) substrate particles. The simulation timestep was 0.012 τ, which corresponds to 5 fs if the unit of mass m 0 is taken to be the molecular mass of perylene. Temperature was controlled with a Nosé-Hoover thermostat 66,67 and set to values of 0.56, 0.62, 0.73, or 0.84ε/k B . Most systems were constrained to a constant overall volume fraction φ av (that is, the volume fraction calculated from the number density of all particles between z = σ sub /2, the surface of the substrate, and z = z w , the position of the harmonic wall) of 0.39, corresponding to a real density on the order of 1.1 g/cm 3 if ε = 1.8 kcal/mol and σ = 3.2 Å. The z coordinate of the harmonic wall, chosen to achieve the desired overall volume fraction for the different values of κ, are given in the ESI Table S1. Additional systems were examined at lower (0.28) and higher (0.49) φ av values to determine the dependence (if any) of interfacial orientation on the system density. The pressure in each system was measured as the normal force per unit area acting on the solid substrate. The average pressure at equilibrium was generally < 200ε/σ 3 , but was up to 1200ε/σ 3 for φ av = 0.49.
To initialise the simulations, fluid particles were packed with random positions and orientations into a large box (periodic in x, y, and z, with x and y dimensions 10× the target values and z dimension equal to the target value). For systems to be simulated with the graphene substrate, the target x and y dimensions were 34.48σ and 34.125σ , respectively, to enable periodic packing of the substrate particles; for the silicon substrate, the final dimensions were x = y = 33.94σ for the same reason. The size in the z dimension was used to control the volume fraction of fluid particles and so depended on fluid particle size; the values for various systems can be found in the ESI Table S1. A soft potential U soft (r) = A 1 + cos πr r c for r < r c , where r c is is a cutoff distance and A an energy pre-factor that varies the "hardness" of the potential, was applied to remove particle overlaps (r c = 5σ , A increased linearly from 0 to 16.7ε over 20,000 timesteps). Interactions were changed from the soft potential to the GB potential, and the box dimensions shrunk linearly over 50,000 timesteps at a higher temperature of T = 0.88ε/k B to give the final dimensions outlined above. To initialise the system in an isotropic phase, the fluid particles were then allowed to equilibrate in the bulk system at T = 0.88ε/k B for another 50,000 time steps. The periodic boundary conditions in the z direction were then removed, and the box extended in the z direction to give a region of vacuum above the bulk fluid. The solid substrate was introduced below the fluid, and the harmonic wall at the vapour interface was positioned at the z coordinate required to give the desired overall volume fraction (ESI Table S1). The energy of the system was then minimised to remove any overlaps between fluid and interfaces. The system was simulated for a further 510,000 timesteps at the specified temperature, with the equilibration and correlation time for the orientation of the top and bottom fluid layers determined using pymbar's timeseries module, 68,69 which estimates the equilibration time as the time that maximises the number of uncorrelated samples (see ref. 70 for details). Only the data after the equilibration time was used for further analysis. Although a number of systems were simulated that were anisotropic in the bulk (typically for the most anisotropic shapes and interactions; see ESI Table S6 for a list of systems that formed anisotropic bulk phases), the slow dynamics in these systems did not allow equilibrium properties to be measured reliably over the time scales that were simulated. Results are therefore only reported in the main paper for systems with an isotropic bulk fluid phase.
Purely repulsive particle simulations
As their interactions have, to a good approximation, no energetic component, purely repulsive particles were used to determine the contribution of entropy to molecular alignment at the solid and vapour interfaces. For these purely repulsive particles, the shape parameters described above (with κ = 0.30, 0.35, 0.40, 0.45, 0.50) were used, and κ was fixed at 0.2 (ε S = 1, ε F = 5). The GB potential was cut-off and shifted to 0 at the minimum for each orientation to obtain a purely repulsive potential, 71 analogous to the Weeks-Chandler-Andersen (WCA) truncation of the LJ potential. 72 Likewise, a cut-off and shifted form of the fluid-substrate interactions, using the mixing rules for combining GB parameters of two dissimilar particles 62 discussed previously, was used for the interactions with the solid substrate. These simulations were conducted at T = 0.62ε/k B at overall volume fractions between 0.1 and 0.5 using either the silicon or graphene substrate. Although these systems are expected to be athermal, simulations were also carried out for the silicon substrate at T = 0.8ε/k B to verify this fact. A full list of the systems studied is given in ESI Table S5. Data for the different substrates and temperatures was combined to determine the scaling of interface orientation with system parameters (see ESI, section S4).
Results and discussion
We have examined the orientation of anisotropic particles with both attractive and repulsive interactions at solid and vapour in-terfaces. The average molecular orientation as a function of the coordinate, z, perpendicular to the interface was quantified by the orientational order parameter where θ is the angle between the short (unique) particle axis and the z axis, and the average · · · z is over all fluid particles whose centres were in the histogram bin centred at position z. A value of 1 indicates fully face-on alignment, and − 1 2 indicates fully sideon. For isotropic orientations, s(z) = 0. Throughout this work, the terms face-on and side-on will be used to describe any orientation that is either predominantly face-on (s(z) > 0) or side-on (s(z) < 0), not just the fully aligned extremes. Representative density and orientation profiles are shown in Fig. 2, and selected additional profiles are given in the ESI, section S5.
For a bulk nematic GB fluid, the orientation at the vapour interface has previously been shown to scale with the ratio of the shape and interaction anisotropy parameters, κ/κ , with a faceon orientation observed for κ/κ < 1, and an side-on orientation otherwise. 36 For the systems studied in this work, in which the bulk phase is isotropic rather than nematic and in which solid interfaces are also considered, Fig. 3 shows that κ/κ is not a good predictor of orientation at either the solid or vapour interface. At the vapour interface, the orientation does not appear to depend on κ/κ for κ/κ > 1, although there seems to be a transition to very marginally face-on orientations at κ/κ < 1, as predicted previously for bulk nematic fluids. 36 The systems that display a slightly face-on orientation are those with the most isotropic interactions studied here (κ = 0.7), resulting in very slight orientation preference, while those with the more face-on orientation are where side-side interactions are stronger than face-face (κ = 1.2). Stronger alignment could be achieved by decreasing κ (increasing interaction anisotropy), but the corresponding increase in the shape anisotropy κ required to maintain κ/κ < 1, coupled with the more anisotropic interactions, would likely result in systems that are anisotropic in the bulk. At the solid interface, there is a weak trend towards a more face-on orientation as κ/κ decreases within each set of substrate/temperature conditions, but there is a strong dependence on temperature and κ/κ = 1 does not correspond to the transition from side-on to face-on orientation. This dependence on temperature points to a significant entropic contribution, which is only accounted for implicitly in the κ/κ parameter through the shape anisotropy κ.
Thus, we have considered an alternative metric for predicting the orientation at both solid and vapour interfaces based on an estimate of the free-energy difference between the face-on and sideon orientations, which explicitly accounts for the roles of temperature and fluid-substrate interactions. The overall free energy includes an entropic term, ∆S, described in section 3.1 and calculated based on the shape anisotropy, κ, and bulk volume fraction, φ , of the system, and an energetic term, ∆Ū, described in section 3.2 and based on the interactions between nearest-neighbour particles. The free energy difference per interface particle be- Interface orientational order parameter, s, versus shapeanisotropy:interaction-anisotropy ratio, κ/κ , at the (a) solid and (b) vapour interfaces for graphene-structure (graphene or "strong"; lled symbols) and silicon (unlled symbols) substrates and various reduced temperatures, T * ≡ k B T /ε. tween face-on and side-on orientations is calculated as where overbars indicate per-particle quantities. A predominantly face-on orientation is predicted for ∆F < 0, and side-on for ∆F > 0.
Entropic component
Although a number of theories exist to quantify the entropy difference between face-on and side-on interface configurations, 33,37,40-42 a simple, accurate, analytical expression for the entropy as a function of system parameters is lacking, and a nematic bulk phase is often assumed. 37 Instead, we have taken a semi-empirical approach, in which the entropy difference is measured in a comparatively small number of MD simulations of purely repulsive particles, to obtain a universal scaling relationship at the solid and vapour interfaces. Due to the absence of attractive interactions, any alignment of the purely repulsive particles at the solid or vapour interfaces can, to a good approximation, be attributed entirely to entropy. The Helmholtz free energy difference between two states of a system can be calculated as ∆F = ∆U − T ∆S, which in the absence of an energetic contribution (∆U = 0) gives ∆F = −T ∆S. The entropy difference per interface particle between states of a system in the NVT ensemble with exclusively face-on and side-on orientations at an interface can then be estimated as where P F and P S are the probabilities of finding a fluid particle in the solid or vapour interfacial region (defined in Fig. 2) in the perfectly face-on (cos(θ ) = 1) and side-on (cos(θ ) = 0) orientations, respectively. Examples of the distribution of cos(θ ) in the interfacial regions are given in the ESI, Fig S6. The orientation at the solid interface was found always to be predominantly face-on, while a predominantly side-on orientation was exclusively found at the vapour interface (Fig. 4). In both cases, the degree of alignment was found to depend systematically on the volume fraction of the bulk fluid and the shape anisotropy of the fluid particles, with higher densities and shape anisotropies giving a greater degree of alignment. It has previously been predicted that a variety of hard anisotropic particles align with their long axis parallel to a solid substrate, [37][38][39][40][41] as this alignment maximises the free volume and hence is the entropically favoured orientation. 37,41 The opposite argument can be applied at the vapour interface, where the particles are able to extend over the interface, increasing the free volume when their long axis is perpendicular to the interface. 33 Studies of hard ellipsoids at penetrable walls also show this behaviour. 56,57 Thus, for oblate particles, a face-on orientation at the solid interface and a side-on orientation at the vapour interfaces are predicted to be entropically favoured, which is qualitatively consistent with the results presented here.
The entropy difference between face-on and side-on interface configurations was found to have a quasi-universal power-law Table S5. (The missing points at low κ and high φ formed anisotropic bulk phases and could not be equilibrated on the simulation time scale, and so are not included.) dependence on φ /κ at the solid interface and on φ /κ 1/2 at the vapour interface, where φ is the bulk fluid volume fraction and κ is the shape anisotropy, as shown in Fig. 5, although there is some scatter in the data, particularly for the solid interface. The scaling relationships that best fit the data in Fig. 5 were at the solid interface and at the vapour, where a negative value favours a predominantly side-on orientation and positive favours face-on.
While the scaling at the vapour interface with φ /κ 1/2 is the same for both substrates, the relationship between the entropy difference and φ /κ for the silicon substrate is not as straightforward as that for the graphene substrate (Fig. 5a). Due to the spacing of the substrate particles (ESI Fig. S1), the fluid-substrate potential has more significant lateral corrugations at the silicon interface compared with graphene. This induces a greater dependence of the potential on the in-plane (x, y) coordinates compared the graphene, which is much closer to an "ideal" planar solid surface for which the potential depends only on the z coordinate. Despite the additional variability, the overall scaling behaviour follows the same trend at both substrates, and the fit to all the data was used.
Overall, if entropy is the only contributing factor and the system is isotropic in the bulk region, the orientation preference at the solid interface is expected to be face-on, and side-on at the vapour interface, consistent with previously published theories and calculations. [33][34][35][36][37][38][39][40][41][42]56,57 Increasing the effect of entropy, such as by using higher temperatures, more anisotropic particles, or higher pressures (to increase density), could be used to enhance this alignment if desired at either interface.
Energetic component
The alignment of particles with attractive interactions at the solid interface is complicated by attractive interactions of the fluid with the substrate and within the fluid itself. Qualitatively, the faceon orientation should be energetically favoured when the facesubstrate interactions are strong (although this will also depend on the relative strength of the fluid face-face, side-side, and sidesubstrate interactions), and the side-on orientation should be energetically favoured when the fluid particle's face-face interactions are stronger than both their side-side and face-substrate interactions. Assuming that the fluid is isotropic everywhere except for the layer adjacent to the interface, and that a fully face-on particle interacts with six nearest neighbours through side-side interactions (hcp packing) while a side-on particle has two faceface interactions and two side-side interactions (cubic packing with inequivalent lattice spacing) with its nearest neighbours (as found in simulations in which the interface layer was close to fully aligned, as shown in ESI Fig. S7), the difference in energy per interfacial fluid particle between a completely face-on orientated Entropy dierence per particle between face-on and side-on orientations as a function of φ /κ n at the (a) solid (n = 1) and (b) vapour (n = 1/2) interfaces, where φ is the bulk volume fraction and κ is the shape anisotropy of the uid particles. The data here includes all sets of conditions for graphene-(lled symbols) and silicon-like (unlled symbols) repulsive substrate at T = 0.62 for 0.84ε/k B . The temperature does not signicantly aect the scaling (ESI Fig. S3). Plots comparing the dierences for dierent temperatures and substrate types are given in the ESI, section S4.
interfacial fluid layer and a completely side-on oriented one can be estimated to be whereŪ F (Ū S ) is the interaction energy of a particle in the face-on (side-on) orientation with its nearest neighbours in the interfacial layer and with the substrate. (Note that the nearest-neighbour fluid interactions are halved in these equations because each such interaction contributes to the energy of two interfacial fluid particles.)Ū i j is the interaction energy in the face-on/face-face orientation (for i = F) or side-on/side-side orientation (for i = S) with the substrate (for j = s) or another fluid particle (for j = f).
For fluid-fluid interactions, the interaction energy in either the face-face or side-side orientation was taken as the minimum of the GB pair potential in the specified orientation.Ū Fs andŪ Ss were calculated as the minimum in the interaction between a single GB ellipsoid and the substrate in the face-face and side-side orientations, respectively. As a particle-based substrate was used in this work rather than being perfectly flat, the energy minimum was averaged over 100 random positions on the substrate. Equation (7) also applies for the energy difference per fluid particle at the vapour interface, but with the fluid-substrate interactions set to zero, i.e.Ū Fs =Ū Ss = 0.
Predicting alignment at solid and vapour interfaces
To predict the behaviour at both the solid and vapour interfaces, we have used the free energy described in eqn (3), with the entropic component determined from the bulk volume fraction and shape anisotropy of each system as described by eqns (5) and (6) for the solid and vapour interfaces, respectively, and with the energetic component determined from the nearest-neighbour interactions in the interfacial layer as described by eqn (7).
Solid interface
The average molecular orientation at the solid substrate, measured by the orientation order parameter s sub , shows excellent universal scaling with the free energy difference defined by eqn (3) for the entire range of systems studied, covering temperatures from 0.56 to 0.84ε/k B , bulk volume fractions from 0.29 to 0.50, strongly and weakly interacting substrates, and shape and interaction anisotropies corresponding to molecules as varied as benzene, perylene, and porphine ( Fig. 6a; see ESI Table S6 for a full list of systems). Interestingly, although systems with a predominantly side-on orientation at the solid interface are observed, they are relatively rare, being obtained in less than 15% of cases. These side-on systems generally correspond to situations for which the fluid particles have low shape anisotropy (high κ), the system is at low temperature, and to a lesser extent the fluid particles have high interaction anisotropy (low κ ). The rarity of the side-on orientation points to a dominant entropic contribution in most systems studied. Orientational order parameter s sub at the solid interface as a function of (a) overall, (b) energetic, and (c) entropic components of the interfacial free energy dierence. A least-squares best t to a logistic function, constrained to go to −0.5 and 1 at high and low ∆Ū − T ∆S respectively, is shown in (a). The energetic component is coloured by the interaction anisotropy parameter, the entropy component by the shape anisotropy parameter, and the overall free energy by the ratio of the two parameters. Filled symbols correspond to the graphene-structure substrate (graphene or "strong") and unlled symbols to the silicon substrate. Shaded grey regions highlight where the points should fall if obeying the predicted scaling with the free energy or free energy component.
On separating the free energy into its energetic (Fig. 6b) and entropic (Fig. 6c) components, the reason for the preference for the face-on orientation at this interface becomes clear. Over the range of parameters studied, the energetic component varies from favouring the side-on orientation by values typically in the range of approximately 2ε, to the face-on orientation by a similar amount. Over the same range of parameters, the entropic component of the free energy always favours the face-on orientation by between 1 and 5ε, increasing in magnitude as the shape becomes more anisotropic. The combined result of these effects is an orientation that is only side-on under conditions where the entropic contribution is lowest (low temperature, low shape anisotropy) or the energetic contribution highest (highly anisotropic interactions). Although highly anisotropic interactions shift the energetic term significantly towards favouring the side-on orientation, high temperatures would be required in many cases to maintain an isotropic bulk fluid, which would increase the mangitude of the entropic term in the opposing direction, giving little change to the overall free energy. This separation of the free energy into its components also highlights that neither the energy nor the entropy is sufficient to completely explain the observed interface orientation, again emphasising the importance of both contributions for accurately predicting the orientation.
When examining the effect of the substrate on the interface orientation, a slight dependence is observed, with the observed orientation at the silicon substrate generally being shifted towards side-on relative to systems at the graphene-structured substrate (with either regular or strong interactions) for the same free energy difference (ESI Fig. S2). As described previously in the context of the purely repulsive systems, this slight substrate dependence is likely due to the looser packing of substrate particles in the silicon substrate compared with the graphene substrate. This increases the dependence of the fluid-substrate potential on the in-plane coordinates of the fluid particle at the silicon substrate, which behaves as a slighly penetrable surface, relative to the graphene substrate, which is much closer to an "ideal" smooth impenetrable solid substrate. No dependence on the strength of the fluid-substrate interactions was observed, however, with the results at the graphene and "strong" (same particle arrangement but double the interaction strength) substrates showing the same behaviour. This result indicates that patterning of the substrate may be an effective method to favour a more side-on alignment, as has been previously observed, 73 although further study would be required to fully understand this behaviour.
Although it appears difficult to significantly influence the interfacial orientation by tuning fluid-fluid interaction strength within the realm of reasonable OSC parameters while the bulk fluid remains in an isotropic liquid phase, the energetic component of the free energy difference does have a strong dependence on the strength of the fluid-substrate interactions. Switching from a strongly (e.g. graphite) to weakly (e.g. silica) interacting substrate has been shown experimentally 19,44,46,[51][52][53][54][55] and computationally 45 to influence the orientation at the solid interface for a range of OSCs. Fig. 7 highlights the relative importance of the fluid-substrate interactions: the difference in energy on switching from a strongly to weakly interacting substrate (e.g. graphene Interface free energy dierence as a function of interaction anisotropy parameter κ at T = 0.62ε/k B for graphene (lled symbols) and silicon (unlled symbols) substrates.
to silicon) is comparable to changing the interaction anisotropy from values close to those representative of benzene to perylene (a significant structural change), and is sufficient to switch the free energy from favouring face-on to side-on orientation in some cases.
Although increasing the strength of the substrate-fluid interactions will increase both the side-substrate and face-substrate interaction strength, which make equal and opposite contributions to the energetic component of the interface free energy different in eqn (7), the face-substrate interactions are expected to dominate for most cases studied in this work due to two factors. First, the stronger face-face interaction compared with sideside interaction (κ < 1) typical of OSCs leads to the strength of face-substrate interactions increasing more rapidly than that of side-substrate interactions when the substrate-substrate interaction strength is increased isotropically, given the mixing rules for fluid-substrate interactions. Second, for oblate (κ < 1) particle shapes typical of OSCs (all the systems studied in this work), a face-on particle interacts with more substrate particles than a side-on particle, and so the strength of the interaction of a fluid particle with the whole substrate increases more rapidly for face-on particles than for side-on particles when the substratesubstrate interaction strength is increased isotropically, even if the fluid particles have no interaction anisotropy (κ = 1). Overall, while changing the fluid-fluid or fluid-substrate interactions are both plausible ways of influencing the energy, large structural changes would likely be required chemically to significantly change the fluid-fluid interaction anisotropy, which will likely also change the shape anisotropy and bulk phase behaviour of the fluid. It should generally be much simpler to change the nature of the solid substrate.
The fluid particles studied here are representative of small OSC molecules such as perylene that do not have any side-chains, but many OSCs feature alkyl side-chains designed to enhance their solubility. The different chemical nature of the backbone and side-chain (one being a conjugated π-system, the other an alkyl chain) means that changing the substrate to favour interactions with either one or the other should enhance either the face-on (strong backbone-substrate interactions) or side-on (strong sidechain-substrate interactions) orientations at the substrate interface. Although we do not consider any molecules with sidechains here, the impact of side-chains can be approximated in the coarse-grained modelling framework used in this work by considering their effect on the overall fluid-substrate interactions: a substrate that is more strongly attracted to the side-chains (particle side) will energetically favour the side-on orientation. This behaviour has been observed experimentally, for example though treatment of a substrate with a self-assembled monolayer of octadecyltrichlorosilane (OTS). 74
Vapour interface
In contrast to alignment at the solid interface, which depends on the fluid-fluid and fluid-substrate interactions, the energetic driver for orientation at the vapour interface is simply expected to be related to the relative strength of the interactions for a face-on and side-on fluid particle with other fluid particles in the interfacial layer. Face-face interactions are maximised in the side-on orientation, so increasing the strength of these interactions is expected to promote the side-on orientation. Compared with the solid interface, the vapour interface does not show as clear universal scaling of the molecular orientation with the interface free energy parameter in eqn (3), but a strong trend with the free energy is still observed (Fig. 8a). Where κ < 1, the orientation at this interface converges towards isotropic at the vapour interface as the free energy approaches zero, and becomes more aligned as the free energy magnitude increases. At high values of κ (low interaction anisotropy), the energetic component (Fig. 8b) shifts towards very slightly favouring a face-on orientation, qualitatively consistent with previous predictions of ordering of bulk nematic fluids at the free interface. 36 Several systems were also examined with κ > 1, corresponding to stronger side-side than face-face interactions, as it is expected to promote the face-on orientation at the vapour interface. Even with reduced face-face interaction strength, bringing them outside the range of typical OSCs, most of these systems did not remain isotropic in the bulk. The two that did were the least anisotropic in shape and are the face-on outliers at ∆Ū − T ∆S ≈ 2ε in Fig. 8a. While this result shows that it is possible to achieve a slight face-on orientation at the vapour interface for anisotropic particles while still maintaining an isotropic bulk phase, these points do not fall in the region where the simple free-energy predictor presented here predicts a face-on orientation.
The deviation from universal scaling at the vapour interface is likely due to several factors. Firstly, for a number of the systems, particularly those at high density, the pressure, measured as the normal force on the substrate, was relatively high -up to 1200ε/σ 3 in some cases. Examining the dependence of the relationship between s vap and the free energy on system pressure (ESI, Fig. S9b) shows more extreme side-on orientation (lower values of s vap ) at higher pressures for the same value of the free energy. While the orientation at the solid interface also shows a slight pressure dependence (ESI, Fig. S9a), the effect at the vapour interface is more pronounced. As the vapour interface is constrained by a repulsive harmonic wall, interacting with the A least-squares best t to a logistic function, constrained to go to −0.5 and 0 at high and low ∆Ū − T ∆S respectively is shown in (a). The energetic component is coloured by the interaction anisotropy parameter, the entropy component by the shape anisotropy parameter, and the overall free energy by the ratio of the two parameters. Filled symbols correspond to simulations with the graphenestructure substrate (graphene or "strong") and unlled symbols to the silicon substrate. Shaded grey regions highlight where the points should fall if obeying the predicted scaling with the free energy or free energy component. centers of each fluid particle, high pressures correspond to a strong interaction between the fluid particles and this wall, which is likely not representative of a true vapour interface, where fluctuations are not artificially suppressed. Secondly, as the energetic component of the interfacial free energy in eqn (7) is calculated assuming an idealised packing structure for a fully face-on and side-on interfacial layer, variations in the packing structure at the interface reduce the accuracy of the estimated free energy. The positions of fluid particles are less strongly constrained by the vapour interface than the solid substrate, and the side-on packing that predominates at the vapour interface appears to be less robust to disorder than the face-on packing that predominates at the solid interface, due to the tendency of nearest neighbours to be displaced parallel or angled slightly with respect to one another. Weaker ordering at the vapour interface results in molecular packings that deviate from the idealised structures, as shown in the ESI in Fig. S8. Similarly, the calculation of the energetic component of the free energy based solely on the fully face-on and fully side-on configurations means that the predictability of the model is likely to be poorer for systems with interfacial orientation that is closer to isotropic.
Despite these complications, the results indicate that it is difficult to obtain a face-on orientation under equilibrium conditions of isotropic bulk GB fluids for parameters representative of OSCs. For the systems studied here, the interaction energy favours the face-on orientation when the interaction strength is the most isotropic (κ = 0.7) of the values studied (Fig. 8b), though only marginally, and more significantly when κ is increased to > 1. In all cases, however, the entropic term (Fig. 8c) is sufficiently large that the overall free energy still predicts the side-on orientation. While significantly lowering the temperature would help to transition to the face-on orientation by reducing the entropic contribution, this is likely to result in a transition to a liquid-crystal or crystalline bulk phase. On the other hand, reducing the entropic component by reducing the shape anisotropy is likely to lead to a more isotropic interface orientation.
It has previously been shown that the orientation of similar GB particles at a nematic-vapour interface can be easily switched between face-on for κ/κ < 1 to side-on κ/κ > 1, 36 whereas only a very slight preference for the face-on orientation was observed if κ/κ < 1 for the isotropic-bulk systems studied here (Fig. 3b). The greater propensity for a nematic bulk fluid to give the face-on orientation at the vapour interface compared with an isotropic bulk fluid can be understood qualitatively in terms of differences between the interfacial energy of the face-on relative to the side-on orientation in the two cases. For the isotropic fluid, this energy is described by eqn (7), whereas for the nematic fluid, interactions with the adjacent anisotropic fluid layer must also be taken into account. Assuming that, compared with the isotropic fluid, each fluid particle in the interfacial layer of the nematic fluid interacts with an additional particle of the same orientation in the next fluid layer, the face-on orientation is predicted to be more energetically stabilised relative to the side-on orientation in the nematic fluid than in the isotropic fluid for the systems studied here with κ < 1 and in ref 36, for which the face-face interac-tions were always stronger than the side-side ones (κ < 1).
Guidelines for controlling OSC interface orientation
In all manner of OSC-based devices, orientation at interfaces has been correlated with device performance. [9][10][11][12][13][14][15][16][17][18][20][21][22][23] Reliably controlling the orientation at these interfaces is therefore an important step towards designing high-performing molecules for commercial applications. Based on the results presented in the previous sections, we propose several general guidelines, which are supported by empirical evidence, for how the molecular structure, interface, and processing conditions may be tuned to shift towards the desired alignment at these interfaces.
Increased shaped anisotropy promotes a face-on orientation at the solid interface, and side-on at the vapour. As the shape anisotropy has a significant effect on the entropic component of the free energy described in eqn (3), it provides a simple handle for controlling orientation at both the solid and vapour interfaces. At the solid interface, increasing the shape anisotropy results in a shift towards a more face-on orientation (shifting left on the phase diagram presented in Fig. 6a), while the opposite behaviour is observed at the vapour interface. In both cases, this is due to an increase in the entropic driving force. Consistent with this prediction, a number of experimental studies have shown that backbone planarisation (induced by fluorination of the backbone, 16,75,76 or extending the conjugation of the backbone, such as by adding aromatic components between the backbone and side-chains 17 , both of which can be associated with an increase in shape anisotropy), results in a more face-on orientation at the donor-acceptor interface in OPVs. It should be noted that both of these chemical modifications will also influence the energetics, so an interplay between changes in interaction strength and shape will control the orientation at the interfaces.
Stronger fluid face-face (or weaker side-side) interactions promote side-on orientation at both interfaces. Corresponding to a shift to the right on the plots in Figs. 6a,c and 8a,c, stronger face-face (or weaker side-side) interactions favour the side-on orientation in order to maximise these energetically more favourable interactions. As previously mentioned, this strategy is possibly the least effective approach for tuning the interface orientation as the energetic contribution to the interfacial free energy difference is close to zero for the interaction anisotropies studied here. Additionally, methods to tune the interaction anisotropy, such as introducing longer/bulkier side-chains that may block the face of the molecule or deplanarising the backbone, also change the shape anisotropy, which may have the opposite effect on interface orientation. Despite these challenges, enhancing the hydrogen-bonding capabilities of a small molecule in the side-side direction has recently been shown to promote a more face-on orientation at a solid substrate, 77 as predicted.
Increasing the interaction strength of a solid substrate promotes the face-on orientation at the solid interface. Changing the substrate (eg. from silicon to graphite) to one that interacts more strongly with the semiconductor is predicted to shift the orientation towards face-on, corresponding to a shift toward the top left of Fig. 6a. There are many examples in the literature of this behaviour. 19,[44][45][46][52][53][54][55] Choosing a substrate that interacts favourably with either OSC backbone or side-chains would also be an effective means of tuning the orientation at this interface.
Processing conditions that favour entropy promote face-on orientation at the solid interface, and side-on at the vapour. As with the shape anisotropy, increasing the effect of entropy by processing at high temperature and pressure (density), will shift the preferred orientation towards the entropically favoured one (face-on at the solid interface, side-on at the vapour). When considering high temperatures, care should be taken to ensure that the orientation remains in the entropically favoured orientation as the system is cooled down, which can be achieved by rapid quenching, as the degree of alignment will be a function of temperature. In cases where the equilibrium orientation is face-on at the solid interface (which appear to be many), processes such as melt annealing can allow the equilibrium orientation to be obtained, even if deposited through a non-equilibrium method in the opposite orientation. 51,78 Conditions for which the bulk fluid is anisotropic, or sideside interactions are stronger than face-face ones, can give the face-on orientation at the vapour interface at equilibrium. From our predictions based on the behaviour of a bulk isotropic fluid, the face-on orientation at the vapour interface is extremely difficult to achieve, although it can be obtained by systems with stronger side-side interactions than face-face interactions (e.g. by chemical functionalisation of the aromatic core of the OSC). However, with face-face interactions typical of OSCs, such systems are unlikely to form isotropic bulk phases under typical conditions. The interface orientation of liquid crystal fluids has previously been shown to be much more variable. 36 Under appropriate conditions, many OSCs form liquid-crystal phases, meaning the predictions of ref 36 based on the shape and interaction anisotropy can be applied to tune the orientation at the vapour interface.
Non-equilibrium processing can result in different behaviour than that predicted here. The above guidelines apply under equilibrium conditions, but could be circumvented using nonequilibrium methods to potentially reach the more difficult to obtain orientations (face-on at the vapour interface, side-on at the solid interface). Many common device processing techniques are non-equilibrium in nature. In methods such as vapour deposition, the layer-by-layer deposition traps lower layers close to their asdeposited orientation (see refs 79 and 80 for extensive reviews of this method and how processing temperature can influence the orientation). Alternate methods, such as spin or blade coating, rely on the evaporation of solvent to form thin films, which is also inherently a non-equilibrium process and will not necessarily result in the equilibrium structure. Nevertheless, the equilibrium structure is still important for OSCs, since systems tend toward equilibrium over time. For thermal stability of devices it is therefore beneficial for the equilibrium structure to be the one that maximises device performance.
Conclusions
Through the use of classical simulations of oblate ellipsoidal fluid particles, this work provides a general analysis of how particle shape and interaction anisotropy influence the equilibrium orientation of a wide variety of isotropic liquids at interfaces with both a solid and vapour-like phase. For a range of parameters spanning a subset of organic semiconductor (OSC) space, different substrate types, and different thermodynamic conditions, we find remarkable universal scaling of interface orientation with a simple interfacial free energy parameter based on nearest-neighbour pair interactions and the interfacial entropy of purely repulsive particles of the same shape. Based on these results, we propose several practical methods for achieving the desired orientation at both interfaces, which we hope provide a pathway towards the rational design of high performing OSC-based devices.
At the interface with a solid, the entropically favoured orientation is face-on, consistent with previous theories based on excluded volume entropic effects. Even when attractions are introduced, the orientation is largely influenced by the entropy and can be tuned by modifying the shape anisotropy. The side-on orientation is more difficult to obtain, but can be achieved most simply by using a substrate that interacts weakly with the face of the fluid particle (or strongly with its side).
The entropically favoured orientation at the vapour interface is side-on, in contrast to the alignment at the solid substrate, but again can be rationalised through excluded-volume entropic effects. Importantly, in all cases studied here, the orientation at the vapour interface of a fluid that is isotropic in the bulk is predicted, and generally observed, to be side-on. Conditions for which the face-on orientation is favoured appear to be difficult to achieve for an isotropic bulk liquid phase for interactions typical of OSCs, but can be obtained if the system has a bulk liquid-crystal phase. Tuning of molecular shape and interaction parameters can result in a shift towards a more isotropic interfacial layer.
Many open questions remain in this area, such as how sidechains, biaxiality, degree of polymerisation, multiple fluid components, or non-equilibrium processes such as pre-aggregation of molecules in solution during solvent evaporation, influence the orientation, but we hope that this work provides a useful starting point for understanding some simple methods for how interfacial orientation can be systematically tuned. Additionally, although this work focuses on oblate ellipsoidal particles, representative of molecules such as perylene, similar arguments should be applicable to prolate molecules such as pentacene. Further studies that account for interactions with the surrounding layers may allow this analysis to be extended to the non-isotropic systems that may be more typical of certain OSC molecules.
Conicts of interest
There are no conflicts to declare. sity of Adelaide's Phoenix High Performance Computing Service. BJB acknowledges The University of Adelaide for the Joyner and Constance Fraser scholarships and the Playford Memorial Trust for a PhD scholarship. | 2022-01-12T16:26:56.226Z | 2022-01-10T00:00:00.000 | {
"year": 2022,
"sha1": "ea4c051cade195cc4a2bc74f9d94a6c083d0bb3e",
"oa_license": "CCBY",
"oa_url": "https://chemrxiv.org/engage/api-gateway/chemrxiv/assets/orp/resource/item/61d7cc6590fc8a4808ed2b31/original/a-simple-predictor-of-interface-orientation-of-mesogenic-fluids-and-its-implications-for-organic-semiconductors.pdf",
"oa_status": "GREEN",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "b42efa95f08230b5b032c82111d70850efed3d2d",
"s2fieldsofstudy": [
"Materials Science",
"Physics"
],
"extfieldsofstudy": []
} |
118955848 | pes2o/s2orc | v3-fos-license | A General Method for the Determination of Matrix Coefficients for High Order Optical System Modelling
The non-linear transformations incurred by the rays in an optical system can be suitably described by matrices to any desired order of approximation. In systems composed of uniform refractive index elements, each individual ray refraction or translation has an associated matrix and a succession of transformations correspond to the product of the respective matrices. This paper describes a general method to find the matrix coefficients for translation and surface refraction irrespective of the surface shape or the order of approximation. The choice of coordinates is unusual as the orientation of the ray is characterised by the direction cosines, rather than slopes; this is shown to greatly simplify and generalise coefficient calculation. Two examples are shown in order to demonstrate the power of the method: The first is the determination of seventh order coefficients for spherical surfaces and the second is the determination of third order coefficients for a toroidal surface.
Introduction
An optical system can be effectively modelled in paraxial approximation by a product of 4×4 matrices, each representing one elementary transformation of the light rays [1]; the elementary transformations are either translations of the ray in homogeneous media or the effects of surfaces separating different media. A more accurate approach implies the consideration of higher order terms but the fact that Snell's law makes use of the sine function rules out the terms of even order; as a consequence, when one wants to improve on the paraxial approximation, one has to consider third order terms.
Aberrations have already been studied extensively [2,3] but work is still going on in order to design symbolic models of optical systems that computers can use for optimisation purposes and humans can look at to gain a better understanding of systems' performance.
The matrix theory has been extended to deal with higher order terms [4] through the use of a vector basis that incorporates two position and two orientation coordinates as well as all their third or higher order monomials, increasing the overall dimension which becomes 24 for third order approximation. It is possible to apply axis symmetry to reduce the matrix dimension through the use of complex coordinates and their higher order monomials; for instance, in third order the matrices to be considered for axis-symmetric systems are 8 × 8 [4,5].
The set of four coordinates normally used to describe the ray consists of the two rectangular coordinates, x and y along with the two ray slopes u = dx/dz and v = dy/dz. I chose to replace the ray slopes with the direction cosines relative to the coordinate axes, respectively s and t, in order to allow an easier and more elegant formulation of the Snell's law at a surface but at the expense of rendering the translation transformation non-linear.
This work details a general method that can be used to establish all the coefficients needed for any order modelling of optical systems built with surfaces of unspecified shape. The power of the method is exemplified with determination of matrix coefficients for spherical systems in seventh order and for a toroidal surface in third order.
The choice of coordinates
The optical axis is assumed to lie along the z axis, so the position of any point is determined by the two coordinates x and y; when dealing with axis symmetric systems, the two coordinates can be combined in the complex coordinate X = x + iy. The ray orientation is defined by cosines of the angles with the x and y axes, respectively s and t; again a complex orientation coordinate can be used, S = s + it, in order to simplify the rotational treatment of axis symmetric systems.
Using the same notation of reference [4], x& is the vector whose components are the coordinates and their monomials up to order n: where total order j + k + l + m is n or less and odd, with all exponents greater than zero. By convention the vector elements are placed in the order of smaller n and larger four digit number jklm. Any transformation of x& into x ′ & can be represented by a square matrix S of dimension N equal to the size of x&, following the equation: The matrix S has the following structure: .
The four submatrices that form S have the following meaning: P 4×4 contains the paraxial constants of the optical system, H 4×(N−4) has 4(N − 4) high order coefficients from the series expansion of the transformed coordinates, 0 (N−4)×4 is composed of zeroes and finally E (N−4)×(N−4) is obtained from the elements of the first four lines by a procedure called extension whereby the elements of lines 5 to N from x ′ & are calculated and all terms of orders higher than n are dropped. More specifically, if we wish to determine the coefficients for the line corresponding to the monomial x j y k s l t m , we take the polynomial expressions for x, y, s and t from lines 1 to 4 and raise them to the powers j, k, l and m, respectively, making their product afterwards; the result is a polynomial of degree n × (j + k + l + m) from which only the terms of orders up to n should be considered. The submatrix E (N−4)×(N−4) can itself be subdivided into components of different orders and components with all zero elements. Although the ray transformation is described by an N × N matrix, the above considerations show that only the 4(N − 4) elements from the first four lines need be considered, as the extension procedure is standard for all transformations. In cases where symmetry exists the number of independent coefficients can be greatly reduced [4,6]. It is apparent that the matrix for any given transformation can be considered completely defined when the first four lines' coefficients have been evaluated, i.e. when the transformed coordinates have been expressed in a power series of order n.
The coordinate conventions made above are not the same as those made in references [4,5] and indeed by the majority of authors; they may appear to be a poorer choice because, as we shall see, they lead to a non-linear translation transformation; on the other hand they will simplify the determination of refraction coefficients. It should be noted, however, that a coordinate change can be seen as a transformation, itself governed by a matrix; it is not difficult to follow the procedure outlined below and then change from cosines into slopes, if needed.
The translation matrix
The first elementary transformation that has to be considered is just the translation of the ray in an homogeneous medium; the orientation coordinates don't change but the position coordinates change according to the equations where e is the distance travelled, measured along the optical axis.
The series expansion of the equations is rather straightforward; up to the seventh order it is: The previous equations give directly the coefficients for lines 1 and 2 of the matrix T which describes the translation transformation; lines 3 and 4 are made up of zeroes, except for the diagonal elements which are 1; this translates the two relations s ′ = s and t ′ = t. The lines 5 to N can be obtained by the extension procedure described before.
If slopes were used instead of direction cosines the translation matrix would have a much simpler form, corresponding to a linear transformation [4].
Refraction formulae
The influence of the surface power on the ray coordinates must now be considered. It is useful to analyse separately the changes on the ray orientation and the modification of the position coordinates. The ray orientation changes due to the application of Snell's law when the ray encounters the surface of separation between the two media; let this surface be defined by the general formula: Let a ray impinge on the surface on a point of coordinates x and y with direction cosines s and t. The normal to the surface at the incidence point is the vector: the direction of the incident ray is the unit vector: and the refracted ray has a direction which is represented by the vector: where s ′ and t ′ are the direction cosines after refraction. One can apply Snell's law by equating the cross products of the ray directions with the normal on both sides of the surface multiplied by the respective refractive indices: where ν represents the refractive index ratio of the two media.
The algebraic solutions of the equation above can be found for many surfaces by means of suitable symbolic processing software like Mathematica; in certain cases the solution is easier to find after a suitable change of coordinates along the z axis, which will not affect the final result; for instance, for a spherical surface it is convenient to shift the coordinate origin to the center. In order to find the coefficients for the matrix R 1 , representing the change in the ray orientation, the explicit expressions for s ′ and t ′ must be expanded in series; this can again be programmed into Mathematica. The examples given later show this procedure for two different surfaces.
Surface offset
The use of just two position coordinates implies that these are referenced to a plane normal to the axis. When a ray is refracted at a surface, the position coordinates that apply are those of the incidence point; nevertheless the translation to the surface is referenced to the vertex plane and thus the position coordinates that come out of the translation transformation, are those of the intersection with that plane; the difference between the two positions is named surface offset. Figure 1: The ray intersects the surface at a point X 1 which is different both from the point of intersection of the incident ray with the plane of the vertex, X, and the point of intersection of the refracted ray with the same plane, X 2 . The surface is responsible for three successive transformations: 1 -an offset from X to X 1 , 2 -the refraction and 3 -the offset from X 1 to X 2 .
The figure 1 illustrates the problem that must be solved: The reference plane for the position coordinates of the ray in a refraction process is the plane of the surface vertex; the coordinates of the point of intersection of the ray with this plane are not the same before and after the refraction. The surface matrix must account not only for the orientation changes but also for the position offset introduced by the refraction process.
If the origin of the rectangular coordinates is located at the vertex and (x, y, 0) are the coordinates of the ray when it crosses the plane of the vertex, the current coordinates of a point on the incident ray are given by the two equations: As the intersection point has to verify both (11) and (6) it is possible to solve a system of three simultaneous equations to determine its coordinates. In fact there may be several intersection points, as the ray may intercept a complex surface in several points; it is not difficult, however, to select the solution of interest by careful examination of the surface region that has been hit; we are usually interested in the solution that has the smallest |z ′ | value.
As before a series expansion of the exact expressions of x ′ and y ′ must be performed taking only the terms up to the nth order; this will yield the coefficients for the matrix T 1 which describes the transformation from point X to point X 1 and will be called the forward offset.
The offset from point X 1 to point X 2 , designated backward offset, is the reverse transformation of the forward offset, for a ray whose direction coordinates have been modified by refraction. We solve (11) and (6) in terms of x and y and apply the previous procedure to determine the coefficients of the transformation matrix, T 2 .
The matrix describing the transformations imposed by the surface is the product T 2 R 1 T 1 , as the ray has to suffer the three successive transformations: forward offset, refraction and backward offset. The surface matrix is called R.
Stop effects
The entrance pupil of the optical system plays an important role in the overall aberrations; the center of the pupil defines the chief ray, which determines the paraxial image point. The ray fan usually extends equally in all directions around the chief ray and the points of intersection of the various rays in the fan with the image plane determine the ray aberrations. When matrices are used to model the system the image appears described in terms of the position and orientation coordinates of the rays when these are subjected to the first transformation, be it a refraction or a translation; in terms of ray aberrations this corresponds to a situation where the object is located at infinity and the entrance pupil is placed where the first transformation takes place. In fact, the orientation coordinates play the role of object coordinates and the position coordinates are the actual pupil coordinates, if this coincides with the first transformation; the chief ray can be found easily just by zeroing the position coordinate.
Ray aberrations can be evaluated correctly if the image is described in terms of both object and stop coordinates or some appropriate substitutes; an object point is then set by the object coordinates, the paraxial image point is found when the stop coordinate is zero and the aberrations are described by the differences to this point. This can be done adequately when the stop is located before the first surface and so coincides with the entrance pupil. If the object is at infinity the problem is very easy to solve and application examples have already been described [6]. If the entrance pupil is located at the position z s relative to the first surface, this means that the rays must perform a translation of −z s prior to hitting it; this can be accommodated via the right product by a translation matrix T −zs , where the index indicates the amount of translation.
The case of near objects is more difficult to deal with because we are faced with a dilemma: If a translation from the pupil position to the first surface is applied the position coordinates of the incident rays become pupil coordinates but the rays' origins on the object are lost. Conversely we can apply a translation from the object position, ending up with the reverse problem of knowing the rays' origins and ignoring their point of passage through the pupil. We need the position coordinates both at the object and pupil location; this suggests that the ray's orientation should be specified not by its direction cosines but by the coordinates of the point of passage through the pupil. In figure 2 a ray leaves an object point of coordinates (x, y) and crosses the entrance pupil plane at a point of coordinates (x p , y p ); then the direction cosines, s and t can be calculated by: where z p is the position of the pupil relative to the object. The matrix theory developed before was based on direction cosines rather than pupil coordinates; in consequence a coordinate change is needed before the translation from the object to the first surface is applied to the rays. Rather than use equations (12) it is more convenient to look at them as a coordinate change governed by a matrix T p ; for the determination of its coefficients we apply the standard procedure of series expansion. The following is the result for 7th order: a similar expression exists for t. The equation above allows the determination of the coefficients for matrix T p which converts the coordinates (x, y, x p , y p ) into (x, y, s, t) before the rays enter the system at the object position; as a symmetrical conversion is not performed when the rays leave the system, there is no place for the use of matrix T −1 p , as would be the case in a similarity transformation. A case study would involve the following successive steps: • Determine the matrices for all the translation and refraction transformations suffered by the ray from the object point to the image plane.
• Determine the matrix T p , relative to the entrance pupil position.
• Multiply all the matrices in reverse order, starting with the translation from the last surface to the image plane and ending with matrix T p ; the result is matrix S.
• Determine a vector base x& with the set of coordinates (x, y, x p , y p ).
• Make the product Once x ′ & has been calculated, its first four elements are polynomials of nth degree on the independent variables (x, y, x p , y p ), which model the position and direction cosines of the rays on the image plane.
Most optical systems have stops located after the first surface, and so the method described above is not adequate. One can always find the gaussian entrance pupil of any system and so revert to the previous situation; this will work even if the entrance pupil is found to lie beyond the first surface, in which case it will give rise to a negative translation. The problem is that the gaussian entrance pupil is a first order transformation of the stop, whereas the real entrance pupil is an aberrated image of the stop given by the portion of the optical system that lies before it. Ideally one should try to describe the image in terms of the object coordinates and the point of passage through the stop, which can be rather difficult. Alternatively one can find a reference ray, defined as the ray that passes through the center of the stop and is not necessarily the same as the chief ray, and work with position and orientation coordinates on the image plane, relative to the reference ray. The plot of the relative ray position versus its relative direction cosines can be used to describe ray aberrations [7].
Coefficients for a spherical surface
A sphere of radius r with its center at z = r is defined by the equation: this equation will replace (6) in the algorithm for the determination of matrix coefficients. In a previous work [6] we presented the results for third order; these are now extended to seventh order. This is a situation where symmetry must be considered in order to reduce the size of the matrices involved; we use the complex variables X and S previously defined, leading to a complex vector base X& whose elements have the general form X j X * k S l S * m with X * and S * representing the complex conjugates of X and S. Kondo et al. [4] have shown that the powers must obey the condition j − k + l − m = 1; for seventh order the allowable combinations are: Equation (10) can now be solved and the resulting expressions for s ′ and t ′ can be expanded in series up to the seventh order; the results can be combined into one complex variable S ′ . The second column of table 1 lists the coefficients for this expansion. Similarly the forward offset coefficients can be found solving equations (11) and combining x ′ and y ′ expansions into one single complex variable X 1 ; the resulting coefficients are listed in column 3 of table 1. The backward offset coefficients are obtained in a similar way, although now it is the x and y expansions that must be combined into a single complex variable X 2 .
Coefficients for an asymmetric surface
Toroidal surfaces are used to correct eye astigmatism; they can be fabricated easily without resorting to sophisticated machinery. In this example we deem to show that the method is applicable to general surfaces but it is not our purpose to list high order coefficients for this particular shape; therefore we will restrict the study to third order. Also, as there is no rotational symmetry we will not use complex coordinates; there is symmetry relative to two perpendicular planes and if we were to use complex coordinates the terms should obey the condition j − k + l − m = ±1 as explained in reference [5]; we feel that it is clearer not to invoke symmetry at all.
The surface is generated by a circle of radius r 1 normal to the xz plane whose center is displaced along a circle of radius r 2 laying on this plane. This surface has curvature radii of r 1 and r 1 + r 2 respectively on the xz and yz planes. The surface equation can be written as: The solution of equation (10) originates the expressions for s ′ and t ′ ; these are then expanded in series up to the third order and the resulting coefficients are listed in columns 2 and 3 of table 2. The same procedure was applied to the expressions for x ′ and y ′ resulting from equation (11); columns 4 and 5 of the same table list these coefficients. The backward offset coefficients need not be listed, as in third order they can be obtained by sign reversal from the forward offset coefficients.
Discussion and conclusion
Matrices can be used to model optical systems built with surfaces of unspecified shape to any desired degree of approximation. The method that was described differs from those found in literature by the choice of direction cosines rather than ray slopes to specify ray orientations. This results in a very elegant formulation of Snell's law which can be easily programmed into symbolic computation software. In spite of some added complication in translation matrices the algorithm is limited only by computing power in it's ability to determine the matrix coefficients, irrespective of surface complexity and order of approximation. For axis-symmetric systems the matrices have a dimension 40 × 40 for seventh order and can usually be dealt with by ordinary PCs; the influence of stops can also be incorporated in the matrix description as was demonstrated. The unconventional coordinate system that was used can, if needed, be converted to the more usual ray slope coordinate system through the product with conversion matrices of the same dimension. | 2019-04-14T03:13:57.847Z | 1999-03-01T00:00:00.000 | {
"year": 2004,
"sha1": "9c7e122c8e1a5a2977f4617bc6f17f8d4bce44ab",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/physics/0410066",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "3f62ec2c81256f241ea84c47e4d2b60fea346c0b",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
55686133 | pes2o/s2orc | v3-fos-license | Dynamics Behaviors of a Laser Produced Plasma: Theoretical Approach
Assuming that plasma particles are moving on continuous and non-differentiable curves, some dynamic properties in plasma ablation are analyzed via scale-relativity theory: the splitting of plasma plume, multi-peak structures, at various distances from the target surface and plasma oscillations through self-similarity. Our theoretical results are in good agreement with the experimental ones.
Introduction
The interaction of high-power laser radiation with target materials is a topic of current interest in various fields such as material-processing, plasma physics, analytical sciences, etc. [1,2].Laser ablation implies a series of complex phenomena: interaction of laser radiation with a solid target [3], laser beam absorption in ablation plume [4], hydrodynamics and electrical processes in the generated transient plasma [5,6], etc.
Correspondingly, theoretical models describing the dynamics related to these complex phenomena become sophisticated and ambiguous [1,2].However, the situation can be standardized taking into account that the elementary processes induced by laser-matter interaction impose various temporal resolution scales [7,8], and that the pattern evolution imposes different degrees of freedom e.g.: from one, at the initial stages, to three at the final stages of the patterns induced by the laser-produced plasma [9].In the present paper various theoretical aspects of a laser-ablation plasma dynamics (generation of two plasma structures, multi-peak structure for various distances from the target surface, plasma-ablation oscillations through self-similarity, etc.) were analyzed using the SR theory.
Hallmarks of Non-Differentiability
For developing our theoretical model, we take into account that, in plasma-ablation, deterministic chaos arises in association with spatio-temporal structures emergence.For temporal scales that are large with respect to the inverse of the highest Lyapunov exponent, the deterministic trajectories can be replaced by collections of potential trajectories and the concept of definite positions by that of probability density.This concept was introduced in the framework of the scale relativity (SR) theory [10][11][12], which states that the particles movement takes place on continuous but non-differentiable curves (fractal curves).Subsequently, all physical phenomena become dependent not only on the spatio-temporal coordinates but also on the spatio-temporal scales.Thus the non-differentiability become a fundamental characteristic of the plasma-ablation dynamics.
Non-differentiability implies the following [10][11][12][13][14][15][16][17]: 1) A continuous and non-differentiable curve (or almost nowhere differentiable) is explicitly scale dependent.Moreover, its length tends to infinity, when the scale interval tends to zero.Consequently, according to Mandelbrot's concept, a continuous and non-differentiable space will be a fractal space [18]; 2) Physical quantities will be expressed through fractal functions, namely those functions depending on space-time coordinates as well as on resolution scale.The invariance of the physical quantities in relation with the resolution scale generates special types of transformations that are called resolution scale transformations.Let us now explain the above statements through an example.We can choose the fractal normalized currentvoltage characteristic in the form [19]: where is the fractal normalized voltage, I is the fractal normalized current and a is a parameter depending on scale resolution.This relation induces conduction bistability (see Figure 1) as follows: the restriction 8 a implies bistability; the value of a sets the scale resolution through the ionization and recombination rates; once a is fixed (with 8 a ), for values of the fractal normalized current in the interval AB on the characteristic (see Figure 1) the fractal normalized voltage can have two distinct stable values; conduction bistability is associated with the negative differential resistance (or hysteresis); since and I are fractal functions (relation (1)) they can exhibit the property of self-similarity.Consequently, conduction bistability in Figure 1 can occur at any scale resolution (i.e. for different ionization and recombination rates).As a result, a correspondence between multiplicity order of the double layer and the one of conduction bistability is likely to occur.
3) Although the local differential time invariance is broken, it can still be recovered through the complex operator [15][16][17]: where ( is the complex speed field and , Δ are the usual operators.The real part of the complex speed field represents the standard classical speed which is differentiable and does not depend on resolution, , while the imaginary part is a new quantity arising from fractality, which is non-differentiable and resolution-dependent. Quantity D is Nottale's coefficient and corresponds to the fractal-non-fractal transition while D F is the fractal dimension of motion curves; 4) Plasma-ablation particles may be reduced to and identified with their own trajectories (i.e.their geodesics), so that the complex system should behave as a special "fluid" lacking interaction via geodesics in a non-differentiable (fractal) space.The equation of geodesics (a generalization of Newton's first principle for motion of fractal curves) can be written in the form [15][16][17]: This means that the global complex acceleration field ˆt V t depends on the local complex acceleration field, , on the non-linear (convective), and dis- sipative, V terms.Moreover, the fractal fluid becomes viscoelastic or hysteretic, i.e. the fractal fluid will be endowed with memory.Such a result is in agreement with the opinion given in refs.[1,2]: the fractal fluid can be described through Kelvin-Voight or Maxwell rheological models using complex quantities, e.g. the complex speed field, the complex acceleration field etc.In addition, relation ( 4) is a Navier-Stokes type equation with an imaginary viscosity coefficient
Fractal Hydrodynamic Model
If the motions of the fractal fluid are irrotational, i.e.
V V
, we can choose having the form: ˆ2i d ln , x y case in the form: Substituting (6a)-(6c) in ( 4) and separating the real and the imaginary parts up to an arbitrary phase factor which may be zero by a suitable choice of the phase of , we shall obtain: with the fractal potential, and 0 the rest mass of the fractal fluid particle.Equation (7a) is the momentum conservation law, Equation (7b) is the density conservation law.They both define the fractal hydrodynamic model (FHM).Now, certain conclusions are evident: 1) Any particle is in permanent interaction with the "sub-fractal level" through the fractal potential Q; 2) The "sub-fractal level" is identified with a nonrelativistic fractal fluid described by the probability density and the momentum conservation laws-see relations ((7a), (7b)).These equations correspond to the generalised quantum hydrodynamic model (GQHM).Indeed, for motions that can be described via fractal curves in fractal dimension F at Compton scale, 0 with the reduced Planck constant, the FHM reduces to a quantum hydrodynamic model (QHM).In this last case the "sub-fractal level" is identified with "sub-quantum level" [10][11][12].
3) The fractal potential (8) results from non-differentiability and should be considered as a kinetic term and not as a potential one.Moreover, the fractal potential (8) can generate a viscosity stress tensor type [15][16][17].
whose divergence is equal to the usual force density associated with Q
Numerical Simulations
Let us now rewrite the equations of FHM for the two-dimensional e where fractal continuity equation of density energy was added [15][16][17].With non-dimensional vario ables, , t (12a) in the case of the ideal gas and with variation σ being induced by the density and temperature variations , const.
, Equations (11a)-(11d) becomes: In Equations (13a)-(13d) the functional scaling relation, For numerical integration, the initial conditions, (14a) d the boundary on , ,0 are considered.In the boundary condition ((15m), (15o)) we assumed that the laser pulse which "hits" the target induces a plasma source that has a spatial-temporal Gaussian profile, similarly with the laser beam.N 0 is the maximum normalized atom density, while T 0 is the maximum normalized temperature which is assumed to be proportional with the laser beam energy, by preserving the number of total released atoms.
The equation system (13a)-(13d) with the initial con- 15p) was numerically integrated via finite differences [20].In Figures 2(a)-(c) space-dependence of the normalized density, N, normalized temperature, T, and in Figures 3(a)-(c 2) the symmetry of the normalized speed, V ξ , according to the axis symmetry of the space-time Gaussian; 3) shock waves and vertices at the plume periphery for the normalized speed field, V η .
These data are in agre ages at various evolution stages [5,6,21].Moreover, if the current density of the particle is plotted on the symmetry axis -see Figures 4(a)-(c), for various distances fro rget surface, a multi-peak structure as in [5,6] can be noticed.Increasing the value of the control parameter, , we conclude: 1) the arrival time of the first peak decreases; 2) the ratio between the first and the second maximum increases.This is in agreement with the experimental data according to which the particles are gradually transferred from the fast into the slow part [22][23][24]; 3) the magnitude of the first maximum decreases as a consequence of lateral expansion.Therefore we conclude that plume splitting is a hydrodynamic process similar to Gaussian disturbance emission.m the ta
Self-Similarity in Plasma Ablation
(4) takes Neglecting the convection ˆ V V , Equation the form: or, separating the resolution scales, for the differentiable scale, and for the fractal scale.Velocity fields are totally separated, U firstly applying Equations ( 17) and ( 18) to the Δ operator, i.e.
and (1 take the unitary form [26]: For the one-dimensional case, the previous equations w ith the substitutions: We impose "clamping" c 1 and for boundary conditions at 0 -see Figure 5.
Owi g ng to scalin
x L t T , we find a solution for Equatio the form: n where the self-similarity variable is: The boundary condition for Substituting this self-similar form of 23), the following equation for the self-similar solution Accordingly, self-similarity generates oscillations in plasma-ablation.This result is experimentally confirmed in [5,6,21,22].
Conclusions
The m e z built.Fractality is introduced suposing that the fractal potential is e ideal gas pressure, the ablation ain conclusions of the present paper are th folwing: 1) the plasma expansion was theoretically ana-lo ly ed assuming that the particle moves on continuous and non-differentiable curves; 2) a fractal hydrodynamic model containing the density and momentum conservation equations was via fractal potential; 3) connected with th plasma expansion is studied through numerical simulations.The splitting of plasma plume, multi-peak structures at various distances from the target surface can be noticed; 4) in the absence of convection, oscillations via self-similarity are induced in plasma-ablation.
Theoretical data are in good agreement with the experimental ones.
Figure 1 .
Figure 1.Theoretical dependence of the fractal normalized current on the fractal normalized potential.
the scalar potential of the complex speed.with For i e S , with the amplitude and S the phase of , the complex speed field (5) takes the explicit form:
JMP 1017 Figure 2 .
Figure 2. (a)-(c) Space-dependence of the normalized density, N, and normalized temperature, T, resulting from numerical
Figure 5 .Figure 4 .
Figure 5. Numerical solution of the Kirchhoff Equation (23) with "clamped-free" unitary conditions, for a uniform initial relaxes to zero within the first few timesteps.
Figure 6 ,
where the Fresnel sine integral, Relation (32) describes an oscillation via a self-similar solution 0 from the theory of diffraction has b x t .This reflects the disp Equation (23).ersive nature of
Figure 6
Figure 6.Self-similar solution ( n of 32) as a functio . | 2018-12-09T23:59:17.235Z | 2013-07-04T00:00:00.000 | {
"year": 2013,
"sha1": "62e59af65bdf8bc82d02078c0413dfd1a688d36f",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=34827",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "62e59af65bdf8bc82d02078c0413dfd1a688d36f",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
40295319 | pes2o/s2orc | v3-fos-license | Comment on “Fluorotechnology Is Critical to Modern Life: The FluoroCouncil Counterpoint to the Madrid Statement”
Comment on Fluorotechnology Is Critical to Modern Life : The FluoroCouncil Counterpoint to the Madrid Statement
We commend the FluoroCouncil for phasing out long-chain poly-and perfluoroalkyl substance (PFAS) chemistry. However, members of the FluoroCouncil have been producing long-chain PFASs for decades while in possession of research showing adverse health effects in humans and animals. This model of chemical manufacturing needs to change. We recommend implementing the principles of green chemistry (Anastas and Warner 1998) in chemical manufacturing to ensure safer and sustainable chemical products. The scientific consensus of the Madrid Statement authors and signatories is that the use of all PFASs is unsustainable, and can and should be greatly reduced and discontinued where feasible. Short-chain fluorinated alternatives were therefore intentionally included in the scope of the Madrid Statement.
Some of the functionalities provided by fluorotechnology have become part of modern life. However, we disagree that PFASs are critical to modern life. Sustainable and less hazardous alternatives are available for many functionalities, and others will be developed. PFAS-based chemistries are used in many nonessential applications such as clothing, sports equipment, food packaging materials, blooming and dispersion agents, and stain-repellant treatments. We urge the FluoroCouncil to provide as much information as possible on the PFAS chemistries used in different commercial products and technologies.
We are aware that short-chain perfluoroalkyl acids bioaccumulate less than longchain ones. However, some short-chain PFASs have been linked to adverse biological effects , and further systematic, representative studies on additional end points are needed. Given the ongoing release and environmental persistence of short-chain acids, increasing environmental and human exposures such as those documented by Glynn et al. (2012) are expected, for example, via contaminated drinking water aquifers (Xiao et al. 2015). Thus, continuous release of short-chain PFASs can be expected to lead to poorly reversible internal exposures, regardless of their low bioaccumulation potential (Scheringer et al. 2014).
Bowman commented that the Madrid Statement cannot claim insufficient data on the hazards and risks of fluorinated alternatives. However, highlighted the specific data gaps that prohibit conducting hazard and risk assessments for many fluorinated alternatives. An assessment commissioned by the FluoroCouncil (ENVIRON International Corporation 2014) also identified many gaps regarding human health data.
Bowman stated that "decisions on the societal acceptability of strategic materials such as PFASs cannot be wisely made on a single attribute such as persistence." However, persistent chemicals are unsustainable in a world with limited resources. We cannot afford to "lose" portions of resources (water, soil, or food) because potentially harmful and persistent chemicals are accumulating over centuries and causing continuous exposure. Because of their persistence, an enormous inventory of PFASs is being created: Even if all PFAS production and uses were to stop immediately, PFASs would continue to be released for decades during products' use and disposal life-cycle phases (Wang et al. 2014a(Wang et al. , 2014b. One of the 12 principles of green chemistry is "design for degradation: chemical products should be designed so that at the end of their function they break down into innocuous degradation products and do not persist in the environment" (Anastas and Warner 1998). We endorse this principle and urge the FluoroCouncil to follow it also.
We welcome collaboration with the FluoroCouncil to establish informationsharing platforms for PFASs and support all opportunities for dialogue. We ask the FluoroCouncil to take leadership and responsibility for the global management of the PFASs they produce, from manufacturing to end of life.
A 171
Correspondence chemistry worldwide. More than a decade ago, the FluoroCouncil member companies responded to concerns about long-chain poly-and perfluoroalkyl substances (PFASs) by working with regulators to voluntarily phase out those substances and develop alternatives with improved health and environmental profiles. The FluoroCouncil also works with regulators and other stakeholders to support a global transition away from long-chain PFASs. This effort stands as a historic collaboration by government and industry to foster sustainable development.
The claim that all PFASs are problematic is simply not supported by the wealth of data available on both long-and shortchain PFASs. Because of the concerns raised in regard to long-chain PFASs, the U.S. Environmental Protection Agency (EPA) has held the fluorotechnology industry to high standards and increased data requirements to ensure the alternatives are well studied and safer than the substances being replaced. Consequently, short-chain PFASs are some of the most robustly studied new chemicals introduced to the market, having undergone years of toxicity and environmental testing at the request of regulators. Industry continues this collaboration with regulators, developing additional data on the alternatives and working to make those data publicly available, including on the FluoroCouncil website (http://www. fluorocouncil.org/ Resources/Research). Based on this robust body of data, regulators globally have determined the alternatives are safe for their intended use. The "sustainable and less hazardous alternatives" sought by the Madrid Statement authors already exist in the form of short-chain PFASs.
We continue to be perplexed by assertions from the authors of the Madrid Statement that short-chain PFASs present hazards comparable to those of long-chain PFASs, citing publications such as and . These publications suffer from important data gaps, such as the failure to cite key published articles on the toxicity of short-chain PFASs (e.g., Klaunig et al. [2015], which presents animal data indicating perfluorohexanoic acid is not carcinogenic). Furthermore, these publications actually acknowledge or demonstrate that many of the leading short-chain PFASs are less bioaccumulative and less toxic than the long-chain PFASs with which they have been compared, based on the available data taken as a whole. This conclusion, which is well accepted by regulatory agencies such as the EPA, compels a different policy outcome than the Madrid Statement suggests. The first priority for risk management should be phase-out of the long-chain PFASs. Attempting to broaden that phase-out to effective alternatives that are less hazardous can only create a technological impasse that supports the retention of long-chain PFASs in the marketplace.
The authors of the Madrid Statement also contend that PFASs are not critical to modern life. The importance of PFAS chemistry, however, was long ago determined by the market. Industries relying on PFASs evaluated fluorinated and nonfluorinated alternatives, as well as alternative technology, and decided on the products that met their specifications and performance needs. Some decisions involved continuing to use PFASs because they meet performance needs that nonfluorinated alternatives cannot. For example, first-responder protective gear is treated with fluorinated products to help maintain performance in fires; firefighting foam produced with fluorinated surfactants provides shorter extinguishment times and critical burnback resistance when fighting flammable liquid fires; and hospital gowns, drapes, and divider curtains rely on fluorinated polymers to provide protective barriers against transmission of diseases. Because the short-chain PFASs have been reviewed and approved by regulatory authorities globally, all applications relying on these substances can be used without presenting a significant risk.
The largest use of short-chain PFASs is for polymeric products. These productslike other polymers-are quite stable under environmental conditions. The resilience of short-chain PFASs is directly connected to its performance, providing long-lasting, durable properties. First responders, medical personnel, and patients would certainly not want the properties in safety gear or medical garments to quickly become ineffective. Even when the short-chain PFASs, which do not present a significant risk, are used in what some people may characterize as "nonessential applications," such as clothing and furniture, these substances significantly extend the effective lifetime of those products, meaning less waste, infrequent washings, and economic savings. These benefits of short-chain PFASs can be further enhanced by reducing emissions through the adoption of best environmental practices, which the FluoroCouncil has identified and is encouraging in the supply chain.
The FluoroCouncil members remain committed to science-based stewardship activities, including continually enhancing the sustainability of their chemistries and products by improving their environmental, health, safety, and performance profiles. We are open to working collaboratively and constructively with stakeholders on 1) strategies to complete the global transition away from long-chain PFASs, 2) identification of issues that warrant further data development and risk assessment, 3) actions that can foster additional stewardship activities within the supply chain, and 4) best methods for transparently sharing information relevant to the health and environmental impact of PFASs. | 2018-04-03T04:48:28.040Z | 2015-07-01T00:00:00.000 | {
"year": 2015,
"sha1": "25e1d307f99038470c938688803a0ff5ea10cee1",
"oa_license": "CC0",
"oa_url": "https://doi.org/10.1289/ehp.1510207",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "ab2b2071a4dc221de71e0990afb4be30f292639f",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
632655 | pes2o/s2orc | v3-fos-license | Transcranial Focused Ultrasound for BOLD fMRI Signal Modulation in Humans
Transcranial focused ultrasound (tFUS) is an emerging form of non-surgical human neuromodulation that confers advantages over existing electro and electromagnetic technologies by providing a superior spatial resolution on the millimeter scale as well as the capability to target sub-cortical structures non-invasively. An examination of the pairing of tFUS and blood oxygen level dependent (BOLD) functional MRI (fMRI) in humans is presented here.
INTRODUCTION
Transcranial focused ultrasound (tFUS) is a new and promising non-surgical low-energy technique for inducing transient modulation with high spatial resolution, adjustable focus and low tissue attenuation. Ultrasound can noninvasively stimulate the hippocampus and motor cortex of intact mice [1,2], modulate monosynaptic and polysynaptic spinal reflexes in cats [3] and disrupt seizure activity in cats [4], rats [5], and mice [6]. In addition, tFUS has been used safely and effectively for intact neural stimulation in mouse [7], rabbit [8], sheep [9] and monkey [10] and our work has shown tFUS to also be a safe and effective method of transient transcranial cortical stimulation in humans [11,12]. Using EEG, our work has shown tFUS to inhibit the amplitude of somatosensory evoked potentials (SEP) and to affect the power and phase of beta and gamma frequencies [13] that has since been expanded upon by other groups [14], demonstrating ultrasound as an efficacious form of highly focal transient stimulation for use in humans. Because of tFUS' high spatial resolution and ability to stimulate at depth, it is of particular interest to pair tFUS with magnetic resonance imaging (MRI) to take advantage of its superior whole brain resolution. Previous literature has shown in craniotomized rabbits that 3.3 W/cm 2 spatial-peak pulse-average intensity (Isppa), 0.69 MHz focused ultrasound to the sensorimotor cortex resulted in a detectable BOLD response that corresponded well in space to the FWHM sonication pressure [8]. Here we report on the feasibility of performing concurrent tFUS/MRI in humans, the issues inherent in this process and the results from our experiments examining cortical BOLD response at 3T and sub-cortical BOLD response at 7T.
CHARACTERIZATION AND APPLICATION OF TFUS WAVEFORMS
We used two separate focused transducers for this study. For the 3T experiment, we used a 0.5MHz transducer that had an active diameter of 60 mm and focal length of 55 mm producing a focal FWHM intensity volume of 48.64 mm 3 . For the 7T experiment we used a 0.86 MHz focused transducer that had an active diameter of 64 mm and a focal length of 54 mm producing a focal FWHM volume of 35.77 mm 3 . Transcranial ultrasonic neuromodulation waveforms were generated using a two-channel, 2-MHz function generator (BK Precision Instruments, CA, USA). Channel 1 was set to deliver US at a pulse repetition frequency of 1.0 KHz, and channel 2 was set to drive the transducer at the acoustic frequency in a bursting mode, with channel 1 serving as an external trigger for channel 2. The pulse duration of the waveform was set by adjusting the number of cycles per pulse on channel 2. Stimulus duration was set by adjusting the number of pulses on channel 1. The output of channel 2 was sent through a 100W linear RF amplifier (E&I 2100L; Electronics & Innovation) matched to the impedance of the transducer by an external matching network. The waveforms were as described in Legon et al. 2014 [12].
Transcranial Focused Ultrasound for BOLD fMRI Signal Modulation in Humans
Leo Ai, Jerel K. Mueller, Andrea Grant, Yigitcan Eryaman, and Wynn Legon 180 cycles of the 0.5 MHz acoustic frequency were pulsed at 1 KHz resulting in a 36% duty cycle. 500 cycles of the pulse repetition frequency were delivered resulting in a 500 ms duration stimulus. For the 7T experiment, 420 cycles of the 0.86 MHz acoustic frequency (50% duty cycle) were pulsed at 1 KHz with the same 500 ms duration as above. Stimulus timing was controlled by custom written Matlab (Mathworks, MA, USA) scripts timed according to TR onset of the MRI scanner. For transducer mounting on the head, the hair was parted to expose the scalp, and the transducer thoroughly coated with acoustic coupling gel and secured in place using surgical meshing and athletic pre-wrap. For the 3T experiment, the transducer was offset from the head 3cm and coupled using a coupling cone filled with acoustic gel. Positioning of the transducer was performed using a stereotaxic neuronavigation system (Rogue Research Inc., CAN) that has been adapted to work with the ultrasound transducers.
Ultrasound acoustic intensity profiles from the above waveforms were measured in degassed, deionized free water and through hydrated human cranium in a 300L acoustic test tank (Precision Acoustics, UK) using a calibrated hydrophone (HNR-0500, Onda Corporation, CA). Scans were first performed without bone and then performed using a 6-mm-thick fragment of human parietal bone (degassed and rehydrated for 48 h) between the transducer and the hydrophone. From this data, we adjusted the driving voltage for each transducer to produce a spatial peak pulse average intensity I sppa after bone transmission of 6W/cm 2 .
TFUS WITH MRI tFUS Inducing BOLD Signal at 3T
Six neurologically healthy volunteers (20 -36 years) were scanned at the Virginia Tech Carilion Research Institute on a Siemens (Erlangan, Germany) Tim Trio 3T scanner using a 12 channel head matrix coil. Prior to scanning, volunteers were fitted with the transducer targeted at the primary motor cortex hand knob of the dominant hemisphere. Prior to the functional scans, a T1weighted anatomical image was acquired to align with each subject's BOLD contrast data. Functional images of BOLD contrast signals were acquired using a gradient echo echo-planar imaging (GRE EPI) sequence (TR = 2000 ms, TE= 30 ms, flip angle = 90, FOV = 190 mm, slices = 33, slice thickness = 3 mm, 450 images). An event-related design was employed where 90 0.5 sec pulsed ultrasound stimuli were delivered every 6-7 TRs (12-14 secs). The resulting time courses were analyzed using Statistical Parametric Mapping (SPM8; www.fil.ion.ucl.ac.uk/spm/). The functional data was slice-time corrected, realigned, preprocessed by linear trend removal, temporal high-pass filtered (128 sec) and threedimensional motion corrected using a trilinear interpolation. Functional data sets were transformed into MNI space and co-registered with anatomical data for each subject. The resulting time courses were filtered using an 8-mm Gaussian kernel at full width half-maximum (FWHM). Statistical analysis was performed by fitting the signal time course of each voxel using the general linear model (GLM). The onset of stimulation was used as a regressor and modeled using the canonical hemodynamic response function. Additional regressors included head motion parameters generated during fMRI preprocessing. Individual analysis was for US on/off using one-sample t-tests.
Testing for significance at the group level (N = 6) was conducted for condition US on/off using a one-sample t-test. Both whole-head and region of interest (ROI) analysis were conducted. The ROI included the precentral gyrus to the depth of the central sulcus as well as the post-central gyrus to the depth of the central sulcus as determined from the T1 scans. Data is presented at p = 0.001 uncorrected with a cluster threshold of 5 voxels. There was no statistically significant area of activation on the whole head group level or on the group level ROI of the sensorimotor region. Further exploration of individual subject data revealed that 3 of the 6 participants showed no discernable BOLD response. However, the remaining 3 participants' revealed very focal BOLD response in the sensorimotor region in good accordance with the focus of the ultrasound beam ( Figure 1). The lack of effect for 3 of the 6 participants is the likely reason for a lack of group effect. The tFUS stimulation was delivered with a block design (5 off/on cycles, stimulation delivered every 6 TRs during on cycles). A GE EPI sequence was used to acquire the functional data (parameters: TR = 600ms, TE = 21.4ms, flip angle = 40, voxel dimensions = 2x2x2mm, 1000 images), along with a T1 anatomical scan. The functional data was analyzed with Analysis of Functional NeuroImages (AFNI) [15] Correlation analysis was used to analyze the fMRI data using a reference function generated by convolving the stimulation paradigm with the hemodynamic response function. As part of the analysis process, motion correction was performed to minimize motion effects, and a Gaussian filter (FWHM = 4mm) was applied for smoothing. A Z score threshold of 3.5 (p = 0.0002, uncorrected) was applied to isolate activations. Initial results indicate a focal BOLD response in the targeted caudate area ( Figure 2).
MR COMPATIBILITY -TEMPERATURE
One of the concerns with concurrent tFUS/MRI is transducer and cable heating. A study was performed to examine surface temperature changes of the transducer operated on a 7T Siemens magnet using a Nova Medical 8x32 Head Coil at UMN's CMRR. A spin echo pulse sequence was used to image the transducer over a period of 20 minutes, with temperatures being recorded, to simulate data acquisition using parameters set to reach the SAR limits set by the FDA (3.2W/kg per 10-minute exposure) [16]. The temperature tests were run within the scanner with the transducers unpowered and powered to deliver a mechanical index of 1 using 0.5 sec pulsed ultrasound stimulation delivered every 5 seconds for the duration of the 20 minute temperature recording session. Two 0.5MHz transducers were used for this examination, one with a diameter of 64 mm and the second with a diameter of 30mm. The transducer faces were first filled with ultrasound gel, and fiber optic temperature probes (Lumasense Technologies, CA, USA) were then placed on the center and along the edges of the transducer faces inside of the ultrasound gel. The transducers, with the temperature probes attached, were then secured to an agar phantom to simulate a real data acquisition. The results with the transducers running showed that the center face recorded the greatest increase in temperature of approximately 1° C for the smaller transducer and 0.75° C for the larger transducer over the recorded period of 20 minutes. In the off state, no temperature increases were detected. One degree centigrade is an acceptable surface temperature increase, indicating minimal risk of surface heating using tFUS/MRI at high field (7T).
MR COMPATIBILITY -ARTIFACTS
Typical ultrasound transducers contain lead zirconate titanate (PZT) for the piezoelectric element. This is a paramagnetic inorganic compound that produces susceptibility artifacts in the MRI scanner, a potentially significant problem that needs to be resolved and needs to be taken into consideration during transducer manufacturing, especially at high field strengths. Using custom designed PZT ceramic focused transducers with normal and half the normal amount of PZT, MR artifact tests were performed on an agar phantom on a 7T Siemens (Erlangan, Germany) magnet using a Nova Medical 8x32 Head Coil (Wilmington, MA) and a GRE EPI pulse sequence (at UMN's CMRR). The normal PZT element produced considerable magnetic susceptibility artifact. The PZT element was then translated in 1cm steps away from the phantom to assess the effect on the artifact. At 4cm from the agar phantom, the artifact resolved ( Figure 3). It should be noted that even with a transducer containing half the PZT, the susceptibility artifacts are not eliminated, but can be reduced to a state that could be acceptable for studies interested in examining cortical activity directly under the transducer with careful shimming (Figure 4). However the issue of susceptibility artifacts still can prohibit studies that involve examinations closer to the surface of the cortex, and more work will have to be done to further reduce artifacts for such studies.
FURTHER DISCUSSION
The results from the preliminary studies indicate that it is possible to detect BOLD fMRI signals from tFUS at both 3T and 7T field strengths in both cortical and subcortical regions of the brain. BOLD activations were detected in the M1/S1 area in the 3T studies, and a BOLD activation was detected in the caudate in the 7T study. There are considerable potential issues with MR artifacts and heating from the transducer, though it looks like these issues can be overcome with proper transducer design.
The 3T study presented results with BOLD fMRI activity detected in the targeted M1/S1 region. The results were not consistent where 3/6 participants showed no discernable BOLD activation. The reasons for this are not currently clear and need to be elucidated. This could be related to the amount of energy, the pulsing strategy, the event design and perhaps an under-powered ability to detect the signal. This study used an event-related design with a long interstimulus interval that may not be efficacious for detecting response. We used an ultrasound waveform that delivered equivalent intensities to the brain as reported by [8] though with differences in pulsing strategy and duty cycle. It is possible that these differences lead to our variability and should be examined in future studies. There is also the consideration that ultrasound does not directly affect neuronal kinetics but rather affects local microvasculature through mechanical perturbation that leads to the BOLD response. We cannot be definitive that this is not the case but, the fact that ultrasound results in behavioral improvements [12,14,17] suggests that ultrasound is mediating a neuronal response.
The 7T pilot study also did present with a BOLD fMRI activation in the targeted caudate area, but there are steps that can be taken to further strengthen the case for deep brain stimulation in humans with tFUS. The acoustic frequency used was not optimal for bone transmission [18], but was used to achieve a better focal volume for specific targeting of small deep-brain structures. Transmission loss can be overcome however by delivering a high initial intensity outside of the head and we have shown that the skull does not serve to distort the acoustic field and may in some cases serve to help focus it [12].
These initial studies provide evidence that detection of a BOLD response is achievable non-invasively with tFUS in humans. However, we realize that optimization is necessary. MRI parameters need to be further examined to maximize the ability to reliably detect tFUS induced BOLD signals, especially shimming strategies and pulse sequence selection (i.e., GRE vs spin-echo sequences). Magnetic susceptibility artifacts caused by placing the transducer in the MR environment will continue to be a concern for all continuing studies involving pairing tFUS with MRI based techniques. It was found that careful shimming can reduce these artifacts to acceptable levels, but this may not always be the case. Transducer materials and design have the largest effect on susceptibility artifacts and there are examinations currently underway to determine the desired transducer characteristics that would be the most beneficial for operating in the MR environment. Once the optimal transducer characteristics are determined, the optimal MRI parameters and procedures will also be thoroughly examined to maximize signal detectability.
CONCLUSION
The studies performed would indicate that it is possible to induce BOLD fMRI signals using tFUS targeting cortical and subcortical structures at 3T and 7T respectively. Optimizations in ultrasound pulse parameters, MRI data acquisition parameters, and ultrasound transducer design needs to be made, but these preliminary studies show promise in using tFUS with BOLD fMRI, offering potentially many advantages over currently used neuromodulation techniques. The tFUS technique is able to target sub-cortical structures with a precision on the millimeter scale. As such, it is of utmost importance to the development of tFUS to examine its effects on BOLD fMRI. Observations from such examinations could have important implications in global brain mapping efforts and in many forms of guided therapies in clinical settings. | 2016-03-01T19:22:12.000Z | 2016-03-01T00:00:00.000 | {
"year": 2016,
"sha1": "0f34e7d5629522847888c1c4a2b2f287408a28af",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1603.00415",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "03599896fe3b6262be341cc8c5c26d273290b34f",
"s2fieldsofstudy": [
"Engineering",
"Medicine"
],
"extfieldsofstudy": [
"Psychology",
"Mathematics",
"Biology",
"Medicine",
"Computer Science"
]
} |
248084859 | pes2o/s2orc | v3-fos-license | Three-point functions of fermionic higher-spin currents in 4D conformal field theory
We investigate the properties of a four-dimensional conformal field theory possessing a fermionic higher-spin current $Q_{\alpha(2k) \dot{\alpha}}$. Using a computational approach, we examine the number of independent tensor structures contained in the three-point correlation functions of two fermionic higher-spin currents with the conserved vector current $V_{m}$, and with the energy-momentum tensor $T_{m n}$. In particular, the $k = 1$ case corresponds to a"supersymmetry-like"current, that is, a fermionic conserved current with identical properties to the supersymmetry current which appears in $\mathcal{N} = 1$ superconformal field theories. However, we show that in general, the three-point correlation functions $\langle Q Q T\rangle $, $\langle \bar{Q} Q V\rangle $ and $\langle \bar{Q} Q T\rangle $ are not consistent with $\mathcal{N}=1$ supersymmetry
Introduction
Correlation functions of conserved currents are among the most important observables in conformal field theory. It is a well known fact that conformal symmetry determines the general form of two-and three-point correlation functions up to finitely many parameters, however, it remains an open problem to understand the structure of three-point functions of conserved currents for arbitrary spin. The systematic approach to study correlation functions of conserved currents was undertaken in [1,2] (see also refs. [3][4][5][6][7][8][9][10][11][12] for earlier results), and was later extended to superconformal field theories in diverse dimensions [13][14][15][16][17][18][19][20][21][22][23][24][25][26]. 1 The most important examples of conserved currents in conformal field theory are the energy-momentum tensor and vector currents; their three-point functions were studied in [1]. However, more general conformal field theories can possess higher-spin conserved currents. As was proven by Maldacena and Zhiboedov in [28], all correlation functions of higher-spin currents are equal to those of a free theory. This theorem was originally proven in three dimensions and was later generalised in [29][30][31] to four-and higher-dimensional cases. The general structure of the three-point functions of conserved higher-spin, bosonic, vector currents was found by Stanev [32] and Zhiboedov [33], see also [34] for similar results in the embedding formalism [35][36][37][38][39][40] (and [41,42] for supersymmetric extensions). There are also some novel approaches to the construction of correlation functions of conserved currents which carry out the calculations in momentum space, using methods such as spinor-helicity variables [43][44][45][46][47][48][49].
The study of correlations functions in conformal field theory has mostly been devoted to bosonic operators with vector indices (except for supersymmetric settings); fermionic operators have practically not been studied. 2 Our interest in studying threepoint functions of fermionic operators is two-fold; first, any conformal field theory possessing fermionic operators naturally breaks the assumptions of the Maldacena-Zhiboedov theorem [28] discussed above. Indeed, the main assumption of the Maldacena-Zhiboedov theorem was that the conformal field theory under consideration possesses a unique conserved current of spin two, the energy-momentum tensor. However, in [28] it was also shown that if a conformal field theory possesses a conserved fermionic higher-spin current then it has an additional conserved current of spin two. Hence, it is not clear whether 1 The approach of [1,2] performs the analysis in general dimensions and did not consider parityviolating structures relevant for three-dimensional conformal field theories. These structures were found later in [27]. 2 Recently, in [34], correlation functions involving fermionic operators were studied, however these operators were not conserved currents.
correlation functions of fermionic higher-spin currents must coincide with those in a free theory. Second, fermionic operators are interesting due to their prevalence in supersymmetric field theories. In fact, there is a natural question: if a conformal field theory possesses a conserved fermionic current, is it necessarily supersymmetric?
The aim of this paper is to study correlation functions of the conserved fermionic higher-spin currents 3 Q α(2k)α ,Q αα(2k) , (1.1) which obey the conservation equations ∂ αα Q αα(2k−1)α = 0 , ∂ ααQ ααα(2k−1) = 0 . (1. 2) The case k = 1 in (1.1) is quite interesting as it corresponds to currents of spin-3 2 which possess the same index structure and conservation properties as the supersymmetry currents. Indeed, one might expect that a conformal field theory possessing conserved spin-3 2 primary operators is supersymmetric. One way to explore this issue is to study the correlation functions involving such operators to see if they are consistent with supersymmetry.
In particular, we must study the general form of the three-point functions involving combinations of the operators Q α(2)α ,Q αα(2) (i.e. (1.1) for k = 1), the energy-momentum tensor T mn and the vector current V m . Recall that in any superconformal field theory the supersymmetry current and the energy-momentum tensor are components of the supercurrent multiplet, J αα (z), where z = (x m , θ α ,θα) is a point in 4D Minkowski superspace. This implies that in supersymmetric theories the three-point functions must be contained in the three-point function of the supercurrent J αα (z 1 ) J ββ (z 2 ) J γγ (z 3 ) , which was shown in [14] to be fixed up to two independent tensor structures. Similarly, in supersymmetric theories the vector current V m is a component of the flavour current multiplet, L(z). Hence, the three-point functions must be contained in the three-point function of the supercurrent and the flavour current J αα (z 1 ) J ββ (z 2 ) L(z 3 ) , which was shown to be fixed up to a single tensor structure [14].
In this paper, we study the general form of the three-point functions (1.3), (1.4) and extend the results to the operators (1.1), using only the constraints of conformal symmetry; supersymmetry is not assumed. The analysis is highly non-trivial and requires significant use of computational methods. To streamline the calculations we develop a hybrid formalism which combines the approach of Osborn and Petkou [1] and the approach based on contraction of tensor indices with auxiliary vectors/spinors. This method is widely used throughout the literature to construct correlation functions of more complicated tensor operators. Our particular approach, however, has some advantages as the correlation function is completely described in terms of a polynomial which is a function of a single conformally covariant three-point building block, X, and the auxiliary spinor variables u,ū, v,v, w,w. Hence, one does not have to work with the spacetime points explicitly when imposing conservation equations. To find all solutions for the polynomial, we construct a generating function which produces an exhaustive list of all possible linearly dependent structures for fixed (and in some cases, arbitrary) spins. The possible structures form a basis in which the polynomial may be decomposed, and are in correspondence with the solutions to a set of six linear inhomogeneous Diophantine equations, which can be solved computationally for any spin.
Using the methods outlined above, we find that the three-point functions (1.3), (1.4), in general, are not consistent with supersymmetry as they are fixed up to more independent tensor structures than the three-point functions JJJ and JJL . This means, based on the constraints of conformal symmetry alone, that the existence of spin-3 2 "supersymmetry-like" conserved currents in a conformal field theory does not necessarily imply that the theory is superconformal. We want to stress that our analysis is based only on symmetries and does not take into account other features of local field theory. We do not know how to realise a local non-supersymmetric conformal field theory possessing conserved spin- 3 2 currents, neither do we have a proof that it is impossible. Our paper is organised as follows: in Section 2, we discuss the general formalism to construct two-and three-point functions in conformal field theory. First, we review the constructions of Osborn and Petkou [1] and introduce our hybrid generating function formalism based on contractions of tensor operators with auxiliary spinors. We construct a generating function which, for a given choice of spins, generates all possible linearly dependent solutions for the correlation function. In Sections 3 and 4, we find the most general form of the three-point functions (1.4). Our conclusions are that the three-point function Q QV depends on three independent tensor structures (here and in all other cases the structures are found explicitly), while the three-point function QQV vanishes in general. In Sections 5 and 6, we find the most general form of the three-point functions (1.3). Our conclusions are that the three-point function Q QT is determined up to four independent tensor structures and the three-point function QQT is fixed up to a single tensor structure. Most of our analysis in Sections 3-6 was performed for an arbitrary k. However, due to computational limitations certain results were proven only for small values k. Nevertheless, we believe that the results stated above hold for all values of k. Finally, in Section 7, we discuss whether our results are consistent with supersymmetry for k = 1, when Q possesses the same properties as the supersymmetry current. We show that, in general, the results obtained in Sections 3-6 are not consistent with supersymmetry. Our four-dimensional notation and conventions are summarised in Appendix A.
Conformal building blocks
In this section we will review the pertinent aspects of the group theoretic formalism used to compute correlation functions of primary operators in four dimensional conformal field theories. For a more detailed review of the formalism as applied to correlation functions of bosonic primary fields, the reader may consult [1]. Our 4D conventions and notation are those of [50], see the Appendix A for a brief overview.
Two-point functions
Consider 4D Minkowski space M 1,3 , parameterised by coordinates x m , where m = 0, 1, 2, 3 are Lorentz indices. Given two points, x 1 and x 2 , we can define the covariant two-point function Next, following Osborn and Petkou [1], we introduce the conformal inversion tensor, I mn , which is defined as follows: This object played a pivotal role in the construction of correlation functions in [1], as the full conformal group may be generated by considering Poincaré transformations supplemented by inversions. However, in the context of this work, we require an analogous operator for the spinor representation. Hence, we convert the vector two-point functions (2.1) into spinor notation using the conventions outlined in appendix A: In this form the two-point functions possess the following useful properties: Hence, we find (2.5) We also introduce the normalised two-point functions, denoted byx 12 , From here we can now construct an operator analogous to the conformal inversion tensor acting on the space of symmetric traceless tensors of arbitrary rank. Given a two-point function x, we define the operator The spinor indices may be raised and lowered using the standard conventions as follows: Now due to the property we have the following useful relations: The objects (2.7), (2.8) prove to be essential in the construction of correlation functions of primary operators with arbitrary spin. Indeed, the vector representation of the inversion tensor may be recovered in terms of the spinor two-point functions as follows: Now let Φ A be a primary field with dimension ∆, where A denotes a collection of Lorentz spinor indices. The two-point correlation function of Φ A and its conjugateΦĀ is fixed by conformal symmetry to the form where I is an appropriate representation of the inversion tensor and c is a constant complex parameter. The denominator of the two-point function is determined by the conformal dimension of Φ A , which guarantees that the correlation function transforms with the appropriate weight under scale transformations. For example, in the case of the fermionic current field Q α(2k)α , the two-point function is uniquely fixed to the following form: , (2.14) where in this case ∆(Q) is fixed by conservation of Q, (Q) at x 1 , (x 2 ). It is not too difficult to show that ∆(Q) = k + 5 2 .
Three-point functions
Given three distinct points in Minkowski space, x i , with i = 1, 2, 3, we define conformally covariant three-point functions in terms of the two-point functions as in [1] where (i, j, k) is a cyclic permutation of (1, 2, 3). For example we have X 32 m , (2.17b) and the differential identities The three-point functions also may be represented in spinor notation as follows: These objects satisfy properties similar to the two-point functions (2.4). Indeed, it is convenient to define the normalised three-point functionsX ij , and the inverses (X −1 ij ), (2.20) Now given an arbitrary three-point building block X, it is also useful to construct the following higher-spin operator: These operators have properties similar to the two-point higher-spin inversion operators (2.7), (2.8). There are also some useful algebraic identities relating the two-and threepoint functions at various points, such as I αα (X 12 ) = I αγ (x 13 )Īγ γ (x 12 ) I γα (x 23 ) ,Īα γ (x 13 ) I γγ (X 12 )Īγ α (x 13 ) =Īα α (X 32 ) .
(2.23) These identities (and cyclic permutations of them) are analogous to (2.17a), (2.17b), and also admit higher-spin generalisations, for examplē (2.24) In addition, similar to (2.18), there are also the following useful identities: These identities allow us to account for the fact that correlation functions of primary fields obey differential constraints which can arise due to conservation equations. Indeed, given a tensor field T A (X), there are the following differential identities which arise as a consequence of (2.25): (2.26b) Now concerning three-point correlation functions, let Φ, Ψ, Π be primary fields with scale dimensions ∆ 1 , ∆ 2 and ∆ 3 respectively. The three-point function may be constructed using the general ansatz where the tensor HĀ 1Ā2 A 3 encodes all information about the correlation function, and is highly constrained by the conformal symmetry as follows: (i) Under scale transformations of Minkowski space x m → x m = λ −2 x m , the threepoint building blocks transform as X m → X m = λ 2 X m . As a consequence, the correlation function transforms as which implies that H obeys the scaling property This guarantees that the correlation function transforms correctly under scale transformations.
(ii) If any of the fields Φ, Ψ, Π obey differential equations, such as conservation laws in the case of conserved current multiplets, then the tensor H is also constrained by differential equations. Such constraints may be derived with the aid of identities (2.26a), (2.26b).
(iii) If any (or all) of the operators Φ, Ψ, Π coincide, the correlation function possesses symmetries under permutations of spacetime points, e.g.
where (Φ) is the Grassmann parity of Φ. As a consequence, the tensor H obeys constraints which will be referred to as "point-switch identities". Similar relations may also be derived for two fields which are related by complex conjugation.
The constraints above fix the functional form of H (and therefore the correlation function) up to finitely many independent parameters. Hence, using the general formula (2.31), the problem of computing three-point correlation functions is reduced to deriving the general structure of the tensor H subject to the above constraints.
Comments regarding differential constraints
An important aspect of this construction which requires further elaboration is that it is sensitive to the configuration of the fields in the correlation function. Indeed, depending on the exact way in which one constructs the general ansatz (2.31), it can be difficult to impose conservation equations on one of the three fields due to a lack of useful identities such as (2.26a), (2.26b). To illustrate this more clearly, consider the following example; suppose we want to determine the solution for the correlation function All information about this correlation function is encoded in the tensor H, however, this particular formulation of the problem prevents us from imposing conservation on the field Π in a straightforward way. To rectify this issue we reformulate the ansatz with Π at the front (2.32) In this case, all information about this correlation function is now encoded in the tensor H, which is a completely different solution compared to H. Conservation on Π can now be imposed by treating x 3 as the first point with the aid of identities analogous to (2.25), (2.26a), (2.26b). What we now need is a simple equation relating the tensors H andH, which correspond to different representations of the same correlation function. If we have equality between the two ansatz above, after some manipulations we obtain the following relation: where is either 0 or 1 depending on the Grassmann parity of the fields Φ, Ψ, Π; since the overall sign is somewhat irrelevant for the purpose of this calculation we will absorb it into the overall sign ofH. In general, this equation is quite impractical to work with due to the presence of both two-and three-point functions, hence, further simplification is required. Let us now introduce some useful definitions; suppose H(X) (with indices suppressed) is composed out of a finite basis of linearly independent tensor structures P i (X), i.e H(X) = i a i P i (X) where a i are constant complex parameters. We definē H(X) = iā iPi (X), the conjugate of H, and also H c (X) = i a iPi (X), which we will call the complement of H. As a consequence of (2.23), the following relation holds: This equation is an extension of (2.14) in [1] to the spinor representation, and it allows us to construct an equation relating different representations of the same correlation function. After inverting this identity and substituting it directly into (2.33), we apply identities such as (2.23) to obtain an equation relating H c andH It is important to note that this is now an equation in terms of a single variable, X, which vastly simplifies the calculations. Indeed, onceH is obtained we can then impose conservation on Π as if it were located at the "first point". However, as we will see in the subsequent examples, this transformation is quite difficult to carry out for correlation functions of higher-spin primary operators due to the proliferation of tensor indices.
To summarise, in order to successfully impose all the relevant constraints on the fields in the correlator, we will adhere to the following three step approach: 1. Using ansatz (2.31), construct a solution for H that is consistent with the algebraic/tensorial symmetry properties of the fields Φ, Ψ and Π.
2. Impose conservation equations on the first and second point using identities (2.25), (2.26a) and (2.26b) to constrain the functional form of the tensor H.
3. Reformulate the correlation function using ansatz (2.32), which allows one to find an explicit relation forH in terms of H c . Conservation of Π may now be imposed as if it were located at the first point.
Generating function formalism
To study and impose constraints on correlation functions of primary fields with general spins it is often advantageous to use the formalism of generating functions to stream-line the calculations. Suppose we must analyse the constraints on a general spin-tensor represent sets of totally symmetric spinor indices associated with the fields at points x 1 , x 2 and x 3 respectively. We introduce sets of commuting auxiliary spinors for each point; U = {u,ū} at x 1 , V = {v,v} at x 2 , and W = {w,w} at x 3 , where the spinors satisfy Now if we define the objects then the generating polynomial for H is constructed as follows: There is in fact a one-to-one mapping between the space of symmetric traceless spin tensors and the polynomials constructed using the above method. The tensor H can then be extracted from the polynomial by acting on it with the following partial derivative operators: The tensor H is then extracted from the polynomial as follows: Let us point out that methods based on using auxiliary vectors/spinors to create a polynomial are widely used in the construction of correlation functions throughout the literature (see e.g. [19,27,[32][33][34]38]). However, usually the entire correlator is contracted with auxiliary variables and as a result one produces a polynomial depending on all three spacetime points and the auxiliary spinors. In our approach, however, we contract the auxiliary spinors with the tensor H A 1 A 2 A 3 (X), which depends on only a single variable.
Our approach proves to be essential in the construction of correlation functions of higher-spin operators. It also proves to be more computationally tractable, as the polynomial H, (2.38), is now constructed out of scalar combinations of X, and the auxiliary spinors U , V and W with the appropriate homogeneity. Such a polynomial can be constructed out of the following scalar basis structures: subject to cyclic permutations of linear dependence relations such as There can be more general linear dependence relations for more complicated combinations of the basis structures (2.41), however, such relations can be obtained computationally.
In general, it is a non-trivial technical problem to come up with an exhaustive list of possible solutions for the polynomial H for a given set of spins. Hence, let us introduce a more convenient labelling scheme for the building blocks (2.41) Now if we also define the objects then the generating function for the polynomial H(X; U, V, W ) may be defined as follows: where the non-negative integers, Γ = {k i ,k i , r i ,r i , s i }, i = 1, 2, 3, are solutions to the following linear system: where a I are a set of complex constants. Hence, constructing the most general ansatz for the generating polynomial H is now equivalent to finding all non-negative integer solutions Γ I of (2.46), where i 1 , i 2 , i 3 and j 1 , j 2 , j 3 are arbitrary non-negative integers. The solutions correspond to a linearly dependent basis of possible structures in which the polynomial H can be decomposed. Using computational methods, we can generate all possible solutions to (2.46) for fixed (and in some cases arbitrary) values of the spins.
In the remaining sections of this paper we will construct solutions for the three-point functions of the fermionic current field Q α(2k)α with the vector current and the energy momentum tensor using the formalism outlined above. We use a combination of the method of systematic decomposition and the generating function approach to reduce the number of possible linearly dependent structures in each case. We present most of our results in terms of the scalar basis structures (2.41), however, the generating function (2.45) underpins most of the calculations. general results of subsection 2.2 is where H is a homogeneous tensor field of degree q = 3 − 2(k + 5 2 ) = −2(k + 1). It is constrained as follows: ing blocks transform as X m → X m = λ 2 X m . As a consequence, the correlation function transforms as 2) which implies that H obeys the scaling property This guarantees that the correlation function transforms correctly under scale transformations.
(ii) The conservation of the fields Q at x 1 and x 2 imply the following constraints on the correlation function: Using identities (2.26a), (2.26b) we obtain the following differential constraints on the tensor H: where ∂α α X = (σ a )α α ∂ ∂X a . There is also a third constraint equation arising from conservation of V at x 3 , however, there are no identities analogous to (2.26a), (2.26b) that allow the partial derivative operator acting on x 3 to pass through the prefactor of (4.1), hence, we use the procedure outlined in subsection (2.3). First we construct an alternative ansatz with V at the front as follows: Since the correlation function possesses the following property: we can now computeH in terms of H. After some manipulations one finds the following relation: This is quite impractical to work with due to the presence of both two-point functions and three-point functions, therefore we will make use of the following relation derived from (2.33): After substituting this relation directly into (3.9), and making use of (2.23), we obtain the following equation: The equation relatingH to H c is now expressed in terms of a single variable, the building block vector X. Conservation on the third point is now equivalent to imposing the following constraint on the tensorH: The correlation function is also constrained by the following reality condition: which implies the following constraint on the tensor H: (3.14) Hence, we have to solve for the tensor H subject to the above constraints. This is technically quite a challenging problem due to the complicated index structure of the tensor H. Instead we will streamline the calculations by constructing a generating function as outlined in subsection 2.4. We introduce the commuting auxiliary spinors u,ū, v,v, w,w, which satisfy u 2 = 0,ū 2 = 0, etc, and define the generating function for H as follows: The tensor H is then obtained from the generating polynomial by acting on it with partial derivatives Again, the generating function approach simplifies the various algebraic and differential constraints on the tensor H. In particular, the differential constraints (3.5a) and (3.5b) become while the homogeneity and reality condition (3.14) become: Our task now is to construct the general solution for the polynomial H consistent with the above constraints.
The general expansion for the polynomial H is formed out of products of the basis objects introduced in (2.41). Let us start by decomposing the polynomial H, we have: where have used the fact that H is homogeneous degree 1 in both w andw. The vector object F is now homogeneous degree 0 in X, homogeneous degree 1 inū, v, and homogeneous degree 2k in u,v. It may be decomposed further by introducing the following basis vector structures: We then have where the F i are polynomials that are homogeneous degree 0 in X, with the appropriate homogeneity in u,ū, v,v. It is not too difficult to construct all possible polynomial structures for each Z i, αα : However, not all of these structures are linearly independent. In particular it may be shown that F αα (X; u,ū, v,v) = 0 for the choice a 2 = −a 1 , a 3 = a 1 , a i = 0 , i = 4, ... , 9 .
(3.23) Therefore we can construct a linearly independent basis of polynomial structures by removing the a 1 structure, which leaves us with 8 independent structures to consider. We now impose the differential constraints and point switch identities using Mathematica. After imposing (3.17a), we obtain the following k-dependent relations: in addition to a 5 = a 6 = 0. Next we impose (3.17b), from which we obtain Hence, the correlation function is determined up to three independent complex parameters, a 2 , a 3 and a 8 . We now must impose the reality condition (3.18b). Using Mathematica, we find that a 2 = iã 2 , a 3 = iã 3 , a 8 = iã 8 , whereã 2 ,ã 3 ,ã 8 are three real constant parameters.
It remains to demonstrate that this correlation function is conserved at x 3 in accordance with conservation of the vector current. First we compute the tensorH using (3.11). This may be written more compactly in the generating function formalism; to do this we introduce the following differential operators: The relation (3.11) is now equivalent tõ H(X; w,w, v,v, u,ū) = 1 (2k)! X 2k−1 (vX∂s) (∂sXv) 2k H c (X; u,ū, s,s, w,w) .
(3.27)
Conservation on the third point (3.12) is equivalent to imposing the following constraint onH: It may be shown using Mathematica that this is satisfied up to k = 4. Beyond k = 4 the calculations for (3.11) seem to become very computationally intensive; however, we have no reason to expect that the result will change for higher values of k. Hence, we are reasonably confident that Q QV is fixed up to three independent real parameters.
In this section we will compute the correlation function QQV . The ansatz for this correlator consistent with the general results of subsection 2.2 is where H is a homogeneous tensor field of degree q = 3 − 2(k + 5 2 ) = −2(k + 1). It is constrained as follows: (i) Under scale transformations of spacetime x m → x m = λ −2 x m the three-point building blocks transform as X m → X m = λ 2 X m . As a consequence, the correlation function transforms as
2) which implies that H obeys the scaling property
This guarantees that the correlation function transforms correctly under scale transformations.
(ii) The conservation of the fields Q at x 1 and x 2 imply the following constraints on the correlation function: Using identities (2.26a), (2.26b), we obtain the following differential constraints on the tensor H: There is also a third constraint equation arising from conservation of V at x 3 : Similar to the previous example, we use the procedure outlined in subsection 2.3 and find the following relation between H c andH: Conservation on the third point is now tantamount to imposing the constraint ∂γ γ XHγγ,ββ(2k),α(2k)α (X) = 0 . Hence, we have to solve for the tensor H subject to the above constraints. Analogous to the previous example in section 3, we streamline the calculations by constructing a generating function, which is defined as follows: (4.11) The tensor H is then extracted from the generating polynomial by acting on it with partial derivatives H αα(2k),ββ(2k),γγ (X) = ∂ ∂U αα(2k) ∂ ∂V ββ(2k) ∂ ∂W γγ H(X; u,ū, v,v, w,w) . (4.12) As will be seen shortly, the generating function approach simplifies the various algebraic and differential constraints on the tensor H. In particular, the differential constraints (4.5a) and (4.5b) become while the homogeneity and point switch constraints become: H(X; u,ū, v,v, w,w) = −H(−X; v,v, u,ū, w,w) . (4.14b) Our task is now to construct the general solution for the polynomial H consistent with the above constraints.
The general expansion for the polynomial H is then formed out of products of the basis objects above. Let us start by decomposing the polynomial H, we have: (4.15) where have used the fact that H is homogeneous degree 1 in both w andw. The vector object F is now homogeneous degree 0 in X, degree 1 in u, v, and degree 2k inū,v. It may be decomposed further using the structures defined in (3.20): where the F i are polynomials that are homogeneous degree 0 in X, with the appropriate homogeneity in u,ū, v,v. It is not too difficult to construct all possible polynomial structures for each Z i, αα , we find: However, not all of these structures are linearly independent. In particular it may be shown that F αα (X; u,ū, v,v) = 0 for the choice a 3 = −a 2 , a 4 = −a 1 + a 2 , a 5 = a 1 − a 2 , a 6 = a 1 − a 2 , a 7 = −a 1 + a 2 , (4.18) Therefore we can construct a linearly independent basis of polynomial structures by removing the a 1 and a 2 structures, hence, there are only 5 independent structures remaining. We now impose the differential constraints and point switch identities using Mathematica. After imposing (4.13a), we obtain the following k-dependent relations between the coefficients: At this stage only two independent coefficients remain. Next we impose (4.13b), from which we obtain (4.20) Hence, the correlation function is determined up to a single complex parameter, a 3 = a. However, it may be shown that this solution is not compatible with the point switch identity (4.14b), hence, this correlation vanishes in general.
Correlator
In this section we will compute the correlation function Q QT , where T is the energy momentum tensor T γ(2)γ(2) with scale dimension 4. The ansatz for this correlator consistent with the general results of subsection 2.2 is where H is a homogeneous tensor field of degree q = 4 − 2(k + 5 2 ) = −2k − 1. It is constrained as follows: ing blocks transform as X m → X m = λ 2 X m . As a consequence, the correlation function transforms as which implies that H obeys the scaling property This guarantees that the correlation function transforms correctly under scale transformations.
(ii) The conservation of the fields Q at x 1 and x 2 imply the following constraints on the correlation function: Using identities (2.26a), (2.26b) we obtain the following differential constraints on the tensor H: There is also a third constraint equation arising from conservation of V at x 3 , Using the same procedure as the previous examples, we construct an alternative ansatz for the correlation function as follows: ×H γ (2)γ (2),β β (2k),αα(2k) (X 32 ) .
(5.7)
Now due to the property × H α (2k)α ,β β (2k),γ (2)γ (2) (X 12 ) . (5.9) We now make use of the following identity derived from (2.34): After substituting this equation into (5.9), we obtain the relatioñ Conservation at x 3 is now equivalent to imposing the following constraint on the tensorH: The correlation function is also constrained by the reality condition (5.13) This implies the following constraint on the tensor H: (5.14) Hence, we have to solve for the tensor H subject to the above constraints. Analogous to the previous examples we streamline the calculations by constructing a generating function, which is defined as follows: The tensor H is then obtained from the generating polynomial by acting on it with partial derivatives Again, the generating function approach simplifies the various algebraic and differential constraints on the tensor H. In particular, the differential constraints (5.5a) and (5.5b) become while the homogeneity and point switch constraints become: H(λ 2 X; u,ū, v,v, w,w) = (λ 2 ) q H(X; u,ū, v,v, w,w) , (5.18a) H(X; u,ū, v,v, w,w) = −H(−X; v,v, u,ū, w,w) .
Let us now construct the general solution for the polynomial H consistent with the above constraints. We start by decomposing the polynomial H as follows: where have used the fact that H is homogeneous degree 1 in bothū and v. The vector object F is now homogeneous degree 0 in X, homogeneous degree 2 in w,w, and homogeneous degree 2k in u,v. It may be decomposed further by defining the following basis vectors: We then have where the F i are polynomials that are homogeneous degree 0 in X, with the appropriate homogeneity in u,v, w,w. The complete list of possible polynomial structures for each Z i, αα is: + a 3 (wXw)(uw)(vw)(uXv) 2k−1 + a 4 (wXv)(uXw)(uw)(vw)(uXv) 2k−2 + a 5 (wXv) 2 (uXw)(vw)(uXv) 2k−2 + a 6 (uXw)(wXw)(wXv)(uXv) 2k−1 , (5.22a) Z 2 structures: There are also the additional "higher spin" structures, which appear only for k > 1: Hence, we will need to treat the cases k = 1 and k > 1 separately. First we will consider k = 1, which corresponds to a field with the same properties as the supersymmetry current, Q α(2),α .
Analysis for k = 1
In this subsection we will determine the constraints on the coefficients for general k. First we must determine any linear dependence relations between the various polynomial structures. Using Mathematica it may be shown that F αα (X; u,v, w,w) = 0 for the following relations between the coefficients: a 3 = a 1 + a 2 − a 10 + a 11 , (5.24a) a 5 = −a 1 − a 4 + a 10 + a 12 , (5.24b) Therefore a linearly independent basis may be obtained by neglecting the structures corresponding to the coefficients a 1 , a 2 , a 4 , a 7 , a 10 , a 11 , a 12 , a 13 , a 16 . There are only nine structures remaining, corresponding to the coefficients a 3 , a 5 , a 6 , a 8 , a 9 , a 14 , a 15 , a 17 , a 18 respectively. Now that we have identified any possible linear dependence between the polynomial structures, we impose the differential constraints and point-switch identities using Mathematica. After imposing the conservation equations (5.17a) and (5.17b), we obtain the following relations between the coefficients: Hence, the differential constraints are sufficient to fix the correlation function up to four independent complex parameters, a 3 , a 5 , a 17 and a 18 . The next constraint to impose is the reality condition (5.18b), from which we determine that the remaining four parameters must be purely real.
Finally we must check that the correlation function satisfies the differential constraint (5.12) in accordance with conservation of the energy-momentum tensor. We begin by computingH using (5.11); in the generating function formalism this may be written as Conservation of the energy-momentum tensor at x 3 (5.12) is now equivalent to imposing the following differential constraint on the tensorH: At this point we set k = 1 and proceed with the analysis. Using Mathematica it may be shown that this constraint is automatically satisfied for the coefficient constraints above, hence, the correlation function Q QT is determined up to four independent real parameters.
Analysis for general k
Now let us carry out the analysis for general k; we must determine any linear dependence relations between the various polynomial structures. Indeed, we find that introducing the higher-spin contributions (5.23a) results in the following supplementary linear dependence relation for k > 1, i.e. F αα (X, u,v, w,w) = 0 for the coefficient relations Therefore the complete list of independent structures corresponds to the coefficients a 3 , a 5 , a 6 , a 8 , a 9 , a 14 , a 15 , a 17 , a 18 , a 21 . We now impose the differential constraints and point switch identities using Mathematica. After imposing the differential constraints arising from requiring conservation on the first and second point, that is (5.17a) and (5.17b) we obtain the k-dependent relations The remaining free coefficients are a 3 , a 5 , a 17 and a 21 ; the relations are also defined only for k > 1. Next we must impose the reality condition (5.18b), from which we find that the remaining coefficients must be purely real. Hence, we find that the correlation function is determined up to four independent real parameters.
Finally, we must impose the differential constraint on x 3 which arises due to conservation of the energy-momentum tensor, that is, (5.27). Indeed, we have shown using Mathematica that (5.27) is satisfied up to k = 4, for higher values of k the computations ofH seem to be beyond our computer power. However, we believe that the results will hold for higher values of k, so we can be reasonably confident that the correlation function is determined up to four independent real parameters for general k.
In this section we will compute the correlation function QQT . The ansatz for this correlator consistent with the general results of subsection 2.2 is where H is a homogeneous tensor field of degree q = 4 − 2(k + 5 2 ) = −2k − 1. It is constrained as follows: (i) Under scale transformations of spacetime x m → x m = λ −2 x m the three-point building blocks transform as X m → X m = λ 2 X m . As a consequence, the correlation function transforms as which implies that H obeys the scaling property This guarantees that the correlation function transforms correctly under conformal transformations.
. (6.9) This implies the following constraint on the tensor H: H αα(2k),ββ(2k),γ(2)γ(2) (X) = −H ββ(2k),αα(2k),γ(2)γ(2) (−X) . (6.10) Hence, we have to solve for the tensor H subject to the above constraints. Let us now streamline the calculations by constructing the generating function: The tensor H is then obtained from the generating polynomial by acting on it with partial derivatives as follows: Let us now convert our constraints on the tensor H to constraints on the generating function. In particular, the differential constraints (6.5a) and (6.5b) become H(X; u,ū, v,v, w,w) = 0 , (6.13b) while the homogeneity and point-switch constraints become: H(λ 2 X; u,ū, v,v, w,w) = (λ 2 ) q H(X; u,ū, v,v, w,w) , (6.14a) H(X; u,ū, v,v, w,w) = −H(−X; v,v, u,ū, w,w) . (6.14b) Our task is now to construct the general solution for the polynomial H consistent with the above constraints. The general expansion for H is formed out of products of the basis objects (2.41). Let us start by decomposing the polynomial H, we have: where have used the fact that H is homogeneous degree 1 in both u and v. The tensor F is now homogeneous degree 0 in X, homogeneous degree 2 in w,w, and homogeneous degree 2k inū,v. It may be decomposed into symmetric and anti-symmetric parts as follows: F αβ (X;ū,v, w,w) = ε αβ A(X;ū,v, w,w) + B (αβ) (X;ū,v, w,w) . (6.16) It is straightforward to identify the possible structures in the expansion for A. We find However, identifying all possible structures for the tensor B is more challenging. To this end we introduce a basis of spinor structures, Y i, α : From these basis spinors, we construct a set of symmetric objects, Y ij, αβ , defined as follows: These objects are symmetric in α, β, hence, they form a basis in which the tensor B may be decomposed. However, since these objects are also symmetric in i, j, only 10 of them are unique, therefore we form the list Z i, αβ out of the unique structures. We then have the decomposition where the polynomials B i are homogeneous degree 0 in X, with the appropriate homogeneity inū,v, w,w. We now construct all possible polynomial structures for each Z i, αβ : 21e) There are also additional structures that are defined only for k > 1. Such structures will be denoted byB.
Therefore we must analyse the k = 1 and k > 1 cases separately.
Analysis for k = 1
First we must determine any linear dependence relations between the various polynomial structures. In this case, since there are many structures, the linear dependence relations are rather complicated. For the A structures, we find A(X;ū,v, w,w) = 0 for the choice of coefficients a 4 = −a 1 + a 2 , a 5 = −a 1 − a 3 , a 6 = 2a 1 − a 2 + a 3 . (6.23) Hence, the structures corresponding to a 1 , a 2 and a 3 may be neglected, and we are left with only the structures with coefficients a 4 , a 5 , a 6 . Next we find linear dependence amongst the B structures, we find B (αβ) (X;ū,v, w,w) = 0 for the choice Therefore a linearly independent basis may be constructed out of the structures corresponding to the coefficients b 6 , b 7 , b 8 , b 9 , b 21 and b 22 . Overall there are nine independent structures to consider. We now impose the differential constraints and point switch identities using Mathematica. After imposing (6.5b),(6.13b) we obtain the following relations between the coefficients: Hence, the differential constraints fix the correlation function up to three parameters. Next we must impose the point switch identity (6.14b), from which we obtain a 5 = a 4 , hence, we are left with the free complex parameters a 4 and a 6 .
We must now impose (6.8) in accordance with conservation of the energy momentum tensor. First we computeH using (6.7), which in the generating function formalism may be written as H(X; w,w, v,v, u,ū) = 1 (2k)! X 2k−3 (vX∂s) (∂sXv) 2k H c (X; u,ū, s,s, w,w) , (6.26) while the differential constraint (6.8) is equivalent to (X; w,w, v,v, u,ū) = 0 . (6.27) At this point we can freely set k = 1 and check whether our solution is consistent with conservation at x 3 . Using Mathematica, it may be shown that (6.27) is satisfied provided that a 6 = − 12 5 a 4 , hence, the correlation function QQT is determined up to a single complex parameter.
Discussion on supersymmetry
In this section we will concentrate on the case k = 1, which corresponds to a "supersymmetry-like" current Q αβα = (σ m ) αα Q m,β of dimension-7 2 satisfying the conservation equation ∂ αα Q αβα = 0 . (7.1) However, our analysis in the previous sections did not assume supersymmetry. The question that naturally arises is whether the "supersymmetry-like" current actually is the supersymmetry current. That is, whether a conformal field theory possessing a conserved fermionic current of spin-3 2 is superconformal. In any supersymmetric field theory the supersymmetry current is a component of the supercurrent J αα (z), which also contains the energy-momentum tensor. As was explained in the introduction, this implies that the three-point functions QQT and Q QT must be contained in the three-point function of the supercurrent JJJ . It is known that the general form of JJJ is fixed by superconformal symmetry up to two independent structures [14]. Hence, this implies that in any superconformal field theory, QQT and Q QT must also be fixed up to at most two independent structures. Moreover, the threepoint function QQT must actually vanish. Indeed, in a supersymmetric theory Q carries an R-symmetry charge and, hence, the entire correlator QQT carries an R-symmetry charge. However, by performing a simple change of variables in the path integral it then follows that QQT = 0. In addition, our analysis in Section 6 showed that, in general, conformal symmetry fixes QQT up one overall parameter, which is inconsistent with supersymmetry. We also found in Section 5 that the three-point function Q QT is fixed up to four rather than two independent parameters, which, in general, is also inconsistent with the general form of JJJ .
Similarly, we can examine the three-point functions Q QV and QQV studied in Sections 3 and 4 respectively. In supersymmetric theories, the vector current V m belongs to the flavour current multiplet L(z). Hence, the correlation functions Q QV and QQV are contained in the three-point function JJL . It is known [14] that JJL is fixed by superconformal symmetry up to an overall real coefficient. Hence, Q QV must also be fixed up to an overall coefficient. As for QQV , it must vanish just like QQT . However, our analysis in Sections 3, 4 showed that Q QV is fixed up to three independent coefficients while QQV vanishes. The result for Q QV is therefore inconsistent with the general form of JJL . define: (σ m )α α ≡ εαβε αβ (σ m ) ββ . (A.5) It can be shown that the σ-matrices possess the following useful properties: (σ mσn + σ nσm ) α β = −2η mn δ β α , (A.6) (σ m σ n +σ n σ m )αβ = −2η mn δα β , (A.7) Tr(σ mσn ) = −2η mn , (A.8) (σ m ) αα (σ m )β β = −2δα α δβα . (A.9) The σ-matrices are then used to convert spacetime indices into spinor ones and vice versa according to the following rules: | 2022-04-12T01:16:20.238Z | 2022-04-11T00:00:00.000 | {
"year": 2022,
"sha1": "aefe5296dd59c33a7e0dc60c8991b970785fb55c",
"oa_license": "CCBY",
"oa_url": "http://link.aps.org/pdf/10.1103/PhysRevD.105.125004",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "aefe5296dd59c33a7e0dc60c8991b970785fb55c",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
53047303 | pes2o/s2orc | v3-fos-license | Dynamic Changes in the Global MicroRNAome and Transcriptome Identify Key Nodes Associated With Ovarian Development in Chickens
The analysis of gene expression patterns during ovarian follicle development will advance our understanding of avian reproductive physiology and make it possible to improve laying performance. To gain insight into the molecular regulation of ovarian development, a systematic profiling of miRNAs and mRNAs at four key stages was conducted, using ovarian tissues from hens at 60 days of age (A), 100 days (B), 140 days-not yet laying (C), and 140 days-laying (D). Comparisons of consecutive stages yielded 73 differentially expressed miRNAs (DEMs) (14 for B vs. A, 8 for C vs. B, and 51 for D vs. C) and 2596 differentially expressed genes (DEGs) (51 for B vs. A, 20 for C vs. B, and 2579 for D vs. C). In addition, 174 DEMs (22 for C vs. A, 74 for D vs. A, and 78 for D vs. B) and 3205 DEGs (118 for C vs. A, 2284 for D vs. A, and 2882 for D vs. B) were identified between nonconsecutive stages. Some DEGs are involved in the Wnt and TGF-beta signaling pathways, which are known to affect ovarian development and ovulation. An integrative analysis of the miRNA and mRNA profiles identified 3166 putative miRNA-mRNA regulatory pairs containing 84 DEMs and 1047 DEGs. Functional annotation of the networks provides strong evidence that the miRNA regulatory networks may play vital roles in ovarian development and ovulation. Ten DEMs and 10 genes were validated by real-time quantitative PCR. The candidate miRNA-mRNA pairs gga-miR-200a-3p-SFRP4, gga-miR-101-3p-BMP5, gga-miR-32-5p-FZD4, and gga-miR-458b-5p-CTNNB1 potentially associated with ovarian development.
INTRODUCTION
Decades of traditional selective breeding by the poultry industry have resulted in marked improvement in commercial egg-laying breeds . The recent development of highthroughput genotyping platforms is likely to enable additional gains in egg production traits through molecular breeding, especially in the indigenous chicken breeds found in developing countries.
Ovarian development and folliculogenesis in chicken is a complex and highly coordinated process, resulting in the maturation of oocytes and the differentiation and proliferation of granulosa and theca cells (Qin et al., 2015;Lyu et al., 2016). Since this biological program is expected to exert a major influence on egg production traits, the cell signaling and associated transcriptional mechanisms that regulate ovarian development are of particular interest. A wide variety of local intraovarian molecular regulation factors participate in folliculogenesis, growth, and development of the ovarian follicles, such as BMP15, STAR, CCND2, CYP11A1, SAV1, and GDF9. Members of the glycoprotein hormone family of gonadotropins are also involved, such as FSH and FSHR (Dong et al., 1996;Bentsi-Barnes et al., 2010;Persani et al., 2014;Zhu et al., 2015). Cell signaling systems like the Hippo/MST signaling pathway are part of the developmental process as well . During the laying period, follicles are recruited into the preovulatory hierarchy from a cohort of small (6-8mm) yellow follicles approximately once a day, a process termed follicle selection (Johnson, 2015). The selected follicle develops rapidly through five (F5 to F1) stage prior to ovulation (Johnson and Woods, 2009).
MicroRNAs (miRNAs) function as reversible regulators of transcription and translation (Cannell et al., 2007) and contribute to at least 60% of the human transcriptome. (Di and Croce, 2013). They have been implicated in a wide range of biological and metabolic processes, including organogenesis, development, cell proliferation, differentiation, and apoptosis, as well as in many diseases (Bartel, 2004;Alvarezgarcia and Miska, 2005;Kloosterman and Plasterk, 2006). Because ovarian development and egg production are a complex traits regulated by many genes, it would be not surprising if they are also regulated through specific miRNA-mRNA interactions. Although some transcriptome studies have focused on mRNA and miRNA expression, the molecular mechanisms involved in the regulation of ovarian development at different stages remains unclear (Kang et al., 2013;Zhu et al., 2015;Wang et al., 2017). Highthroughput sequencing of mRNAs and miRNAs together offers a strategy to identify additional genes involved in ovarian development and to understand their interactions with miRNAs in detail.
The present study focused on identifying DEGs and miRNAs at four critical time points during ovarian development in Jining Bairi hens, an indigenous chicken breed that is well known for early sexual maturity at around 100 days. Using the Illumina MiSeq/HiSeq TM 2000 high-throughput sequencing platform, we analyzed miRNA and gene expression in ovarian tissues at ages of 60, 100, 140 days (but not yet laying eggs), and 140 days (laying eggs). A wide variety of ovarian development-related genes were identified. After comparing expression among the four groups, stage-specific and DEGs were matched to corresponding regulatory pathways by GO and KEGG analyses. mRNA and miRNA interaction networks were also predicted based on the integrated analysis of mRNA and miRNA expression profiles.
miRNA Expression Profiling and Screening for Differentially Expressed miRNAs
In order to visualize the pattern of miRNA expression during ovarian development, we conducted a time-course analysis using four samples, obtained at 60 days (A), 100 days (B), 140 days (egg laying has not yet begun; C), and 140 days (egg laying has commenced; D). 8,275,891, 7,091,297, 9,022,100, and 10,724,020 raw reads were collected from the four libraries, respectively. After discarding low quality reads, reads containing 3 and 5 adaptors, and reads with other defects, over 97% of the sRNA reads were retained in each of the data sets (Supplementary Table S1). Length distributions of the cleaned reads varied slightly (Supplementary Figure S1), with 21-24 nt and 22 nt as the most abundant lengths in all libraries, consistent with previous results (Kang et al., 2013).
Clean reads were mapped to the NCBI chicken reference genome (galGal4; NCBI) using BowTie. The fraction of mapped reads varied from approximately 46 to 55% per library (Supplementary Table S2 and Supplementary Figure S2). After redundant reads (rRNA, tRNA, snRNA, and snoRNA) were excluded, 507 known miRNAs and 35 novel miRNA candidates were identified by alignment to the miRBase and Rfam databases. 399,383,407,and 372 known miRNAs,and 24,21,26,and 22 novel miRNAs, were detected in the stage A, B, C, and D libraries, respectively (Supplementary Tables S3, S4).
To evaluate the development-dependent miRNA activities, we performed a time course differential miRNA expression analysis by comparing temporally adjacent stages after normalizing the mapped reads (Wang et al., 2010) and applying two filtering criteria, q-value < 0.01 and |log 2 (fold change)| > 1 (Supplementary Figure S3). The results are summarized in Figure 1A and detailed in Supplementary Table S5. Figure 1C) and contained the most DEMs when compared with other samples (Figure 1A).
24 miRNAs with sustained increasing expression and 38 miRNAs with sustained decreasing expression in the four consecutive samples were also identified using standardized reads data ( Figure 1D).
Expression Profiling of Ovarian Development-Dependent Genes
To correlate miRNA expression with gene expression, we analyzed transcriptional changes in the ovaries at four developmental stages in chickens using high-throughput mRNA sequencing (for bulk sequencing statistics, quality control, and assembly results, see Supplementary Table S6 and Supplementary Figure S4). Hierarchical cluster analysis shows that the gene expression profiles differ greatly (Figure 2A), with dramatic mRNA expression changes accompanying the transition to egg laying (stage D). Notably, a very significant proportion of DEGs at all four developmental stages are novel. As was the case for the miRNAs, comparisons between nonsequential developmental stages revealed larger numbers of differently expressed mRNAs (3205) than sequential pairwise comparisons (2596) ( Figure 2B and Supplementary Figure S5). Comparisons between the four consecutive time points (B vs. A, C vs. B, and D vs. C) reveal that only 1 differentially expressed mRNAs was common to all three comparisons ( Figure 2C, top). In contrast, 45 differentially expressed mRNAs were common to the three nonconsecutive time point comparisons (C vs. A, D vs. A, and D vs. B) ( Figure 2C, middle), similar to the bias observed when miRNA expression was compared between time points (cf. Figure 1B). 1702 DEGs were in common in comparisons between sample D vs. A, B, and C ( Figure 2C bottom). Of these, some were well-characterized genes involved in follicular development, such as WNT4 (Hernandez-Gonzalez et al., 2006), FZD1 (Hsieh et al., 2002), and FZD4 (Ricken et al., 2002). 133 genes with sustained increased expression, and 553 genes with sustained decreased expression, were identified ( Figure 2D and Supplementary Table S7). A complete list of differently expressed mRNAs is provided in Supplementary Table S8.
To obtain a better understanding of DEG functional roles, we performed a gene over-representation test (Boyle et al., 2004) using KEGG and GO for biological theme comparison (Figure 3). Several follicular growth-and ovulation-related pathways and GO terms were identified. For example, expression of components of the Wnt signaling pathway (gga04310) (Boyer et al., 2010), the TGF-beta signaling pathway (gga04350) (Fan et al., 2010), and the Wnt-protein binding biological process (GO: 0017147) increased significantly in stage D chickens. Other pathways such as Ribosome (gga03010), Focal adhesion (gga04510) and ECM-receptor (gga04512) were also enriched in genes with increased expression at stage D, suggesting that these pathways function in follicle development. In order to categorize the genes whose expression changed after the onset of egg laying (stage D), we conducted GO and KEGG analyses for the 1702 DEGs common to stage D and the other stages. Altogether, 91 significant GO terms and 13 pathways were associated with follicular development, including the Wnt signaling pathway (gga04310) and TGF-beta signaling pathway (gga04350) (Figures 4A,B and Supplementary Table S9). To summarize and visualize the findings, we constructed a graph using the GO biological process terms. As shown in Figure 4C, genes are primarily associated with the GO_BP terms translation and system development. We also analyzed complex situations in which a gene belongs to multiple annotation categories (Figures 4D,E). The results show that some genes in the Focal adhesion (gga04510) pathway also participate in the regulation of the actin cytoskeleton (gga04810) and the ECMreceptor interaction process (gga04512). Some genes in the Wnt signaling pathway (gga04310) were also found in Melanogenesis (gga04916) and the TGF-beta signaling pathway (gga04350). Overall, the transition to egg laying has a profound impact on gene expression in the chicken ovary.
Identification of miRNA-mRNA Regulatory Interactions Associated With Ovarian Development
The integrated analysis of miRNA and mRNA expression profiles should make it possible to identify functional miRNA-mRNA interaction pairs involved in specific biological processes. We applied a multi-step approach to identify and characterize the dynamically co-regulated miRNA-mRNA network related to ovarian development. First, DEMs and their potential targets were investigated based on sequence complementarity and the free energy of the predicted RNA duplex. Since most miRNAs negatively regulate the expression of their target mRNAs by RNA cleavage, we then searched for negative relationships between the levels of miRNAs and their predicted targets, as most studies have done Zhang et al., 2016). However, these statistical relationships are not necessarily causal, and can also be the result of the mRNA regulating the miRNA, or a third molecule regulating both the miRNA and the mRNA. To tackle this problem, we applied the R script miRCausality, which implements the IDA method to capture causal regulatory relationships from expression data (Le et al., 2013). After combining pairs having negatively related expression levels with the regulatory effects calculated by miRCausality (ef < 0.9), we identified 3166 putative miRNA-mRNA regulatory pairs containing 84 DEMs and 1047 DEGs (Supplementary Table S10).
To assess the functions of the putative miRNA targets, we conducted a GO enrichment analysis. The results showed that the targets are associated with cell development and tissue development (Figure 5A), similar to the results obtained in the whole transcriptome GO analysis, implying that these miRNAs define a core regulatory network for ovarian development. Key miRNAs were used to nucleate the ovarian development miRNA/mRNA interacting networks shown in Figures 5B,C, and the topologies were analyzed using the NetworkAnalyzer Cytoscape plugin (see Supplementary Table S11).
DISCUSSION
Efficient ovarian development and ovulation is necessary for high-volume commercial egg production. Despite the identification of many genes and regulatory factors related to folliculogenesis and ovarian development, the molecular mechanisms and pathways underlying the transition from one developmental stage to the next remain unclear. Elucidating these will not only help us understand the program of ovarian follicular development, but will also provide better tools for chicken breeding.
The Jining Bairi chicken, an indigenous Chinese chicken breed, reaches sexual maturity more rapidly and lays earlier than commercial laying lines. Here, we present a systematic study of miRNA and mRNA profiles to analyze ovarian development in this breed. Four developmental stages are represented, using ovary tissue samples from sexually immature (60 days; stage A) (100 days; stage B), pre-laying (140 days; stage C), and laying/mature (140 days; stage D) animals.
Our study focused on these miRNAs and their targets, BMP5, SFRP4, FZD4, and CTNNB1. The functions of these genes within the context of the Wnt/TGF-beta pathway are summarized in Figure 9. Wnt signaling first caught the attention of ovarian biologists after a landmark study by Vainio et al. (1999), which demonstrated that WNT4 null mice lose most of their oocyte reserves in the days prior to birth. Subsequently, WNTs and Wnt signaling pathway components were found in the postnatal ovary, where many are differentially expressed throughout follicle development, suggesting potential roles in processes such as follicle formation, growth, and ovulation/luteinization (Hsieh et al., 2005;Parma et al., 2006;Chassot et al., 2008;Harwood et al., 2008;Hernandez et al., 2009). Many components of the Wnt/CTNNB1 pathway (one of three Wnt signaling pathways) are expressed during follicle development (Ricken et al., 2002;Hernandez-Gonzalez et al., 2006;Harwood et al., 2008;Wang et al., 2009). As shown in Figure 9, the linchpin of the canonical pathway is CTNNB1, a subunit within a large multiprotein complex that also includes the scaffold proteins APC and AXIN FIGURE 6 | Comparison of expression levels for 10 miRNAs and 10 mRNAs determined using qRT-PCR and RNA-seq. Four samples at each stage were pooled for RNA-seq and four chickens from an independent cohort of animals were used for qRT-PCR. qRT-PCR values are shown as mean ± SD of four measurements, using U6 snRNA and GAPDH as internal standards. (Giles et al., 2003;Logan and Nusse, 2004). When bound to these proteins, CTNNB1 is phosphorylated by other components of the complex, resulting in subsequent ubiquitination and degradation ( Figure 9I). The Wnt/CTNNB1 signaling pathway is activated by the binding of a WNT to the FZD and LRP co-receptors (Figure 9II), engendering the destruction of the complex and allowing CTNNB1 to escape and translocate to the nucleus. In granulosa cells, CTNNB1 enhances the expression of the CYP19A1 gene, the key mediator of estrogen biosynthesis, through direct interaction with the transcription factor NR5A1 (Parakh et al., 2006).
Crosstalk between the Wnt signaling pathway and TGFbeta family occurs at multiple points. Wnt signaling inhibits phosphorylation of a multiprotein complex (Figure 9I), leading to Smad stabilization (Fuentealba et al., 2007). In response to stimulation by Wnt, the canonical Wnt and Smad pathways can synergize to activate transcription of target genes. Smad3 facilitates CTNNB1 nuclear translocation and coordinates with CTNNB1 at specific promoter sequences to regulate gene expression. Other components of the TGF-beta pathway, including Smurf1, Smurf2, and Smad7, modulate the Wnt signaling pathway (Luo, 2016). While the details of the canonical Wnt signaling and TGF-beta pathways are well understood in many cell types, whether the signal occurs in vivo as described, and whether these pathways are regulated by miRNAs during ovarian development, remain unresolved questions.
Our study predicted a negative relationship between gga-miR-200a-3p and SFRP4. SFRP4 is a modulator of Wnt signaling that functions by binding WNTs directly and preventing their interaction with FZDs ( Figure 9I). Based on our observations, we hypothesize that decreased levels of gga-miR-200a-3p in stage D ovaries enhance the expression of SFRP4. Although the role of SFRP4 in the ovary is unclear, it may be involved in modulating follicle cell survival (Drake et al., 2003). Similarly, levels of gga-miR-32-5p and gga-miR-458b-5p decreased dramatically in stage D, consistent with the increased expression of FZD4 and CTNNB1. FZD4 has been detected throughout follicle development (Hsieh et al., 2002;Ricken et al., 2002). In addition, expression of the putative gga-miR-101-3p target BMP5 increased in stage D ovaries. In this case, decreasing miRNA levels may help boost the Wnt signaling pathway. The increased expression of LRP and DKK favors this model, and further examination of these miRNA/mRNA pairs is warranted.
To conclude, deep sequencing of the transcriptome and microRNAome of the chicken ovary at four developmental stages reveals significant shifts in miRNA and mRNA expression. We identified 3166 putative miRNA-mRNA regulatory pairs containing 84 DEMs and 1047 DEGs. These are involved in various processes, including the Wnt signaling and the TGF-beta pathways. Understanding their relevance to ovarian development will be the focus of future research.
Animals
One hundred one-day-old Jining Bairi chickens were obtained from Jining Datang Chicken Breeding Co., Ltd. The chickens were raised under uniform environmental conditions at the conservation farm of the Shandong Agricultural University. Randomly selected individuals were sacrificed at 60 days (stage A), 100 days (stage B), 140 days but not yet laying eggs (stage C), and 140 days with egg laying having commenced (stage D). Eight chickens were sacrificed at each stage, of which four were used for miRNA and mRNA Illumina sequencing, and four for validation of the high-throughput sequencing data. Whole ovaries from stage A, B and C hens were collected. For stage D hens, follicles of different sizes, F1-F6 (10-34 mm in diameter), Y (yellow follicle, 6-8 mm in diameter), and W (white follicles, 2-4 mm in diameter), as well as the remaining ovarian tissues were collected. For large follicles, yolks were carefully removed. After being quick-frozen in liquid nitrogen, these tissues were ground into powder for isolation of total RNA.
Library Construction and Deep Sequencing
Eight RNA libraries (four for sRNAs, and four for mRNAs) were constructed using pooled RNA from each stage (A, B, C, and D). For all libraries, total RNA (3.0 µg for each sample) was extracted using TRIzol Reagent (Invitrogen, Carlsbad, United States) according to the manufacturer's protocol. The concentration and quality of RNA were determined using a NanoPhotometer R spectrophotometer (IMPLEN, CA, United States) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, United States) (Supplementary Figure S6).
Identification of miRNAs and mRNAs
After trimming adaptor sequences and removing contaminated reads, clean miRNA reads classified using computational methods. In brief, the clean reads were mapped to the NCBI chicken genome galGal4 with BowTie (Langmead et al., 2009). Reads corresponding to rRNA, tRNA, snRNA, scRNA, and snoRNA, identified using RepeatMasker and the Rfam database, were discarded. Novel miRNAs were predicted using miRDeep2 (Friedlander et al., 2012) and miREvo (Wen et al., 2012). Normalized miRNA reads were quantitated as TPM (Zhou et al., 2010). mRNA sequence reads were aligned to the chicken genome assembly version galGal4 using TopHat v2.0.9 with default parameters. Sequence segments were spliced and annotated using Cufflinks v2.1.1. Gene expression was quantitated as RPKM using HTSeq v0.5.4p3 and the Ensemble database as a reference (Mortazavi et al., 2008). miRNA and mRNA raw counts were further normalized as TMM (trimmed Mean of M-values). DE analysis of miRNA and mRNA sequence data was performed with the Bioconductor package DEGseq R (Wang et al., 2010), using parameters P < 0.01 and |log 2 (fold change)| > 1.
Quantitative Real-Time PCR
Total RNA was extracted using TRIzol Reagent and reverse transcribed using the One Step PrimeScript R miRNA cDNA Synthesis Kit (Tiangen Biotech Co., China) according to the manufacturer's instructions. The quality of RNA samples for qRT-PCR were detected by 1% agarose gel electrophoresis (Supplementary Figure S7). Real-time quantitative PCR was performed using an Mx3000p TM SYBR R Green Real-time quantitative PCR Analyzer (Stratagene, United States). The primer sequences for qRT-PCR are shown in Supplementary Table S12. The relative amounts of miRNA and mRNA were normalized against U6 snRNA and GAPDH, and the fold change for each miRNA and mRNA was calculated using the 2 − Ct method.
Bioinformatic Analyses
GO term analyses and KEGG enrichment analyses for DEGs and target DEGs were performed using the R packages "clusterProfiler" (Yu et al., 2012) and "GOstats" (Falcon and Gentleman, 2007), with p-values calculated using right-sided hypergeometric tests. To prevent high false discovery rate (FDR) in multiple testing, q-values were also estimated for FDR control (Storey, 2003). Figures were prepared using the R packages "ggplot2" (Wilkinson, 2011) and'UpSetR' (Gehlenborg, 2016).
Identification of miRNA-mRNA regulatory relationships was based on computational target predictions and negative regulation relationships. TargetScan and miRanda were used for computational target prediction. An R script (miRCausality.R) was used to identify miRNA-mRNA regulatory relationships (Le et al., 2013). The method learns a causal structure from expression data, and applies do-calculus to infer regulatory effects (reported over the range −1 to 1). We calculated pairwise effects between each DE miRNA and mRNA based on their expression across all samples. A regulatory effect less than −0.9 was considered to represent a negative regulatory relationship. Cytoscape was used to visualize and analyze the miRNA-mRNA networks, and network topology was analyzed using the NetworkAnalyzer Cytoscape plugin (Assenov et al., 2008).
ETHICS STATEMENT
This study was carried out in accordance with the recommendations of Guidelines for Experimental Animals established by the Ministry of Science and Technology (Beijing, China). The protocol was approved by the Institutional Animal Care and Use Ethics Committee of Shandong Agricultural University.
AVAILABILITY OF DATA
The datasets for this study can be found in the NCBI SRA database under accession number SRP130184 (https://www.ncbi. nlm.nih.gov/sra/SRP130184).
AUTHOR CONTRIBUTIONS
HT, QZ, WW, and KW conceived this study and carried out the data analyses and wrote the manuscript. LK and SS performed the library construction and Illumina sequencing. MJ conducted the experimental validation. All authors reviewed and approved the final manuscript. | 2018-10-23T13:04:15.186Z | 2018-10-23T00:00:00.000 | {
"year": 2018,
"sha1": "b9489540598780b6fce49defacc67590829cfcb2",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fgene.2018.00491/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "b9489540598780b6fce49defacc67590829cfcb2",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
52080950 | pes2o/s2orc | v3-fos-license | Successful systemic thrombolytic therapy for massive pulmonary embolism in a patient with breast cancer, brain metastasis, and thrombocytopenia: A case report
Key Clinical Message A female with massive PTE and absolute contraindication for thrombolytic did not meet guidelines due to unavailability of catheter or surgical embolectomy and giving thrombolytic as a last resort to save a life. In such cases, physicians should consider to act outside of the guidelines to save a life.
| INTRODUCTION
Venous thromboembolism (VTE) is a common complication of cancer and is a major cause of morbidity and one of the leading causes of mortality in terminal cancer patients.
Treatment of thromboembolic disease especially lifethreatening conditions such as massive pulmonary embolism in cancer patients is challenging.
It is common for clinicians to be faced with the dilemma of how to manage cancer patients with massive pulmonary embolism in life-threatening settings. In the present case, the problem was further complicated by brain metastasis and thrombocytopenia both of which have a dramatic impact in the risk of intracranial hemorrhage after thrombolysis.
| CLINICAL CASE REPORT
A 65-year-old female, known case of metastatic breast cancer with brain involvement, presented to our emergency department because of dyspnea, tachypnea, tachycardia and hypotension since 2 hours ago.
She had history of triple positive (ER+, PR+, HER2+) breast cancer and left radical mastectomy 15 years ago after which systemic chemotherapy and hormone therapy (Letrozole/Tamoxifen) was implemented. Three years before current presentation, bilateral hystero-salpingoovoforectomy, was performed. About 2 months ago, routine clinical examination by her oncologists revealed left axillary lymphadenopathy (2 × 3 cm) with fixed, firm and nontender node and an additional fixed mass (1 × 3 cm) in the anterior left forearm. Both lesions were excisionally biopsied. Further Imaging showed bilateral hilar lymphadenopathy and multiple metastatic lesions in the liver & abdomen and chest. PET scan showed 2 metastatic lesions in brain.
Pathologic examination of excised node revealed metastatic breast cancer (ER+/HER2+) and salvage chemotherapy with Paclitaxel 100 mg and Gemcitabin 1000 mg weekly plus Herceptin (Trastuzumab) 440 mg and Perjecta (Pertuzumab) 14 mg/kg tri weekly was initiated. One day after the first dose of Herceptin and Perjecta and the fourth dose of Gemcitabin and Paclitaxel, the right lower extremity swelling developed and subsequent Compression ultrasound and Doppler studies revealed acute extensive thrombus formation in the superficial femoral vein of the right leg. The patient was admitted and anticoagulation therapy with LMVH (60 mg BID) was initiated under the supervision of an oncologist at a private hospital. Preliminary laboratory tests revealed mild pancytopenia with a platelet count of 60 000 that was attributed to recent chemotherapy.
Two days after anticoagulation initiation she was discharged home.
Shortly after discharge, she presented to our emergency department with complaint of severe shortness of breath beginning abruptly 2 hours ago.
Her vital signs upon arrival to the emergency department showed a systolic blood pressure of 70 mm Hg, a heart rate of 155 beats per minute, a respiratory rate of 40 per minute & oxygen saturation of 70% in ambient air. She was apparently pale, agitated, diaphoretic and unable to speak in full sentence. Jugular veins were distended and cardiac exam showed moderate tachycardia and a right ventricular heave.
Pulmonary examination was unremarkable except for decreased breath sounds at the mid-zone of the left lung and the base of right lung.
Radial pulses were very weak bilaterally with no palpable pulse in lower extremities. The initial electrocardiogram showed sinus tachycardia at a rate of 135 per minute, right bundle branch block with a QRS duration of 130 ms and right axis deviation of 110 degrees ( Figure 1).
Immediately a 500 cc bolus of normal saline was administered and norepinephrine was initiated as vasopressor of choice.
Pulmonary CT angiography (CTA) confirmed large bilateral pulmonary thromboemboli with radiological evidence of RV strain (Figure 2).
According to hemodynamic compromise resulting from the massive pulmonary embolism and absolute contraindication for thrombolytic therapy due to brain metastases and concurrent thrombocytopenia, she became candidate for emergent catheter or surgical embolectomy. Surgical consult with cardiothoracic surgeon was requested but due to critical condition of the patient complicated by refractory hypoxia and unstable hemodynamic parameters surgery and anesthesiology team declared the patient as inoperable and there was no equipment and specialist for catheter embolectomy in our center.
Ultimately the critical decision was made, despite absolute contraindications, to treat the patient with systemic thrombolysis. Alteplase was selected as thrombolytic of choice and was prescribed according to approved dosage for massive PTE (100 mg infusion over a 2 hours' period).
Approximately 1 hour after termination of thrombolytic therapy, the norepinephrine was discontinued. Blood pressure returned to 110/70 mm Hg, with a heart rate of 100 beats per minute and a respiratory rate of 20 per minute with an oxygen saturation of 90% in ambient air. Electrocardiogram now shows sinus tachycardia at a rate of 100 per minute, normal QRS duration and normal axis, Figure 3.
Heparin infusion was initiated with half-dose because of concurrent thrombocytopenia.
After 30 hours, heparin infusion stopped due to a drop in platelet count (platelet count: 23 000/mm 3 ), consulted with hematologist colleagues, they believed it was due to recent chemotherapy, so 10 units of platelets were transfused. One day later platelet count reached 123 000/mm 3 , heparin infusion was resumed, and after 24 hours it changed to a therapeutic dose of subcutaneous enoxaparin (LMWH with a dose of 60 mg twice a day). The patient continued to improve clinically and was discharged from hospital after 5 days with LMWH. Last echocardiography performed immediately before discharge demonstrated mild RV enlargement & dysfunction and pulmonary artery pressure of 20-25 mm Hg. One week after discharge, laboratory findings showed Hemoglobin: 11.5 (g/dL), Platelet count: 227/ mm.
| DISCUSSION
Venous thromboembolism (pulmonary embolism and deep venous thrombosis) is a common complication of various cancers and their treatments. 1,2 Malignancies are associated with a 4-fold increase in VTE incidence 3 and VTE affects up to 20% of cancer patients, such that it is one of the major causes of morbidity and mortality in these population. 4 Metastatic disease at the time of diagnosis is considered to be the strongest predictor of VTE within the first year of F I G U R E 2 Pulmonary CT angiography demonstrates bilateral pulmonary thromboemboli with RV strain diagnosis and is associated with a 1.4-21.5-fold higher risk of VTE according to the type of cancer. [5][6][7][8][9][10] Cancer-related VTE is associated with higher rates of mortality or morbidity, in part due to increased bleeding complications imposed by anticoagulation/fibrinolytic therapy compared with VTE in patients without cancer 4,11 it is also a marker of more advanced disease and portends a poor prognosis. 12 Metastatic brain lesions are more prone to spontaneous intracranial hemorrhage compared with primary brain tumors. 13 However, the risk of intracranial hemorrhage from breast cancers brain metastases is relatively low (1%-5%). 14,15 In recent decades, owing to striking developments in the field of treatment of metastatic breast cancer, long-term survival can be achieved in many patients. 16 Furthermore, the increased prevalence of brain metastases is becoming a major impediment to improve quality of life for many breast cancer patients.
International society guidelines addressing management of pulmonary embolism, generally propose 2 viable options in the management of massive PTE: (i) thrombolysis, (ii) surgical embolectomy. Systemic thrombolysis leads to rapid resolution of thrombi and improves hemodynamic instability. 17,18 Despite this recommendations, in a study of 108 patients with massive pulmonary embolism, 2-thirds of the patients did not receive thrombolysis or embolectomy. 19 Improvement in the survival of cancer patients may lead to an increase in malignancy-associated VTE episodes, many of which are complicated by concomitant metastatic involvement of brain. In conditions such as massive pulmonary embolism, the patients need emergency treatments. Systemic thrombolysis has absolute contraindication in the cases of brain metastases and emergency embolectomy is not available in all hospitals -and if availablemany surgeons and anesthesiologists are reluctant to accept the risk of surgery. Act outside of the guidelines to save a life, Therefore, in life-threatening conditions, clinicians may be faced with the question of how to manage patient with massive pulmonary embolism and they have to act outside of the guidelines to lifesaving. 20 In the current case, we encountered a management crisis in a patient with massive pulmonary embolism who was at extremely high risk for bleeding complications of thrombolytic therapy and at the same time, no other acceptable recourse was available. Ultimately the patient survived the jeopardy of this double-edge remedy. By this case report, we reported successful outcome of thrombolytic therapy as a last resort in massive PTE despite its contraindication. | 2018-08-28T09:51:17.020Z | 2018-06-05T00:00:00.000 | {
"year": 2018,
"sha1": "0209f35a49f9bc38f53d887de56f906aa8f93dac",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/ccr3.1629",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "0209f35a49f9bc38f53d887de56f906aa8f93dac",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
234081286 | pes2o/s2orc | v3-fos-license | Optimization of E-Commerce Logistics Distribution Path Algorithm under the Background of Big Data
In recent years, online shopping has become a major development trend of shopping. The traditional e-commerce logistics distribution mode has played e-commerce by taking advantage of the advantages of the distribution path algorithm. However, with the increase of the order quantity, the previous path optimization algorithm is no longer efficient. This paper makes an in-depth investigation on the main problems existing in China’s e-commerce logistics. Through the analysis of the survey results, we can see that there are mainly problems such as low degree of data application and imperfect function of logistics information platform. Aiming at these problems, this paper optimizes the logistics distribution routing algorithm according to the actual demand. Through the establishment of three-dimensional constraint model, the problem of low efficiency in the traditional model is effectively improved. In contrast with the genetic algorithm, the distribution mileage of the optimization algorithm is reduced by more than 25% compared with the genetic algorithm, which improves the comprehensive performance of the model. Constraint model established in this paper makes up for the shortcomings of the current e-commerce logistics distribution path algorithm to a certain extent, and makes a contribution to the development of China’s e-commerce logistics distribution field.
Introduction
With the rapid increase of orders, the corresponding information data is also increasing, but the previous information processing methods cannot make better use of data [1][2][3]. However, big data technology can do a good job in data collection and analysis, help logistics enterprises make scientific decisions, optimize and improve the logistics distribution mode, so as to reduce costs and improve service quality [4][5][6].
The so-called big data refers to the large amount of information, good real-time performance, which cannot be processed by conventional technical methods for a period of time, and has the characteristics of large amount, high value, fast generation and diversification. Social network, cloud computing, Internet of things and other Internet tools are gradually applied in people's life and work, and various data are also in explosive growth [7][8]. According to a survey by International Data Corporation (IDC), the average annual growth rate of global data is multiple. It is estimated that by 2020, the total amount of global data will reach 30 times as much as it is now. It can be seen that we have entered the era of big data [9][10].
In this paper, the current situation distribution is deeply investigated. Through the analysis of the survey results, it can be seen that China's e-commerce is in a period of rapid development. With the rapid growth of orders, it puts forward a huge test for China's e-commerce logistics distribution mode, especially in the aspect of path optimization, which needs to be upgraded urgently. Therefore, this paper establishes the optimization research of e-commerce logistics distribution path algorithm under the background of big data. In the research process, according to the development requirements of e-commerce logistics distribution, and combined with the actual distribution situation, this paper constructs a new three-dimensional constraint model. By simplifying the algorithm and optimizing the structure, the model can effectively improve the calculation accuracy and the competition ratio of the calculation results. The advantages of the algorithm are further verified, especially in the calculation of mileage has been greatly reduced. Analysis shows that the path optimization algorithm not only reduces the distribution cost [11].
Connotation of Big Data
The so-called big data refers to large information, strong real-time, cannot be processed by conventional technology in a certain period of time, with high value, rapid generation and diversification. Internet tools are gradually applied to people's life and work, and all kinds of data are growing. According to the survey of International Data Corporation (IDC), global data is a multiple of growth every year. It is estimated that by 2020, the total amount of global data will be 30 times as much as it is now.
Definition of E-Commerce Distribution
This shows that the logistics and distribution enterprises need to use the network technology, equipped with modern hardware equipment, software system and advanced management means to deliver goods according to customers' order requirements. The system will collect the information data of each sub point to form a large centralized database. In order to shorten the mileage of distribution and reduce the cost, scientific decision-making should be made in the statistical transfer.
Questionnaire Survey on the Main Problems of E-Commerce Logistics in China under the Background of Big Data
In the previous e-commerce distribution mode, e-commerce logistics enterprises usually only deal with the orders with personal experience and management habits, lacking of data analysis. However, after using big data analysis technology, it can effectively predict the market demand and prospect, understand the development direction of logistics distribution, so as to make better and scientific decisions, select the optimal distribution mode, provide better services, and maximize the profits of enterprises.
This survey adopts the method of questionnaire survey and on-the-spot interview, and conducts a survey on 500 e-commerce enterprises. The purpose of the survey is to deeply understand the main problems existing. A total of 500 questionnaires were distributed and 495 questionnaires were collected. The main problems were sorted out according to the survey results. The results are shown in Table 1. According to the survey results in Table 1, under the background of big data, China's e-commerce logistics mainly has low data application degree, imperfect logistics information platform function, lack of cooperation between logistics enterprises, unreasonable resource allocation, easy 3 disclosure of user information, and lack of big data logistics talents. The analysis shows that the development of big data e-commerce logistics in China is still in the early stage, and more efforts should be made in core technology and personnel training. Table 1. Survey results of main problems in China's e-commerce logistics under the background of big data.
Serial number
Main problems 1 Low degree of data application 2 The function of logistics information platform is not perfect 3 There is a lack of cooperation between logistics enterprises and the allocation of resources is unreasonable 4 User information is easy to leak 5 Lack of big data logistics talents
Comparative Analysis of the Simulation Experiment between the Algorithm and Genetic Algorithm in this Paper
In order to verify the actual effect of the optimization path, this paper compares the genetic algorithm with the optimization path algorithm in this paper, and carries out simulation test on the e-commerce logistics task of enterprise a. According to the test results in Figure 1, compared with the genetic algorithm, the data model in this paper has not changed in the number of vehicles. However, due to the reduction of distribution mileage, the distribution mileage is shortened by more than 25% compared with the genetic algorithm, resulting in the reduction of vehicle depreciation costs, maintenance costs and personnel costs. In addition, the company can increase the distribution business volume and stimulate the logistics demand, Therefore, this path optimization can not only reduce the logistics cost, but also promote the sustainable development of the company. Figure 1. Comparative analysis of simulated path test distance between the algorithm and genetic algorithm in this paper According to the comparison results of competition ratio in Figure 2, when the car capacity decreases, the traditional genetic algorithm has the advantage, the optimization algorithm in this paper is more advantageous. Data show that the traditional genetic algorithm is weak in dealing with large orders, and its disadvantages become more obvious with the increase of orders. The optimization path algorithm in this paper can better improve this shortcoming, and the allocation in time and space is more reasonable.
Three-Dimensional Constraint Model of Distribution Route Optimization
Suppose there are m distribution centers, n customer nodes and k distribution vehicles in a certain range. Then the number of vehicles available for distribution center p is Ap . However, customers often need a variety of goods, so on the premise of meeting customer distribution requirements, the distribution cost should theoretically be the minimum of the calculation results as the objective function. Based on the characteristics of e-commerce logistics distribution, a two-dimensional constrained logistics distribution path optimization model with distribution time, distribution space and quality evaluation are established as follows: The objective function formula (1) consists of two parts: the first part is the distribution cost of vehicles, the second part is the one-time consumption cost and personnel cost of vehicles; formula (2) represents the time constraint, that is, the delivery time required by customers; and (3) represents the space constraint, that is, the total amount of customer goods delivered by the same vehicle should not be greater than the total amount of unit vehicle; Equation (4) indicates the quality constraint, that is, the total mass of customer goods delivered by the same vehicle shall not be greater than the carrying capacity of unit vehicle. The data processing framework and related processing strategies relatively low, and the universality is not strong. Therefore, in the context of big data, it is necessary to apply the current big data technology to the logistics distribution and transportation of e-commerce, and play its own value. Establish a complete and mature strategic framework for big data processing. According to the deficiencies of the existing framework for big data processing, we can transform and realize innovative development. The big data existing in logistics distribution can provide effective basis for e-commerce enterprises. Therefore, in order to promote the improvement of logistics service level, it is necessary to open up and use special network to query some information to promote the effective acquisition of logistics distribution information. Through the analysis of information elements, logistics distribution information is based on data, and realizes different forms and different storage modules.
Impact of Big Data on E-Commerce Logistics Distribution (1) Realize information docking
Using big data technology to process a large number of logistics information can meet the demand of order distribution in e-commerce logistics distribution work. Moreover, after the establishment of big data sharing platform, the distributed data in e-commerce information can be combined with each other and processed.
(2) Improve customer loyalty In the process of continuous development and growth of e-commerce enterprises, the development of big data can improve customer loyalty. Therefore, in order to improve customer loyalty, promote brand image optimization, and promote the optimization of enterprise service quality, it is necessary to improve the service quality of e-commerce. In the traditional sense, there is still a lot of room for the improvement of logistics distribution services. After improving the corresponding customer service, customers' evaluation of logistics distribution will also be improved, and from then on.
(3) Increase the use value of data With the continuous progress and development of modern society, e-commerce has been rapid development. There are a lot of data in every link of e-commerce. We need to transform the data structure from unstructured to structure. In the case of e-commerce, the value of distribution will be affected. However, in the context of big data, the realization of special data management can avoid the occurrence of information failure and improve the value of e-commerce logistics distribution data. The potential value of data is huge. If enterprises want to develop for a long time, they must pay attention to the use value of data.
Conclusions
Through the results of big data technology, as well as the current development trend of e-commerce in China. In the aspect of path algorithm, this paper establishes a new three-dimensional constraint model, which can better meet the current requirements of e-commerce logistics distribution for path optimization. The analysis shows that at present, e-commerce logistics enterprises should pay attention to the realization of information docking and data sharing, and enhance their data analysis ability while improving data transparency. Finally, in order to verify the actual effect of the three-dimensional constrained path optimization model in this paper, the traditional genetic algorithm is used as the comparison, and the simulation experiment is carried out. The experimental results show that the optimization path method in this paper has a significant advantage in both mileage and competition ratio. | 2021-05-10T00:03:28.084Z | 2021-02-01T00:00:00.000 | {
"year": 2021,
"sha1": "755e66d5b5c19acab7a08ac841e4597f840c8800",
"oa_license": "CCBY",
"oa_url": "https://iopscience.iop.org/article/10.1088/1742-6596/1744/4/042214/pdf",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "590f39fba94bef1c0b34404827218532e695b5cd",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": [
"Computer Science",
"Physics"
]
} |
13384906 | pes2o/s2orc | v3-fos-license | Extracellular HSPs: The Complicated Roles of Extracellular HSPs in Immunity
Extracellular heat-shock proteins (HSPs) interact with the immune system in a very complex manner. Many such HSPs exert powerful effects on the immune response, playing both stimulatory and regulatory roles. However, the influence of the HSPs on immunity appears to be positive or negative in nature – rarely neutral. Thus, the HSPs can act as dominant antigens and can comprise key components of antitumor vaccines. They can also function as powerful immunoregulatory agents and, as such, are employed to treat inflammatory diseases or to extend the lifespan of tissue transplants. Small modifications in the cellular milieu have been shown to flip the allegiances of HSPs from immunoregulatory agents toward a potent inflammatory alignment. These mutable properties of HSPs may be related to the ability of these proteins to interact with multiple receptors often with mutually confounding properties in immune cells. Therefore, understanding the complex immune properties of HSPs may help us to harness their potential in treatment of a range of conditions.
Hsp90 Context ++ (14)(15)(16)(17)(18) Hsp110 Pro +++ (19)(20)(21)(22) Grp94 Pro + (23,24) Grp170 Pro +++ (25)(26)(27)(28)(29) Calreticulin Pro + (23,24) and immunological balance in tissues ( Table 1). The hypothesis of a pro-immune function for extracellular HSPs was derived primarily from studies utilizing molecular chaperone vaccines in cancer treatment (10,23,33). It was shown that HSPs from a number of chaperone families could be extracted from cancer cells while they were associated with a range of tumor peptide antigens (11,33,34). These HSP-peptide complexes could then be injected into hosts as anticancer vaccines, delivering a range of tumorderived antigens to the immune system and promoting antitumor immunity (19)(20)(21)(22)(25)(26)(27)(28)(29). HSPs were, by the proponents of this approach, conventionally, viewed as playing a dominant role as promoters of immunity (32). In addition, a number of studies showed them to be pro-inflammatory mediators, and extravagant claims were made for molecular chaperones as activators of multiple facets of immunity. However, other investigators have demonstrated powerful anti-inflammatory roles for HSPs that we will discuss more fully, later in this manuscript (12,35,36). In addition, the properties of extracellular HSPs have now expanded to include powerful roles in processes outside the immune response. For instance, secreted Hsp90 has been shown to mediate wound healing and tumor metastasis (36,37). Thus, extracellular HSPs appear to have come of age as major intercellular signaling molecules in biology and medicine. Some of the issues discussed here, particularly the role of HSPs in antigen presentation, have been mentioned in a previous review (38). Here, however, we focus mainly on the potentially confound pro-and anti-inflammatory roles of HSPs and discuss how these properties can be manipulated toward clinically useful outcomes in both treatment of autoimmune conditions and in the deployment of chaperone anticancer vaccines.
ReLeASe OF HSPs iNTO THe eXTRACeLLULAR MiCROeNviRONMeNT
Structural considerations would tend to make one skeptical regarding the possibility of HSP secretion into the extracellular milieu. HSP family proteins lack an N-terminal hydrophobic signal sequence, characteristic of most secreted proteins, and thus, cannot be released from cells by the conventional secretion pathways. However, a number of non-canonical secretion pathways exist, many of which are employed by cytokines to gain access to the extracellular milieu. These eccentric mechanisms include release of the polypeptides via secretory lysosomes, a pathway utilized in the release of IL-1β from inflammatory cells (39).
Hsp70 has been shown to be secreted from a number of cells in free form by a similar pathway, through a mechanism requiring the lysosomal pH gradient (31,40). Indeed, Hsp70 is cosecreted from cells along with the lysosome resident protein LAMP1 (31). Hsp70 is also released from a range of other cells including tumor cells, reticulocytes, peripheral blood mononuclear cells, B cells, and dendritic cells in various types of lipid vesicles [reviewed by De Maio (41) and Vega et al. (42)]. These vesicles may include a variety of lipid-bounded structures, including ectosomes that are vesicles derived from the plasma membrane and that may contain cytosolic proteins as well as exosomes. Formation of exosomes is a complex process including the internalization of portions of the plasma membrane and subsequent release of exosomes containing a variety of previously intracellular proteins, including HSPs (43). The exosomal pathway is also utilized by some cells for IL-1β secretion (44). HSP-containing exosomes have a wide array of properties including both immunostimulatory and immunosuppressive functions, depending on the protein content of the exosome, cell of origin, and target cell (45)(46)(47).
Heat-shock proteins, therefore, can be secreted from a variety of cells in free form and in membrane-bounded particles. In addition, they can be released from cell undergoing necrotic death when membranes are disrupted, and the HSP can leak passively out of the cells (48). Hsp70 released in such a way has been shown to be strongly immunostimulatory.
HSPs AS CARRieRS OF TUMOR ANTiGeNS AND MeDiATORS OF iMMUNiTY
Adaptive immunity Molecular chaperones are unique immune modulators in that they can associate with a wide range of antigenic peptides and facilitate their delivery to antigen-presenting cells (APCs) (11,13,23,33,34). This property has proven to be desirable in the preparation of anticancer vaccines. Only a relatively small number of tumor antigens have been characterized, and we presume that this group represents a small minority of the real repertoire of unique cancer-derived antigens. Thus, chaperones, such as Hsp70, can be considered to "sample" the antigenic milieu of the malignant cell on encountering processed peptides in vivo and can be used to carry this sample into the APC during immunization (Figure 1). Such HSP-containing vaccines have proven to be highly effective in studies in experimental tumor systems in mice, in which they can lead to tumor regression associated with the generation of specific immunity (10,13,14,20,(49)(50)(51). Issues in the preparation of the vaccines, which may influence the clinical effectiveness of vaccines, include the degree to which antigens can be retained by the chaperone and the affinity for the peptide during immunization and entry into APC (11,14,52).
Cross-presentation is a process by which extracellular antigens can gain access to the MHC class I pathway, a mechanism normally reserved for processing and presenting endogenous antigens (38). Efficient antigen cross-presentation is very important for vaccine effectiveness as MHC-I-peptide complexes permit recognition of cells bearing the complexes and killing by CD8 + cytotoxic The HSP-peptide complexes that comprise the anticancer vaccine are shown to interact with APC after vaccination of host. The vaccines can efficiently (1) cause cross-presentation of tumor antigen and, thus, prime CD8 + T lymphocytes as well as activating CD4 + T cells through the (2) class II pathway. However, many investigations suggest that HSPs may not have major effects on (3) inflammatory signaling and may require combination with agents with adjuvant activity or inflammatory cell killing. Gray spheres indicate nuclei.
T lymphocytes (53). Interestingly, Hsp90 appears to protect the integrity of internalized antigens associated with it, to trigger cross-presentation, and to carry antigens deep into the cell, penetrating the plasma membrane and endosomal membranes, and delivers chaperoned peptides to cytoplasmic proteasomes for processing (15,16).
Although CD8 + T cells can be triggered by DC to recognize antigens after cross-presentation, in the absence of further signals, such T cells are unable to kill their targets. Other inputs are required for full activation (54,55). The principal pathway used by APC for sampling external antigens is the MHC class II pathway. MHC class II molecules are found only on the surfaces of immune cells. The class II pathway involves the uptake of antigens by receptors on DC, processing of such antigens in the lysosomal compartment, transport of vesicles containing antigen-MHC-II complexes to the cell surface, and presentation to CD4 + T lymphocytes. Hsp90 is able to carry associated antigens into APCs and direct them into the class II pathway as well as facilitate crosspresentation. The choice of direction regarding entry of antigens into the class I/cross-presentation or class II pathways appears to usually depend upon the antigen-binding receptor that mediates triage between the two presentation systems (56,57). However, Hsp90 appears neutral in this regard and increase penetration of associated peptides into either pathway (17,18). One important activity governed by the MHC class II pathway is a process called dendritic cell licensing (54). In this mechanism, CD4 + T cells that recognize the antigen on a particular DC produce a reaction in the APC that permits it to activate CD8 + cells that interact with the same APC. Interaction of the T cell receptor on the CD4 + T cell triggers the expression of CD40 ligand (CD40L) that can bind the CD40 counter receptor on the DC and induce expression of inflammatory cytokines, such as TNFα and IL-12 as well as stimulatory coreceptors, like CD80 and CD86 (58,59). These coreceptors cooperate with MHC class I in fully activating the CD8 + T cell through the T cell receptor. Thus, HSP-peptide complexes become internalized and trigger both the MHC class I and II pathways and may permit DC licensing to occur (17,18). Our findings that HSPs can facilitate uptake of individual Ova antigens through the MHC-I and MHC-II pathways suggest the possibility of HSP-antigen complex could mediate DC licensing, although this has not yet been formally proven. Homing of CD8 + T cells toward licensed DC may involve surface chemokine receptor CCR5, a process strongly stimulated by chemokines, CCL3 and CCL4 (54). It has been shown in Lewis lung carcinoma cells, in vivo, that antitumor immunity was activated along with release of chemokines CCL2, CCL5, and CCL10, by a mechanism dependent on Hsp70 and TLR4 (60).
inflammation and innate immunity
On exposure to prokaryotic cells or cell products, a separate branch of immunity known as innate immunity is stimulated. In this process, molecules characteristic of individual pathogens including contrasting types of viruses and bacteria, known as pathogen-associated molecular patterns (PAMPs) herald the infection and prime the immune response (61). Then, PAMPs interact with specific receptors on macrophages or DC, known as pattern recognition receptors (PRR), and trigger innate immunity. Best known among the PRR are the toll-like receptors (TLR) that can couple binding of individual PAMPs to intracellular signaling pathways and gene expression programs (62,63). Most notable among the mechanisms triggered by PRR occupation are the NKK and MAP-kinase pathways that influence inflammatory transcription through activation of factors, such as NFκB and IRF3 (64). This process can lead to synthesis of costimulatory molecules, such as CD80, and activating cytokines such as TNFα and IL-12 that synergize with MHC class I signaling in generation of active and long lived CTL (54). It is not clear to what extent HSPs derived from prokaryotes might function as PAMPs, although their extreme conservation across all cellular species would seem to argue against this. HSPs derived from mycobacteria are, however, recognized by the mammalian immune response and invoke powerful immunity to the extent that they have been described as superantigens (65). The mechanisms by which prokaryotic Hsp60 activates immunity are not clear but could involve PRR, such as TLRs, or other mechanisms.
It has also been shown that some molecules released from damaged and dying cells, such as uric acid and high mobility group box 1 protein (HMGM1), may trigger a form of sterile innate immunity, and such molecules are referred to as damage-associated molecular patterns (DAMPs), in order to suggest a functional similar to PAMPs (66,67). Thus, DAMPs are thought to trigger innate immunity by binding to PRR and triggering inflammatory signaling cascades. Hsp70 was widely reported to function as a DAMP and to trigger innate immunity through the TLR2 and TLR4 pathways (32). Although this field has run into some controversy, the majority of findings in studies carried out in vivo over the past 15 years suggested that Hsp70, through interaction with TLR4, could potentially act as a DAMP (68). This area has been recently reviewed in depth (69). In addition to Hsp70, extracellular Hsp27 has recently been shown to cause both inflammatory and anti-inflammatory effects (5).
SOMe ANTi-iNFLAMMATORY AND iMMUNOReGULATORY eFFeCTS OF HSPs
Although prokaryotic HSPs can trigger a powerful immunodominant response in animals, most reports indicate that their effects are generally not pro-inflammatory, and the antibodies and T cells activated in the response had anti-inflammatory properties (69)(70)(71). Curiously, the epitopes that T cells respond to in mycobacterial Hsp60 were conserved with mammalian HSPs, and such cells recognized and responded to epitopes in the mammalian proteins. Prokaryotic Hsp60, therefore, did not seem to act as a PAMP (7). In addition, although HSPs were shown to interact with TLRs, such PRR often provoked anti-inflammatory signaling (8). For instance, Hsp60-derived peptides interacted with TLR2 on regulatory T cells (Tregs) leading to an immunosuppressive response. In addition, purified mycobacterial Hsp70 inhibited the maturation of DC (72,73). Intracellular HSP levels were shown to increase in inflamed tissues and HSP-derived peptides expressed on the cell surface and appeared to activate Treg responses, thus mediating immunoregulatory functions (74). No studies, to date, have shown direct interactions between HSPs and TLRs, and in fact, attempts to show such binding have been negative (73,75). The extracellular influences on TLR activity that have been reported may, therefore, be indirect and likely dependent on primary interactions of the HSP with other receptors on immune cells, such as the scavenger receptors (SR), followed by recruitment of TLR (64). The powerful immune effects of non-mammalian Hsp60 may also involve mechanisms independent of TLRs, and it has been suggested that the immune response may be genetically programmed to respond to such chaperones (76).
Interestingly, some of the studies applying HSP vaccines to cancer therapy indicated that, although there was significant activation of antitumor CTL by these agents, these were followed by a delayed Treg response. These findings suggest contrasting effects of the vehicle (HSP) and cargo (antigenic peptide) components of chaperone vaccines on immunity. These data might be interpreted as, suggesting that, while tumor antigens chaperoned by the HSPs trigger antitumor immunity, processed peptides from the HSP component of the vaccine led to a suppression of immunity. Using the chaperone vaccines at lower doses appeared to favor induction of CTL over the immunoregulatory response, perhaps by reducing the levels of HSP-derived peptides below a threshold (77,78). It would seem that most chaperone vaccines, although efficiently triggering external tumor antigen presentation, do not deliver the inflammatory signal required to overcome antigenic tolerance (Figure 1). Such vaccines might be improved by use of adjuvants or pro-inflammatory forms of therapy, as discussed below.
wHeN HSPs BeCOMe PRO-iNFLAMMATORY FACTORS One notable finding observed in multiple investigations of HSPs was that, even in studies where an inflammatory response to HSPs was not detected, the chaperones could strongly amplify responses to PAMPs, such as LPS (79). Mycobacterial Hsp65, a protein discussed in the last section as provoking generally immunomodulatory responses, when covalently fused to antigenic polypeptides produced a potent vaccine that generated effective CTL even in the absence of adjuvant (80). In addition, the combination of Hsp70 elevation in target tissues with therapies leading to necrotic cell killing led to a profound stimulation of inflammation and CTL killing that could lead to tumor rejection (81). This approach involved, after elevation of tissue Hsp70, targeting the normal tissues of origin with treatments that led to inflammatory modes of cell killing. This combined treatment resulted in the regression of distant, transplanted tumors (81). The findings observed in these studies were that, for instance, in prostatic tissue, cell killing by fusogenic viruses in the presence of elevated Hsp70 led to induction of the cytokines IL-6 and TGF-β, resulting in generation of highly inflammatory IL-17 and tumor rejection by antigen-specific CTL (82). This effect seemed to depend on generation of IL-6 by the combination of high tissue levels of Hsp70 and inflammatory death. However, in similar studies on pancreatic tissues, combination of Hsp70 and lytic virus failed to generate IL-6 and led to generation of a Treg response and continued growth of pancreatic carcinoma (82). Thus, the balance between immunoregulatory and immunogenic responses of Hsp70 appears to be poised on a knife-edge, influenced by the tissue type and mode of cell killing.
It is well known that the mode of cell death has a powerful influence on inflammation and immunity in interacting APC (83). For instance, when cells die by an apoptotic mechanism, their intracellular contents remain enveloped by an external membrane and, thus, are not released into the environment to trigger inflammation (Figure 2). In addition to this, many apoptotic cells expose "eat me" signals, such as the phospholipid phosphatidylserine on the surface, triggering engulfment by macrophages and leading to immunosuppression (84,85). Additional anti-inflammatory signals emanating from apoptotic cells may include the release of AMP from the apoptotic cell (86). In necrotic cell death, cell membranes become permeabilized, the intracellular milieu becomes externalized, and DAMPs, such as HMGB1, urate, and nucleic acids, released in this way become accessible to detection by neighboring macrophages or DC (85). It should be noted that the response of phagocytic cells to apoptotic bodies is complex and depends on the nature of the dead or dying cells and the surface densities of "eat me" or "don't eat me" signaling molecules that are cell specific. In addition, in late apoptotic cells that have failed to be phagocytosed at an early stage, membranes become permeabilized. Therefore, such late apoptotic cells acquire some of the properties of necrotic cells, permitting release of DAMPs and switching the effects of the cell corpses on engulfing phagocytes toward a more immunogenic influence (85,87). It is not clear in which way extracellular HSPs might synergize with the extruded contents of necrotic cells in activating immunity, although increased capacity for engulfment by immune cells which is triggered by Hsp70 might amplify inflammatory functions (88)(89)(90). The earlier studies of Todryk et al., in fact, indicated that extracellular Hsp70 decrease DC maturation and increase capacity for engulfment of antigenic materials by the immature APC (91). Phagocytosis is a complex process and involves the formation of a broad synapse type structure between the engulfing cell and the mammalian cell body of prokaryotic cells. Multiple receptors appear to make contact between the cells. Interestingly, in C. elegans this process has been shown to require the surface receptor CED-1 (92). Both, mammalian scavenger receptor associated with endothelial cells (SREC-I) and cluster of differentiation 91 (CD91) possess sequence similarities to this protein, suggesting some involvement in phagocytosis and a potential mechanism for HSPs in this process (93). In addition, necrotic cells can supply signals to augment the immune effects of HSPs through release of HMGB1, a ligand for TLR4, and externalized nucleic acids that might interact with TLR3 or TLR9 and trigger inflammatory signaling (85) (Figures 2 and 3).
POTeNTiAL MeCHANiSMS -CeLL SURFACe ReCePTORS AND CeLL SiGNALiNG PATHwAYS
Most evidence suggests that the biological effects of extracellular HSPs are mediated through cell surface receptors (24,94) ( Figure 3). Thus, some of the clues to the properties of extracellular HSPs may lie within the interactions of these chaperones with these receptors and the unleashing of their embedded signaling cascades. The first surface receptor reported to bind to HSPs was CD91 also known as Low density lipoprotein receptorrelated protein 1 (LRP1), the alpha-2-macroglobulin receptor (A2MR), and apolipoprotein E receptor (APOER) (24). This high molecular weight receptor was shown originally to bind to LDL and, as its possession of multiple names suggests, a wide range of (over 30) other extracellular structures (94). CD91 is expressed most abundantly in vascular smooth muscle cells and in hepatocytes. There has been some skepticism regarding its expression in APC, such as DC, and, therefore, its potential role in immune responses involving the HSPs (95). However, there seems little doubt that CD91/LRP1 is a bona fide HSP receptor, and this protein appears to play key roles in, for instance, responses to Hsp90α in the wound healing response and in tumor metastasis (36,37).
It was next shown that Hsp70 could associate with lectintype oxidized LDL receptor (LOX-1), a receptor found on the surfaces of human DC and that such binding could mediate cross-presentation of peptide cargo associated with this HSP, leading to CD8 + lymphocyte-mediated immunity (75). LOX-1 is the product of the oxidized low-density lipoprotein receptor 1 (ORL1) gene expressed most abundantly in vascular endothelial cells, macrophages, and DC (96). Although LOX-1 belongs to the C-type lectin receptor family, it also clusters with the SR, a group of proteins that have, in common, the ability to bind to covalently modified LDL molecules, such as oxidized LDL and acetylated LDL (97). Furthermore, exploration of the SR family indicated significant interaction of Hsp70 with at least two other members, including SREC-I, encoded by the SCARF1 gene and Stabilin-1/FEEL-1 (encoded by the STAB1 gene) (73,98). SREC-I/SCARF1 was able to bind to Hsp70 and Hsp90 in DC and mediated cross-presentation of associated tumor antigens, leading to activated CTL (13,17). Murine bone marrow derived DC appeared to utilize both LOX-1 and SREC-I/SCARF1 in interacting with Hsp90-peptide complexes (17,75). In addition, SREC-I was also shown to mediate the uptake of antigens chaperoned by Hsp90 into the Class II pathway and stimulate activation of CD4 + T lymphocytes (18). The SR are regarded as receptors that respond to cellular debris including cell bodies, the remains of endogenous proteins spilled into the extracellular milieu, and the residue from invading pathogens. As such, the SR might be regarded as good choices for receptors with which APC might sample the extracellular environment (93).
Stabilin-1 appears to be only sparsely expressed on the surface of DC and appears to function mostly in intracellular trafficking (99). Thus, there appeared little doubt that that the SR could bind to HSP-PC and mediate presentation of peptide cargo (17,18). However, the nature of the receptors that might respond to HSPs and modify the inflammatory response still required investigation.
Many reports suggested that TLRs, particularly TLR2 and TLR4, might mediate inflammatory responses to HSPs in an up or down manner (32,76). However, as mentioned above, attempts to demonstrate direct binding of HSPs to TLRs have not been generally successful (73,75). Thus, HSP interaction with TLRs is likely to be an indirect one. It was shown, recently, that SREC-I/ SCARF1 could interact with TLR4 on the surface of mouse macrophages (64). Interestingly, both LPS and Hsp90 could mediate this complex interaction and lead to the sequestration of TLR4 in lipid raft domains (64). In the case of LPS, encasement of SREC-I-TLR4 complexes within lipid rafts was required for activation of signaling through the NFκB and MAP-kinase cascades, resulting in inflammatory cytokine secretion. Significantly, exposure to Hsp90, although leading to SREC-I-TLR4 colocalization, failed to trigger the inflammatory cascade through this mechanism. Alternative pathways of inflammatory suppression by Hsp90 may, thus, be involved (64).
Recent studies have also suggested that another receptor family, sialic acid-binding immunoglobulin-like lectins (Siglecs) may participate in inflammatory responses resulting from binding to HSPs (100). These receptors have been shown to bind to conjugated sugar residues in the cell coats of adjacent cells and generally suppress inflammation. Suppression of inflammation by these receptors involves immunoreceptor tyrosine-based inhibitory motif (ITIM) sequences in the intracellular domain that recruit anti-inflammatory proteins after ligand-triggered tyrosine phosphorylation (100). Upon activation, Siglec receptors have been shown to directly associate with TLRs and inhibit TLRmediated activation of inflammatory signaling cascades, such as the NFκB pathway (101). Interestingly, recently, it was shown that two human Siglecs, Siglec-5 and Siglec-14 each, bind to Hsp70 (102). What was most intriguing about these interactions was that, while Siglec-5 contained the expected ITIM domain and repressed anti-inflammatory signaling through quenching NFκB, Siglec-14 appeared to have acquired a domain that binds proteins, such as DAP12, containing an intracellular immunoreceptor tyrosine-based activating motif (ITAM) that stimulated release of inflammatory cytokines (102,103). Differential expression of members of the Siglec family might, thus, either amplify or suppress the HSP regulated activities of cell surface TLR and inflammatory signaling through mediation of the NFκB pathway. A further property of extracellular HSPs mentioned above, that may play a key role in their immune functions, is their ability to trigger phagocytosis. However, the mechanisms and receptors involved in this process have not been well characterized and further experiments will be required to understand the mechanisms involved (Figure 3).
Thus, responses to extracellular HSPs appear to involve the cooperative or confounding outputs of a range of cell surface receptors that together may determine their influence on immune reactions within tissues. To date, however, we have not been able to identify a dedicated high-affinity HSP receptor. Dedicated receptors, such as the insulin receptor, are often able to pick up tiny concentrations of ligands, often with affinities in the range of 10 −9 -10 −10 M, with exquisite selectivity. For HSPs, we have identified, so far, only "hand me down" receptors that also bind a large range of other ligands with moderate affinity. It remains to be determined if dedicated HSP receptors will be found in immune cells of mammalian species.
CONCLUSiON
The intracellular roles of HSPs in protein folding have been conserved since the dawn of cellular life (3). However, the HSPs also appear to have acquired key roles in the immune systems of animals early during evolution, and these roles are preserved in modern animal species (104). Such HSPs can capture intracellular antigens and present them to APC, mediating the crosspriming of recipient cells. In addition, HSPs, when processed and presented on the APC cell surface, can activate Treg cells and inhibit immunity and inflammation. Their effects on the immune system are, thus, bivalent. In the presence of PAMPs or tissues undergoing necrosis, Hsp70, in particular, becomes a strong inflammatory agent. The precise nature of the responses elicited by extracellular HSPs may, therefore, depend upon the particular tissue milieu within which they are released and the identities of the receptors on the surfaces of immune cells that encounter them.
OUTSTANDiNG QUeSTiONS AND POTeNTiAL ADvANCeS
(1) Many of the immune-active HSPs are members of different gene families, with only a modicum of structural similarity ( Table 1). However, each chaperone appears to be recognized by similar families of receptors, including SR and the LDL receptors, despite their sequence dissimilarity (94). Understanding the structural basis for extracellular HSP recognition by these and potentially new classes of receptors would be an advance in determining the roles of the chaperones in immunity and predicting novel recognition structures. However, this has been an open question for a number of years now and research is ongoing. (2) It is becoming apparent that members of individual HSP gene families have distinct properties. For instance, only Hsp90α is secreted from cells and the other major isoform, Hsp90β, is mostly retained within the cell (36). It will be enlightening to learn whether such specificity in extracellular function holds for other HSP families. (3) We have, in recent years, acquired some understanding of how these "leaderless" proteins gain access to the extracellular microenvironment in intact cells (see Release of HSPs into the Extracellular Microenvironment), although it will be invaluable to gain a more concerted understanding of mechanisms of HSP secretion. (4) We will continue the search for HSP receptors in immune cells, their expression patterns and their connection to inflammatory and anti-inflammatory responses to both endogenous HSPs and prokaryotic paralogs (94). It is likely that a major key to understanding how chaperones trigger immune responses and the direction of such responses depends on understanding the combinatorial effects of the multiple receptor families. (5) It seems clear that many HSPs, although capable of efficient transport of antigenic peptides within chaperone vaccines, do not on their own deliver a second signal for APC maturation, thus reducing their effectiveness as stand-alone agents. In order to break immune tolerance to tumor antigens, chaperone vaccines might be best deployed in cancer therapy in combination with PAMPs, such as double stranded RNA or unmethylated CpG motifs or checkpoint inhibitors such as anti-CTLA4 (105,106). Alternatively, vaccines could be used with agents, such as ionizing radiation, that can cause necrotic killing and subsequent inflammatory effects in tumors (including release of intracellular DAMPs) thus leading to immune rejection of the cancers (10).
AUTHOR CONTRiBUTiONS
SC, JG, and AM contributed in writing the text of the review. | 2016-06-17T03:46:20.825Z | 2016-04-25T00:00:00.000 | {
"year": 2016,
"sha1": "8633571bd18301b8ca84e27de80f43ec211450b0",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fimmu.2016.00159/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a3df9155bd94b1ec0338c5216c9232222616451c",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
236413335 | pes2o/s2orc | v3-fos-license | Securing Malaysia ’ s Borders from Covid-19: Legal Considerations to the Influx of Rohingya Migrants
managing this humanitarian disaster should be shared with other nations, particularly those party to the United Nations Convention relating to the Status of Refugees 1951 .
Introduction
Malaysia occupies terrestrial and maritime territories stretching from the Malay Peninsula to Sabah and Sarawak on the island of Borneo, sharing its frontiers with other South-East Asian nations. 1 Thailand, Brunei Darussalam, Singapore, Indonesia and the Philippines are among Malaysia's immediate neighbours. 2 The northern portion of Peninsula Malaysia shares common frontiers with Thailand, extending 646.5 km from Perlis in the west to Kelantan in the east. 3The Thai and Malaysian governments have erected eight immigration checkpoints along the length of their common border: Wang Kelian and Padang Besar in Perlis; Bukit Kayu Hitam and Durian Burung in Kedah; Pengkalan Hulu in Perak and Bukit Bunga; and Rantau Panjang and Pengkalan Kubor in Kelantan (Figure 1). 4 Securing Malaysia's Borders from Covid-19 117 XIV JEAIL 1 (2021) Two international gateways link Malaysia and Singapore at the southern end of Peninsula Malaysia: Tanjung Kupang-Tuas and Johor Bahru-Woodlands. 5The Malaysian Borneo states of Sabah and Sarawak share an uninterrupted land border -2,019.5 kms from Tanjung Datu all the way up north to Tawau, with Indonesian provinces of West, Central and North Kalimantan. 6Sarawak surrounds its tiny but affluent neighbour, the Sultanate of Brunei Darussalam, sharing a land boundary of approximately 481.3 km. 7alaysian frontiers are mostly isolated hilly areas and are thickly forested.Extensive surveillance to track down illegal activities such as smuggling, human trafficking and illicit trade in narcotics may be difficult. 8For instance, the discovery of a shallow Rohingya grave in 2015 in areas near the Malaysia-Thailand border in Perlis was and still is a manifestation of the grim effects of human trafficking involving illegal Rohingya migrants. 9ohamed Farid Noh, All Quiet at World's Busiest Causeway, new straits tiMes, Mar. 18, 2020, https://www.nst.com.my/news/nation/2020/03/575714/all-quiet-worlds-busiest-causeway.In addition, the vast maritime areas of northern Strait of Malacca-possessing hundreds of islands on both Thai and Malaysian sides-provide perfect hideouts for human traffickers, making it challenging for authorities to conduct extensive surveillance and monitoring.Recently, it has been reported that the outlying Thai island of Adang has been used as a transit point for Rohingya migrants to be smuggled into Malaysia by traffickers. 10Since the implementation of the Movement Control Order, Malaysia has conducted Ops Benteng to strengthen border control and crackdown against illegal migrants, including the Rohingyas. 11n this essay, the authors will examine the main reason behind the influx of Rohingya illegal migrants and discuss whether the Malaysian government should tighten its borders to safeguard against COVID-19.In the following part, we will analyze the background of the Rohingyas' immigration, and the legal considerations to end such an influx in view of the Covid-19 pandemic that is still ravaging the nation.
The Rohingyas
The Rohingyas are a minority Muslim community residing in the Arakan or Rakhine provinces in north-eastern Myanmar, bordering neighbouring Bangladesh. 12They are ethnically and linguistically distinct from other ethnic groups in Myanmar. 13There are conflicting views in regard to the place of origin of the Rohingya community. 14Some argue that the Rohingyas are originally from Rakhine, while others contend that they are migrants from Bengal who came to the region in the fifteenth century. 15he Burmese empire reached its height as a regional power, with far-flung territories in the seventeenth and eighteenth centuries. 16However, this former formidable www.forbes.com/sites/realspin/2017/01/23/the-rohingya-exodus-is-malaysia-in-a-catch-22-situation/#20732a403667.empire fell to British dominance and was annexed as part of the British Raj in 1885. 17he Rohingyas were brought into Myanmar by British administrators in the nineteenth century to toil on farms as labourers during the era of the British Raj; the fertile lands of Rakhine were sparingly populated then. 18Some were brought in as military recruits.Under British administration, the Rohingya community contributed to the development of Rakhine province. 19This migration, however, ceased after Myanmar gained independence and became a sovereign nation in 1948. 20he Malayan population possesses a somewhat similar history.Malays have established kingdoms and built empires on the Malay Peninsula and across the Malay Archipelago for centuries. 21Langkasuka was believed to be the first regional kingdom to rule the Malay Peninsula as early as the third century AD. 22 In subsequent centuries, the first migration of a non-Malay community, the Baba-Nyonya Chinese community, came to the Malay Peninsula during the era of the Malacca Sultanate in the fifteenth century.The number was small, however. 23hen Malaya was ruled by the British, mass migration of both Chinese and Indian communities into Malaya was encouraged in the nineteenth and the twentieth centuries. 24They were employed as cheap labourers to satisfy the demands of the lucrative tin mining and rubber industries. 25ohingya immigrants in Myanmar and non-Malay immigrants in Malaysia (then Malaya) were treated differently and had separate destinies.While the Chinese and Indian immigrants were granted citizenship when Malaya gained independence in 1957, in return of special rights provided to Malays, Rohingyas were not regarded as compatriots. 26In contrast, the Chinese and Indian communities are treated with respect and enjoy rights as citizens provided by the Federal Constitution. 27Many prominent politicians, successful entrepreneurs and respected scholars hail from Chinese and Indian communities in Malaysia.
The government of Myanmar generally respects all religions and allows freedom of religion, despite Buddhism being the main religion of the majority of its population. 28There is a mosque situated just next to the famed Sule Pagoda in central Yangon, which could be viewed as a symbol of religious tolerance in Myanmar. 29n addition, there is a large Malay Muslim population in the southern Tanintharyi region in Kampong Tengah, Kampong Ulu and Pase Panjang. 30These Malay Muslim communities did not face discrimination and enjoy rights as citizens along with other ethnic groups in Myanmar.
Some argue that the oppression against the Rohingyas is not born of religion, but of race. 31Most of the Myanmar population views the Rohingya community as illegal immigrants from Bengal who have nothing in common with other ethnic groups in Myanmar, and as such, should be expelled from the country. 32
Persecution
There are around 2 million Rohingyas in the world; most are in Bangladesh, where about 1.3 million live in camps in Cox's Bazaar and a Rohingya settlement on Bhasan Char in the Bay of Bengal. 33About 150, 000 Rohingya migrants live in Malaysia as of 2020.International media and human rights organisations have described Rohingyas as "among the world's least wanted" and "one of the world's most persecuted minorities." 34he Rohingyas have been struggling for recognition of their rights since World War II.When the Japanese surrendered in 1945, Rohingyas attempted to merge the Mayu region in Rakhine as part of East Pakistan (now Bangladesh), which ended in failure. 35In the 1950s, the Rohingyas planned to establish an autonomous Muslim state in Rakhine.This effort was opposed by Burmese forces led by General Ne Win, who launched military operations against the Rohingyas as early as 1962. 36In 1982, the Burmese government, through the Citizenship Act 1982, denied Rohingyas Burmese citizenship, resulting in millions of Rohingyas becoming stateless in their own homeland. 37iolence against the Rohingyas still occurs, causing an exodus of refugees from Myanmar into neighbouring countries. 38Meanwhile, the government of Myanmar, in its capital city, Naypyidaw has officially denied committing the alleged atrocities against the Rohingyas. 39alaysia is one of the founding state members of the Association of Southeast Asian Nations (ASEAN).Thus, it respects the principle of non-interference-an underlying fundamental principle that an ASEAN member should not interfere in the domestic affairs of another. 40ASEAN embraces the concept of "prosper thy neighbour." The Rohingya issue was originally considered a domestic question of Myanmar.However, as more and more Rohingyas flee persecution in Myanmar and arrive in neighbouring countries, this humanitarian disaster should no longer be viewed or treated as a domestic issue, but as rather an inter-regional catastrophe.
Humanitarian Disaster
Although Malaysia is relatively young compared with other developed Western countries, it has been shouting its views against global persecution.This can be observed from the fact that Kuala Lumpur refused to engage in diplomatic relationships with the apartheid government of South Africa in Pretoria until 1994. 41Malaysia has been very actively fighting for the rights of Bosnians fleeing the country during war in 1994. 42Malaysia's position remains the same today; it shelters thousands of Rohingya and Syrian migrants at present. 43ven though being globally praised and acknowledged as a nation that aids Rohingya migrants, Malaysia has been criticised for not ratifying the United Nations conventions on refugees. 44The choice of whether to ratify certain international conventions is solely the right of the sovereign State.Despite the US being one of the largest maritime nations in the world, its government has yet to become party to the United Nations Convention on the Law of the Sea 1982 (UNCLOS). 45This inaction does not mean that the US government has no say in international law of the sea or global maritime matters because these are part of customary international law applicable to all states. 46l Jazeera in November 2014 published a video entitled "Malaysia's Unwanted."It severely criticized the Malaysian government for neither stopping human trafficking, nor protecting refugees suffering from various kinds of abuse. 47Malaysia has done its best accommodating the influx of migrants, particularly Rohingyas.Without Malaysia's steady economic growth and relative peace, these migrants would not be flocking into Malaysia. 48Before the blame is attributed to Malaysia, it should be noted that developed countries such as the UK and those of the European Union face great challenges and difficulties in handling Syrian refugees and African economic migrants swarming into Europe. 49
Protecting Malaysia's Frontiers
Illegal immigration to Malaysia is the cross-border movement of people to Malaysia under the conditions in which official authorisation is lacking, breached, expired, fraudulent or irregular. 50The cross-border movement of workers has become well established in South-East Asia, where Malaysia is a major labour-receiving country and Indonesia and the Philippines are the region's main labour-sending states. 51s mentioned above, Rohingya migrants have fled their hostile home country of Myanmar and overcrowded camps in Bangladesh's Cox's Bazaar; they have been illegally trafficked into Malaysia, which is regarded as their "land of dreams" by most Rohingyas. 52Therefore, managing cross-border migration (labour, refugee and human trafficking) has become an issue of increasing concern in Malaysia and its international contemporaries. 53Human traffickers engaging in this illegal activity make billions of dollars every year, profiting from the suffering of others. 54alaysian law stipulates that any individuals who enter Malaysia without valid documents are deemed "illegal immigrants." 55The offender, if found guilty, may be liable to a fine of RM10,000.00and/or imprisonment, not exceeding 5 years and whipping of no more than 6 strokes. 56Illegal migrants shall be liable to be removed from Malaysia, as stated in Section 32(1) of the Immigration Act 1959/1963 (Act 155).
Since the outbreak of COVID-19, the virus has claimed more than 500 lives in Malaysia, and there were more than 14,000 active cases by mid-December 2020. 57y January 2021, the number had increased to 25,000.Malaysian Director General of Health Tan Sri Dr Noor Hisham Abdullah has repeatedly mentioned in his daily press conferences that the government should shut and secure Malaysia's international gateways to stop foreign importation of the virus. 58In October 2020, Malaysian Prime Minister Tan Sri Muhyiddin Yassin also expressed concern about the increasing COVID-19 infections caused by illegal migration into Malaysia. 59onsequently, Malaysia, via Ops Benteng, has enhanced surveillance, not only on its official gateways, 60 but also on hundreds of "back alleys" along the thousands of kilometres of land and maritime boundaries encircling the nation. 61In this regard, the recent arrival of the Rohingyas by boat in Malaysian waters near Langkawi indicates that efforts should be undertaken to further safeguard the nation's maritime frontiers. 62These boats are not in any way exercising the right of innocent passage because their intention was to illegally bring Rohingya migrants. 63 illegal migrants are believed to have been trafficked into Malaysia by numerous agents. 64ome were even funded by Rohingya communities that already reside in this country. 65According to Malaysian Minister of Home Affairs Datuk Seri Hamzah Zainuddin, traffickers 66 could be charged up to RM 15,000.00 per migrant they traffic into the country. 67Malaysian law stipulates that consent of trafficked persons is irrelevant for the purpose of convicting an accused person committing such an offence. 68In other words, although an individual has given his or her permission to be trafficked, the traffickers could still be prosecuted if caught committing this offence in Malaysia.
The Rohingyas are stateless people, so that they do not possess valid documents 69 when entering Malaysia. 70There have been reports that Rohingya migrants purposely damaged their boats to compel Malaysian authorities to bring them ashore. 71There have also been calls by some that the Malaysian government should continuously do except at an authorised landing place, airport or point of entry. 64Naimul Karim, Traffickers Demand Ransoms for Rohingyas Held at Sea in SE Asia, reuters, June 15, 2020, https:// de.reuters.com/article/us-bangladesh-malaysia-rohingya-traffick/traffickers-demand-ransoms-for-rohingyas-held-atsea-in-se-asia-idUSKBN23M1AC. 65Based on the interview with the Enforcement Division of the Immigration Department of Penang, the Rohingya community already residing in Malaysia funds traffickers to bring in the rest of their family to Malaysia from Cox's Bazaar, Bangladesh.They would be brought in via a sea route through the transit island of Koh Adang in Satun, Thailand, or overland via the Malaysia-Thai border.According to the Immigration Department of Penang, there are no cases of Rohingya migrants being smuggled in via the gazetted entry and exit points of Malaysia.If traffickers somehow managed to bring them in, their activities would be repelled through roadblocks set up by police or the Malaysian Border Security Agency.Rohingya migrants are illegally transported into Malaysia via a number of "back alleys," which are created by the traffickers.This interview was conducted on October 30, 2020.The lead researcher, Associate Professor Dr. Mohd Hazmi bin Mohd Rusli interviewed three immigration officers, Dian Karmilla, Mohd Shaafily bin Abdul Latif and Wan Mohd Zunaidi bin Wan Deraman at 3pm at the Immigration Department of Penang Building in Seberang Jaya, Penang, Malaysia.See Figure 2-a photo taken with the Immigration Department of Penang. 66Malaysia enforces strict laws against trafficking in people.Section 12 of the Anti-Trafficking in Persons and Anti-Smuggling of Migrants Act 2007 (Act 799) stipulates that any person who traffics in people for the purpose of exploitation will, upon conviction, be punished with imprisonment for a term not exceeding 15 years, and shall also be liable to a fine.This would also apply to any person who profits from the exploitation of a trafficked person as mentioned in Section 15 of the same Act. 67Etnik Rohingya bayar sindiket RM15,000 [Rohingyas paid RM15,000 to trafficking syndicates], harian MetrO, Apr. 10, 2020, https://www.hmetro.com.my/mutakhir/2020/04/564961/etnik-rohingya-bayar-sindiket-rm15000. 68Anti-Trafficking in Persons and Anti-Smuggling of Migrants Act 2007 (Act 799), §16. 69Section 2(1) of the Passport Act 1966 (Act 150) provides that every person entering Malaysia from any place beyond Malaysia shall produce to an immigration officer a passport and a valid visa for Malaysia (if required). 70 Nonetheless, as Malaysia is not a party to the 1951 Refugee Convention, it has consistently categorised them not as refugees, but as illegal immigrants. 73In addition, Malaysia is not a party to the Convention Relating to the Status of Stateless Persons 1954 (1954 Stateless Person Convention), either.However, as a responsible nation, Malaysia has always respected the spirit of the 1951 Refugee Convention by sheltering thousands of Rohingya migrants for decades. 74ccording to research published by the Humanitarian Policy Group in 2016, the main priorities for Rohingya migrants building a new life in Malaysia are to financially support themselves and their families and ultimately to ensure better future for their children. 75n facing COVID-19, Malaysia should amend its priorities to protect and improve the lives and livelihood of Malaysians first, which will be the utmost priority. 76owever, this does not mean ignoring in totality the interests of the Rohingya migrants already residing in this country.
Malaysia's healthcare system could still accommodate the rising number of COVID-19 patients.However, this might not be the case if Malaysia opens its gates and continuously receives Rohingya migrants coming to its shores.Malaysia is a sovereign nation having exclusive rights and jurisdictions to protect its territories.
Cramped in dilapidated boats for months, these Rohingya migrants may not only have COVID-19, but perhaps also other diseases once they arrive in the country.This would make it more difficult and complicated for the Malaysian authorities to control and contain the outbreak, ultimately affecting the wellbeing of other Malaysians.
Malaysia has been doing its best to provide protection and shelter for thousands of Rohingyas for years. 77However, conditions are no longer the same.This is the time for the government to prioritise Malaysians not only by tightening all gazetted entry and exit points, but also by identifying and destroying the "back alleys" often created by traffickers.In 2021, Malaysia is required to increase its efforts to disconnect the chain of transmission of COVID-19, so that it may rebuild its economy, which has been shattered by the pandemic.
Conclusion
As a member of ASEAN, Malaysia respects the principle of non-interference embraced by all ASEAN parties.Nonetheless, the Rohingya dilemma is no longer a domestic issue of Myanmar; it has exploded to become one of the worst humanitarian disasters in Asia as well as the world.These migrants were evicted from their hometowns in Rakhine and resettled in Cox's Bazaar, Bangladesh.Some have been relocated to Bhasan Char, an isolated island in the Bay of Bengal, by the Bangladeshi government.As COVID-19 is infesting the nation, the Malaysian government has taken the necessary steps to seal its frontiers to curb further unwarranted importation of this deadly virus.The front line-the Royal Malaysian Police, the Malaysian Maritime Enforcement Agency (MMEA), the Malaysian Armed Forces and the Malaysian Border Security Agency-have been working tirelessly to safeguard the nation's frontiers through Ops Benteng.
Malaysia has strict laws on border protection to combat trafficking of people.Although security at the gazetted points of entry and exit all over Malaysia has been enhanced, there are still encroachments in a number of so-called "back alleys," where traffickers illegally operate.Therefore, the Malaysian government must take extra measures to invest in the assets of its front line, so that its frontiers, which stretch for hundreds of kilometres both in Peninsula and Borneo, can be better monitored, supervised and protected.Smuggling of people and other illegal activities taking place along the frontiers, particularly in the "back alleys," could then be suppressed to avoid the unwarranted entry of illegal migrants.Despite Malaysia's strict laws on this matter, it may be advisable for the Malaysian government to increase the penalty prescribed in the Immigration Act 1959/1963 (Act 155), Passport Act 1966 (Act 150) and Anti-Trafficking in Persons and Anti-Smuggling of Migrants Act 2007 (Act 670), among others, as a deterrent against future offences.For example, Section 15 of Act 670 prescribes a fine of no less than RM 5,000.00 but not exceeding RM 1,000,000.00against a person who has been convicted of profiting from exploitation of a trafficked person.In comparison with billions of ringgit of profits reaped by these human traffickers, a fine of RM 1,000,000.00 is a trivial amount that would not stop traffickers from continuously committing this offence.
In fact, Myanmar regards Rohingyas residing in Rakhine as illegal immigrants.Millions now remain stateless in a number of countries, such as Bangladesh, Thailand, Indonesia, Malaysia and nations in the Middle East.As a nation that values human rights and defends world peace, Malaysia has played its part in easing the exodus of Rohingya migrants.
Nevertheless, conditions are yet improving at all.In light of the COVID-19 pandemic, Malaysia has to secure its borders and stop receiving any more illegal migrants from Rohingya that may be carriers of COVID-19.The virus is spreading easily and quickly.It is time to think of new policy towards the influx of Rohingya
Figure 1 .
Figure 1.The Eight Immigration Checkpoints at the Thai-Malaysia border
Figure 2 :
Figure 2: Officers of the Immigration Department of Penang
75
Caitlin Wake & Tania Cheung, Livelihood Strategies of Rohingya Refugees in Malaysia 'We Want to Live in Dignity' (2016), https://www.refworld.org/pdfid/57922bbb4.pdf. 76M'sians Come First during Crisis, Not Rohingyas: Dons, daily exPress, Apr. 25, 2020, https://www.dailyexpress.Malaysia has received Rohingya migrants for more than a decade.Although Malaysia is neither a party to the 1951 Refugee Convention nor to the 1954 Stateless Person Convention, it has provided shelter to them; at the moment, more than 150,000 Rohingyas are seeking refuge in Malaysia. | 2021-06-29T13:12:26.864Z | 2021-05-30T00:00:00.000 | {
"year": 2021,
"sha1": "24f721c709b24347e7f828578ca95037003fd501",
"oa_license": "CCBYNC",
"oa_url": "http://journal.yiil.org/home/pdf/publications/2021_14_1_pdf/jeail_v14n1_06.pdf",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "985e3912690e75716edfdfd491c67e8eba61de54",
"s2fieldsofstudy": [
"Law",
"Political Science"
],
"extfieldsofstudy": [
"Political Science"
]
} |
86776524 | pes2o/s2orc | v3-fos-license | Performance Optimization of Wireless Sensor network for Multimedia services
WSN multimedia services are currently being used by several types of applications. These services experiences performance related issues in the form of slow data rates, packet loss etc. due to stringent resources. Lots of solutions are contributed to develop a system that optimizes the resources utilization, performance and packet loss. Performance evaluation for enhancing QoS of WSN is also required where packets arrive and depart the service facility (sink) in burst. In this paper, novel analytical approach is developed for packets arrival from sensor node in burst and departure of services after getting the facility, in burst. The evaluation has been performed by shifting coefficient of variance for the time when services are provided in burst and when services are arrived on different scales of utilization of the sink node. At the end of this paper quantitative results are compared after applying priority to data packets to check credibility of proposed solution.
Introduction
Rapid Technological progress in micro-electro-mechanical systems (MEMS) and radio components enabled a new generation real-time embedded Wireless Sensor Networks (WSNs) [1,2]. These devices communicate with each other over physical environment, perhaps for a short duration, with common capabilities of distributed sensing, storing and for collaboration in a more convenient manner. These devices can be disseminated under extreme environment and can act as monitoring shield at a safer distance. The research is motivated by the fact that utilization efficiency of wireless sensor networks (WSN) is enormous in many fields such as surveillance, industrial and consumer applications, health monitoring, and so on [3,4,5] whereas on the other hand the lack of proper thoroughly evaluated communication protocols is a major factor in its slow and restricted growth. Due to remote deployment of WSNs, the major challenge is the power supply [6,7] since lifetime of the sensor nodes will shorten, if there is any imbalance in energy consumption across the network due to premature failure of devices [8][9][10][11][12]. The primary focus of WSNs network is different from traditional wireless networks such as Adhoc and cellular wireless networks. The necessary goal of interest is to optimize constraints on quality of services such as computational speed, bandwidth, signal-noise distortion ratio, delays and loss. Transmission of data through multimedia WSNs demands high communication bandwidth and reliability for better collection of largely chunked data. Though, some applications of WSNs beside aforementioned characteristics require high quality. In this scenario, major factor in quality of service (QoS) metrics in WSNs is in reliability of network and quality. For interfacing between sensors and energy, selecting relevant microcontroller component and transceiver plays an important role, where data controls should be organized in such a way that incoming data packet loss is minimized. The scheme of multimedia traffic in WSN with Quality of Service (QoS) assures the required performance and analysis modeling of nodes of Queuing Network Models (QNMs). This paper emphasizes on performance evaluation of a non-preemptive scheme for high data rates in WSN, given Blocking probability (PB) will illustrates clearly that how system specification be selected so that blocking state could be minimized. To evaluate the WSN the packets arrive in burst and depart in burst, we have first derived analytical expressions for performance evaluation, followed by numerical results.
Related work
To minimize packet loss in [1] buffering mechanism was improved in MAC layer so that the arriving packets may be prevented by minimizing retransmission of packets. It was proposed that relevant transceivers and microcontroller equipment be designed to minimize packet loss data controls without compromising QoS. In [2] an adaptive queue management scheme is proposed that constraints the average queuing delay in a router to a required value through the mapping relationship between queuing delays and queuing thresholds. Packets are dropped when the queue length reaches the threshold. In [26] a priority based scheme is proposed where the packets with higher priority are transferred at higher data rates and low priority packets are transferred with low data rates to minimize end to end delay. The model is for single arrival and single departure of packets and follows poisson distribution. In [25] the authors carried out performance evaluation for buffer less queue for WSN, an analytical solution was devised to analyze the blocking of packets. The packet arrival process is poison and follow single arrival single departure phenomena. The results shows that the blocking of packets increase with the increase in arrival rate. We have proposed an analytical solution followed by numerical results where traffic arrival and departure is in burst. The packets arrive from different nodes to sink node in batches and leave the sink node after getting processed in batches. The blocking of packets is calculated by varying Squared co efficient of variance for inter arrival and inter service times and by increasing the utilization of sink node.
Model description
To derive the analytical solution the behavior of traffic is modeled. We assume the case that no packets leave without service when the system is empty and packets arrive at the sink node get service. Also when packet arrive at the sink node and the queue is full; the packets will be blocked and these lost packets leave the system and never return to the system again. Due to high band width and high data rates the packets do arrive at the sink node in burst and are serviced in burst. These packets follow generalized exponential distribution for arrival process and batch poisson distribution for service. The analytical expressions for blocking, number of packets processed, number of packets arrived and number of packets in service are derived by taking the limit of first branch of receiver and server to infinity of Two stage hyper exponential model. The description of the model is as follow The inter arrival and inter service time for two state hyper exponential model are independent and exponentially distributed with mean 1/λ_2 at receiver centre branch and 1/μ_2 at service centre branch respectively. A packet that enters the system can either choose branch 2 with probability 1-σ or with probability σ can choose second branch, and at the server centre can choose branch 2 or 1 with probability 1-γ or with probability γ respectively as shown in Figure 1(a).
GE Inter arrival and service time are independent and identically distributed according to GE distribution and geometrically distributed batches are of renewal one [5], therefore mean bulk arrival rate is σλ and mean bulk departure rate is γμ. By substituting λ_2 with σλ and μ_2 with γμ, we get solution for censored GE/BPP/1/N as follows: State transition diagram is given in figure
Results
We have compared priority solution with non priority system and results are generated for Blocking and Service time for data, voice and video packets using Matlab. The results are shown in figures 3(a)-3(f). The performance of the aforementioned model has significantly increased the performance by reducing service time required for packets to get through the system and has reduced the blocking.
In figure 3(a) blocking of priority and non priority data packets is shown; from figure we can see that there is a significance difference between the two blocking probabilities with the increase in arrival rate. The blocking of non priority packets is above 0.8 whereas of priority data packets remain between 0.0-0.4. In figure 3(b) blocking of voice packets against that of non priority packets is shown. The blocking for voice is nearly equal to zero whearas that of the non priority system is above 0.8. In figure 3(c) blocking of video packets against that of non priority packets is shown. The blocking for video is nearly equal to zero whearas that of the non priority system is above 0.8. The blocking of video packets is much less than the blocking of packets in priority scheme. That eventually improve the QoS, as shown in figure 3( c ). The time it takes to enter system, get service and leave system is higher for non priority data packets as compared to priority scheme and same is true for video and voice packets as shown in figure 3(d)-3(f).
Conclusion
In this paper analytical solution was devised for WSN where packets arrive in burst, get service and depart the service facility in burst. The proposed priority model was compared with the nonpriority model and it has showed that the priority scheme has significantly increased the performance of the system by decreasing the blocking of packets and service time of packets.
In future we will work on simulation model of the proposed analytical model to investigate the accuracy of analytical model with simulated model. | 2019-03-28T13:14:41.348Z | 2018-10-07T00:00:00.000 | {
"year": 2018,
"sha1": "0b90259422cdb575da32ea06a7ac32a264fc789d",
"oa_license": "CCBY",
"oa_url": "https://sciencepubco.com/index.php/ijet/article/download/23021/11460",
"oa_status": "GREEN",
"pdf_src": "Anansi",
"pdf_hash": "ea5b76bdfc89a4bbfa56725132777d4074130ca2",
"s2fieldsofstudy": [
"Computer Science",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
239882089 | pes2o/s2orc | v3-fos-license | Tribological characterization of Fused Deposition Modelling parts
The customisation or redesign of parts for Additive Manufacturing (AM) to meet the design requirements is an increasing trend. In this context, numerous studies related to the improvement of the mechanical properties of Additive Manufacturing parts used for static applications are emerging. However, the use of these parts in dynamic applications or in relative movement situations has not been deeply developed. Some studies have been focused on the fatigue properties of AM parts while few authors have analysed the sliding behaviour of these parts. This paper presents a characterisation of the wear behaviour of Fused Deposition Modelling (FDM) parts using Pin-on-Disc techniques with the aim of studying their possible implementation in dynamic applications.
Introduction
Traditionally, the manufacturing processes have been understood as the transformation of a starting material into a product through the application of energy [1,2]. The different industrial revolutions have modified this concept up to date, when a new technological revolution is taking place. This revolution is supported by the development of a new forms of manufacturing and the development of new processes, such as Additive Manufacturing (AM). This process, opposed to the traditional subtractive technologies so widely used in the industry, improves the energy and environmental aspects without renouncing the possibility to produce highly complex parts and dispensing with any tooling (clamping, moulds, etc.) [1,3].
According to ASTM F2792-12a, AM can be considered as a process of adding or joining materials, usually layer by layer, to create objects from 3D CAD computer models, as opposed to its subtractive opposite, machining, where material is cut or removed.
While conventional manufacturing is governed by processing constraints related to industrial mass production, AM is inherently agile, allowing faster changes in design. This fact makes easier the manufacturing of customised objects designed to meet the demands of specific people and applications [4,5]. Highly developed in recent years, AM not only provides parts with a certain geometric accuracy, but also with mechanical or surface properties. One particular process, Fused Filament Fabrication (FFF) also known as Fused Deposition Modelling (FDM), has become a very important process with a high economic and low environmental impact [4,6].
Despite the AM maturity has acquired in such a short time, certain aspects of the AM parts behaviour are not well studied. This is even more pronounced in the case of thermoplastic polymers, which [4,7]. To date, the studies in AM have focused on static analyses referring to the improvement of tensile strength [7][8][9], compressive strength [8,10,11] or process control to obtain customised parts by modifying their internal structure with different patterns [12]. Nevertheless, far too little attention has been paid to moving parts or dynamic analyses of AM parts. A search of the literature revealed few studies referred to classical applications such as vibration on [11] or fatigue [13], but the effects of the tribological properties have not been closely examined.
The use of polymers in anti-friction applications is not so rare [14]. They are usually doped with some type of additive, such as reinforcing agents or lubricants, to improve their tribological properties. This fact is not usually compatible with the improvement of mechanical strength. For this reason, a compromise solution has to be achieved to simultaneously improve mechanical and wear properties [14,15]. Therefore, achieving parts that combine both properties is a challenge that can be transferred to the case of polymeric AM, where FDM is the main exponent of this technology.
Initially, the study of wear and friction of polymeric materials presents many similarities with to metallic materials. Nevertheless, there are significant differences in the involved wear mechanisms and, as a consequence, in the level of friction that occurs. These differences can be exploited to produce materials usable in friction parts, which can change the commonly accepted expectations of their tribological performance [14].
Mohamed et al. [16] studied the case of PC-ABS (Polycarbonate (PC), Acrylonitrile Butadiene Styrene (ABS) bonding) and determined that the wear caused by sliding can produce delamination, deformation of the material and structural damage at a high number of cycles. Likewise, cracks can be caused at the junction of the various filaments that make up the part, which can progress beneath the stressed surface. Later studies, such as those presented by Sood et al [17] found that ABS alone shows similar behaviour. However, it was also determined that this behaviour is very sensitive to the process parameters and they decisively influenced the wear rate.
Poly Lactic Acid (PLA) has a similar behaviour. According to Roy et al. [18] under similar manufacturing conditions there do not seem to be large variations in tribological properties. As mentioned above, additives can improve this behaviour. In the case of PLA, there are a large number of filaments with PLA matrix and particles of different materials and it has been proved as a easy to additive material. For example, bronze, which has good tribological properties, improves the behaviour of PLA [19].
Additionally, as for metallic materials, an improvement of the surface design can control the wear process. Surface treatments can be applied to customized AM parts with a considerable surface improvement using PLA [20]. Similarly, the contact between the filament layers and the air gap parameters have a high importance on the surface behaviour [17,18]. Likewise, the external shape, usually semi-cylindrical, of the deposited yarns makes a difference between printed parts and continuous parts even for the same material. This fact complicates the modelling of the process [21], in which the material, the manufacturing process and the strength of the part have to be considered. Gurrala et al. [22] observed that the normal load, sliding speed and part orientation have a significant influence on the wear rate for Pin-on-disk tests of AM parts.
Finally, the wear behaviour of other materials with very interesting characteristics such as PETG (Polyethylene terephthalate glycol) or ASA (Acrylonitrile styrene acrylate) have not been properly analysed. This work presents a study of the tribological properties of these materials manufactured in FDM to fill this gap of the literature and to analyse them as possible substitutes for certain industrial applications.
Experimental Procedure
The studied samples were Fused Deposition Modelling flat specimens manufactured in different materials. The specimen geometry was 45x45 mm 2 and 3 mm thickness. They were manufactured on a standard commercial Fused Deposition Modelling equipment. The materials selected for the study were PLA, PETG, ASA and Nylon. PLA and PETG were supplied by the manufacturer SUNLU and ASA IOP Publishing doi:10.1088/1757-899X/1193/1/012068 3 was supplied by Winkle. The Nylon specimen was used for comparative purposes due to its antifriction properties and PLA specimens were selected since it is the most used material in FFF.
Full filled parts were obtained using a fixed set of parameters for every material. Extrusion speed was set up at 30 mm/s, overlap at 55%, environmental temperature at 60º and filling of 100%. Likewise, rectilinear trajectories without perimeters were selected to avoid the effect of the perimeters trajectories on the tribological behaviour of the specimens.
Nevertheless, the printing parameters were adapted to the material (table 1). Particularly, the extrusion temperature was selected based on the manufacturer's recommendations. It is understood that by using a constant temperature the mechanical parameters will not be altered. Three different layer thicknesses were used for PETG and ASA (0.15, 0.25 and 0.35 mm) to analyse their effect on the friction behaviour. Only one layer thickness was selected for PLA and NYLON (0,25 mm), since these samples were used comparison purposes. The specimens were processed using Repetier Host® software. After printing, the surface of the specimens was characterised using a stereo optical microscopy (Nikon SMZ 800) and a roughness measurement station (Mahr Perthometer PGK 120). Then, the specimens tribological behaviour was studied using a Pin-on-Disc test equipment from Microtest MT series. The surfaces were tested against a stainless steel pins. All the tests were carried out with a under 15N load, 100 rpm angular speed and 250 m sliding length.
Wear measurement were taken using the same roughness measurement station used to analyse the specimens surface. The geometric data of the groove pattern were obtained using this machine and then they were exported and processed by a CAD software. Since it is not possible to apply any existing standards, these characteristic profiles were studied measuring the minimum distance between the lowest point of the groove and the least squares regression line of the profile.
Results
Traditionally, a rough surface is related to bad wear behaviour. Therefore, it is expected to obtain high wear rates on surfaces obtained by FFF. The geometrical characteristics of these surfaces have a decisive influence on the tribological behaviour of the specimens [21]. The topography generated is not a perfect cylinder but an ovalised profile, which depends on the extrusion parameters and the material [23]. For this reason, the initial surface was characterised. Figure 1 shows the images obtained for PETG and ASA for different layer thickness. Both materials presented similar patterns, with the characteristic motor pitch effect [23]. It was also observed an effect of the layer thickness on the generated surface. For lower, layer thickness, the wave profile of the surface is transferred to the next layer, obtaining complex pattern that disappear for 0.35 mm layer thickness. However, this behaviour was not noticeable for PLA or NYLON samples ( figure 2). Particularly, the NYLON topography seems less periodic. This was probably due to the plasticity of the material, which causes problems during its extrusion.
This behaviour was checked analysing the roughness of each specimens ( figure 3). PTEG and ASA presented better roughness results than PLA and NYLON. The worst roughness results were observed for NYLON, verifying the optical analysis. PETG and ASA showed stable average behaviour for the different layer thickness. Although as it can be seen in the error bars, there is a large variability in terms of roughness. This behaviour is much more pronounced in the case of PETG. This fact may be produced by the plasticity of the material which has a tendency to create bubbles. Regarding the tribological results, on the one hand, ASA samples presented a high stability in its wear behaviour. Its friction coefficient tended to stabilize at 0.4 ( figure 5). This value is close to the one obtained for ABS parts [17,18] and lower than PLA parts, even they are mixed with metallic additives [19]. This stabilization was achieved in the first instants for the highest layer thickness (0.25 and 0.35 mm). This fact proved that for ASA samples, the roughness of the process does not have a negative impact. On the other hand, PETG samples had a different behaviour. Although the coefficient of friction is the lowest for the studied materials, it had a clear dependence on the layer thickness. There was found no clear relation between the roughness and the average friction coefficient. Nevertheless, the tests for 0.25mm layer thickness, the sample with the highest roughness, presented aperiodic peaks. These picks were probably caused by the material dragging phenomena, where particles were detached from the surface and deposited on both the periphery of the groove and on the pin ( figure 6 and 7). It should be considered this phenomena had not damaged the piece or broke the fibres as in ASA samples (figure 7) and it was not observed in in PETG or NYLON samples. This fact may be related to the material properties. For low friction coefficient, particles do not detach from the sample but the sliding phenomenon occur anyway. This is clearly seen when comparing the four materials under the same layer thickness conditions (figure 8), where the best results were obtained for NYLON and PETG. PLA ASA Figure 6. Pin adhered material for PLA and ASA for 0.25 mm layer thickness. The materials with lower layer thicknesses do not show particle detachment and those with higher layer thicknesses do. This will also be related to the depth of the groove that appears. Low friction materials had no significant grooves, while for materials with a higher coefficient the groove is noticeable and measurable ( figure 9). However, this had not a direct relationship. For instance, PETG plasticity prevented particles from detaching easily, on the contrary dragging created jumps and even burrs on the wear track. In fact, it was precisely on the most visible jumps where the deeper grooves appeared ( figure 9). This proved that the material characteristics combined with the printing parameters IOP Publishing doi:10.1088/1757-899X/1193/1/012068 7 influenced the wears behaviour. Therefore, processing parameters such as extrusion temperature had an impact on the final surface behaviour and its optimisation will be critical for industrial applications.
Conclusions
This paper studied the tribological behaviour of Fused Deposition Modelling samples manufactured in PETG; ASA, PLA and NYLON. It has stablished an interesting base to study the dynamical friction behaviour for industrial applications.
An approach to the friction behaviour of PETG and ASA had been made. This two interesting materials were selected due to their special characteristics and their good environmental performance. PETG friction coefficient results were comparable to NYLON under low load tests. Whereas, ASA presented debris on both pin and track wear. Either ways, it could be an alternative for wet applications due to its good stability and resistance in chemical environments.
Further studies are recommended to determine the possible applications of these materials and different ones and enhance the industry competitiveness. | 2021-10-26T20:08:46.014Z | 2021-10-01T00:00:00.000 | {
"year": 2021,
"sha1": "0ae2df29f2507210e29d227dbaf1a5f8e18d7903",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1757-899x/1193/1/012068",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "0ae2df29f2507210e29d227dbaf1a5f8e18d7903",
"s2fieldsofstudy": [
"Engineering",
"Materials Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
16091142 | pes2o/s2orc | v3-fos-license | Flexibility in metabolic rate confers a growth advantage under changing food availability
Phenotypic flexibility in physiological, morphological and behavioural traits can allow organisms to cope with environmental challenges. Given recent climate change and the degree of habitat modification currently experienced by many organisms, it is therefore critical to quantify the degree of phenotypic variation present within populations, individual capacities to change and what their consequences are for fitness. Flexibility in standard metabolic rate (SMR) may be particularly important since SMR reflects the minimal energetic cost of living and is one of the primary traits underlying organismal performance. SMR can increase or decrease in response to food availability, but the consequences of these changes for growth rates and other fitness components are not well known. We examined individual variation in metabolic flexibility in response to changing food levels and its consequences for somatic growth in juvenile brown trout (Salmo trutta). SMR increased when individuals were switched to a high food ration and decreased when they were switched to a low food regime. These shifts in SMR, in turn, were linked with individual differences in somatic growth; those individuals that increased their SMR more in response to elevated food levels grew fastest, while growth at the low food level was fastest in those individuals that depressed their SMR most. Flexibility in energy metabolism is therefore a key mechanism to maximize growth rates under the challenges imposed by variability in food availability and is likely to be an important determinant of species’ resilience in the face of global change.
Introduction
Organisms often live in variable environments where conditions such as food availability or temperature can change considerably over their lifetime. The ability of individuals to alter their physiological, morphological and behavioural traits, that is their phenotypic flexibility, can have important consequences for their own fitness as well as the extent to which whole populations can adapt to or cope with new and changing environments (Nussey, Wilson & Brommer 2007;Bolnick et al. 2011). Given recent climate change and the degree of habitat modification currently experienced by many organisms, it is therefore critical to quantify the degree of phenotypic variation present within populations, individual capacities to change and what their consequences are for fitness (Hofmann & Todgham 2010;Tuomainen & Candolin 2011).
Flexibility in metabolic rate may be particularly important since metabolism reflects the energetic cost of living and is one of the primary traits underlying organismal performance. Standard metabolic rate in ectotherms (SMR; equivalent to basal metabolic rate -BMRin endotherms) represents the baseline energetic expense of maintaining the systems and processes critical to sustain life (Fry 1971). SMR and BMR (referred to as SMR hereafter) are known to vary extensively among individuals within a population, even after controlling for differences in mass, age and sex (Kvist & Lindstr€ om 2001;Labocha et al. 2004;Steyermark et al. 2005). This intraspecific variation in SMR, in turn, has been linked to individual differences in fitness through its association with various behavioural and life-history traits (reviewed in Careau et al. 2008;Biro & Stamps 2010;Burton et al. 2011;Mathot & Dingemanse 2015) and, as such, is considered a key physiological trait.
Standard metabolic rate is generally repeatable over time when measured under constant conditions (Nespolo & Franco 2007). However, there is also increasing evidence that it is flexible and can change in response to food availability in as little as a few days to weeks (Guppy & Withers 1999;McKechnie 2008;McCue 2010); studies now show that SMR increases when food availability is raised and decreases when food levels decline in many different taxa, including insects (Roark & Bjorndal 2009), fishes (Van Leeuwen, Rosenfeld & Richards 2012), amphibians (Hervant, Mathieu & Durand 2001), reptiles (McCue 2007), birds (Wiersma, Salomons & Verhulst 2005) and mammals (Ostrowski, M esochina & Williams 2006). However, individuals can differ in the degree to which their SMR changes as a function of food level (O'Connor, Taylor & Metcalfe 2000;Fu, Xie & Cao 2005). An individual's metabolic flexibility and therefore its capacity to respond to changing food availability may have important consequences for fitness, but to our knowledge, the impacts of such flexibility have not been quantified.
Metabolic flexibility may be especially critical for somatic growth during the juvenile stage, a key determinant of fitness in many animals because of its effects on body size (Sogard 1997;Gebhardt-Henrich & Richner 1998;Blanckenhorn 2005). There is some evidence suggesting that individuals with a higher SMR grow faster under high food levels, while a lower SMR is associated with faster growth at low food levels (Derting 1989;Killen, Marras & McKenzie 2011;Reid, Armstrong & Metcalfe 2011). A higher SMR may therefore be beneficial at high food levels but can consume a disproportionate amount of an individual's daily energy budget when food becomes scarce (Burton et al. 2011). If these are causal relationships between SMR and growth, there could be significant growth advantages for those individuals that show the biggest upregulation of SMR when food becomes abundant and the greatest decrease in SMR when food becomes scarce.
We examined individual variation in metabolic flexibility in response to changing food levels and its consequences for somatic growth. We chose juvenile brown trout (Salmo trutta) as our model species because larger body size often confers an advantage in their competitive interactions (Johnsson, N€ obbelin & Bohlin 1999) and survival (Einum & Fleming 1999;Carlson, Olsen & Vøllestad 2008), so early growth rates can have important consequences for fitness. However, food availability can exhibit marked spatial and temporal variation in the freshwater streams they inhabit (Elliott 1970;Lagarrigue et al. 2002;Martin-Smith & Armstrong 2002), so flexibility in their metabolic rates may be critical to growth.
feeding regime and growth measurements
Young-of-the-year brown trout (n = 120) were collected from a small tributary to the River Endrick, Scotland, and brought to the University of Glasgow where they were measured for body mass and fork length and then housed in individual compartments in two separate stream tank systems in a temperaturecontrolled room (11Á5 AE 0Á5°C) with a 12L:12D cycle. Fish were fed an intermediate ration of Inicio Plus trout pellets (Bi-oMar Ltd, Grangemouth, UK) twice daily based on their body size (Appendix S1a). Fish mass and length were measured after 2 weeks to adjust feeding rations to changes in their body size.
Standard metabolic rates were first measured in each fish after they had been on the intermediate ration for 28 days (see respirometry section below). Equal numbers of fish were then switched to one of three rations (low food, ad libitum and the same intermediate ration as before) until their metabolic rates were measured again 28 days later. The three ration levels were determined using equations from Elliott (1976) that describe the growth of brown trout as a function of caloric intake, temperature and initial body size (Appendix S1a). Fish were then fed twice daily, once in the early morning and once in the late afternoon, and allowed 1 h to consume each meal before leftover food and faecal matter were siphoned from their tanks. The fork length and body mass of each fish were remeasured each week, and rations were adjusted to account for increases in body size. Daily growth (in mm) for each fish over the 28-day period was calculated as (length day28length day0 )/28 days. respirometry Standard metabolic rate was measured as the rate of oxygen consumption using continuous flow-through respirometry. The experimental set-up consisted of 16 respirometry chambers (volume 400 mL) arranged in parallel and submerged in a water bath. A peristaltic pump (Cole Parmer, London, UK) pumped water from an upper aerated bin, through the chambers, passed an oxygen sensor, and into a lower bin before it was recirculated back to the upper bin. A UV sterilizer (v2 Vecton 600, Tropical Marine Centre, Bristol, UK) helped to minimize background respiration rates. A chiller (Teco Tr5, Ravenna, Italy) maintained the water bath at 11Á5 AE 0Á5°C. Sheets of insulation were wrapped around the bin containing the respirometry chambers to prevent light penetration and thereby reduce fish activity. Multichannel oxygen meters (FireStingO2, PyroScience, Aachen, Germany) and attached sensors recorded oxygen levels in the chambers every 2 seconds using FireSting software version 3.0 (PyroScience). Flow rate of water through the chambers was set at 1Á47 L h À1 for the first metabolic measurements and then increased to 1Á68 L h À1 for the second measurement 28 days later (since the fish had grown) to ensure that oxygen consumption rates were detectable but oxygen levels remained above 80% saturation.
Fish were not fed for 48 h prior to measurements. This is an appropriate time interval to allow gut evacuation at the test temperature and thereby prevent the energetic costs of digestion, tissue synthesis and deposition from influencing measurements of standard metabolic rate (Higgins & Talbot 1985;Secor 2009;Rosenfeld et al. 2014). Fish were then placed in individual respirometry chambers, and their oxygen consumption was measured continuously over a 20-h period (from roughly 1400 to 1000 the next day). The system permitted the simultaneous continuous measurement of oxygen consumption rates of 15 fish each day (a total of 8 batches over 9 days), with a fish-free chamber serving as a control measure of background respiration. Standard metabolic rate (mg O 2 h À1 ) was measured as follows: where V w is the flow rate of water through the respirometry chamber (L h À1 ), and C wO2 control and C wO2 fish are the concentrations of oxygen (mg L À1 ) in the outflow of the chambers lacking and containing fish, respectively (Clark, Sandblom & Jutfelt 2013). SMR for each fish was calculated by taking the mean of the lowest 10th percentile of oxygen consumption measurements over the 20-h measurement period and then excluding outliers, that is those measurements below 2 standard deviations from this mean (Clark, Sandblom & Jutfelt 2013).
statistical analyses
We first examined the relationship between double log-transformed SMR and body mass. Since body mass can influence both metabolic and growth rates, mass-independent estimates of individual SMR (rSMR), that is residuals from the regression, were used in subsequent analyses (Appendix S1b). We then tested whether rSMR changed as a function of food regime over the 28day period. Food regime, measurement time (i.e. initial and final measurement of rSMR) and their interaction were included as categorical effects. The interaction between food regime and measurement time was statistically significant, so changes in rSMR were further evaluated by testing whether the final metabolic measurement was significantly different from the initial measurement for each food regime. We then used this same approach to test whether initial rSMR was a good predictor of the change in rSMR (DrSMR: final rSMRinitial rSMR), after correcting for regression to the mean (Kelly & Price 2005).
Finally, we tested whether DrSMR over the 28-day food treatment period had consequences for growth and whether those consequences depended on food regime. The model included food regime as a categorical effect, initial length as a continuous predictor since growth is often a function of body size, DrSMR as a continuous predictor and initial rSMR as a continuous predictor since this could also influence growth rates. Interactions among initial rSMR and DrSMR with food regime were also included. Error variances of growth (v 2 = 37Á5, d.f. = 3, P < 0Á001) differed among food treatments, so their error variances were modelled separately. The interaction between DrSMR and food regime was statistically significant, so the effects of DrSMR on growth were further evaluated for each food regime. Growth rates at each food level may be a function of the final rSMR rather than DrSMR. However, final rSMR and DrSMR were highly positively correlated (Pearson's r = 0Á76, P < 0Á001), so the effects of final rSMR and DrSMR were evaluated in separate models to avoid problems associated with their multicollinearity. Their Akaike information criteria (AIC) values were used to compare the models. The structure of the model including final rSMR was the same as described above for DrSMR.
In all analyses, batch number (15 fish per group) was included as a random effect to account for the order in which fish entered the experiment. Fish identity was also included as a random effect when repeated measures of an individual's SMR were included in the same analysis. Analyses were conducted using the proc MIXED procedure in SAS version 9.3 (SAS Institute, Cary, NC, USA). Growth rate, final rSMR and DrSMR were modelled as continuous traits, but we refer to their extremes in the results and discussion section as a means of comparing individuals with contrasting responses. Effects were considered significant when P < 0Á05. All means given are AE 1 SE. Of the 120 fish, only 116 were included in analyses because 2 fish regularly consumed less than half their daily rations, and the final standard metabolic rate measurements of 2 fish were compromised by faulty tubing.
Discussion
Standard metabolic rate changed a considerable degree in response to food availability. While the general trend was for SMR to increase when food levels were raised and decrease when food levels declined, individuals exhibited markedly different responses to the same change in food availability. The result was that those individuals who increased their SMR more had a higher final SMR than those whose metabolic rates were inflexible or decreased. These differences in flexibility and their subsequent effects on final SMR, in turn, had consequences for their growth rates. Individuals that increased their SMR more in response to elevated food levels grew fastest, while growth under low food levels was faster in individuals that depressed their SMR most. These differences in metabolic flexibility were associated with over 3-fold differences in growth among some individuals even when accounting for food level.
Both final SMR and growth were higher at higher food levels, but links between SMR and growth were not always positive and changed as a function of food level. Together, these results suggest that the relationship between SMR and growth may be causal and not just an artefact of their mutual dependence on food availability (Van Leeuwen, Rosenfeld & Richards 2012;Rosenfeld et al. 2014). Furthermore, they point to the processes underlying SMR as being the cause and not the effect of individual variation in growth. If the latter were true, then we would expect a positive relationship between SMR and growth regardless of food level rather than the negative one we observed at low food levels. Indeed, shifts in SMR are thought to reflect underlying changes in digestive and assimilative processes (Armstrong & Bond 2013), mitochondrial efficiency (Monternier et al. 2014) and/or respiratory substrate use (McCue 2010) that may be critical for growth and explain observed relationships between metabolic rates and growth across the spectrum of food availability. The changes in SMR we observed also highlight the need to control for variation in food intake among individuals in studies aimed at quantifying individual variation in metabolic rates and their links to other traits and ultimately fitness under common conditions. Studies across a wide range of taxa have often used animals that are reared in communal tanks or cages (e.g. McCarthy 2000; Terblanche, Klok & Chown 2004;Steyermark et al. 2005) where the establishment of dominance hierarchies can lead to asymmetries in food intake (Sloman & Armstrong 2002;Herberholz, McCurdy & Edwards 2007) that may affect subsequent measurements of standard metabolic rate. Isolating individuals and feeding each a standardized ration, as we have done here, should help to alleviate these confounding effects of food intake.
An individual's initial SMR had no effect on the degree to which its SMR changed nor its subsequent rates of growth once food levels changed. This suggests that knowledge of an individual's SMR can allow us to predict its performance under the current conditions, but not necessarily if those conditions change. Furthermore, it argues that patterns of food availability, that is constant versus variable, and the time an organism has spent under those conditions need to be taken into account since relationships between metabolic rates and performance, much like when evaluated at different temperatures (Gr€ ans et al. 2014), may depend on whether an organism is acutely exposed or acclimated to current food conditions. The link between the change in SMR, but not an individual's initial SMR, and growth under changing food levels also points to an organism's flexibility in metabolic rate (and thereby its capacity to respond to changing food availability) as a potentially more important target of selection than its metabolic rate at only one point in time. Individuals with a higher degree of metabolic flexibility may have a selective advantage in variable environments. However, it was not just the degree of flexibility per se but also the direction of change and the subsequent final SMR that were important for growth. Although the majority of individuals in our study shifted their SMR in the direction that promoted higher growth, some increased and decreased their SMR at low and high food levels, respectively, even at expense to their growth. These latter responses may be maladaptive or represent alternative strategies to maximize other aspects of fitness. For example, increasing SMR may serve to preserve somatic maintenance and repair when food becomes scarce, while decreasing SMR in response to high food levels may help to maximize metabolic scope (aerobic capacity above that of SMR) for other functions such as digestion and activity that SMR might otherwise impinge on. However, further study is needed. Additionally, it is unclear whether some individuals are more flexible in general or only within a limited range of food availability.
Flexibility in SMR can allow individuals to maximize their growth under the constraints imposed by variability in food availability, as we show here. Given that SMR is a fundamental trait underlying many aspects of organismal performance, it likely plays an equally important role in allowing organisms to cope with changes in many environmental factors in addition to food availability. Indeed, there is increasing evidence that SMR can change in response to factors such as diet quality (Naya, Lardies & Bozinovic 2007), temperature (McKechnie, Chetty & Lovegrove 2007), hypoxia (Hochachka et al. 1996) and salinity (Allan, Froneman & Hodgson 2006). In addition to its effects on growth, this flexibility may also be a key determinant of how other components of fitness such as reproduction and survival are affected by environmental change. However, the costs, benefits and ultimate fitness consequences of these shifts in metabolism are rarely quantified explicitly. Given the pressing need to gauge the capacity of species and populations to cope with current rapid environmental change, we therefore need more studies that evaluate not just how metabolic rates change in response to different environmental factors, but also what their consequences are for individual fitness and subsequent population vital rates and dynamics. Fig. 3. Growth in fork length as a function of change in standard metabolic rate in juvenile brown trout. SMR was measured after fish had been on an intermediate ration for 28 days and then measured again after they were switched to a low, intermediate or ad libitum ration for an additional 28 days. Plotted are partial residuals of growth rate evaluated at mean fork length = 91 mm and mean initial rSMR = 0 as a function of back-transformed values for the change in SMR standardized for the mean body mass of 10Á2 g. three anonymous reviewers and an associate editor whose comments improved the manuscript. All procedures were carried out under the jurisdiction of the UK Home Office project license (PPL 60/4292) governed by the UK Animals Scientific Procedures Act 1986. This research was supported by an ERC Advanced Grant (number 322784) to NBM.
Data accessibility
Data have been archived in Appendix S1d of the online Supporting Information. | 2016-05-04T20:20:58.661Z | 2015-06-10T00:00:00.000 | {
"year": 2015,
"sha1": "fde604e218f48e671a308780f5aa11150e394fc5",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.1111/1365-2656.12384",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "fde604e218f48e671a308780f5aa11150e394fc5",
"s2fieldsofstudy": [
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
119333111 | pes2o/s2orc | v3-fos-license | Projective Representations Construction for Different Points of Brillouin Zone. Application for Space Symmetry Groups $P4_1 2_1 2$ and $P4_3 2_1 2$
It was suggested method of constructing of irreducible projective representations in different points of Brilluin zone. Points \Gamma, \Lambda, Z, S, A, \Sigma, M, V, R, and X of space symmetry groups $P4_{1}2_{1}2$ and $P4_{3}2_{1}2$ were examined. At each of these points one- and two-valued irreducible projective representations of wave vector groups were constructed for two enantiomorphous modifications. Influence of time inversion at these points also was taken into consideration by means of Herring criterion.
. Schematic drawing of Brillouin zone which corresponds to tetragonal crystals of space symmetry groups Р4 1 2 1 2 and Р4 3 2 1 Space symmetry groups P4 1 2 1 2 and P4 3 2 1 2, are very interesting in practice, because they describe two enantiomorphous modifications of crystals, which characterised by appreciable anisotropy of the lattice and have quite high symmetry at the same time. So it is good opportunity to develop the general group theory method, that can allow to analyse energy states in such crystals.
In this paper we propose method of constructing of irreducible projective representations in different points of Brilluin zone of mentioned space groups. Here we examine points Γ, Λ, Z, S, A, Σ, M, V, R, and X, (Fig.1) but detailed calculation we will do only for points Γ, Λ, and Z, for other points procedure is similar, and we will represent final result (see appendix can be defined as following: is the matrix of irreducible representation of projective class the same as corresponding quotient system.
For points Γ, Λ, and Z we will take the following wave vectors as a canonical: So, when centre of Brilluin zone corresponds to point (0,0,0), we consider points with negative value. Wave vector groups are equal for points Γ and Z, and are equal to full space group G with elements g. Point Λ will be examined separately, because wave vector group for this point is differ from G and is isomorphous to point group 4.
Primitive elements h і = g і they determine wave vector groups, which can include only non-trivial translations, are defined as following ( Fig.2): for group P4 1 2 1 2, and for group P4 3 2 1 for points Γ and Z can be constructed with formula: (2) It is obvious that all elements of ) ,r (r ω Note: for group P4 3 2 1 2 quotient system ) ,r (r ω 1 2 1 can be constructed in the same way. These quotient systems belong to class K 1 , because for commutating elements of this group the ratio ) , ( , n is fold of main axis (in our case n = 4). For group P4 1 2 1 2 m = 1, and for group P4 3 2 1 2 m = 3.
Formulas (3), (4) is used for generating elements b р and a κ . For elements r = b р a κ following formula is used: (5) Table 2 Functions This quotient systems, which is equal to standard ) , ( The quotient system ) , r r ω is defined as following: where θ is the rotation angle, which corresponds to product of elements 1 2 r r . This quotient system is the same for points Γ and Z, because it is determined only by symmetry group of wave vectors directions.
Procedure of reducing the quotient system ) , The characters of irreducible projective representations, which correspond to standard quotient system of K 1 class [1] are presented in tab.5. Characters of irreducible projective spinor representations (tab.6) can be obtained by multiplication of characters of tab.5 by corresponding functions ) ( 2 r u (these characters are equal to characters of double group (422)').
It is necessary to calculate values of ) ( a k i e + − α to construct characters of irreducible projective representations for points Γ and Z. Values of vectors a (tab.7) for elements of groups P4 1 2 1 2 and P4 3 2 1 2 can be calculated using above mentioned vectors of non-trivial translations. We calculate characters of irreducible projective representations for points Γ and Z (tab.8 and 9) by substitution matrix of irreducible projective representations ) (r D on their characters ) (r D χ , which correspond to quotient systems of K 0 and K 1 classes for states with integer and half-integer spin. It should be noted, that characters of irreducible projective representations for space groups, which describe two enantiomorphous modifications are the same.
For points S, A, Σ, M, V, R, and X characters of irreducible projective representations is shown in appendix. The merging and doubling of representations are caused by degeneracy of states due to invariance with respect to time inversion. Ganging and doubling were made according to the Herring criterion [1].
There is a principal possibility to examine transformation characters of irreducible projective representations along Oz axis from point Γ up to point Z. Therefore we have to determine characters of irreducible projective representations for point Λ as a function of wave vector Λ k and 2 0 1 b k − < < Λ (tab.10 and 11). As it was mentioned the wave vector group of point Λ is isomorphous to point group 4. Group 4 is sub-group of limiting symmetry group (or Curie group). The limiting symmetry groups include the infinite-fold axis as symmetry element, and these groups are enantiomorphous, subordinate point groups are enantiomorphous too [2]. So, it is necessary to examine right point group 4 (it has screw tetrad axis 4 1 ) and left one 4 (it has screw tetrad axis Table 11. Characters of irreducible projective representations of wave vector group of group Р4 3 2 1 2 for point Λ Thus, the method, we suggest, gives the possibility to examine any point of Brilluin zone, to find characters of irreducible projective representations, and to determinate functions of wave vector, which describe transformations representations from one point to another. All these data can be used for determination of selections rules of different types of energy excitations, for analysing whether examined points are points of zero slope and for construction of dispersion curves for different directions of wave vector. | 2019-04-14T02:05:52.285Z | 2004-06-22T00:00:00.000 | {
"year": 2004,
"sha1": "3a4dcf01692d71a9a28453d11e5d26b9fe953ec9",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "3a4dcf01692d71a9a28453d11e5d26b9fe953ec9",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Physics"
]
} |
17517839 | pes2o/s2orc | v3-fos-license | Tri-nucleotide threading for parallel amplification of minute amounts of genomic DNA
Efforts to correlate genetic variations with phenotypic differences are intensifying due to the availability of high-density maps of single nucleotide polymorphisms (SNPs) and the development of high throughput scoring methods. These recent advances have led to an increased interest for improved multiplex preparations of genetic material to facilitate such whole genome analyses. Here we propose a strategy for the parallel amplification of polymorphic loci based on a reduced set of nucleotides. The technique denoted Tri-nucleotide Threading (TnT), allows SNPs to be amplified via controlled linear amplification followed by complete removal of the target material and subsequent amplification with a pair of universal primers. A dedicated software tool was developed for this purpose and variable positions in genes associated with different forms of cancer were analyzed using sub-nanogram amounts of starting material. The amplified fragments were then successfully scored using a microarray-based PrASE technique. The results of this study, in which 75 SNPs were analyzed, show that the TnT technique circumvents potential problems associated with multiplex amplification of SNPs from minute amounts of material. The technique is specific, sensitive and can be readily adapted to equipment and genotyping techniques used in other research laboratories without requiring changes to the preferred typing method.
INTRODUCTION
Following the complete sequencing of the human genome, efforts have intensified to decipher the genetic code and ascertain the relationships between genetic variations and phenotypic differences, e.g. variations in disease susceptibility, drug responses and cancer ontogeny. Such analyses of genetic variations also have applications in diverse fields, such as genealogy, epidemiology and forensics. In some cases single nucleotide variants have been demonstrated to be associated with specific disease states, but in the vast majority of cases it is unlikely that a single polymorphism can be causally linked to a certain trait. Instead, certain combinations of variants among the millions of single nucleotide polymorphisms (SNPs) in the genome are likely to give rise to traits, such as increased risk for cancer. Consequently, accurate and sensitive analysis of SNPs may play an important role in future DNA diagnostics. Whole genome SNP typing platforms are already available to enable such high throughput scoring (1).
To facilitate these efforts, equally efficient methods for target amplification are needed. Padlock probes and linker ligated genomic tags, accompanied by amplification with universal probes, have demonstrated utility for such purposes (2)(3)(4), while use of multiplex PCR (i.e. PCR with multiple specific primers) has limitations since only a few loci can be amplified in parallel. Previous works have demonstrated that multiplex PCR with > 20 genomic fragments is problematic, in both theory and practice, due to its high complexity and complications associated with multiple, unspecific, competing annealing and extension events (5,6). These problems are circumvented in the approach presented here by reducing the number of nucleotides included in the extension reactions from four to three, leading to a predefined termination close to the extending probe. The central objective of this work is to provide a system for the simultaneous preparation and amplification of minute amounts of genomic DNA at a large number of genetic loci by a specific and sensitive tri-nucleotide threading procedure (TnT) that can be applied to large-scale scoring experiments in a single automated reaction sequence (see Figure 1).
Overall, the TnT approach involves the following steps: (i) database searching and filtering; (ii) polymerase-mediated extension with just three dNTPs (tri-nucleotides); (iii) ligation of the 3 0 end of the extended probe to the 5 0 end of a second probe (thread-joining primer), which increases the specificity of the assay; (iv) cycling of the extension-ligation steps leading to linear amplification; (v) capture of the resulting ligated products (the DNA threads), resulting in complete removal of the genomic DNA; (vi) amplification of the ligated products by a pair of universal PCR primers and (vii) genotyping of the amplified fragments. In this study, the genotyping of the amplicons was carried out using a microarray-based variant of an allele-specific extension assay denoted proteasemediated allele-specific extension (PrASE) (7). However, since TnT represents a general solution to parallel amplification of genomic DNA, any other typing method capable of analyzing a single reaction harboring several different DNA fragments can be employed.
SNP selection
The SNPs were identified in genes linked to various types of human cancers using the dbSNP (http://www.ncbi.nlm.nih. gov/SNP/), Ensembl (www.ensembl.org) and SNP500Cancer (http://snp500cancer.nci.nih.gov) databases. This was followed by a filtering procedure, which primarily involved manual selection of candidate SNPs with relatively high allele frequencies in the Caucasian population and removal of SNPs with neighboring polymorphisms (<40 residues). The sequences were automatically retrieved in Fasta format from dbSNP (build 121) using a Perl script. The script also extracted the sequences 40 bases upstream and 40 bases downstream of each SNP. Note that SNPs with shorter surrounding sequences were removed. Subsequently, the human genome sequence was screened with the candidate SNP sequences using blastn (Genome build 35) to remove SNP sequences spanning duplicated gene segments or repetitive regions as well as low-complexity regions (see the SNP selection box in Figure 2).
Algorithm and scoring
A software tool implemented in Java was designed to rank the possible tri-nucleotide combinations (AGT, AGC, ATC and GTC) and subsequently suggests two optimal tri-nucleotide reactions for a given set of SNPs (see Figure 2 the Software box and Supplementary Figure 1a).
The algorithm for ranking different tri-nucleotide combinations consists of the following steps (see also the Software box in Figure 2). (i) Extend the first SNP in silico with all possible tri-nucleotide sets for the SNP under study. The four possible tri-nucleotide sets (ACG/ACT, ACG/AGT, ACT/CGT and AGT/CGT) can be used for R, Y, K and M, while for W and S only two sets (AGT/TCA CTG/GAC) can be employed (Supplementary Figure 1b). (ii) Assign scores of 2, 4, 8, 16, 64, 64, 64, 32, 32, 32, 16 and 16 to 1-, 2-, 3-, 4-, 5-, 6-, 7-, 8-, 9-, 10-, 11-and 12-base extensions, respectively. (iii) Allocate the SNP to each of the possible tri-nucleotide reaction combinations so that each combination's total score is maximized. (iv) Repeat steps (i) to (iii) for all SNPs in the dataset. The trinucleotide combination with the highest score (rank) is the optimal choice. (v) The user confirms the selection of the combination with the highest score, or a preferred combination, and the software designs primers given amplification tags and a user-defined Tm window (see Supplementary Figure 1a). Note, in this study, the defined cutoff was 75 SNPs, and since the number of input SNPs was 88, the SNPs included in a certain combination may differ from those in another. However, the software tool is flexible, so the number of input SNPs may be exactly 75 or any number exceeding 75. The optimal trinucleotide combination for this dataset was dCTP, dTTP, dGTP (39 SNPs) and dATP, dGTP, dTTP (36 SNPs) with a total score of 2968. The 75 SNPs were located in 61 different genes, distributed amongst all but four chromosomes and included all of the different single nucleotide alterations (IUPAC codes: K, S, R, Y, W, M), as listed in Supplementary Table 1. (vi) The lengths of the final primers are adjusted to avoid PCR bias. Since the longest fragments were 115 bases long in this case, none of the DNA threads was allowed to be shorter than 104 bases (giving a maximum difference of 10%). The sequences of the primers are listed in Supplementary Table 2. (vii) In addition, the software designs genotyping primers. Given a set of signature tags, the software designs 3 0 -termini allele-specific primers (fused with signature tags at the 5 0 end) for genotyping by PrASE (Supplementary Table 3).
Primers and template
All oligonucleotides were synthesized by MWG-Biotech AG (Ebersberg, Germany). For each primer pair (later forming a DNA thread) the extension primer was biotinylated at the 5 0 end and the other was phosphorylated at the 5 0 end. The signature tags of the generic tag arrays (anti-tags) were ordered with an amino group at the 5 0 -terminus for the array preparation (2) (Supplementary Table 4).
The genomic DNA used was an ECACC Human Random Control DNA Panel (SIGMA) isolated from lymphoblastoid cells from UK population with Caucasian origins (96 samples with DNA concentrations of 8 ± 2 ng/ml, treated according to the manufacturer's instructions). In addition, genomic DNA was isolated from blood samples from ten healthy anonymous Swedish Caucasian individuals.
TnT
Two reaction mixtures containing 0.2 mM of the dXTPs 'C, T and G' and 'A, T and G', 0.01 mM of each extension primer, 0.05 mM of each thread-joining primer and 8 ng genomic DNA were prepared in 20 mM Tris-HCl (pH 8.3), 25 mM KCl, 10 mM MgCl 2 , 0.5 mM NAD and 0.01% Triton X-100 together with 2 U Ampligase (Epicentre) and 0.5 U Stoffel fragment DNA polymerase (Applied Biosystems). For each set of ten individuals to be analyzed, one reaction without genomic DNA and one reaction without ligase were included as negative controls.
The reactions were performed in a GeneAmp 9700 PCR System (Applied Biosystems) with a temperature profile consisting of an initial pre-cycling step (20 C for 5 min, 95 C for 5 min) followed by 99 annealing, extension and ligation cycles (95 C for 15 s and 65 C for 12 min).
Thread clean-up
In order to remove genomic DNA and excess primer, a cleansing procedure was set up robotically with a Magnatrix 1200 (Magnetic Biosolutions) liquid handling robot running custom made scripts. Prior to immobilization of the DNA threads on to streptavidin-coated Dynabeads DP (Dynal), each reaction tube was subjected to Proteinase K treatment (20 mg) for 20 min at 60 C and the contents of the two tri-nucleotide reaction tubes were pooled. After immobilization of the DNA threads, the beads were magnetically collected and washed with 1· TE [10 mM Tris-HCl (pH 7.5), 1 mM EDTA] followed by thorough washes with water (Millipore) and 0.1 M NaOH with constant mixing. Finally, the DNA threads were released by heating to 80 C for 1 s in water (Magnetic Biosolutions) (8).
Parallel thread amplification
Following release, the 75 DNA threads were amplified in a single PCR tube. The PCR mixture consisted of 10 mM Tris-HCl (pH 8.3), 2 mM MgCl 2 , 50 mM KCl, 0.1% (v/v) Tween-20, 0.2 mM dNTPs, 0.2 mM of each universal primer and 1 U of AmpliTaq Gold DNA polymerase (Applied Biosystems) in a total volume of 50 ml. One of the amplification primers was biotinylated. After activating the polymerase at 95 C for 12 min the reaction was cycled 60 times using a profile consisting of 95 C for 30 s, 68 C for 30 s and 72 C for 30 s, followed by a final elongation step at 72 C for 2 min.
Array preparation, PrASE-genotyping and data analysis
Generic signature tag microarrays (with 16 sub-arrays per array) were designed and prepared using a QArray system (Genetix). Each of the 16 sub-arrays printed on a slide accommodated a duplicate of 150 tags, allowing the analysis of 75 SNPs by PrASE-genotyping (7). In brief, PrASE is an alternative approach to conventional allele-specific primer extension (ASE) assays but has increased specificity features. The ASE assay uses the inherent ability of DNA polymerases to extend 3 0 -termini mismatched primers discriminately. However, certain 3 0 -termini mismatched sequences cannot be distinguished by the DNA polymerases, resulting in comparable extensions of matched and mismatched primer complexes. This problem can be circumvented by exploiting the fact that mismatched primers have slower reaction kinetics, by adding a protease in the extension reaction.
Following the PrASE reaction, data were obtained by scanning the slide with an Agilent scanner (Agilent Technologies) and were analyzed using GenePix Pro 5.1 software (Axon Instruments). The median local background intensities were subtracted from the median intensities of the spots by GenePix Pro 5.1 and the data were analyzed in Microsoft Excel, where the mean values of fluorescence intensities of the duplicates for each signature tag were used to calculate the allelic fractions of the 75 SNPs for each individual. Allelic fractions were calculated as intensity of spot A/(intensity of spot A + intensity of spot B) and were plotted against the logarithm of the total signal intensities of spot A + spot B (see Figure 4).
RESULTS AND DISCUSSION
The TnT assay analyzes all six classes of SNPs (substitution polymorphisms): K (G/T), M (C/A), Y (C/T), R (G/A), W (A/ T) and S (C/G). In order to analyze all of these SNPs by tri-nucleotide extensions, four tri-nucleotides (AGT, TCA, CTG and GAC) may be required. For example, a K polymorphism may be extended by nucleotide combinations AGT or CTG. Since the polymorphism can be scored on either strand, the complimentarity of the tri-nucleotide mixtures can be exploited to reduce the number of extension reactions required to cover all genetic variants, from four to two. Figure 1 illustrates four extension scenarios for a polymorphic position with a C to T base variation. As shown in the figure, all four tri-nucleotide combinations may be employed to extend over this SNP. When the DNA strand with the C/T (Y) polymorphism is used as the target template, the combinations AGT (indicated by red) or AGC (indicated by green) can be used to generate a complementary strand-a DNA thread. Note that the nucleotides G and A have to be present in order to extend over and cover both allelic variants. When the AGT nucleotide combination is used in the reaction, the extension primer is extended by 3 nt when either allele variant provides the template. In this case, if the polymorphic position in the target template is homozygous C/C, the nucleotides 'gtt' will be incorporated, and since the nucleotide C is not included, the extension will be terminated when C has to be incorporated and the extended product can be ligated to a downstream probe (thread-joining primer). Similarly, the extension will be terminated at the same position with both homozygous and heterozygous variants. However, if the polymorphic position in the target template is homozygous T/T the polymerase will incorporate 'att' nucleotides, and when the sample is heterozygous C/T, the target alleles with the C and T variants will induce the incorporation of 'gtt' and 'att', respectively. Thus, both allele products can be ligated to the second primer and amplified in the PCR that follows. Accordingly, when the AGC nucleotide combination is used, only one base can be incorporated into the extension primer. As shown in Figure 1, the extension primer in both chosen cases (AGT or AGC nucleotide combinations) is the same and its 3 0 end is situated next to the polymorphic site. However, the thread-joining primer has to be relocated depending on the number of incorporated bases in the gap. As illustrated in Figure 1 and described above, the user of the technique can select any of the four tri-nucleotide sets. However, in order to compensate for possible unspecific extension/ligation events, an extension-length scoring system has been implemented, favoring extensions with more than 1 nt. After a cyclic extension/ligation reaction and complete removal of genomic DNA, the products are subjected to PCR amplification, using a pair of universal primers, followed by microarray-based PrASE to score the SNPs. Thus, the genotyping primer functions not only as a scoring probe but also as an additional specificity factor. Note that almost all steps in this method are automated by using a robotic workstation capable of handling magnetic beads.
The bioinformatic steps involve screening the human genome with the candidate SNP sequences to remove SNPs with sequences spanning duplicated gene segments sharing high sequence similarities, repetitive regions, low-complexity sequences and those with neighboring polymorphisms (see the SNP selection box in Figure 2). In addition, we developed a software tool to make in silico extensions with the four different tri-nucleotide combinations. An exponential extension-length scoring system that awards high scores to the most theoretically favorable extension-lengths is implemented in the software. Since two reactions have to be conducted to produce the DNA threads, the software calculates the total score for two tri-nucleotide combinations (AGC/CTA, AGC/AGT, CTG/CTA and CTG/AGT) and selects the combination with the highest score, then automatically designs primers (with desired Tm) for the multiplex reaction and the PrASE reaction (see the Software box in Figure 2 and Supplementary Figure 1). The extension-length scoring criteria have been set to favor long extensions (but not too long, see below), since formation of some unspecific DNA threads may be unavoidable even though the TnT technique should prevent the extension step entering an exponential phase and subsequent specific joining. However, if an unspecific event takes place, the likelihood that the unwanted thread will be discriminated by the genotyping primer(s) is dramatically increased if the number of nucleotides incorporated in the specific threads is raised. Consequently, DNA threads formed with a single nucleotide incorporation are given the lowest score (2) and extensions spanning 5-7 bases the highest score (64). The reason for implementing an exponential scoring system was to clearly differentiate between the different extensions lengths. However, to avoid PCR bias associated with product length, long extensions, such as 11 and 12 are given moderate scores (16). In addition, the software adjusts the length of the primers of short DNA threads, ensuring that the differences in length between the longest and shortest threads do not exceed 10% (see the Software box in Figure 2). Since the majority of the steps in the TnT technique are performed automatically using a bead-handling robotic workstation, a custom made script dedicated to the project was developed.
After a successful pilot experiment with the TnT approach, comparing array based and Pyroseqeuncing genotyping technology (data not shown) we chose to proceed with the array platform. Hence, SNPs located in genes related to different forms of cancer were identified and randomly selected using the dbSNP, SNP500Cancer-and Ensembl databases. Our current array of generic tag arrays has the capacity to genotype 75 SNPs (using the PrASE technique, which requires two allelespecific primers/SNP), so the software was designed to select 75 SNPs. The nucleotide mixture selection was based on the scoring system where the tri-nucleotide combinations of AGT and CTG received the highest total score of 2968. Using this combination, 39 SNPs would be in the CTG set and 36 in the AGT set. The scores (and SNP distributions) of combinations AGC/AGT, AGC/CTA and CTG/CTA were 2736 (36/39), 2086 (46/29) and 2396 (54/21), respectively. The user has the option to make a final decision to select a more suitable combination. Such a scenario may occur when two or more combinations have obtained close scores and the software selects the combination with the highest score, although the SNP distribution suggests that one of the other combinations would be better. Nevertheless, the 75 SNPs selected in this study (combinations of AGT/CTG) are represented on all chromosomes except chromosomes 12, 21; X and Y (see Supplementary Table 1).
Genomic DNA from a panel of ten Swedish individuals was subjected, at two concentrations (200 and 10 ng/ml), to 75-plex amplification using the TnT approach, and the individuals were genotyped using the PrASE protocol. Genotyping results obtained with the two concentrations of genomic DNA showed complete correlation, indicating that the approach can be used for parallel amplification of very small amounts of target genomic DNA as starting material (Figure 3).
Consequently, the DNA threading was carried out using sub-nanogram amounts of genomic DNA as starting material. The SNPs are subsequently sorted to each of the four possible pairs of tri-nucleotide mixtures and the extension score is added to the combinations' total score. The combinations are then ranked according to their total scores and two primers, the extension primer and the thread-joining primer, are designed for each SNP. The length of the primers is adjusted so that the differences in the lengths of the threads do not exceed 10% (Length compensation). In vitro: the mixture of primers, enzymes, nucleotides and DNA template are subjected to cyclic extension-ligation reactions followed by an automated purification step to remove genomic DNA and excess primer. The two sets of reaction products are merged and the threads are amplified with a general primer pair. Genotyping: the different threads are decoded by a genotyping method, in this case PrASE, and allelic fractions are calculated for correct genotype scoring.
DNA from a total of 96 Caucasian individuals (UK population) was then amplified with the TnT protocol and genotyped with PrASE. 68 out of the 75 assayed SNPs (93%) rendered partitioned clusters of homozygous and heterozygous genotypes ( Figure 4). The SNPs that did not give satisfactory clusters (mostly due to very weak signals) were SNPs 16, 34, 40, 52, 60, 70 and 71. The SNPs 52 and 60 did however give relatively good signals but cluster partitioning was not satisfying.
To investigate the cause of the weak signals for the remaining SNPs, these were subjected to the TnT protocol, but were amplified and genotyped in a duplex fashion (by including a control SNP as a reaction control). Despite the lowcomplexity in terms of multiplexing, the array signals from Table 5).
In addition, in order to investigate the sensitivity of the approach, a series of dilutions of two samples was prepared including 8, 4, 2, 1 and 0.5 ng of genomic DNA, which was subjected to parallel DNA threading. The genotyping results were then included in the cluster diagrams obtained from the analysis of the 96 samples with 8 ng of genomic DNA. Identical genotyping scores were obtained for all 68 of the working SNPs in concentrations down to 1 ng per reaction, corresponding to 150 cells. With 0.5 ng DNA, some of the SNPs did not give reliable scores due to weak signals.
To our knowledge, the TnT technique provides a unique means to amplify any given polymorphism in small amounts of starting material. A number of techniques, based on fragmentation of the genomic DNA and subsequent amplification (9-11) and direct typing on whole genome amplified DNA (12), have recently been described. In addition, a recent study demonstrated co-amplification of more than 1000 selected SNPs by general PCR (13). But all these techniques have a major disadvantage since they lack the ability to genotype any specified polymorphic position. There are however two techniques, the Golden Gate assay (Illumina Inc) (14) and the MIP assay (Parallele Biosciences) (15), that are capable of parallel amplification of specified polymorphisms. The Golden Gate assay is based on ASE followed by ligation, but unspecific 3 0 end extensions of certain mismatches (16) tend to occur, and the technique requires careful optimizations. In addition, unspecific primer hybridization in combination with uncontrolled extensions, due to the presence of all dNTPs, makes it impossible to initiate linear amplification so relatively large amounts of starting material, 0.2 to 1 mg of genomic DNA, are required. The MIP assay uses specific pre-circle probes that are locked (padlocked) to the target molecule after gap-filling and ligation steps. However, the padlock formation may prevent cycling of the gap-fill/ligation procedure and consequently 0.4 to 2 mg of genomic DNA may be required in the assay, while the TnT technique provides reliable results with as little as 1-9 ng (100-1000 cells) of genomic DNA without employing biased whole genome amplification (17,18). An apparent advantage of the Golden Gate and MIP assays is that they integrate amplification with the genotyping, while the TnT approach focuses solely on amplification. However, despite the need to add a genotyping step, this may be advantageous because the technique can be applied in any genotyping laboratory without changing the typing routines. In addition, as noted above, the genotyping primers in combination with the incorporation of more than one base in the formation of the DNA threads gives high specificity, as reflected in the cluster diagrams and the fact that 68 out of 72 SNPs worked in the assay. A possible drawback that the TnT technique shares with most genotyping and amplification techniques, in particular with the Golden Gate and MIP assays, is the negative effect of neighboring SNPs on the assay performances. As some SNPs have a neighboring sequence variation with a distance closer than 40 bases, these could affect the probe hybridization efficiency and are recommended to be avoided.
In conclusion, we present a specific and sensitive technique for the multiplex amplification of DNA sequences with variable positions (SNPs). A 75-plex approach may represent a high degree of complexity and cost (approximately USD 0.056/SNP), for academic facilities with modest financial resources, but the TnT technique has great potential to specifically amplify thousands of variable segments. The sensitivity of analytical techniques is a very important issue, as well as their specificity, when the numbers of cells available is limited, as they often are in analyses of tumor materials, forensic investigations and attempts to identify biomarkers using biobanks for instance (19). Since the TnT technique can give identical results with 1, 8 or 200 ng of genomic DNA, we believe that by taking advantage of signal amplification strategies (12,20) it can retain much of its sensitivity when a 1500-plex approach is designed. | 2014-10-01T00:00:00.000Z | 2006-03-31T00:00:00.000 | {
"year": 2006,
"sha1": "16132694d2b29899444d91714a7f7cd1eac15b6b",
"oa_license": "CCBYNC",
"oa_url": "https://academic.oup.com/nar/article-pdf/34/6/e49/7128527/gkl103.pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "191d86660f50ba606f34180dab76e7e5db44e661",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
235908421 | pes2o/s2orc | v3-fos-license | The potential role of Transient Elastography in assessing patients with Primary Budd Chiari Syndrome
Background: Budd-Chiari syndrome (BCS) is a rare disease de�ned as hepatic venous out�ow obstruction at any level from the hepatic venules up to the cavo-atrial junction. Transjugular Intrahepatic Portosystemic Shunt (TIPS) is performed as a decompressive treatment in some patients. Aim: To evaluate the potential role of Transient Elastography (TE) in assessing liver stiffness in patients with primary BCS. Methods: Twenty one BCS patients and 10 patients with liver cirrhosis with different underlying etiologies underwent abdominal ultrasound and TE. Results: Ninety-ve percent of BCS patients had liver stiffness compatible with F4 with a median of 21 kPa, values which are usually obtained in patients with liver cirrhosis. Ten BCS and 10 cirrhotic patients underwent repeated TE with a median of 320 days between exams for BCS and 4.5 years for cirrhotic patients. The change of liver stiffness in BCS patients was 5.75 kPa (range − 0.4 to 26.6), compared with − 4.85 kPa (range − 15.6 to 15.0) in cirrhotic patients (p-value = 0.0029). Change in liver stiffness from baseline to follow-up in BCS patients who underwent TIPS (n = 4) was 0.2 kPa (range − 0.4 to 15.3), whereas in patients without intervention (n = 6) it was 6.75 kPa (range 1.3 to 26.6). The difference was not statistically signi�cant. Conclusion: Liver stiffness in BCS patients is a dynamic progressive process with parameters of TE resembling liver cirrhosis. Even if TIPS seem to slow down the increment of liver stiffness, because of decreased liver congestion, it kept most patients with high score. The TE in BCS patients may be considered for monitoring for stable or upfront disease deterioration.
Introduction
Budd-Chiari syndrome (BCS) is de ned as hepatic venous out ow obstruction at any level from the hepatic venules up to the cavo-atrial junction, in the absence of congestive or restrictive heart disease.It occurs in 1 of 100,000 in the general population worldwide (1) and with annual incidence of 0.8 per million per year inhabitants (2) .The underlying etiologies for primary BCS vary.While intravascular thrombosis is more common in the West, membranous webs obstructing the inferior vena cava (IVC) are noticeable in Asia and association with Behcet's disease is common in Turkey (3) .Thrombotic conditions commonly associated with BCS include polycythemia vera (PV), paroxysmal nocturnal hemoglobinuria, inherited de ciency of natural anticoagulants, e.g. protein C, protein S and antithrombin or presence of factor V Leiden or prothrombin G20210A mutation (4,5) .Currently, myeloproliferative diseases (MPD) with Janus kinase 2 (JAK2) V617F gene mutation are the leading cause for BCS accounting for 25%-50% of cases (6,7) .
The clinical course of BCS varies from asymptomatic to fulminant hepatic failure.The latter is reported in 7% of cases, whereas subacute or chronic types occur in 45%-48% of patients.The occlusions of the hepatic veins lead to increased sinusoidal blood pressure with reduced sinusoidal and portal blood ow along with increased arterial blood ow in the microcirculation and centrilobular necrosis and brosis (8) .
Prolonged occlusion results in progressive brosis and liver cirrhosis with portal hypertension (2) .Whether decompression leads to regression of liver brosis is still a matter of debate.Doppler ultrasonography is the rst line imaging modality for BCS diagnosis with a sensitivity of up to 85% (9,10,11) .
Transient Elastography (TE) is a noninvasive tool for measuring liver stiffness (12,13,14) .It is based on calculating the speed of a pressure-induced shear wave propagation through tissue subject to different biomechanical properties.Use of TE in the context of BCS was for the rst time reported in assessing hepatic congestion 24 hours after endovascular interventions (15) .
In the present study, we study the potential role of TE in assessing and monitoring patients with BCS.
Patients
Twenty-one patients with BCS were recruited from the liver units of several hospitals and Maccabi health maintenance organization (HMO) in Israel.Exclusion criteria were secondary BCS attributable to hepatic cell carcinoma.Patients with ascites were not included since the elastic waves of TE do not penetrate liquids (16) .All patients had contrast-enhanced CT before recruitment to ascertain diagnosis of primary BCS, to con rm the number of occluded hepatic veins and to distinguish them from collateral vessels.In 12/21 patients, all 3 hepatic veins were occluded, 7 patients had 2 veins occluded and in 2 patients only one vein was occluded.Ten of 21 patients had also an MRI, and in 4 of them had repeated MRI before the beginning of the study.Most of the MRI examinations were performed at subacute or chronic phase of BCS and assessed the morphological changes of the liver with enlarged caudate lobe.Regenerative nodules were described in 4 BCS patient that underwent MRI.Gastroscopy was performed in 20 of 21 participants and 11 patients (55%) had esophageal varices.
The control group comprised 10 patients with liver cirrhosis with diverse underlying etiologies and splenomegaly but without ascites that were part of the group served as controls in previous studies (17,18) .The underlying causes for liver cirrhosis were: hepatitis C virus (HCV) (n = 5), hepatitis B virus (HBV) (n = 2), primary sclerosing cholangitis (PSC) (n = 2) and non-alcoholic steatohepatitis (NASH) (n = 1).
The study was approved by the Ethics Committee of the Tel Aviv Sourasky Medical Center, and all participants gave written informed consent according to the Declaration of Helsinki (0394-13-TLV).
Ultrasound and Doppler analysis
Ultrasound and Doppler examination were performed in all participants as described previously (17,18) .Special attention was given to liver and spleen size, enlargement of caudate lobe and liver echogenicity, presence of nodules in the liver and spleen and presence of ascites.Doppler examination assessed the ow in the portosystemic system, hepatic veins, inferior vena cava, in the intrahepatic or extrahepatic collaterals and in the TIPS, when present.
Transient Elastography (TE)
The TE was carried out by Fibroscan® (Echosens; Paris, France).Liver stiffness was expressed in kilopascals (kPa).and measurement translated for different degrees, e.g., mild brosis (F1), moderate brosis or more (F2), severe brosis (F3), and cirrhosis (F4) as reported elsewhere (14) .The results were given as the median of 10 subsequent measurements and de ned as valid if the ratio of the interquartile range and median was less than 30%.All patients fasted at least 8 hours before the examination.Liver stiffness was measured 1.5-3 cm below the liver capsule while the patient was supine in right lateral decubitus with the right arm placed overhead in maximal abduction to extend intercostal space for placing the transducer in the right intercostal space.
Statistical analysis
Data are summarized by median and range.Pearson correlations were computed to assess the strength of the relationships between the continuous variables.The Wilcoxon test was used to compare groups with respect to continuous variables.
Clinical characteristics
The clinical characteristics of the BCS patients are presented in Table 1.Risk factors were found in 67% of the patients.All BCS patients received anticoagulants unrelated to etiology and only one patient received direct oral anticoagulant, e.g., rivaroxaban.Median duration of the disease was 5 years (range 0-16 years) at the time of the rst TE examination.Median spleen size was 14.5 cm with a range of 7.5-22 cm at the time of the rst abdominal ultrasound during the study.
Gender
Median The median liver stiffness in BCS patients was 21 kPa (range 5.7 to 69.1 kPa) with a compatible score of F4 similar to patients with liver cirrhosis.
Spleen size was weakly related to liver stiffness (r = 0.31) and duration of disease (r = 0.14).
Ten of 21 BCS patients that were rst included in the study had follow-up TE with a median of 320 days apart (range 90 to 608 days) between exams.We then compared the change in liver stiffness from baseline to follow-up exam in patients with or without a previous endovascular intervention.As seen in Fig. 1, the 4 patients with TIPS had a median stiffness change of 0.2 kPa (range − 0.4 to 15.3), whereas the 6 patients without TIPS had a median change of 6.75 kPa (range 1.3 to 26.6).In those 4 BCS patients, endovascular interventions including TIPS or stent were previously carried out 2 months to16 years (median of 58.3 months) before participation in this study.The difference is not statistically signi cant (p-value = 0.17 using the Wilcoxon test) and the sample is too small to provide a statistically reliable result.The other 11 patients did not have follow-up exams since the elapsed time was too short.
Median liver stiffness in the control group of cirrhotic patients was 19.7 kPa (range 12.2 to 35.3 kPa).The 10 cirrhotic patients had follow-up TE with a median of 4.5 years between exams (range 0.5 to 8 years).
In 5 cirrhotic patients with hepatitis C, the repeat TE was performed between 3 months to 2 years after antiviral treatment was given and when no virus load was detected.
We next compared the change in liver stiffness in the 10 BCS patients and the 10 cirrhotic patients who had a repeat TE.As seen in Fig. 2, the BCS patients had a median stiffness change of 5.75 kPa (range − 0.4 to 26.6), whereas the cirrhotic patients had a median stiffness change of -4.85 kPa (range − 15.6 to 15.0).The difference is statistically signi cant (p-value = 0.0029 using the Wilcoxon test).
Albeit reduced liver stiffness in 8 of 10 cirrhotic patients following speci c therapy for underlying disease, only in 2 patients there was a change from F4 to F3.
Discussion
In this work, we found that 95% of the patients with BCS, unrelated to underlying etiology, had high values of liver stiffness measured at TE, similar to patients with liver cirrhosis.
Four of 21 BCS patients (19%) had PV with JAK2 V617F mutation.Notably, treatment with cytoreductive therapy, e.g., hydroxyurea along with anticoagulants and antiaggregants was not able to halt the progressive increment in liver stiffness measured by TE.In patient number 4, the high level of liver stiffness measured at TE anticipated clinical deterioration with liver failure and indeed this patient is programmed for liver transplantation.
BCS has rarely been associated with in ammatory bowel disease (19,20,21,22) .However, 5 of our 21 BCS patients (24%) had Crohn's disease where 3 of them were on TNF alpha antagonist treatment.Such association can be explained by the prothrombotic state caused by uncontrolled in ammation during ares in Crohn's disease and perhaps by brosis reported to be related to treatment with anti TNF-α antagonist (23,24) .The later reported to be associated with pulmonary brosis in patients with rheumatoid arthritis.
Two patients in our cohort had antiphospholipid syndrome.The association between antiphospholipid syndrome and BCS is not clear (25) .Factor V Leiden mutation was present in one of the patients and may have been a risk factor, as reported previously (10) .In 29% of our BCS patients, the etiology was still unknown.
In 8 of the 10 BCS patients that had a follow up TE, liver stiffness increased.The later could be attributed to increased congestion and perhaps brosis.It is assumed that the increased liver stiffness developed during the rst few weeks after hepatic vein occlusion and probably depends on the extent of initial vein occlusion.Though it is still not clear why the course of BCS is progressive when hepatic decompression happens through collaterals or when endovascular procedures are performed.It is hypothesized that the collaterals that formed are not e cient at impeding chronic ischemia (26) .In our study, in BCS patients with TIPS the progression of liver stiffness did occur but was slower in different to other groups.In the Hong-Wei Wang study, liver stiffness measured at 2 days after TIPS insertion was signi cantly decreased when assessed by Real-time Shear Wave Elastography but remained stable at 3 months though still in the cirrhotic range (27) .Another group reported that repeated TE measurement years after TIPS detect decreased value (28) .
In sharp contrast with BCS, in 80% of cirrhotic patients, liver stiffness declined following treatment for underlying disease even if 70% of the cirrhotic patients remained with F4.In 3 control patients with HCV cirrhosis, a follow-up values of TE was compatible with F2-F3 after treatment.Indeed, liver brosis was shown to reduce following suppression of HCV and HBV (29) , likewise in treated patients with autoimmune hepatitis (30) or abstinence of alcohol in alcoholic liver cirrhosis which could be related to reduce in ammation.
The question that one would ask is whether the increased liver stiffness in patients with BCS is irreversible.It may occur that in response to injury due to vascular thrombosis, the sinusoidal endothelial cells activate quiescent hepatic stellate cells HSC) with proliferative contractile myo broblast phenotype that produce extracellular matrix in the liver, producing bronectin, TGF-β1 and PDGF (31) .
The contribution of increased liver congestion during the years, albeit collateral development should not be ignored as it is known that congestion by itself could increase liver stiffness (27) .
We are aware of the major limitation of the work by the lack of liver biopsies.Yet liver biopsy in BCS is not routinely performed as it does not change management and it is still invasive procedure.Another limitation is the small group of patients.However, the fact that the liver stiffness was high in almost all our BCS patients gives us con dence that a larger group would not alter our results.The sample size of the cirrhotic patients is small but it re ects the parameters of liver stiffness known in cirrhotic patients worldwide (13,14) .
To note that only 48% of BCS patients had a second TE and at variable time gaps.However, we did nd that in 80% of patients, the liver stiffness increased even when the repeat examination was done after a short interval.This raises a doubt about the effectiveness of the treatment given today to the patients with BCS.
In summary, liver stiffness in BCS patients is a dynamic progressive process despite the application of speci c therapy to the underlying disease.
TE can serve as a surrogate modality to assess progression of liver stiffness in BCS patients and in cases where target therapy is implemented.
Figure 1 Follow
Figure 1
Table 1 .
Clinical characteristics of patients with BCS | 2020-09-10T10:17:48.997Z | 2020-08-04T00:00:00.000 | {
"year": 2020,
"sha1": "73c0023b965b21276d4afeecb6a252340aae5233",
"oa_license": "CCBY",
"oa_url": "https://www.researchsquare.com/article/rs-45623/v1.pdf?c=1596589994000",
"oa_status": "GREEN",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "a849e014db5b0551c363a3193c7011e32bbd497a",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
235348593 | pes2o/s2orc | v3-fos-license | Accuracy of COVID-19 rapid antigenic tests compared to RT-PCR in a student population: The StudyCov study
Objective There is a lack of data evaluating performance of antigenic test (AT) for SARS-CoV-2 diagnosis (Ag-RDT) in clinical practice, especially in asymptomatic subjects. The main objective of this study was to evaluate the diagnostic performance of AT compared to Reverse Transcription Polymerase Chain Reaction (RT-PCR) for SARS-CoV-2 diagnosis. Methods StudyCov is a monocentric cross-sectional study. A SARS-CoV-2 screening facility was set up in the Bordeaux University health campus from October 28th to November 20th 2020. Students willing to have a RT-PCR test (ARGENE SARS-CoV-2 R-GENE, BioMérieux, France) for SARS-CoV-2 diagnosis were also offered the Abbott Panbio™ SARS-CoV-2 antigenic rapid test. All participants attending the screening facility with an AT in addition to RT-PCR and having signed an informed consent were included in the study. The main objective was to assess performance of AT as compared with RT-PCR in the recruited population. Secondary objectives dealt with the analysis of the main objective stratified by current symptoms and risk exposure. A sensitivity analysis with different RT-PCR cycle thresholds was included. Results RT-PCR and AT results were available for 692 subjects. Overall sensitivity and specificity of AT tests were respectively 63.5% (95% confidence interval (CI): 49.0 – 76.4) and 100% (95% CI: 99.4 – 100). In the asymptomatic sub-group, they were respectively 35.0% (95% CI: 15.4% - 59.2%) and 100% (95% CI: 99.3 - 100). Conclusions This study shows the poor sensitivity of AT in asymptomatic subjects, specificity being however excellent. The performance results fall below the World Health Organization recommendation of 80% sensitivity and question using AT in general population, especially when asymptomatic.
Introduction
Rapid identification of SARS-CoV-2 cases is a cornerstone in the management of the COVID-19 pandemic. Reverse Transcriptase Polymerase Chain Reaction (RT-PCR) is the reference test for SARS-CoV-2 diagnosis. In addition, various antigenic tests (AT) have been developed. They provide test results in about 15 min, which is much shorter than for RT-PCR tests.
Some studies have compared both RT-PCR and AT performances. Regulatory health agencies performed meta-analyses of the latter in order to establish testing recommendations [1][2][3][4]. Concerning AT, the Haute Autorité de Santé (HAS) -the French National Authority for Health -concluded to an homogeneous good specificity of 98.7% (95% confidence interval (95% CI): 97.3 -99.4). However, sensitivity was extremely heterogeneous, from 17 to 97%, depending on the study and the targeted population, with a pooled sensitivity of 90% (95% CI: 73 -96) [4]. Therefore, the World Health Organization (WHO) recommended a minimum sensitivity of 80% and a minimum specificity of 97% for AT effective use [3]. However, to date, few studies have been performed in the general population and in asymptomatic people. This is of concern as some countries are promoting the use of AT without solid knowledge about their actual performance [4][5][6][7].
From October 28th to November 20th 2020, a screening center has been operated in the Bordeaux University health campus (France). Students willing to be tested for SARS-CoV-2 by RT-PCR were also offered AT. The objective of this cross-sectional study was to assess the diagnosis performance of a nasopharyngeal antigenic rapid test compared to RT-PCR in this population of students and in the subgroup of asymptomatic subjects.
Methods
This study is a monocentric diagnosis cross-sectional retrospective study. All participants older than 18 years and tested with both AT and RT-PCR at the screening facility of the Bordeaux University health campus were included in the study. The index test was the nasopharyngeal Abbott Panbio™ SARS-CoV-2 Ag rapid test. The reference test was the RT-PCR of N and RdRp genes (ARGENE SARS-CoV-2 R-GENE, BioMérieux, France) with a maximum cycle threshold (Ct) of 45, the current standard of care in the Bordeaux University Hospital. Nasopharyngeal samplings for both tests were performed simultaneously, in each nostril. AT and RT-PCR readings were blinded to each other thanks to different logistic tracks. Testing and reading were performed by trained teams and the reading was carried out by a single reader following the manufacturer's instructions.
The main objective was sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of AT compared to RT-PCR in the whole study population. RT-PCR was considered positive when both nucleocapsid (N) and RNA-dependent RNA polymerase (RdRp) genes were detected. Secondary objectives were performances of AT compared to RT-PCR stratified by current symptoms and risk situation exposure (over the previous week). Risk situation exposure was defined as attending a party, practicing team sports or having contact with another person outside home without a mask over the previous week. Sensitivity analysis with different RT-PCR Ct was included. We also compared the Ct of N and RdRp SARS-CoV-2 gene amplifications among subjects with a positive SARS-CoV-2 RT-PCR test stratified by AT results and current symptoms.
Subjects with missing RT-PCR or AT results were excluded from the analysis. Sensitivity, specificity, PPV and NPV confidence intervals were computed using the exact binomial method. Ct values for N and RdRp viral genes were compared using Wilcoxon ranked test. Analyses were performed using R v4.0.0 and the epiR package [8].
Results
Subjects' characteristics are described in Table 1. There was no falsepositive case when considering RT-PCR positive if both genes were detected. When considering a subject as positive if both genes were detected, one of which with a RT-PCR Ct ≤ 30, 3 subjects proved falsepositive. Table 2 shows the performance of AT compared to RT-PCR by subgroup and by positive SARS-CoV-2 RT-PCR definition. Overall specificity and sensitivity were respectively 100 (95% CI: 99.4 -100) and 63.5 (95% CI: 49.0 -76.4). In the asymptomatic group, they were respectively 100% (95%CI: 99.3 -100.0) and 35.0% (95% CI: 15.4 -59.2). Results were similar when considering asymptomatic patients exposed to a risk situation. Fig. 1 shows the association between AT result, current symptoms and Ct.
In this cohort of 692 subjects, 52 had a positive RT-PCR, 47 of which were considered likely contagious (i.e. 2 genes detected, one of which with a Ct ≤ 33) [9,10]. Using AT instead of RT-PCR, 19 (36.5%) subjects would have been missed, including 14 considered likely contagious.
Among the 20 asymptomatic RT-PCR positive students, 17 were considered likely contagious. Using AT instead of RT-PCR, 13 (65%) subjects would have been missed, including 10 considered likely contagious.
DISCUSSION
This study highlights that AT lack sensitivity compared to RT-PCR in the context of SARS-CoV-2 diagnosis and screening. For the asymptomatic sub-group, two thirds of SARS-CoV-2 cases would be considered negative by AT even among the likely contagious ones. However, there was no false-positive in this study thus making AT reliable when positive. False negative in the asymptomatic people might be explained by high Ct, as shown in Fig. 1.
Few studies comparing AT and RT-PCR have been conducted in the general population and in asymptomatic subjects. The Abbot Binax-NOX™ was compared to RT-PCR in a walk-up population: 84% were asymptomatic and 3% had a positive RT-PCR [11]. Sensitivity and specificity were respectively 93.3% (95% CI: 68.1-99.8%) and 99.9% (95% CI: 99.4-99.9%) for a Ct < 30 case definition. Compared to our study, the swab was done in both nostrils for both tests, the population was older and the AT was different. Those differences might partly explain the performance differences. Another study compared Panbio™ SARS-CoV-2 AG Rapid Test Device with RT-PCR for emergency units-referred patients, 72.1% of which were asymptomatic [12]. Sensitivity was 73.3% (95% IC: 62.2-83.8%) which is consistent with our results.
This study is one of the first studies focusing on asymptomatic subjects. The reference test is the one used in the standard of care and both RT-PCR and AT evaluations were done blinded to each other.
As it is a retrospective study, extensive data about subjects' characteristics such as comorbidities were not available. Included subjects are exclusively healthcare students which is relevant as it is a strategic population in contact with patients but limits the extrapolation of those findings.
StudyCov highlights lack of antigenic tests sensitivity in a student population and in a field setting for the use of AT: it falls far below the WHO performance recommendations of a minimum sensitivity of 80% [3]. It suggests the need for contextualizing AT use and its higher relevance for symptomatic cases. More studies are needed to find out if those findings are replicable in different populations. Meanwhile, this study should help to reevaluate testing policies.
OTHER information
StudyCov was conducted by the Bordeaux University Hospital, the Agence régionale de santé Nouvelle-Aquitaine (Regional Health Agency of Nouvelle-Aquitaine) and Bordeaux University without any specific funds. This study received an IRB approval (IRB00003888, 7th December 2020).
Transparency declaration
Conflict of interest: Pr. Dehail reports personal fees from Allergan, outside the submitted work. Other authors have nothing to disclose.
Access to data
Patrick Dehail has full access to the data.
Declaration of competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. | 2021-06-06T13:17:31.430Z | 2021-06-05T00:00:00.000 | {
"year": 2021,
"sha1": "405435c7e1c11817afdfcb61a78283933e5fec58",
"oa_license": null,
"oa_url": "https://doi.org/10.1016/j.jcv.2021.104878",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "8f66f7bdf0d3446e8ae6c2faaf2933016b80fe51",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
219094057 | pes2o/s2orc | v3-fos-license | A synthesis of bacterial and archaeal phenotypic trait data
A synthesis of phenotypic and quantitative genomic traits is provided for bacteria and archaea, in the form of a scripted, reproducible workflow that standardizes and merges 26 sources. The resulting unified dataset covers 14 phenotypic traits, 5 quantitative genomic traits, and 4 environmental characteristics for approximately 170,000 strain-level and 15,000 species-aggregated records. It spans all habitats including soils, marine and fresh waters and sediments, host-associated and thermal. Trait data can find use in clarifying major dimensions of ecological strategy variation across species. They can also be used in conjunction with species and abundance sampling to characterize trait mixtures in communities and responses of traits along environmental gradients.
Background & Summary
Several research groups have advocated for a trait-based approach to ecology of bacteria and archaea [1][2][3][4][5][6][7][8][9] , but so far this has remained at the level of conceptual discussion or interpretation of particular study systems. Here we describe a scripted workflow that generates a unified microbial trait dataset suitable for investigating which traits are correlated across species versus which vary independently. The dataset spans the full range of bacterial and archaeal habitats, including fresh and marine waters, soils and sediments, animal and plant hosts, and thermal environments. Data sources include well-established repositories, such as GenBank 10 , Bergey's Manual of Systematics of Archaea and Bacteria 11 , and a number of compilations published in the literature (Online-only Table 1). (2020) 7:170 | https://doi.org/10.1038/s41597-020-0497-4 www.nature.com/scientificdata www.nature.com/scientificdata/ We believe this data product will prove useful to other research groups in several ways. Some may use the current version of the dataset for their own data analyses. They may adjust the scripted workflow to adopt different merger rules; for example, about how data sources are aggregated or prioritized when multiple records are available. Some may choose to update the dataset, since among the contributing data sources several are continuing to receive new data. Some may choose to add further data sources or merge their own data sources, which should be made easier by the scripted structure we provide. Once scripted into the workflow, new or updated data sources can be merged with the current data product in GitHub resulting in a new version of the data product.
Trait data can have a variety of research purposes. Correlations among traits can be investigated to elucidate the main dimensions of variation across species 12 . Species lists and their abundances in communities can be interpreted, for example whether communities have similar trait mixtures despite different taxonomy. Responses of traits along environmental or geographical gradients can be described 13 . If relevant traits are available to combine with species identifications and abundances, aspects of ecosystem function can be inferred.
Synthesizing trait data is a continuing process rather than a finite project. During the time taken to add any particular data source to the merger, new data sources continue to appear. The data merger in its current form and as reported here emphasizes quantitative genomic traits (such as genome size and number of rRNA gene copies) and phenotypic traits (such as potential rate of increase, cell radial diameter and growth temperature).
We have included information from culture on metabolic pathways and carbon substrates. However, we have not yet included metabolic pathways inferred from genomes, and consequently the question of reconciling genome-inferred pathways with culture-observed pathways does not arise. Also we have not yet included presence or absence of specific genes as qualitative traits, for a combination of reasons. First, there are potentially a very large number of such traits. Second, the number of complete genomes available continues to increase rapidly, and so such data will be out of date quickly. Third, there exist a number of databases (MIST 14 , MACADAM 15 , ANNOTREE 16 for example, and more emerging all the time) that specialize in annotations from genomes. When users wish to ask questions involving these genome-derived traits it will be better for them to link those databases to ours, which can be done using NCBI Taxon IDs.
Methods
The scripted workflow was developed to reproducibly (a) prepare datasets to be merged; (b) combine datasets; (c) condense similar or the same traits into columns; and (d) condense rows into species based on either the NCBI taxonomy 17 or the Genomic Taxonomy Database (GTDB) taxonomy 18 (Fig. 1, Online-only Table 1). This workflow generated five data products 17 for the 23 phenotypic, genomic and environmental traits shown in Online-only Table 2. The first two products are record level, which includes taxonomic levels below species (e.g., strain) and based on the NCBI taxonomy and GTDB taxonomy, respectively. A reference table was generated to track provenance of raw data through the workflow. The last two products are aggregated at species-level for the NCBI taxonomy and GTDB taxonomy, respectively. Trait coverage across the phylogenetic tree is shown in Fig. 2 and the trait distributions are shown in Fig. 3. Table 1 shows species-level trait data derived from original datasets.
Prepare. The preparation steps removed unwanted columns from raw datasets, ensured standard trait (column) naming, and established that each record (row) had an NCBI taxon ID and reference. In cases where NCBI taxon IDs were not provided in the raw dataset, taxon mapping tables were created using the NCBI taxonomy API, which could retrieve IDs by fuzzy searches of name or accession number, depending on what was available 10,17 . In cases where the API did not resolve to a single taxon, the NCBI taxonomy browser was used to manually look-up parts of names in case of misspellings or name fragments (e.g., strain names that were truncated to species level). DOIs or full text citations were used for referencing where possible, but in some cases only NCBI BioProject or accession numbers were available and were used to track provenance instead. All changes in the www.nature.com/scientificdata www.nature.com/scientificdata/ preparation stage were scripted and commented in dataset-specific preparation scripts. Other dataset-specific steps included splitting number ranges into different components (e.g., 10-20 µm to 10 [min], 20 [max] and µm [unit]), and any general data translation issues (e.g., spreadsheet software issues that manipulated characters, dates, and other inconsistencies). Only the traits summarised in Online-only Table 2 were retained for the steps where data are combined (next).
Combine. All the raw datasets were placed into a single sparse matrix with zero overlap (Fig. 1b). A column was added with the name of the dataset (Online-only Table 1) to keep track of dataset provenance. All columns containing referencing information (reference and reference type) and NCBI taxon IDs were moved into dedicated columns. The basic taxonomic hierarchy was mapped onto each row using either of the NCBI or GTDB taxonomies, which added columns for species, genus, family, order, class, phylum and superkingdom.
Condense traits.
Condensing trait data involved moving values for the same trait from different datasets into one column (Fig. 1c). The inherent assumption is that data for the same taxon from different datasets were observed independently (e.g., cell sizes for a given strain or species that occurred in multiple datasets were considered different observations, and so are included as multiple rows). This assumption had little influence on the data following the condense species step (next). During the condense traits step, columns with categorical values were mapped into a predefined nomenclature using manually defined lookup tables (e.g., sporulation values were mapped to either "yes" or "no"; Online-only Table 2).
Isolation source or habitat information for prokaryotes follows different schemes in different data sources, and often is unstructured, consisting of a string of words or sentences. With a view to making possible investigation of species and trait distributions across environments, we have developed for this data synthesis a scheme consisting Fig. 2 A graphical representation of data coverage and gaps for the 21 core traits mapped onto a phylogeny (black tree). The phylogeny was created by grafting star phylogenies (NCBI species to phylum) onto a recent molecular phylogeny 20 (phylum and above) and was created here purely for illustrative purposes. To avoid clutter, only the six most speciose phyla are delineated at the outer rim (>100 species). Coloured bands represent the presence of traits in the dataset for 14,884 species. In order for the centre outwards, green are habitat traits (isolation source, optimum pH, optimum temperature, growth temperature), blue are organism trait (gram stain, metabolism, metabolic pathways, carbon substrate, sporulation, motility, doubling time, cell shape, any cell diameter), and red are genomic traits (genome size, GC content, coding genes, rRNA16S genes, tRNA genes).
www.nature.com/scientificdata www.nature.com/scientificdata/ of approximately 100 environment labels. The scheme is hierarchical using up to four levels of specificity, for example a one-term label is "host", a two-term is "host_animal", a three-term is "host_animal_endotherm", and a four-term is "host_animal_endotherm_intestinal"). This allowed us to be relatively specific or relatively vague depending on the information available. To translate environment information into this new scheme, all columns Table 2. Barplots are used for categorical traits and frequency histograms for continuous traits. Due to the high number of distinct metabolic pathways (>80) (d) and carbon substrates (>100) (e) included in this data, to simplify presentation each of these were grouped into major categories; pathways were grouped by the primary compound involved or distinct processes where no primary compound exists, and carbon substrates were grouped by chemical classification. (2020) 7:170 | https://doi.org/10.1038/s41597-020-0497-4 www.nature.com/scientificdata www.nature.com/scientificdata/ in each data-source that contained environment information were concatenated into one comma-separated string, thus capturing as much information as was available in the data source. These concatenated strings were then manually translated into their most appropriate label in terms of our scheme and saved in a translation table. Given the large number of unique strings created in this way, only the most prevalent strings have at this stage been translated (>3,000), covering approximately 65% of the species in the species condensed dataset. These environmental labels were annotated with terms from the Environmental Ontology (ENVO) and stored in the "environments.csv" table in the GitHub project; however, ENVO annotations do not currently appear in the data products 19 because most environmental terms required the union of multiple ENVO terms.
A step was also included to correct datum-specific errors. Some of these likely occurred during original data entry, such as wrong units or misspellings. Others were values that seemed surprising, and also stronger or newer evidence was available from other sources. These corrections were scripted as a translation table that contained the original dataset, taxon, trait and value where the error occurred, and then the new, corrected value as well as a comment and reference as to why the change was made (see Technical Validation). The condense trait step generated three files 19 : "condensed_traits_NCBI.csv", "condensed_traits_GTDB.csv" and "references.csv".
Condense species. At this stage, rows in the dataset represented both strains and species, and each strain and species could have multiple replicate rows for a given trait. Because every row could be mapped to species (but not vice versa), data were aggregated at either the NCBI 10,17 or GTDB 18 species level. That is, all records for a given species, and strains of that species, were condensed into one record. All rows not resolved to species using these taxonomies were excluded (e.g., those with "sp. " instead of a recognised species name).
For numerical traits, aggregation consisted of calculating the average, standard deviation and number of records for a given species/trait combination. These derived values were saved as columns labelled by the trait name and then the trait name with ".stdev" and ".count" appended, respectively. The script for species condensation can be altered to calculate other derived values, like median, minimum, maximum, and so on.
For categorical traits, the majority rule was used, where terms for a given trait were tallied and the term with greater than 50% of the tally was assigned as the species aggregate. For binary categorical variables (e.g., gram stain, sporulation), and also cell shape, only the dominant term (>50% of total) was assigned and, in the case of ties, no term was assigned (i.e., the value was left blank). For categorical variables with multiple terms and levels of specificity (e.g., metabolism and motility), the following logic was employed: www.nature.com/scientificdata www.nature.com/scientificdata/ • If no single term dominated, a simple logic was used to select the most appropriate term based on grouping of terms into main categories of resemblance (e.g., aerobic vs. anaerobic, motile vs. non-motile) and specificity level (e.g., "aerobic" was considered less specific than "obligate aerobic"; for motility, "yes" was considered less specific than "flagella"). • If all terms belong to the same category, the most specific term was selected (e.g., "obligate aerobic" selected instead of "aerobic"). • If all terms belong to the same category and all have the same level of specificity (e.g., "facultative aerobic" and "obligate aerobic"), the term is converted to its least specific form (i.e., "aerobic"). • If terms belong to different categories (e.g., "aerobic" vs. "anaerobic"), then no term was assigned (i.e., the value was left blank).
Due to the hierarchical nature of the naming schemes for isolation sources, selecting the most representative term was done on a per-level basis. Each isolation source term potentially contained up to 4 levels of detail (e.g., level 1: host, level 2: animal, level 3: endotherm and level 4: blood). For each level (starting at level 1 and proceeding through levels 1 to 4), the occurrence of each term amongst all observations for a given species was counted, and the dominant term chosen and combined with the dominant term in the next level. If no dominant term could be found at a given level (not resolved), the process was stopped at that level. As such, an isolation source may contain 1 to 4 levels of information with increasing specificity.
Bergey's Manual of Systematics of Archaea and Bacteria 11 contains a large amount of useful phenotypic trait detail, such cell size, sporulation, gram, metabolism and more, across the whole of Archaea and Bacteria, but is not stored as a dataset. Therefore, this data source was used at the final stage of the species condense step to fill in data gaps, especially for traits that were easily extractable using text matching (e.g., cell size and metabolism; see scripted workflow for details). The condense species step generated two files 19 : "condensed_species_NCBI.csv" and "condensed_species_GTDB.csv".
Data Records
1. "condensed_traits_NCBI.csv": A trait condensed data record containing all focal trait data (Online-only Table 2) from original datasets using the NCBI taxonomy 19 . Rows represent strain-or species-level measurements, and there can be more than one row per taxon. On the whole, this is a strain-level, non-aggregated data record. 2. "condensed_traits_GTDB.csv": Same as "condensed_traits_NCBI.csv" but using the GTDB taxonomy 19 .
This trait condensed data record is smaller, because the GTDB protocol does not accept all NCBI taxa. 3. "references.csv": A table containing reference information for the data 19 . Each row in the trait condensed data ("condensed_traits_NCBI.csv" and "condensed_traits_GTDB.csv") has a unique ID that points to a reference in the reference table for that particular data record. Species condensed data (below) have multiple reference IDs. 4. "condensed_species_NCBI.csv": A species condensed data record contained all focal traits (Online-only Table 2) aggregated so that there is one row per NCBI-defined species 19 . 5. "condensed_species_GTDB.csv": Same as "condensed_species_NCBI.csv" but using the GTDB taxonomy 19 . However, this species condensed data record is smaller, because the GTDB protocol does not accept all NCBI taxa.
technical validation
Approximately 80% of the time spent developing this bacteria and archaea trait data pipeline was consumed by searching for and fixing errors and inconsistencies in the raw datasets that were ultimately combined. When inconsistencies across datasets could not be resolved, the data were removed. These fixes necessarily involved human judgment, hence the large time expense. All fixes to datasets have been recorded into a data correction table (in "data/conversion_tables/data_corrections.csv") that is implemented by the script so that the decision-making process is transparent. In addition to basic error checking (e.g., looking at unique lists of controlled terms, removing whitespace, etc.), we paid particular attention to outliers, which sometimes (though certainly not always) turned out to be problematic. We located outliers by inspecting distributions of the continuous traits, and also bivariate plots (e.g., by sorting residuals from model fits), or boxplots where one variable was categorical. Users who find and wish to correct further errors, or who wish to apply a different judgment about anomalous and outlier traits, can readily implement this through the same data correction and other data translation tables in the GitHub repository.
Usage Notes
The data records are available at figshare 19 . The script that generated the data records is available at GitHub (https://github.com/bacteria-archaea-traits/bacteria-archaea-traits/releases/tag/v1.0.0). Two large files were not included with the GitHub project: the NCBI taxonomy translation table and PATRIC dataset. These files are automatically downloaded to their correct directories the first time the workflow script is run. If download problems occur, instructions for where to place these large files manually can be found in the project readme file. Please note that several of the raw datasets entering into the workflow were sourced from dynamic, growing databases (see Online-only Table 1). Therefore, users of the Data Records may consider obtaining fresh versions of the different sources from the links or data providers in Online-only Table 1, and then re-applying the scripted workflow to build an updated data synthesis. Additionally, the datasets we merge contain additional traits that we do not collect in our workflow, given our broader research goals. Adding these traits requires adjusting the project settings and editing dataset specific preparation files. Instructions for doing so are in the project readme file and (2020) 7:170 | https://doi.org/10.1038/s41597-020-0497-4 www.nature.com/scientificdata www.nature.com/scientificdata/ dataset specific readme files ("data/raw"). Translation tables created to map trait variables, including isolation source, are in the "data/conversion_tables" directory. Additional quality control will be necessary following the addition of new or updated datasets and traits to the workflow.
We encourage other groups who update or add new data sources to this data product to do so using our procedure outlined in the Methods (above) and in more detail at the GitHub project readme. This project uses GitHub's standard fork and pull request workflow, which is well documented at GitHub. Such changes would follow this general pattern: • Forking the GitHub project. • Updating the existing or adding the new dataset in its raw form to the "data" repository.
• Writing a data preparation script ("R/preparation"), which includes appending NCBI taxon IDs if not already in the dataset. • Identifying the traits to be merged ("R/settings.R"), and writing a conversion table if the trait is not in the same units of categories as the present dataset version ("data/conversion_tables"). • Looking for outliers and other errors, which can be removed or altered using the corrections table ("data/ conversion_tables/data_corrections.csv") • Running and testing the merger ("workflow.R").
• Submitting a pull request via GitHub, at which point we will review and test the changes.
• Once the pull request is accepted, the project version will be updated.
Code availability
The complete data workflow was scripted in the programming language R (https://www.R-project.org) and instructions for generating the merged data sets accompanying this data descriptor can be found at GitHub (https://github.com/bacteria-archaea-traits/bacteria-archaea-traits/releases/tag/v1.0.0). | 2020-05-30T23:02:02.285Z | 2020-06-05T00:00:00.000 | {
"year": 2020,
"sha1": "6a8688efc78dfc86a5cea14f080a60a923622c8d",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41597-020-0497-4.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "290c106c18d3d0bb0a11c6432362ab7cd9f526b6",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
52029480 | pes2o/s2orc | v3-fos-license | Simultaneous Sensing of Temperature and Bi-Directional Strain in a Prismatic Li-Ion Battery
Thermal and pressure stability of Li-ion batteries (LiB) are the most important parameters for safety. In abuse operating conditions, the rapid increase of temperature and pressure can cause the appearance of hot-spots, which may lead to an increasing degradation rate or even to the battery’s explosion and/or combustion. A sensing network of fiber Bragg gratings is attached to the surface of a prismatic LiB to monitor its temperature and bi-directional strain variations through normal charge (0.70 C) and two different discharge rates (1.32 C and 5.77 C) in the xand y-directions. More significant variations are registered when the LiB operates in abnormal conditions. A maximum temperature variation of 27.52 ± 0.13 ◦C is detected by the sensors located close to the positive electrode side. Regarding strain and consequent length variations, maximum values of 593.58 ± 0.01 με and 51.05 ± 0.05 μm are respectively obtained by the sensors placed on the y-direction. The sensing network presented can be a solution for the real-time monitoring, multipoint and in operando temperature and bi-directional strain variations in the LiBs, promoting their safety.
Introduction
Lithium-ion batteries (LiB) are extensively used as power sources for a wide range of electronic devices such as smartphones and laptops, ensuring optimized conditions in the perspective of power, energy, long cycle-life, and slow self-discharge [1,2].Prismatic LiBs, characteristically used on smartphones, are more disposed to thermal and mechanical abuses from external actions.
The induced strain can be a question that affects the LiB stability and safety, making it the principal cause of material cracking and other forms of performance degradation [3].Analogous to other electrochemical energy storage systems, the chemical compositions of the active materials change under the charge/discharge processes, which induces strains in electrode particles and causes changes in LiB volume.The aim of thickness reduction of smartphones can be a problem for the users, because these new designs of the smartphone do not integrate a dually protective device which relieves at a set pressure, thus avoiding the overpressure of the LiB.Additionally, LiB companies pursue higher energy density and thinner devices at the cost of safety, which moves against the inherently safer design of a commercial LiB [4].
In addition to strain, thermal runaway is also an essential issue, with impact in the global LiB performance, which is reproduced by the fast increase of temperature.The internal structures of LiBs are made-up of multiple layers, forming a 'jelly roll' structure, where each layer consists of the anode, cathode, electrolyte, and polymer film separators.Under abnormal conditions, such as temperature exceeding the separators melting point, mechanical deformation, or breakdown of the layered materials, an internal short circuit can happen.The hot spot generated by internal short circuit can ignite thermal runaway, leading to fire or explosion of the LiB [5,6].
The ability to quantify and evaluate the mechanism of strain and thermal runaway generated during the electrochemical processes that the batteries can operate will be beneficial information regarding their behavior as well as an active tool to promote their safety.The temperature and strain sensing of LiB is typically performed using thermocouples [7,8], pyrometers [9], electro-mechanical sensors [10], and 3D digital image correlation [6].Due to their ease in multiplexing, fast response, immunity to electromagnetic interference advantages when compared to electronic sensors [11], recent works showed that fiber Bragg gratings (FBG) are an effective method to perform temperature and strain measurements in LiBs [12][13][14][15][16].
In this work, through a sensing network of FBGs, quantitative temperature, bi-directional strain shifts, and correspondent longitudinal and transversal variations are provided, as a function of the respective voltage signal of a prismatic LiB, during different cycling protocols.
FBG Sensors: Mechanism of Operation
Typically, a FBG sensor consists of a short segment of a single-mode optical fiber (with a length of a few millimeters) with a photoinduced periodically modulated index of refraction in the core of the fiber.The FBG resonant wavelength is related to the effective refractive index of the core mode (n eff ) and to the grating period (Λ).When the grating is illuminated with a broadband optical source, the reflected power spectrum presents a sharp peak (with a full width at half maximum of a few nanometers), which is caused by interference of light with the planes of the grating and can be described through Equation (1) [11] where λ B is the so-called Bragg wavelength.When the fiber is exposed to external variations of a given measurand (such as strain, temperature, stress or pressure, among others), both n eff and Λ can be altered, causing a shift in the Bragg wavelength.
The FBG sensitivity towards a given parameter is obtained simply by subjecting the sensor to pre-determined and controlled variations of such parameter and measuring the Bragg wavelength for each step.In the case of a linear response, the sensitivity (K) is given by the slope of the linear fit obtained from the experimental data.The effects of temperature are accounted for in the Bragg wavelength shift by differentiating Equation ( 1) where α and ξ are the thermal expansion and thermo-optic coefficient of the optical fiber material, respectively.On the other hand, if the fiber is subjected to strain variations, its response can be determined by differentiating Equation ( 1) where p e is the photoelastic constant of the fiber (~0.22) and ∆ε is the applied strain.The strain variations can be determined using the equation ∆ε = ∆L/L where ∆L is the length variation and L the fiber length over which strain is applied.
Strain and Temperature Discrimination: Reference FBG Method
The temperature and bi-directional strain discrimination were performed using the reference FBG method.On a single measurement of the Bragg wavelength-shift, it is not possible to discriminate the effect of changes in strain and temperature.Therefore, a reference is required for temperature measurement.Several methods have been addressed in the literature for an accurate discrimination between these two parameters [17][18][19].The most straightforward way is to use a separated, strain-free FBG, which acts as a temperature sensor.In this case, the wavelength shift, ∆λ B1 , is related to temperature variations, ∆T 1 , according to where K T1 is the temperature sensitivity of the non-fixed FBG.This reference FBG is in the same thermal environment as the other sensors, which simultaneously detect strain and temperature.Assuming that the wavelength-shifts to strain and temperature are linear, a response to a strain change, ∆ε 2 and a temperature change, ∆T 2 , is given by where K ε2 and K T2 are the strain and temperature sensitivities of the fixed FBG [11], determined in the calibration procedure.Therefore, using Equation ( 4), the wavelength variations detected by the strain-free FBG can be converted into temperature variations.These temperature variations are then considered in Equation ( 5), allowing the determination of strain.
Experimental Setup
A network of five FBG sensors, recorded in two different fibers, was placed in direct contact with the LiB surface on the xand y-directions, as schematized in Figure 1.The FBGs network used in the monitoring system, with estimated gratings length of ~3.0 mm, was pre-fabricated in our lab and written in commercial photosensitive single mode fiber (125 µm of cladding diameter) (FiberCore PS1250/1500) by the phase mask method.The UV radiation system used was a pulsed excimer laser (KrF) (Industrial Bragg Star, Coherent, CA, USA), emitting at a wavelength of 248 nm, 4 mJ/pulse (20 ns duration), and 500 Hz repetition rate.
Batteries 2018, 4, x FOR PEER REVIEW 3 of 9 separated, strain-free FBG, which acts as a temperature sensor.In this case, the wavelength shift, ΔλB1, is related to temperature variations, ΔT1, according to where KT1 is the temperature sensitivity of the non-fixed FBG.This reference FBG is in the same thermal environment as the other sensors, which simultaneously detect strain and temperature.Assuming that the wavelength-shifts to strain and temperature are linear, a response to a strain change, Δε2 and a temperature change, ΔT2, is given by where Kε2 and KT2 are the strain and temperature sensitivities of the fixed FBG [11], determined in the calibration procedure.Therefore, using Equation ( 4), the wavelength variations detected by the strain-free FBG can be converted into temperature variations.These temperature variations are then considered in Equation ( 5), allowing the determination of strain.
Experimental Setup
A network of five FBG sensors, recorded in two different fibers, was placed in direct contact with the LiB surface on the x-and y-directions, as schematized in Figure 1.The FBGs network used in the monitoring system, with estimated gratings length of ~3.0 mm, was pre-fabricated in our lab and written in commercial photosensitive single mode fiber (125 µm of cladding diameter) (FiberCore PS1250/1500) by the phase mask method.The UV radiation system used was a pulsed excimer laser (KrF) (Industrial Bragg Star, Coherent, CA, USA), emitting at a wavelength of 248 nm, 4 mJ/pulse (20 ns duration), and 500 Hz repetition rate.As can be seen in Figure 1, the FBG1 was placed on the x-direction, FBG4 on the y-direction and FBG3 out of the battery to monitor the external room temperature.These sensors were only used to measure temperature variations, as they were not fixed to the battery.The simultaneous detection of the strain and temperature variations was accomplished by fixing a second fiber side by side to the first one, with two sensors: FBG2 and FBG5.This fiber was pre-tensioned and fixed to the battery on its extremities, along the LiB x-and y-directions.Thermal paste was placed along both fibers to increase thermal conductivity between the battery and the FBGs.To calculate the length variations As can be seen in Figure 1, the FBG1 was placed on the x-direction, FBG4 on the y-direction and FBG3 out of the battery to monitor the external room temperature.These sensors were only used to measure temperature variations, as they were not fixed to the battery.The simultaneous detection of the strain and temperature variations was accomplished by fixing a second fiber side by side to the first one, with two sensors: FBG2 and FBG5.This fiber was pre-tensioned and fixed to the battery on its extremities, along the LiB xand y-directions.Thermal paste was placed along both fibers to increase thermal conductivity between the battery and the FBGs.To calculate the length variations on the two directions, the distance between the two fixed points was considered.As the glue was placed as close to the edge of the LiB as possible, the length variations are related to the battery dimension.
The LiB used in this work was a commercial hard prismatic rechargeable smartphone LiB (Iphone 5 G Battery, Singapore) with an open-circuit voltage of 4.30 V, nominal capacity of 1440 mAh, cut-off voltage of 3.20 V, dimensions of 8.6 (length) × 3.0 (width) × 0.3 cm (thickness), and mass of 23.47 g.Inside the aluminum-plastic pouch, the internal structure is constructed by winding the ribbon-like electrode and separator to form a 'jelly-roll' structure.The negative and positive active materials are coated on each side of the separator materials, and thin layers of copper and aluminum foils are used as the current collectors [4].
Prior to the fiber sensor network's attachment to the LiB, a calibration to strain and temperature was carried out.The strain characterization was performed using a micrometric translation stage between 0 µε and 2000 µε.The sensors thermal calibration was made on a thermal chamber (Model 340, Challenge Angelantoni Industrie, Massa Martana, Italy).The temperature range was between 20.0 • C and 50.0 • C, in steps of 5.0 • C. The temperature and strain sensitivities obtained are presented in Table 1.The reflected Bragg wavelengths were measured with an optical interrogator (sm125-500, Micron Optics Inc., Atlanta, GA USA) operating at 2.0 Hz and wavelength accuracy of 1.0 pm.The voltage signal was monitored using a 12-bit resolution data acquisition system (DAQ) (USB6008, National Instruments, Austin, TX, USA).The acquisition modules were controlled by a LabVIEW ® customized application, allowing the real-time monitoring of the acquired data.The LiB was cycled twice under two different discharge rates of 1.32 C and 5.77 C (where 1.00 C is the rate of the ideal capacity charge/discharge in 1 h) through two power resistors of 2.0 Ω and 0.47 Ω until a voltage of 1.95 V was reached.A normal charge rate of 0.70 C was applied, using a commercial battery charger with output voltage of 5.00 V (ETA0U83EWE, Samsung, Vietnam).
Results and Discussion
The temperature variations registered by the FBG sensors placed on the two sides of the battery are presented in Figures 2 and 3.The positive and negative electrodes indicated in the Figures correspond to the sensors placed on the yand x-directions, respectively.The cycling protocol, where the discharge rate of 1.32 C was applied, is shown in Figure 2.Over the CC charge step (1), a maximum temperature variation of 1.29 ± 0.13 • C was measured by the sensors placed on the positive electrode side, whereas on the negative electrode side, the maximum temperature shift was of 0.86 ± 0.13 • C. Between the constant current (CC) charge/discharge steps, constant voltage (CV) steps (2) of 10 min and 15 min were selected to stabilize the temperature and relax the battery.
registered by the FBG sensors located on the positive and negative electrodes.As expected, the main difference was in the temperature variations achieved during the CC discharge step (3).In all discharge steps, a monotonic increase can be observed, until the LiB reaches 1.95 V.A maximum temperature variation of 27.52 ± 0.13 °C was recorded on the positive electrode side by the FBG4.On the negative electrode side, an inferior ΔT value of 27.25 ± 0.13 °C was obtained, however, this difference is not significant, taking the measurement error into account.The bi-directional strain variations detected by the FBG sensors on the x-and y-directions, during the experimental cycling protocol with the discharge rate of 1.32 C, as well as their voltage signal, are presented in Figure 4.Over the CC charge steps (1), mean strain variations of 122.49± 0.01 positive electrode negative electrode registered by the FBG sensors located on the positive and negative electrodes.As expected, the main difference was in the temperature variations achieved during the CC discharge step (3).In all discharge steps, a monotonic increase can be observed, until the LiB reaches 1.95 V.A maximum temperature variation of 27.52 ± 0.13 °C was recorded on the positive electrode side by the FBG4.On the negative electrode side, an inferior ΔT value of 27.25 ± 0.13 °C was obtained, however, this difference is not significant, taking the measurement error into account.The bi-directional strain variations detected by the FBG sensors on the x-and y-directions, during the experimental cycling protocol with the discharge rate of 1.32 C, as well as their voltage signal, are presented in Figure 4.Over the CC charge steps (1), mean strain variations of 122.49± 0.01 positive electrode negative electrode The starting of the CC discharge step (3) causes an instantaneous increase on the battery surface temperature, however, the maximum temperature shift was achieved only at the end of the discharge process.These fast and significant increases are related to the larger electrochemical reactions produced by the flow of Li+ ions on migration to the positive electrode.During all the discharge steps, two different moments of temperature variations can also be observed.The first moment occurs when the LiB voltage crosses the cut-off voltage, and the second between the cut-off and the end of the discharge process.The maximum temperature variations detected were very similar on both electrodes, around 11.50 ± 0.13 • C.
Figure 3 shows the cycling protocol where a discharge rate of 5.77 C was applied.Attending to the CC charge steps (1) and comparing with previous cycling protocol, the same range of values was registered by the FBG sensors located on the positive and negative electrodes.As expected, the main difference was in the temperature variations achieved during the CC discharge step (3).In all discharge steps, a monotonic increase can be observed, until the LiB reaches 1.95 V.A maximum temperature variation of 27.52 ± 0.13 • C was recorded on the positive electrode side by the FBG4.On the negative electrode side, an inferior ∆T value of 27.25 ± 0.13 • C was obtained, however, this difference is not significant, taking the measurement error into account.
The bi-directional strain variations detected by the FBG sensors on the xand y-directions, during the experimental cycling protocol with the discharge rate of 1.32 C, as well as their voltage signal, are presented in Figure 4.Over the CC charge steps (1), mean strain variations of 122.49± 0.01 µε and 29.61 ± 0.01 µε were measured on the yand x-directions, respectively, when the battery reached the 3.95 V.As already mentioned, during the 10 min CV step (2), between the CC charge (1) and CC discharge (3) steps, a relaxation occurs on the battery.This relaxation was higher on the y-direction, translating into a strain diminishing of 25.58 ± 0.01 µε.
Batteries 2018, 4, x FOR PEER REVIEW 6 of 9 µε and 29.61 ± 0.01 µε were measured on the y-and x-directions, respectively, when the battery reached the 3.95 V.As already mentioned, during the 10 min CV step (2), between the CC charge (1) and CC discharge (3) steps, a relaxation occurs on the battery.This relaxation was higher on the y-direction, translating into a strain diminishing of 25.58 ± 0.01 µε.
When the battery is subjected to the CC discharge step (3), at 1.32 C, three different behaviors can be observed on the two different directions.In the first one, until the LiB reached the 3.30 V, there is a sudden increase of strain.In the y-direction this increase is nearly two-times higher than the one measured in the x-direction.After this, a strain decreases of 7.39 ± 0.01 µε and 6.46 ± 0.01 µε was measured in the y-and x-directions, respectively, until the cut-off voltage was reached.This was followed by a fast increase of strain, as the battery starts to operate under abusive conditions, between 3.20 V and 1.95 V.In this case, the y-direction strain increased up to 213.53 ± 0.01 µε.On the x-direction, a 3.4-times lower strain value was recorded.
These values can be converted to length variations (ΔL), by multiplying the strain by the length or the width of the LiB on the y-or x-directions, respectively.Thus, maximum ΔL of 18.36 ± 0.05 µm and 1.86 ± 0.05 µm on the y-and x-directions were calculated, respectively.These length variations can be translated to a longitudinal and transversal variation in the battery of 0.021% and 0.006% in the y-and x-directions, respectively.The same LiB was submitted to a different cycling protocol test, where a different discharge rate of 5.77 C was applied.This higher discharge rate was selected so that the battery would operate under abnormal conditions.However, the same charge rate was used to charge the battery at 3.95 V.The bi-directional strain variations obtained are presented in Figure 5.
Comparatively to the cycling protocol presented previously and attending the CC charge steps (1) and CV steps (2), the same range of Δε were obtained, causing maximum ΔL of 10.53 ± 0.05 µm and 0.89 ± 0.05 µm in the y-and x-directions, respectively, and longitudinal and transversal shifts of 0.012% and 0.003%, respectively, on the battery.
When the battery is subjected to CC discharge steps (3), a monotonic increase can be observed in both directions.However, a different increasing time occurs, before and after the battery reached the cut-off voltage, following the fast voltage decrease.On the y-direction, a maximum strain variation of 593.58 ± 0.01 µε was measured and a correspondent ΔL of 51.05 ± 0.05 µm was calculated.This high elongation translated in a battery longitudinal variation of 0.060%.The sensors located in When the battery is subjected to the CC discharge step (3), at 1.32 C, three different behaviors can be observed on the two different directions.In the first one, until the LiB reached the 3.30 V, there is a sudden increase of strain.In the y-direction this increase is nearly two-times higher than the one measured in the x-direction.After this, a strain decreases of 7.39 ± 0.01 µε and 6.46 ± 0.01 µε was measured in the yand x-directions, respectively, until the cut-off voltage was reached.This was followed by a fast increase of strain, as the battery starts to operate under abusive conditions, between 3.20 V and 1.95 V.In this case, the y-direction strain increased up to 213.53 ± 0.01 µε.On the x-direction, a 3.4-times lower strain value was recorded.
These values can be converted to length variations (∆L), by multiplying the strain by the length or the width of the LiB on the yor x-directions, respectively.Thus, maximum ∆L of 18.36 ± 0.05 µm and 1.86 ± 0.05 µm on the yand x-directions were calculated, respectively.These length variations can be translated to a longitudinal and transversal variation in the battery of 0.021% and 0.006% in the yand x-directions, respectively.
The same LiB was submitted to a different cycling protocol test, where a different discharge rate of 5.77 C was applied.This higher discharge rate was selected so that the battery would operate under abnormal conditions.However, the same charge rate was used to charge the battery at 3.95 V.The bi-directional strain variations obtained are presented in Figure 5.
In all cycles, it was observed that the maximum strain variations are registered when the battery discharge process ends.A relation between the different discharge rates which are applied and the longitudinal and transversal variations on the LiB can be performed by considering the battery dimensions and the ΔL values obtained in the two directions.For instance, the ratio of the LiB dimensions is 2.87, whereas dividing the longitudinal by the transversal variation measured over the higher discharge rate, a ratio of 2.73 is calculated.Making the same analogy but for the discharge rate of 1.32 C, a ratio value of 3.50 is obtained.This means that the expansion and contraction of the internal 'jelly roll' structure that constitutes the prismatic battery behaves differently on each charge/discharge process which is subject, and it is not proportional to the LiB dimensions.Through a complete analysis of the cycling test, it is evident that, in both cases, the maximum values obtained for the temperature and strain variations agree with the sudden decrease of the voltage.Considering all strain and temperature values measured, different behaviors in terms of direction of expansion and temperature variations on the two sides of the battery were detected.These results are in good agreement with the ones found in the literature [6,9,10].
The monitoring of the internal and external parameters during charge and discharge cycles, combined with battery management systems, seems to be adequate to improve overall safety.They also could be used to determine realistic operating conditions and geometries, towards more stable cell pack designs.
Conclusions
A network of FBG sensors was successfully used to simultaneously monitor temperature and bidirectional (x-and y-direction) strain in a prismatic rechargeable LiB, under an experimental cycling protocol with normal CC charge and different CC discharge steps (1.32 C and 5.77 C).When the battery was subjected to abnormal operating conditions, as fast discharge and operating below the cut-off voltage, it is evident that higher temperature and strain variations occur, which are promoted by the rapid Li+ transport between the positive and negative electrodes.Over the CC charge step, maximum temperature and strain variations were reached at the end of the process, with values of 1.29 ± 0.13 °C in the positive electrode side and 122.49± 0.01 µε and 29.61 ± 0.01 µε on the y-and xdirections, respectively.Comparatively to the cycling protocol presented previously and attending the CC charge steps (1) and CV steps (2), the same range of ∆ε were obtained, causing maximum ∆L of 10.53 ± 0.05 µm and 0.89 ± 0.05 µm in the yand x-directions, respectively, and longitudinal and transversal shifts of 0.012% and 0.003%, respectively, on the battery.
When the battery is subjected to CC discharge steps (3), a monotonic increase can be observed in both directions.However, a different increasing time occurs, before and after the battery reached the cut-off voltage, following the fast voltage decrease.On the y-direction, a maximum strain variation of 593.58 ± 0.01 µε was measured and a correspondent ∆L of 51.05 ± 0.05 µm was calculated.This high elongation translated in a battery longitudinal variation of 0.060%.The sensors located in the x-direction followed a similar behavior as in the previous cycling protocol and detected lower strain variations.In this case, a maximum transversal variation of 0.022% was obtained.
In all cycles, it was observed that the maximum strain variations are registered when the battery discharge process ends.A relation between the different discharge rates which are applied and the longitudinal and transversal variations on the LiB can be performed by considering the battery dimensions and the ∆L values obtained in the two directions.For instance, the ratio of the LiB dimensions is 2.87, whereas dividing the longitudinal by the transversal variation measured over the higher discharge rate, a ratio of 2.73 is calculated.Making the same analogy but for the discharge rate of 1.32 C, a ratio value of 3.50 is obtained.This means that the expansion and contraction of the internal 'jelly roll' structure that constitutes the prismatic battery behaves differently on each charge/discharge process which is subject, and it is not proportional to the LiB dimensions.
Through a complete analysis of the cycling test, it is evident that, in both cases, the maximum values obtained for the temperature and strain variations agree with the sudden decrease of the voltage.Considering all strain and temperature values measured, different behaviors in terms of direction of expansion and temperature variations on the two sides of the battery were detected.These results are in good agreement with the ones found in the literature [6,9,10].
The monitoring of the internal and external parameters during charge and discharge cycles, combined with battery management systems, seems to be adequate to improve overall safety.They also could be used to determine realistic operating conditions and geometries, towards more stable cell pack designs.
Conclusions
A network of FBG sensors was successfully used to simultaneously monitor temperature and bi-directional (x-and y-direction) strain in a prismatic rechargeable LiB, under an experimental cycling protocol with normal CC charge and different CC discharge steps (1.32 C and 5.77 C).When the battery was subjected to abnormal operating conditions, as fast discharge and operating below the cut-off voltage, it is evident that higher temperature and strain variations occur, which are promoted by the rapid Li+ transport between the positive and negative electrodes.Over the CC charge step, maximum temperature and strain variations were reached at the end of the process, with values of 1.29 ± 0.13 • C in the positive electrode side and 122.49± 0.01 µε and 29.61 ± 0.01 µε on the yand x-directions, respectively.
During the CC discharge step, the higher strain and temperature values of 593.58 ± 0.01 µε and 27.52 ± 0.13 • C were respectively register for the higher discharge rate and when the LiB voltage was lower.These values also relate to the y-direction and positive electrode side and correspond to a battery longitudinal expansion of 0.060%.As expected, there is a deformation increase when the temperature also increases, due the thermal expansion of the materials that compose the battery.Thus, the internal structure of the battery is an important parameter to have in consideration and can influence the behavior of battery materials in terms of expansion and contraction over its operation.
The sensing network presented proved to be an effective, precise, alternative solution to real time monitor, multipoint and in operando temperature and bi-directional strain changes in the LiBs promoting their safety.
Figure 1 .
Figure 1.Experimental setup diagram of the temperature and bi-directional strain monitoring system of the battery: (a) down view; (b) top view.
Figure 1 .
Figure 1.Experimental setup diagram of the temperature and bi-directional strain monitoring system of the battery: (a) down view; (b) top view.
Figure 2 .
Figure 2. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 1.32 C.
Figure 3 .
Figure 3. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 5.77 C.
Figure 2 .
Figure 2. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 1.32 C.
Figure 2 .
Figure 2. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 1.32 C.
Figure 3 .
Figure 3. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 5.77 C.
Figure 3 .
Figure 3. Voltage and temperature variations recorded during two cycling tests, in the sides of the positive and negative electrodes of the battery.(1) CC charge; (2) CV; (3) CC discharge at 5.77 C.
Figure 4 .
Figure 4. Voltage and strain variations recorded during two cycling tests in the battery in the x-and y-directions.(1) CC charge; (2) CV; (3) CC discharge at 1.32 C.
Figure 4 .
Figure 4. Voltage and strain variations recorded during two cycling tests in the battery in the xand y-directions.(1) CC charge; (2) CV; (3) CC discharge at 1.32 C.
Figure 5 .
Figure 5. Voltage and strain variations recorded during two cycling tests in the battery in the x-and y-directions.(1) CC charge; (2) CV; (3) CC discharge at 5.77 C.
Figure 5 .
Figure 5. Voltage and strain variations recorded during two cycling tests in the battery in the xand y-directions.(1) CC charge; (2) CV; (3) CC discharge at 5.77 C. | 2019-04-09T13:10:42.910Z | 2018-05-10T00:00:00.000 | {
"year": 2018,
"sha1": "3651f1ba2e080fbeaecb7182c1ab152b42d7074e",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2313-0105/4/2/23/pdf?version=1525946027",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "3651f1ba2e080fbeaecb7182c1ab152b42d7074e",
"s2fieldsofstudy": [
"Materials Science",
"Engineering",
"Physics"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
236640179 | pes2o/s2orc | v3-fos-license | Insights Into Patient Variability During Ivacaftor-Lumacaftor Therapy in Cystic Fibrosis
Background: The advent of cystic fibrosis transmembrane conductance regulator protein (CFTR) modulators like ivacaftor have revolutionised the treatment of cystic fibrosis (CF). However, due to the plethora of variances in disease manifestations in CF, there are inherent challenges in unified responses under CFTR modulator treatment arising from variability in patient outcomes. The pharmacokinetic (PK) data available for ivacaftor-lumacaftor cystic fibrosis (CF) transmembrane conductance regulator (CFTR) modulator drug combination is limited. Methods: Secondary objectives were to identify (1) patient characteristics and (2) the interactions between ivacaftor-lumacaftor responsible for interindividual variability (IIV). Results: Peak plasma concentrations (Cmax) of ivacaftor - lumacaftor were >10 fold lower than expected compared to label information. The one-way ANOVA indicated that the patient site had an effect on Cmax values of ivacaftor metabolites ivacaftor-M1, ivacaftor-M6, and lumacaftor (p < 0.001, p < 0.001, and p < 0.001, respectively). The Spearman’s rho test indicated that patient weight and age have an effect on the Cmax of lumacaftor (p = 0.003 and p < 0.001, respectively) and ivacaftor metabolite M1 (p = 0.020 and p < 0.001, respectively). Age (p < 0.001) was found to effect on Cmax of ivacaftor M6 and on Tmax of ivacaftor M1 (p = 0.026). A large impact of patient characteristics on the IIV of PK parameters Cmax and Tmax, was observed among the CF patients. Conclusion: Understanding the many sources of variability can help reduce this individual patient variability and ensure consistent patient outcomes.
STUDY SUMMARY:
• What is already known about this subject: The pharmacokinetic (PK) data available for ivacaftor-lumacaftor cystic fibrosis (CF) transmembrane conductance regulator (CFTR) modulator drug combination is limited. • What this study adds: ○ Peak plasma concentrations significantly lower than expected compared to label information. ○ A high variability in ivacaftor-lumacaftor pharmacokinetics based on e.g. age and weight was observed. • Limitations: Small patient collective with a number of individual variables.
INTRODUCTION
Cystic fibrosis (CF) is an autosomal recessive genetic disorder that affects chloride transport throughout the epithelial cells of the body, resulting in abnormalities in the respiratory, endocrine, gastrointestinal, and reproductive systems (Condren and Bradshaw, 2013;Tan et al., 2020). The dysfunction of the cystic fibrosis transmembrane conductance regulator (CFTR) channel causes dehydration of mucosal surfaces subsequently increasing viscous mucus that obstructs luminal compartments in lung, pancreas and intestine (Elborn, 2016;Schneider et al., 2017a). The discovery and development of CFTR modulators that directly influence the dysfunctional chloride channel has had a significant impact on CF treatment world-wide (Ghelani and Schneider-Futschik, 2019;Schneider-Futschik, 2019;Allobawi et al., 2020). Ivacaftor-lumacaftor is a fixed-dose tablet containing a corrector (lumacaftor) and potentiator (ivacaftor) of the CFTR (Deeks, 2016). Ivacaftor monotherapy is approved for children 6 more older suffering from gating mutations. While ivacaftor-lumacaftor therapy is approved for children 2 yr and older with a homozygous F508del mutation and for patients 6 yr and older with homozygous F508del mutation or heterozygous and with one residual function mutation. The ivacaftor-lumacaftor combination is prescribed as a tablet (200 mg lumacaftor and 125 mg ivacaftor, Table 1) dosed every 12 h. Following its oral administration, both ivacaftor and lumacaftor are readily absorbed from the gut; however, they have low solubility in water (<0.05 μg/ml). It has been reported that high fat meals can improve the absorption of (Fohner et al., 2017). As a CYP3A4 substrate ivacaftor undergoes extensive liver metabolism and is, therefore, affected when administered with CYP3A4 inducers such as lumacaftor (Schneider, 2018). Furthermore, based on in vitro data lumacaftor and one of the major metabolites of ivacaftor, ivacaftor-M6 are strong inducers of CYP3A4, impacting the PK profile of ivacaftor by potentially reducing ivacaftor concentrations in vivo (Schneider et al., 2016;Schneider, 2018). Additionally, ivacaftor-lumacaftor induces several cytochrome P450 enzymes including 3A4, 2B6, 2C9, 2C19, and p-glycoprotein (EMA, 2015;Robertson et al., 2015;Schneider, 2018). Other factors that are known to cause variability in response across patients are the concomitant intake of high fat containing foods as they result in increased absorption of ivacaftor by ∼2.5 to 4-fold.
This paper presents data on the pharmacokinetics (PK) of ivacaftor-lumacaftor. The objectives of this study were to evaluate the PK interactions between ivacaftor and lumacaftor and to evaluate the effects of patient characteristics on their pharmacokinetics. We observed that patient characteristics such as age and weight resulted in large IIV in PK among CF patients.
MATERIALS AND METHODS
The study protocol was approved by Monash University Human Research Ethics Committee (project number 0426) and carried out in conformity with the declaration of Helsinki (amended 1986).
Patients 35 female and 25 male patients from five different treatment centers in Melbourne (n 1), Berlin (n 8), Innsbruck (n 9), Munich (n 17) and Paris (n 25): currently taking ivacaftorlumacaftor were included in this study after providing written informed consent. All patients experiencing acute infections or exacerbations were excluded. Demographic data collected for these patients included age, gender, and weight. Patients included did not receive any co-medications and were on standard high fat diets.
Study Design and Drug Administration
This study was a multiple dose, multi-center, open, observational trial reflecting a "real-life" clinical scenario. Patients received ivacaftor-lumacaftor combination administered as an oral tablet twice daily as a fixed dose combination of 125 mg ivacaftor + 200 mg lumacaftor therapy (Supplementary Table 1). The tablet was recommended to be taken with a high fat meal to ensure maximal absorption.
Blood Sampling
Blood samples (5 ml) were collected into Vacutainer tubes (Becton Dickinson, Rutherford, NJ, United States) containing ethylene-diamine tetra acetic acid as the anticoagulant. Blood samples for measurements of ivacaftor-lumacaftor were taken over the weekly dosing interval immediately before and over a 12h interval after administration of ivacaftor-lumacaftor. Blood was equilibrated at 20°C for 10 min and then centrifuged (1500 × g at 20°C for 10 min) to separate the plasma. The plasma aliquots were stored at −80°C until analysis.
Assay Data
The 90% confidence limits around the mean assay biases of duplicate quality control samples were determined at low, medium and high concentrations from pooled batches using the LC/MS. The closeness of agreement between a series of measurements obtained from multiple sampling of the same homogeneous sample under the prescribed conditions is expressed as the precision of the analytical procedure. We considered precision at three levels: repeatability, intermediate precision and reproducibility. Herein, the precision of the analytical procedure is expressed as relative standard deviation or coefficient of variation of a series of measurements. The relative standard deviation was calculated by taking the standard deviation of the sample set multiplied by 100% and dividing it by the sample set average. The relative standard deviation is expressed as percent: %RSD s/x * 100, where s is the standard deviation and x is the average of three independent measurements. The coefficient of variation (CV) acceptable for both ivacaftor and lumacaftor is a reflection of between-batch precision for peak response data.
Pharmacokinetic Analysis
PK samples were obtained 4-6 h post administration to characterize C max and T max , other samples were around 2.5 and 10 h time post administration. Plasma samples were analyzed as previously described to characterize Ivacaftor, its metabolites M1 and M6 and lumacaftor concentrations (Schneider et al., 2016;Schneider et al., 2017b).
Statistical Analysis
One-way Analysis of Variance (ANOVA) with Tukey post-hoc test was preformed to evaluate the effect of categorical demographic data (sex, site) on PK parameters (C max and T max ) of ivacaftor, ivacaftor-M1, ivacaftor-M6, and lumacaftor. Spearman's Rho non-parametric test was performed to measure the strength of the association between the continuous demographic data (age, weight, and height) and PK Frontiers in Pharmacology | www.frontiersin.org August 2021 | Volume 12 | Article 577263 parameters (C max and T max ) of ivacaftor, M1, M6, and lumacaftor.
Pharmacokinetic Data
A total of 667 pharmacokinetic samples were obtained from the 60 patients (166 ivacaftor, 169 ivacaftor-M1, 163 ivacaftor-M6, and 169 lumacaftor samples). Mean and median plasma PK data for ivacaftor -lumacaftor are summarized in Supplementary Table 3 and illustrated in Figure 2. C max concentrations of ivacaftor -lumacaftor were >10 fold lower compared to the concentrations reported in the drug approval report (EMA, 2015).
Patient Characteristic Effects on Pharmacokinetics
The one-way ANOVA indicated that the patient site had a significant effect on the C max values of ivacaftor-M1, ivacaftor-M6, and lumacaftor (p < 0.001, p < 0.001, and p < 0.001, respectively). Additionally, the site has a significant effect on T max for ivacaftor-M1 and ivacaftor-M6 metabolites of ivacaftor (p 0.001 and p 0.004, respectively). The Spearman's rho test indicated that patient weight and age have a significant effect on the C max of lumacaftor (p 0.003 and p < 0.001, respectively) and ivacaftor-M1 (p 0.020 and p < 0.001, respectively). Age (p < 0.001) was found to have a significant effect on C max of ivacaftor-M6 and on T max of ivacaftor M1 (p 0.026). Patient characteristics had no significant impact on ivacaftor PK. For ivacaftor-M1 site, weight and age had an effect on C max , respectively (p < 0.001, p 0.02, p < 0.001) and for site and
DISCUSSION
Drug action is the function of the net absorption, distribution, metabolism, excretion and interactions with target sites which are influenced by genetic factors (Dubovsky, 2015). In CF care, screening for potential drug-drug interactions is of the paramount importance to identify and potentially substitute co-medications that alter drug bioavailability via cytochrome P450 (CYP450) induction or inhibition of CYP450. Clinically relevant drug interactions are metabolized by CYP450 enzymes and are divided into families (Condren and Bradshaw, 2013;Tan et al., 2020;Elborn, 2016), sub-families (A-E) and individual gene number. Genetic polymorphisms in a given CYP450 gene result in variations in the enzyme activity leading to poor metabolizers, extensive metabolizers, or ultra-rapid metabolizers. Furthermore, induced cytochrome enzyme activity results in enhanced metabolism of other drug substances, potentially decreasing exposure and reducing therapeutic efficacy. Ivacaftor, a known CYP3A4 substrate, undergoes extensive liver metabolism, and concentrations are likely to be affected when administered concurrently with CYP3A4 inducers (EMA, 2015;VERTEX, 2015). In a recent in vitro study, we have reported strong CYP3A4 induction of lumacaftor (Schneider, 2018). These findings together with the data from the EMEA report between lumacaftor and ivacaftor could be at play, wherein the former induces the metabolism of the latter, effectively reducing its effective plasma concentration (EMA, 2015;Schneider, 2018). Ivacaftor is metabolized in the liver by cytochromes CYP3A4 and -A5 with main metabolites produced by oxidation [hydroxymethyl-ivacaftor (ivacaftor-M1) and ivacaftor-carboxylate (ivacaftor-M6)] (Condren and Bradshaw, 2013). Elimination of ivacaftor and metabolites ivacaftor-M1 and ivacaftor-M6 occurs predominantly through the bile (Table 1) (Wainwright, 2014). Lumacaftor is not heavily metabolized with the majority being excreted unchanged in the faeces (EMA, 2015;Sponsor, 2015). Further metabolising enzyme interactions have been reported for ivacaftor including on sensitive substrates of CYP-3A4, -2C8, -2D6 and P-glycoprotein (Robertson et al., 2015). Hence, careful monitoring is recommended when ivacaftor is co-administered with substrates of CYP2C9, CYP3A, and/or P-glycoprotein, particularly drugs with a narrow therapeutic index (Robertson et al., 2015;Paulin and Schneider-Futschik, 2020). Combination therapy of CFTR potentiator and -corrector combinations has progressively transformed quality of life for the majority of patients. Current CFTR combination treatment is selected based on the CFTR mutation e.g. ivacaftor-lumacaftor combination for F508del CFTR mutations. However, clinical relevance of CYP polymorphisms related to dose, effectiveness and/or toxicity are key issues when prescribing other drugs like warfarin, tricyclic antidepressants or proton pump inhibitors. Despite a large number of poor and non-responders under ivacaftor-lumacaftor therapy genomic profiling including cytochrome profiling is not yet considered standard of care (Figure 4, Supplementary Figure 1).
There is an urgent need for well-designed clinical studies aimed at collecting the necessary data to perform population PK analysis in patients undergoing ivacaftor-lumacaftor therapy once they attained steady-state concentrations. The objective of our study was to assess the impact of patient characteristics on the PK of ivacaftor-lumacaftor following multiple doses administered to patients with CF at five different treatment sites. The regimen of 125 mg ivacaftor + 200 mg lumacaftor twice daily is the standard maintenance dose for the treatment of CF (with halflive of 12-13 h for ivacaftor alone; 26 h for lumacaftor alone and 9 h for the combination). Sample collection was performed after patients were on the medication for adequate time to ensure that near steady-state concentrations for both drugs was achieved. We observed that the patient site had a significant effect on the C max values of ivacaftor-M1, ivacaftor-M6, and lumacaftor as well as on T max for the ivacaftor-M1 and ivacaftor-M6. Bioavailability and absorption of ivacaftor are positively correlated with high-fat containing foods (EMA, 2015). A limitation of our study was that the dietary intake of patients at the different sites based on culturally or other choices could not be monitored and could have influenced the absorption of ivacaftor-lumacaftor. Furthermore, we observed that patient weight and age have a significant effect on the C max of lumacaftor and ivacaftor-M1. In particular, age was found to have a significant effect on C max of ivacaftor-M6 and on T max of ivacaftor-M1.
The steady-state exposure of ivacaftor was lower than that seen following one day of therapy and this could be likely due to the CYP3A induction effect of lumacaftor. We observed that C max concentrations of ivacaftor-lumacaftor were >10 fold lower than the values reported in the approval report [9]. Our findings are in line with the respective biomarker study by Masson et al. (2019) (Masson et al., 2019). The authors investigated the biomarker profile and PK/pharmacodynamics (PD) on ivacaftor-lumacaftor in 41 French patients utilizing potential screening tools for predicting PK/PD and patient outcomes utilizing CF biomarkers such as sweat chloride, β-adrenergic peak sweat, lung function (percentage predicted FEV1 [ppFEV1] and residual volume) and CFTR activity. A 5% improvement in Frontiers in Pharmacology | www.frontiersin.org August 2021 | Volume 12 | Article 577263 6 FEV1 after 6 months of ivacaftor-lumacaftor therapy was determined as the clinical response threshold. The authors found that only 15 patients had an increase in ppFEV1 of at least 5% (average improvement 13.5% [2.1]; p < 0.0001) and were classified as responders; however a majority of patients had a ppFEV1 below 5% and were classified as non-responders (n 21, average change of ∼1% [0.8]; NS). Taken together, reliable in vivo biomarkers which predict non-, poor and normal responders for ivacaftor-lumacaftor therapy are urgently required, emphasizing the importance of comprehensive genetic and pharmacogenomic studies in patients using CFTR modulators.
Further patient factors that influence patient outcome in clinical practice ranging from PK/PD parameters to concomitant medications or diets. A decrease in bioavailability or reduction in protein binding from displacement interactions is known to be able to produce decreases in total exposure with no changes in half-life (MacKichan, 1989). In a study looking at human liver microsomes Robertson et al (2015) reported the potential of ivacaftor to inhibit p-glycoprotein (pgp) (Robertson et al., 2015). Inhibition of intestinal pgp is known to decrease systemic bioavailable of drugs (Seden et al., 2010;Dubovsky, 2015). As most CF patients receive pancreatic supplementation with meals, fat malabsorption is also a potential factor that could affect the absorption of lipophilic drugs such as ivacaftor.
Based on the complicated PK/PD and drug-drug interaction issues and our data presented in this study we hypothesise whether a personalized approach combining the identification of CFTR mutations with PK/PD and cytochrome identification studies before initiating CFTR modulators. Personalized medicine approaches and personalized biomarker identification have successfully been applied in other disease models including anaplastic lymphoma kinase biomarker for lung cancer (Cutter and Liu, 2012) or pharmacogenomic approaches in multiple sclerosis (Zhou et al., 2019). Carefully constructed PK/PD models together with biomarker identification and pharmacogenomic profiling for CF are becoming more and more indispensable for various reasons: Firstly, by anticipating how a subset of patients e.g. CFTR mutation and cytochrome status, might react to CFTR modulator treatment, better clinical trials stratifying the patient subset, can be designed to evaluate patient outcomes. Secondly, these models can be used to investigate a number of co-variates including age, weight or gender. An additional difficulty with CF disease is that with more than 2,000 different disease-causing mutations, clinical manifestations vary (pancreatic insufficiency, liver impairment, CF-related diabetes) which can complicate the interpretation of scientific and clinical results (Supplementary Figure 1) (Paulin and Schneider-Futschik, 2020). Factors that can influence patient specific outcomes in CF range from patient specific genetic backgrounds (CFTR mutations, CYP metabolism), inter-patient variables (age, BMI, weight, gender) or others including disease severity and co-morbidities (Supplementary Figure 1). Another unique and non-invasive tool to explore inter-individual pharmacokinetic variability is exhaled breath metabolomics. Nuclear magnetic resonance-based metabolomics of exhaled breath condensate which has successfully been employed to recognize biomarkers of respiratory diseases such as asthma or chronic obstructive pulmonary disease could furthermore contribute to the individual tailoring of treatment for CF (Montuschi et al., 2014;Montuschi et al., 2018). In order to illustrate a personalized medicine approach, clinical and genetic biomarkers need to be identified and PK/PD model that express the dynamic behaviour of the CFTR drug effect need to be developed in patients that are classified as non-/poor and normal responders (Figure 4). In a second step, the relationship between the plasma concentration of a given drug, its corresponding clinical effect and reliable biomarkers need to be identified. When investigating the effect of covariates by stratifying based on current data available taking a personalised approach based on mutation-specific modulator treatment e.g. F508del in combination with determining the cytochrome profile e.g. rapid CYP3A4 metaboliser of each patient is best to be employed. Despite the added cost of these pharmacogenomic tests, unwanted drug-drug interactions due to polypharmacy in the CF population will be minimised. Similarly, the implementation of stem cell and organoid models can pave the way to validate novel surrogate biomarkers of clinical response for CFTR modulators as currently employed in the Netherlands.
Given the genetic and clinical heterogeneity in CF patients developing PK/PD models including comprehensive genetic markers can be utilized as a benchmark to ensure accurate prediction of biomarkers that account for inter-patient variability, higher drug efficacy and minimal adverse effects. To demonstrate the true clinical benefit of genotyping (and not just CFTR mutation identification) for predicting treatment responses for patients under CFTR modulator treatment, it would be necessary to conduct prospective comparisons of different treatments chosen by genotype; which has not been done yet (treatment is chosen based on only the mutation e.g. ivacaftor-lumacaftor for F508del FIGURE 4 | Comparison between current and personalized therapeutic approaches for CF therapy. Frontiers in Pharmacology | www.frontiersin.org August 2021 | Volume 12 | Article 577263 CFTR). Funding sufficiently powered controlled studies of combining genotyping with CFTR mutations and clinical biomarkers is likely to be a challenge.
CONCLUSION
In summary, lumacaftor had variable effects on ivacaftor pharmacokinetics including reducing steady-state concentrations. Significant weight and age effects were observed under standard ivacaftor-lumacaftor therapy. Future studies to include both more patients as well as genotyping/PK/ PD approaches are warranted to determine if lumacaftor also has variable effects on the PK of ivacaftor.
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Monash University Human Research Ethics Committee. 0426. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.
AUTHOR CONTRIBUTIONS
PH, GR, and ES-F were involved in the conception and design; PH, IS-G, MG, MK, CS, JW, HE, MT, TV, GR, and ES-F were involved in the analysis and interpretation of the data and the drafting of the paper or revising it critically for intellectual content; GR and EKS-F are responsible for the final approval of the version to be published. All authors agree to be accountable for all aspects of the work. | 2021-08-02T13:28:29.505Z | 2021-08-02T00:00:00.000 | {
"year": 2021,
"sha1": "66ad9123117b0fdf9a7068e8a0f5385da004b1aa",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fphar.2021.577263/pdf",
"oa_status": "GOLD",
"pdf_src": "Frontier",
"pdf_hash": "66ad9123117b0fdf9a7068e8a0f5385da004b1aa",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
260030936 | pes2o/s2orc | v3-fos-license | Hydrogen Evolution Reaction on Ultra-Smooth Sputtered Nanocrystalline Ni Thin Films in Alkaline Media—From Intrinsic Activity to the Effects of Surface Oxidation
Highly effective yet affordable non-noble metal catalysts are a key component for advances in hydrogen generation via electrolysis. The synthesis of catalytic heterostructures containing established Ni in combination with surface NiO, Ni(OH)2, and NiOOH domains gives rise to a synergistic effect between the surface components and is highly beneficial for water splitting and the hydrogen evolution reaction (HER). Herein, the intrinsic catalytic activity of pure Ni and the effect of partial electrochemical oxidation of ultra-smooth magnetron sputter-deposited Ni surfaces are analyzed by combining electrochemical measurements with transmission electron microscopy, selected area electron diffraction, X-ray photoelectron spectroscopy, and atomic force microscopy. The experimental investigations are supplemented by Density Functional Theory and Kinetic Monte Carlo simulations. Kinetic parameters for the HER are evaluated while surface roughening is carefully monitored during different Ni film treatment and operation stages. Surface oxidation results in the dominant formation of Ni(OH)2, practically negligible surface roughening, and 3–5 times increased HER exchange current densities. Higher levels of surface roughening are observed during prolonged cycling to deep negative potentials, while surface oxidation slows down the HER activity losses compared to as-deposited films. Thus, surface oxidation increases the intrinsic HER activity of nickel and is also a viable strategy to improve catalyst durability.
Introduction
One main aim and indispensable development in our society is decarbonizing energy considering the related climate change. Technological progress already allows the generation of sustainable green energy by wind turbines, photovoltaic plants, and solar power for competitive prices. A challenge, however, is the temporal and local dependency of renewable energy sources. Thus, suitable energy storage is needed to efficiently provide green energy independent of its natural fluctuations [1]. Therefore, hydrogen is considered to play a key role in the energy transition, serving as a green energy vector. However, worldwide hydrogen production relies almost entirely on fossil resources, while sustainably-produced hydrogen through water electrolysis still faces high production costs and limited availability [2]. To compete with well-established fossil energy carriers, the generation of green hydrogen needs to become more effective in terms of energy conversion and connected prices.
One crucial field of improvement is the research of electrocatalytic materials to boost electrolytic water splitting. Due to the sluggish reaction kinetics of both half-cell reactions, decreasing the overpotential of the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) is one crucial parameter in lowering the energy barrier and, therefore, decreasing the overall process energy consumption [3,4].
The efficiency of the cathodic HER in alkaline media is strongly connected to the adsorption and desorption behavior of reaction educts, products, and intermediates. The optimum catalytic efficiency is reached with intermediate binding energies. Weak binding of intermediates leads to inefficient reaction activation, while strongly bound intermediates might block the surface and could poison the catalyst [5,6]. For the HER, the free energy of hydrogen adsorption is an indicator of catalyst activity [5] and could be applied as a descriptor in different electrolytes [7].
The HER in the neutral and alkaline electrolytes is described by the following steps [8]: H ads + H 2 O + e − → H 2 + OH − + * (Heyrovsky step) 2H ads → H 2 + 2* (Tafel step) where (*) stands for the adsorption site on the catalyst surface. The cathodic reaction proceeds either via the Volmer-Tafel ( (1) and (3)) or via the Volmer-Heyrovsky route ( (1) and (2)). The Volmer step (1) describes the dissociation of a water molecule and the formation of an adsorbed hydrogen species, H ads . Additionally, during electrolysis, the charge carrier OH − is formed. The role of OH − adsorption on the catalyst surface is a matter of recent intense studies in determining catalytic activity [9][10][11]. In the Tafel step (3), H ads can recombine with another H ads , forming an H 2 molecule which ideally would leave the surface in the form of gas. Another option for the second reaction step is the Heyrovsky reaction, where H ads combines with a water molecule, generating a hydroxide anion in addition to the hydrogen molecule (2). In contrast to acidic water splitting, alkaline electrolysis suffers from more sluggish reaction kinetics due to an additional water dissociation step, creating an extra energy barrier [12]. Not only is the catalyst-hydrogen (M-H) binding energy shown to be relevant, but water and hydroxyl ion adsorption energies are also of major interest [11,13]. Furthermore, factors like the electrochemical double layer and the potential of zero free charge can influence the overall kinetics [14].
To lower the arising energy barriers, three main approaches can be emphasized in developing novel catalytic materials, including manipulating the electronic structure, the design of specific intermediate adsorption sites, and adjusting the catalyst surface [9]. Factors like improved conductivity and an enlarged active surface area can effectively improve the catalytic efficiency [14]. Further, tailoring supported catalysts acknowledging the active role of the support material is a feasible method [15].
Concerning price reduction, the industry is interested in producing non-noble metal, high surface area structures to limit the use of material and maximize efficiency in upscale applications. However, limited work has been done on fundamental aspects of thin layers and flat surfaces, although this is relevant for benchmarking the material's intrinsic catalytic activity and determining layer thickness influences which ultimately provide essential data for necessary material consumption. Therefore, this work aims to study the intrinsic material properties of an abundant element as Ni, rather than roughness and surface area effects. The latter imposes significant problems in evaluating the intrinsic catalytic activity of Ni surfaces as there is no reliable electrochemical method for determining the electrochemically active Ni surface area.
Concerning availability and economic factors, industrial alkaline electrolysis uses either stainless steel or Ni catalysts, as they present the best non-noble alternatives to expensive, platinum group metal catalysts [16][17][18]. Due to its good activity and stability in alkaline solutions, a fundamental understanding of the HER mechanism on Ni surfaces is a relevant topic of current research from both fundamental and application perspectives. It has already been shown that the presence of transition metal oxide species influences the catalytic behavior of pure metal catalysts like Pt and Ni [19][20][21]. Metal|metal oxide heterostructures were shown to possibly have a positive influence on the shift of HER overpotentials to energetically more favorable ones [22]. In the case of Ni, the presence of oxide species on the surface is considered a possible factor in tailoring surface activities and creating potential reduction sites. The interplay of Ni and its oxide species (NiO, α-/β Ni(OH) 2 , NiOOH) has been extensively investigated before, showing that nickel oxide species positively influence the activity and reaction overpotential via tailoring the adsorption behavior of intermediate species [21,23,24].
However, bulk NiO and Ni(OH) 2 show poor electrocatalytic behavior, mainly attributed to the unfavorably high hydrogen adsorption energy [23]. Therefore, a synergistic effect between Ni and NiO/Ni(OH) 2 induces superior catalytic activities. For this effect, surface Ni 0 provides optimized conditions for hydrogen adsorption, while the formation of OH ads should be promoted by surface oxide species (NiO) [11,25]. It should also be highlighted that the release of reaction products is facilitated by oxide species on the surface, thus preventing the occupation of the surface and its blocking [24]. However, a reliable estimation of real activity increases by the formation of NiO|Ni interfaces also suffers from the abovementioned obstacles related to real (electrochemically active) surface area measurements.
This work concentrates on the preparation and evaluation of the HER activity of ultra-smooth polycrystalline Ni thin films and the effect of electrochemical oxidation on the catalytic activity of its modified surfaces towards HER in alkaline media. In this work, we aim to evaluate the intrinsic catalytic properties of Ni and the effect of in-situ electrochemical oxidation on the catalytic behavior simultaneously, while precisely monitoring the evolution of the surface state and the surface roughness of the deposited Ni films. The present experimental work is supplemented by Density Functional Theory (DFT) calculations and Kinetic Monte Carlo (KMC) simulations to obtain a deep atomic-level understanding of the HER on oxide-modified Ni surfaces and the role of the rates of different elementary processes involved in the HER mechanism. Finally, we analyze the impact of prolonged Ni film cycling to deep cathodic potentials while monitoring the evolution of surface topology and local crystal structure.
Preparation of Ni Thin Layers
Low roughness Ni thin film electrodes were produced by DC magnetron sputtering. The metal layer was deposited on various substrates, specifically on Cu substrate with enhanced adhesion, as well as on silicon wafers for layer thickness evaluation. Before deposition, the Cu foil was mechanically cleaned and polished to minimize the original roughness. Subsequently, the substrate was washed thoroughly with deionized water and isopropanol. The substrates were etched in H 2 SO 4 (v/v 10%) for enhanced adhesion and rinsed with isopropanol. The cleaned substrates were used immediately to prevent the formation of an oxidation layer, which negatively influences the adhesion of the sputterdeposited layers.
Magnetron sputter deposition was performed with a HEX series Compact Thin Film Deposition System by Korvus Technology equipped with a Fission DC Sputtering Source. Due to the magnetic properties of Ni, the strong magnetron available from Korvus Technology was required for deposition. The Ni sputtering target with a purity of 99.9%, a diameter of 50.8 mm, and a thickness of 1 mm was obtained from HMW Hauner GmbH & Co. KG (Röttenbach, Germany). The targets were pre-sputtered for three minutes before Nanomaterials 2023, 13, 2085 4 of 18 every deposition. The deposition power was chosen to be 80 W with an argon flow of 20 sccm and a total constant pressure of 1.4 × 10 −3 bar. The sample stage rotation was set to 20 rpm for homogenous deposition on the substrates with a fixed sample stage to target distance of 12 cm. The deposition time was selected to produce Ni samples of about 50 and 300 nm thickness, yielding actual layer thicknesses of 46 ± 8 nm and 287 ± 4 nm. For simplification, the layers will be referred to as 50 and 300 nm films, respectively.
Materials Characterization
Layer thickness, deposition rates, and estimation of the roughness of the electrodes were obtained by laser scanning confocal microscopy with an Olympus LEXT OLS4000 3D laser microscope (Tokyo, Japan). Determination of the thickness was performed for Ni layers on Si wafers, enabling better contrast in the specimens.
Transmission electron microscopy (TEM) imaging was done on a Philips CM 12 TEM with an acceleration voltage of 120 kV. The TEM samples were directly sputter deposited on electropolished Cu support.
Topographic images of the thin films were recorded with a Bruker Dimension Icon atomic force microscope (AFM). The AFM was operated in intermittent contact mode with a TESPA-V2 etched silicon probe (Bruker) with a typical spring constant of 37 N m −1 , a resonance frequency of 320 kHz, and a tip radius of 7 nm. Image processing and evaluation were executed using Gwyddion 2.61 [26] data analysis software. All presented images were processed using step-line correction. Before any roughness evaluation, images underwent flattening to eliminate curvature originating from sample mounting.
X-ray photoelectron spectroscopy was conducted with a Thermo Fisher Scientific Inc. Nexsa G2 photoelectron spectrometer system equipped with a low-power Al K-Alpha X-ray source yielding a 30-400 µm adjustable X-ray spot size. Survey spectra were recorded for binding energies up to 1350 eV for each sample. Additionally, Ni 2p (840-890 eV) spectra and O 1s (525-545 eV) spectra were recorded.
Electrochemical Measurements
Electrochemical measurements were conducted with a PARSTAT4000A potentiostat using VersaStudio 2.63.3 software and a three-electrode setup. The working electrodes were the produced sputter-deposited Ni thin films on the Cu substrate. The electrodes were contacted and electrically insulated to avoid Cu contribution during the analysis. A Pt coil served as the counter electrode. We note that Pt counter electrode dissolution followed by Pt re-deposition on the working electrode was not observed using XPS. All measurements were conducted with an ALS Co., Ltd. (Tokyo Japan) Hg/HgO reference electrode for alkaline solutions and were performed in 1 M KOH at room temperature. The employed electrolyte was prepared from KOH pellets (99.99% purity; Sigma-Aldrich, Darmstadt, Germany) dissolved in deionized water. The iR drop was compensated in all cyclic voltammetry (CV) measurements up to 75% of the electrolyte resistance, which was determined via single-point impedance measurements (E = −0.8 V vs. Hg/HgO, 100,000 Hz).
To analyze the catalytic activity towards the HER and monitor the effect of electrochemical oxidation of the samples, the working electrode was first used as a cathode examining the HER by cyclic voltammetry in the region from −0.6 V to −1.3 V vs. Hg/HgO. For electrochemical oxidation of the films, the samples were cycled in the anodic regime from −0.4 V to 0.7 V vs. Hg/HgO. The electrode was cycled from −0.6 V to −1.3 V vs. Hg/HgO after the electrochemical oxidation process to monitor the influence of oxidation on the catalyst activity. A sweep rate of 5 mV s −1 was chosen for all cyclic voltammetry experiments. For further comparison of acquired data and calculations, the back loop of the third cycle was selected. For reproducibility, all experiments were performed at least three times. Comparative data were collected from the back-cycling curve at −10 mA cm −2 . As benchmark catalysts, electrodeposited Ni and Pt layers from McCrory et al. were chosen [27].
To test the durability of the sputter-deposited Ni thin films during HER, cyclic voltammetry in the range of −0.6 V to −1.3 V vs. Hg/HgO and a sweep rate of 5 mV s −1 was performed. In total, 100 cycles were recorded to monitor the evolution of the catalyst activity and analyze the influence of the surface oxide. The cell was purged with N 2 continuously during the measurements.
The potentials provided in this work were converted from Hg/HgO to the Reversible Hydrogen Electrode (RHE) by the following equation:
Computational Modeling
First-principles Density Functional Theory (DFT) calculations were performed using the Vienna ab initio simulation code (VASP) [28][29][30]. The Generalized Gradient Approximation (GGA) in the parametrization by Perdew, Burk, and Ernzerhof [31] combined with the projector augmented wave (PAW) method was used [32]. Cut-off energy of 600 eV and Gaussian smearing with a width of σ = 0.025 eV for the occupation of the electronic levels were used. Spin-polarization was included in all the calculations. The first irreducible Brillouin zone was obtained using a Γ-centered 4 × 4 × 1 grid utilizing the general Monkhorst-Pack scheme [33].
The Ni(111) surface was modeled using a (4 × 4) supercell with 5 Ni layers. During structural relaxation, two bottom layers were kept fixed in their bulk positions, while other layers were allowed to fully relax until the forces acting on the atoms dropped below 0.01 eV Å −1 . To investigate the interface between metallic nickel and oxidized nickel phase as catalytically active sites, a (NiO) 2 fragment was placed on the surface of the Ni(111) slab. To quantify the interactions between adsorbates (OH and H) with (modified) Ni(111) surface, the binding energies (E b ) were calculated as: where E surf+A , E surf , and E A are the total energies of the surface with adsorbate, the total energy of the surface, and the total energy of an isolated adsorbate, respectively. KMCLib v1.1 [34] was used for all Kinetic Monte Carlo (KMC) simulations. All elementary processes (rare events) with their assumed reaction rates, along with the initial configuration of the system, defined on a regular grid in space, were provided to the code. The systems were then propagated in time on the free energy landscape defined by the provided elementary processes and rates. Simulations were provided enough time to equilibrate, and the statistics were then collected, including integral H 2 production rates and spatial maps of H 2 production (normalized from 0 to 1). The initial configurations of the systems were defined on a 130 × 130-points grid which was first populated with metallic sites (Ni), and then circular islands of oxidized Ni phase (OX) were placed on the surface. The set of elementary processes involved reactions (1-3) and considered possible water dissociation at the Ni|OX interface in such a way that H ads is formed on Ni sites, while OH − does not remain adsorbed at OX (much faster removal of OH − compared to other rates in the system). Thus, the reaction is analogous to the Volmer step but only occurs at the Ni|OX interface. Water dissociation at the Ni|OX interface and clean parts of the Ni surface were considered independent and mutually affected indirectly only by the amount of H ads at Ni sites. In addition, surface diffusion of H ads at non-modified parts of the Ni surface was also included in the model. As explained in previous works [15,35], the rates of the considered elementary processes were 10 n (with n being an integer number), and were then systematically varied to analyze the effects of the rates of the elementary processes on the total H 2 production rate.
Properties of the Deposited Ni Films
Ni thin layers were fabricated via magnetron sputter deposition on a conductive and polished Cu substrate. The deposition time was chosen to produce layers of 50 nm and 300 nm, allowing us to analyze the influence of film thickness on the catalytic behavior. Using laser confocal microscopy (LEXT), real obtained thin film thicknesses were confirmed. Subsequent calculations revealed a deposition rate of 11.6 nm min −1 . Freshly produced films were further characterized via TEM imaging (Figure 1(a1)). The TEM bright-field image shows a contrast sensitive to the grain orientations and therefore reveals that the synthesized Ni thin films are nanocrystalline with an estimated mean grain size of 9 nm. This small grain size and, therefore, high density of grain boundaries (GBs) has already been found to potentially have a positive influence on the catalytic activity towards HER [36,37]. It was shown in a combined DFT and experimental study by Jiang et al. that the GBs efficiently affect the activity of Au catalysts. As a consequence, it is concluded that a high density of GBs can be traced to an increased ability to adsorb hydrogen and stabilize H* during the reaction steps [37]. The accompanying selected area electron diffraction (SAED) pattern and the corresponding profile obtained by azimuthal integration [38] confirm the polycrystalline fcc structure of the sputter-deposited film (Figure 1a). In addition to examining microstructure and crystal structure, the surface topography was also investigated. AFM of the as-deposited films revealed a homogenous surface coverage of a dense Ni layer on the Cu substrate without evidence of periodic pore structures, which would systematically increase the real surface area. As the produced Ni films were tuned in a way to exclude the roughness factor (RF) effect in the evaluation of Ni activity, sputter deposition and substrate preparation were optimized for obtaining low-roughness samples. The root mean square roughness (RMS) of the Ni films was determined to be around 2 nm using AFM (Figure 1b,c), confirming a successful optimization of the thin film properties. The residual roughness mainly originates from the substrate and its preparation technique rather than the deposition settings. Thus, produced Ni layers exhibit roughness factors (RF) practically equal to 1 (Table 1), allowing a precise determination of kinetic HER parameters without the uncertainties related to electrochemical active surface area determination. the rates of the considered elementary processes were 10 n (with n being an integer number), and were then systematically varied to analyze the effects of the rates of the elementary processes on the total H2 production rate.
Properties of the Deposited Ni Films
Ni thin layers were fabricated via magnetron sputter deposition on a conductive and polished Cu substrate. The deposition time was chosen to produce layers of 50 nm and 300 nm, allowing us to analyze the influence of film thickness on the catalytic behavior. Using laser confocal microscopy (LEXT), real obtained thin film thicknesses were confirmed. Subsequent calculations revealed a deposition rate of 11.6 nm min −1 . Freshly produced films were further characterized via TEM imaging (Figure 1(a1)). The TEM brightfield image shows a contrast sensitive to the grain orientations and therefore reveals that the synthesized Ni thin films are nanocrystalline with an estimated mean grain size of ~9 nm. This small grain size and, therefore, high density of grain boundaries (GBs) has already been found to potentially have a positive influence on the catalytic activity towards HER [36,37]. It was shown in a combined DFT and experimental study by Jiang et al. that the GBs efficiently affect the activity of Au catalysts. As a consequence, it is concluded that a high density of GBs can be traced to an increased ability to adsorb hydrogen and stabilize H* during the reaction steps [37]. The accompanying selected area electron diffraction (SAED) pattern and the corresponding profile obtained by azimuthal integration [38] confirm the polycrystalline fcc structure of the sputter-deposited film (Figure 1a). In addition to examining microstructure and crystal structure, the surface topography was also investigated. AFM of the as-deposited films revealed a homogenous surface coverage of a dense Ni layer on the Cu substrate without evidence of periodic pore structures, which would systematically increase the real surface area. As the produced Ni films were tuned in a way to exclude the roughness factor (RF) effect in the evaluation of Ni activity, sputter deposition and substrate preparation were optimized for obtaining low-roughness samples. The root mean square roughness (RMS) of the Ni films was determined to be around 2 nm using AFM (Figure 1b,c), confirming a successful optimization of the thin film properties. The residual roughness mainly originates from the substrate and its preparation technique rather than the deposition settings. Thus, produced Ni layers exhibit roughness factors (RF) practically equal to 1 (Table 1), allowing a precise determination of kinetic HER parameters without the uncertainties related to electrochemical active surface area determination. Overpotentials at −10 mA cm −2 η 10 were evaluated and compared to benchmark catalysts. Tafel slopes (b) and exchange current densities (j 0 ) were extracted using Tafel analysis. Roughness factors were obtained using AFM analysis as the ratio between the real surface obtained by AFM and the geometric surface area of scanned spots. In ref. [27], RFs were obtained using the double-layer charging method.
HER Catalysis by the Deposited Ni Films
Following extensive characterization of the initial state, low-roughness Ni thin films were analyzed according to their electrocatalytic behavior towards the HER. First, the HER activity of as-deposited films was determined. Next, the electrode surface was intentionally oxidized by polarizing the working electrode in the anodic regime. This step was performed immediately after the HER analysis without removing the electrodes from the electrolyte between subsequent steps. This in-situ oxidation treatment was followed by another analysis of the HER activity following the same procedure as described for the first step.
The reason for the application of this exact method is the anodic oxidation of the pure Ni catalysts, as it has been reported that Ni/Ni(OH) 2 and Ni/NiO catalysts develop high efficiencies towards the HER [24,[39][40][41][42]. The formation of the surface oxides starts with the formation of α-Ni(OH) 2 at low anodic potentials. This formation is reversible between 0.30-0.50 V vs. RHE, oxidizing and reducing Ni to Ni(II) and forming α-Ni(OH) 2 and NiO [43][44][45]. In the region of 0.7-1.2 V vs. RHE, α-Ni(OH) 2 is irreversibly transformed to β-Ni(OH) 2 [46]. This process is featureless in the CV of the Ni electrode. Moreover β-Ni(OH) 2 can be derived by direct oxidation of Ni [44]. Alsabet et al. concluded from XPS analysis that the formed oxide layers consist predominantly of β-Ni(OH) 2 and NiO, while the hydroxide presents the outermost layer, and NiO is considered an interlayer between Ni and β-Ni(OH) 2 . With increasing potential, the oxide layer is prone to growth [44]. Above 1.35 V vs. RHE, the reversible formation and reduction of β-NiOOH take place [47]. All those processes considered, the expected oxygen species residing on the surface after the cyclic oxidation and reduction in the anodic regime are assumed to be β-Ni(OH) 2 and, possibly, underlying NiO. Both of the named species can potentially influence the catalytic activity of Ni catalysts. The analysis of the as-deposited thin films in the cathodic regime (Figure 2a) revealed the overpotentials at a current density of −10 mA cm −2 (η 10 ) to be −(0.263 ± 0.003) V for the 50 nm Ni thin film. The 300 nm Ni film developed an overpotential of -(0.256 ± 0.005) V, slightly lower than the overpotential of the 50 nm sample but within the experimental uncertainty (Table 1). After anodic oxidation of the Ni films, as expected, a shift of overpotentials was noticed, yielding −(0.220 ± 0.004) V for the 50 nm Ni films and −(0.215 ± 0.006) V for the 300 nm films. Compared with the initial overpotentials, a clear enhancement of the catalytic activity towards the HER can be observed (Figure 2a). cathodes [48][49][50][51]. These findings could indicate that increasing roughness factors can improve the catalytic activity to a certain point, but pushing boundaries for HER activity requires additional strategies to amplify the intrinsic catalytic activity to approach or even outperform noble Pt catalysts. One possibility to do so is surface oxidation, as shown herein. Despite HER activity being significantly improved upon oxidation, it is still quite far from the Pt benchmark (Table 1). The obtained HER polarization curves were further processed using Tafel analysis. The linear sections of the Tafel plots ( Figure 2b) were extrapolated to η = 0 V to extract exchange current densities (j 0 ). The summarized HER kinetic data (Table 2) indicate a minor effect of surface oxidation on the Tafel slopes, while the exchange current densities are increased by a factor of 3.2 for the 50 nm film and by 4.8 for the 300 nm film. For the HER on Ni metals of various morphologies, there are reports of Volmer-Tafel and Volmer-Heyrovsky reaction mechanisms, as well as mixed forms [48]. However, at higher overpotentials, where Tafel analysis was performed, the Heyrovsky reaction path is considered the ratedetermining step in a successive combination of the Volmer and Heyrovsky mechanism. In contrast, the Tafel step is considered negligible in this potential range [49]. AFM analysis (Figures 2c,d and S3) of the as-deposited thin films showed comparable surface roughness and was only slightly affected by surface oxidation. This finding indicates that the RFs (Table 1) hardly changed upon anodic treatment and remained effectively equal to 1. Thus, the calculated exchange current densities are considered to correspond to those evaluated for the electrochemically active surface area. Moreover, AFM did not indicate any difference in the porosity comparing dense sputter-deposited Ni films before and after oxidation. We believe that the presented exchange current densities are qualified to be considered reliable values for future reference due to the minimization of uncertainties in real surface estimation by careful experimental design. Interestingly, the obtained η 10 goes along with established data presented by McCrory et al. who set the benchmark for Ni catalysts for the HER to η 10 = −0.26 V, despite featuring a much higher RF [27], as well as with some other high surface area Ni-based cathodes [48][49][50][51]. These findings could indicate that increasing roughness factors can improve the catalytic activity to a certain point, but pushing boundaries for HER activity requires additional strategies to amplify the intrinsic catalytic activity to approach or even outperform noble Pt catalysts. One possibility to do so is surface oxidation, as shown herein. Despite HER activity being significantly improved upon oxidation, it is still quite far from the Pt benchmark (Table 1).
Surface Chemical Composition
For the boosted catalytic activity, two main surface properties are assumed to have a decisive influence. Both morphology of the surface and the chemical composition do effectively alter the HER mechanism [48][49][50][51]. Primarily, through the oxidation treatment, roughening of the surface could occur and might impact the overall catalytic activity. However, measured RMS and RFs of the samples were found to be practically unaffected by surface oxidation. Thus, increased roughness enhancing the catalytic activity can be evidently excluded, leaving the surface chemical composition the relevant tuning factor to be monitored closely. XPS measurements were conducted to detect the available surface species on the as-deposited Ni thin films and after the in-situ oxidation process, allowing a better understanding of the origin of the increased HER activity (Figure 3).
The evaluation of XPS spectra was done via curve fitting using Ni 0 , NiO, and Ni(OH) 2 specific fitting envelopes. The sharp peak at a binding energy (BE) of 852.7 eV can be assigned to Ni 0 in all recorded curves [52][53][54]. For both samples, it decreases strongly in relative height upon oxidation.
The peaks occurring at BE 854.4 eV and BE 856.2 eV originate from Ni 2+ , and the peak at the lower BE can be assigned to NiO, while the latter partly originates from Ni in Ni(OH) 2 [44,55]. The satellite peaks at BE 861.4 eV support the assignment and availability of Ni(OH) 2 on the surface [52][53][54][55], while the satellite at BE 858.4 eV is attributed to Ni 0 [53]. Table 2 summarizes the Ni surface species found on the studied catalytic films and their relative contributions.
Analyzing the XPS spectra of the as-deposited Ni thin films reveals that the Ni 0 peak is the dominant species displayed in the Ni 2p spectra. However, peaks for Ni 2+ are also detected, indicating the presence of a native oxide layer on the sputter-deposited Ni surface. Upon intended electrochemical oxidation of the surface, a decrease in the relative intensity of Ni 0 can be monitored. Therefore, Ni(OH) 2 presents the most prominent species upon oxidation and further thin film electrochemical processing. However, the Ni 0 peak is still detectable after oxidation of the surface, indicating that the Ni(OH) 2 and NiO layers might not form a continuous thin film with complete surface coverage. The coexistence of those three species on the catalyst surface seems to contribute simultaneously to lowering the HER overpotential. It is already well established that the presence of elements and structures in addition to pure Ni can lead to a synergistic effect, which can tailor catalytic sites and lead to higher catalytic efficiency [23]. This interplay of Ni and Ni(OH) 2 has been shown by Danilovic et al., who stated the synergistic effects on bifunctional metal/metal hydroxide surfaces [40]. Also, the less prominent NiO is known to be capable of producing a comparable effect [25,41,42,50,56]. Here, we see ( Table 2) that both Ni films have very similar surface composition upon surface oxidation, varying by less than 2 at.%. However, the 300 nm-thick as-deposited films have significantly higher NiO content than the 50 nm-thick counterparts. This finding could explain the slightly higher HER activity of the as-deposited 300 nm Ni films (Figure 2a), which is also visible in the evaluation of survey spectra ( Figure S1). The higher NiO content might result from prolonged deposition time, increasing the probability of contamination by oxygen species. However, the pronounced formation of Ni(OH) 2 in both cases can be associated assuredly with increased HER activity. Additionally, the theory of Ni(OH) 2 formation is supported by collected O 1 s spectra ( Figure S2). The two arising peaks are attributed to NiO (529.6 eV) and NiO in combination with Ni(OH) 2 (531.6 eV), while the asymmetry of the latter is caused by adsorbents and defective sites [52]. The strong decrease of the pure NiO peak at lower energies and the simultaneous relative growth of the mixed oxide and hydroxide peaks indicate Ni(OH) 2 formation on the surface. decisive influence. Both morphology of the surface and the chemical composition do effectively alter the HER mechanism [48][49][50][51]. Primarily, through the oxidation treatment, roughening of the surface could occur and might impact the overall catalytic activity. However, measured RMS and RFs of the samples were found to be practically unaffected by surface oxidation. Thus, increased roughness enhancing the catalytic activity can be evidently excluded, leaving the surface chemical composition the relevant tuning factor to be monitored closely. XPS measurements were conducted to detect the available surface species on the as-deposited Ni thin films and after the in-situ oxidation process, allowing a better understanding of the origin of the increased HER activity (Figure 3).
The evaluation of XPS spectra was done via curve fitting using Ni 0 , NiO, and Ni(OH)2 specific fitting envelopes. The sharp peak at a binding energy (BE) of 852.7 eV can be assigned to Ni 0 in all recorded curves [52][53][54]. For both samples, it decreases strongly in relative height upon oxidation. Although the Ni films were carefully produced to have RF values close to 1 and to eliminate other possible factors that make the understanding of the HER on oxidized Ni surfaces difficult, it is still elusive what the key factors in boosting HER activity are and what is their mutual interplay. Moreover, XPS is a surface-sensitive technique, but it is not limited to the uppermost atomic layer only. It is, thus, challenging to state with certainty the exact fractions of Ni species observed by XPS at the Ni|electrolyte interface.
DFT Modeling and KMC Simulations
Surface oxides and hydroxides can increase the HER reaction rate by boosting the Volmer reaction step via increasing the dissociation of water and adsorption of hydrogen species on the catalyst surface [40,50]. It is believed that in the given surface scenario, water adsorption will occur at Ni species via the interaction of O with Ni oxide species and H with pure Ni on the surface. After the dissociation of the water molecule, Ni facilitates the adsorption and recombination of hydrogen [56]. Moreover, the multifunctional catalyst oxide domains might mediate OH − desorption from the catalyst surface [21,56,57].
To better understand the effect of the Ni surface modification by oxygen-containing species, we performed a series of DFT calculations on Ni(111) and (NiO) 2 @Ni(111) surfaces, comparing how H and OH interact with these surfaces. We found that H ads prefers Ni(111) terraces on both surfaces and that the H ads binding energy is unaffected by the presence of oxide island (NiO) 2 (Figure 4a). OH ads was found to prefer a bridge position between partially oxidized Ni atoms from (NiO) 2 islands and Ni atoms from the Ni(111) surface (Figure 4b). The interaction of OH ads with the (NiO) 2 -modified surface is significantly stronger compared with clean Ni(111) (−4.01 eV vs. −3.25 eV). The linearity of the relation between the activation energy barrier for water dissociation and the enthalpy change of water dissociation [58] leads to an acceleration of the Volmer step (1) on an oxide-modified surface compared to clean Ni(111) surfaces. Importantly, our findings suggest that the effect is solely due to the enhanced interaction of OH ads with the oxidized phase, while the impact of H ads energetics is negligible. Under electrochemical conditions, OH − is formed upon dissociation (1). It is unstable at the negative potentials applied for the HER and is, therefore, desorbed off the surface. Considering the formation of the subsurface hydride on clean and modified Ni(111), we found that H sub (Figure 4c) is energetically more favorable for a modified surface. (d1-d3) Spatial maps of H2 production when the OX|Ni interface is not active for H2O dissociation with increasing Hads diffusion rate, (e1-e3) spatial maps of H2 production when the OX|Ni interface is active for H2O dissociation with increasing Hads diffusion rate, (f1-f5) spatial maps of H2 production, when the OX|Ni interface is active for H2O dissociation and Hads diffusion is fast with increasing coverage of Ni by the OX phase. The numbers above the maps (d-f) give integral H2 production rates (in a.u.).
HER Activity upon Prolonged Cycling-The Crucial Role of Surface Oxidation
In addition to the initial activity of a catalyst, its stability is also a crucial factor to be closely monitored. The stability of the Ni thin films was tested via CV in the cathodic regime. The electrodes were cycled to deep negative potentials, typically reaching ~ −50 mA cm −2 . The electrodes were switched 100 times between intense HER and potentials corresponding to the formation of α-Ni(OH)2 and NiO. Considering the potential scan rate, (d1-d3) Spatial maps of H 2 production when the OX|Ni interface is not active for H 2 O dissociation with increasing H ads diffusion rate, (e1-e3) spatial maps of H 2 production when the OX|Ni interface is active for H 2 O dissociation with increasing H ads diffusion rate, (f1-f5) spatial maps of H 2 production, when the OX|Ni interface is active for H 2 O dissociation and H ads diffusion is fast with increasing coverage of Ni by the OX phase. The numbers above the maps (d-f) give integral H 2 production rates (in a.u.). Moreover, the literature suggests that the availability of surface oxides mediates the amount of H ads and OH ads and promotes H ads recombination via the Tafel reaction (3) [21,56]. To investigate this possibility, we performed a set of KMC simulations for the Volmer-Tafel mechanism of H 2 evolution (Figure 4d-f). In the first scenario, the surface modifier (OX phase) does not contribute to the water dissociation. When the rate of H ads surface diffusion increases, the total H 2 production rate decreases (Figure 4(d1-d3)). This is due to the high mobility of H ads , reducing the probability of H ads recombination. In the second set, the surface modifier (OX phase) allows for fast H 2 O dissociation at the OX|Ni interface. Increasing H ads mobility in this scenario boosts the total H 2 production (Figure 4(e1-e3)), as H ads is allowed to diffuse away from its formation site. Empty OX|Ni interfacial sites allow the creation of more H ads and increase H ads concentration on the surface. This raises the probability of the H ads recombination tremendously. In this case, H 2 is not only produced at the OX|Ni interface but all over the Ni surface (Figure 4(e2) vs. Figure 4(e3)). It can, therefore, be concluded that if the presence of a surface modifier increases the mobility of H ads , both the H 2 O dissociation rate as well as the H 2 production will increase significantly. For comparison (Figure 4(di) vs. Figure 4(ei) (i = 1, 2, or 3)), the rates of all elementary processes were chosen equally (including H ads diffusion rates), except for the H 2 O dissociation at the OX|Ni interface, which was assumed 10,000 times faster in the second scenario. Dissociation acceleration gives H 2 production enhancement factors close to those experimentally observed by comparing j 0 before and after oxidation of the Ni films. In a third scenario, we investigated how the surface coverage by a modifier (OX phase) affects the H 2 production rate. We note that in our simulations, H 2 cannot be produced at the OX phase surface. Therefore, the modifier's only contribution is the mediation of H 2 O dissociation at the OX|Ni interface. In this set of simulations, H ads diffusion was considered as fast as in previous simulations. It was found that an increase in the OX surface fraction leads to an increase in the H 2 production rate. This accelerating effect passes through a maximum governed by Ni surface coverage, unfolding in the finding that an optimal OX phase coverage provides a maximum H 2 production rate. The existence of this maximum is the result of an interplay between the length of the OX|M interface at which H 2 O dissociates and the number of Ni sites at which H ads is formed and can recombine to H 2 . In practice, this means that the concentration of oxidized Ni phase on Ni catalysts can/should be carefully optimized to provide a maximum enhancement of HER activity.
HER Activity upon Prolonged Cycling-The Crucial Role of Surface Oxidation
In addition to the initial activity of a catalyst, its stability is also a crucial factor to be closely monitored. The stability of the Ni thin films was tested via CV in the cathodic regime. The electrodes were cycled to deep negative potentials, typically reaching −50 mA cm −2 . The electrodes were switched 100 times between intense HER and potentials corresponding to the formation of α-Ni(OH) 2 and NiO. Considering the potential scan rate, each experiment took about 8 h. The results (Figure 5a) show a continuous overpotential increase upon cycling. However, the observed activity loss is more pronounced for as-deposited films in comparison to the oxidized films. Even after 50 cycles, the overpotentials of oxidized samples still outperform the corresponding as-deposited films (Figure 5a). AFM analysis shows more prominent topography changes after 100 cycles and confirms an increasing RMS roughness to 4.1 nm (50 nm films) and 12 nm for 300 nm films (Figure 5b,c). Surface roughening occurs after the oxidation step and after 100 cycles, each reaching intense H 2 evolution. Still, the roughness factors remain rather low. For 50 nm films, RF is found to be 1.2, and for 300 nm films, RF is 1.6. Higher RFs for 300 nm Ni films align with higher HER activities. The surface oxidation associated with stable Ni(OH) 2 formation (Table 2) is crucial to maintain HER activity. We have not observed any substantial changes in the surface topology. However, we note that samples were not exposed to very high currents (up to −50 mA cm −2 ) during the testing. The possibility of more pronounced structural changes at industry-relevant currents cannot be excluded at this point and is beyond the scope of the present work. Despite the overall activity decrease trend during cycling, the oxidized thin films preserve superior activity compared to their as-deposited counterparts. For as-deposited films, after 100 cycles, η 10 is shifted by −65 mV and −71 mV for 50 nm and 300 nm-thick films. For oxidized films, η 10 shifts less, specifically by −41 mV (50 nm films) and −34 mV (300 nm films). Moreover, the state of the Ni surface evolves under HER conditions (Table 3), surprisingly increasing the total NiO and Ni(OH) 2 content at the expense of Ni 0 . The evolution of the chemical composition depends on the film thickness, even though the surface chemical composition of both 50 nm and 300 nm films was very similar after the oxidation ( Table 2). At the same time, the surface phases present under electrochemical conditions could be different from those observed in ex-situ experiments such as XPS or SAED. We believe that cycling leads to mechanical stresses within the Ni layer upon oscillating lattice expansion and subsequent contraction, which is most likely responsible for surface roughening. This emphasizes the possibility of either hydrogen incorporation into interstitial sites or even the formation of a NiH phase in the catalyst. The hydrogen uptake could, therefore, lead to oscillating lattice deformation and potentially causes damage in the sputtered layers promoting delamination. That mode of catalyst deactivation via NiH formation has already been described in several studies and can be attributed to a change in electronic surface structure [59,60]. Thus, the activity loss can be linked to the uptake and incorporation of hydrogen into Ni, forming hydrides on the metal surface and within the thin film [45,61], despite the fact that no direct observation of hydrides was possible in this work. Influencing the density of states on the surface, NiH is considered to increase the overpotentials and form a barrier for hydrogen diffusion within the catalyst [45,[59][60][61]. However, we note that the DFT calculations suggest that subsurface hydride is more stable in modified Ni(111) (Figure 4c) compared to clean Ni(111) by 0.09 eV, but this result relates to thermodynamic stability only. Thus, the question remains whether the oxidized Ni phases might induce an additional barrier for H ads diffusion to subsurface sites, which could explain slower activity loss for the oxidized films. Using TEM and SAED analysis, no residual NiH after the cycling stability tests was observed. On the other hand, structural inhomogeneities arise in the films, which present themselves as NiO and Ni(OH) 2 -rich islands on the surface. This inhomogeneous coverage with oxide and hydroxide species supports the XPS results, indicating the coexistence of Ni 0 and its oxide species on the surface. The detection of in-situ-formed NiH might be challenging, especially considering the presence of oxides and hydroxides ( Figure 6). This issue is yet to be addressed, and it is particularly challenging to resolve it as the as-deposited Ni films are also clearly partially oxidized (Table 2). The profile of the as-deposited layer and theoretical reference peaks are displayed for comparison [62].
Conclusions
At the same time, the surface phases present under electrochemical conditions could be different from those observed in ex-situ experiments such as XPS or SAED.
Conclusions
Ni thin films of different thicknesses were prepared via magnetron sputter deposition to minimize roughness and properly evaluate the intrinsic catalytic activity of Ni towards HER in alkaline solutions. TEM investigations revealed a Ni fcc nanocrystalline structure of the films with a grain size of about 9 nm. AFM images confirmed the deposition of smooth films with extremely low roughness factors close to 1. HER activities of 50 nm and 300 nm-thick Ni films are comparable. To study the premise of enhanced HER due to the presence of oxide phases, the surface chemical composition was tuned via electrochemical oxidation performed by cyclic voltammetry in the anodic regime to study the premise of enhanced HER due to the presence of oxide phases. XPS studies confirmed a change of the exposed surface species and, therefore, a successful alteration of Ni surface-active species upon oxidation, resulting in a high surface content of Ni(OH) 2 , irrespective of film thickness. On the other hand, negligible surface roughening is seen, allowing for precise evaluation of exchange current densities for as-deposited and oxidized films. The applied oxidation process shifts the onset of hydrogen evolution towards lower overpotentials and increases the exchange current densities 3-5 times. DFT calculations correlated the boost of HER activity with enhanced OH adsorption at the oxide|nickel interface, while no impact of oxide species on the H ads energetics was observed. Additionally, KMC simulations proved that speeding up H 2 O dissociation at the oxide|nickel interface enhances the H 2 production rate, even if the rates of other processes in the overall mechanism remain unchanged. The H ads surface diffusion rate on the nickel surface also plays a significant role. The maximum H 2 production rate can be reached when the number of oxide|nickel sites for H 2 O dissociation is appropriately balanced with the H ads surface diffusion rate. The results imply that surface oxidation is crucial for retaining high HER activity under a prolonged operation of a nickel catalyst. During this process, the surface roughness increases while structural inhomogeneities arise in the deposited Ni films, which can be associated with Ni hydride formation under HER conditions. The presented results contribute to a better understanding of the HER on nickel in alkaline media and give additional guidelines for boosting the HER performance of Ni-based cathodes.
Data Availability Statement:
The data presented in this study are available on request from the corresponding author. | 2023-07-22T15:48:18.962Z | 2023-07-01T00:00:00.000 | {
"year": 2023,
"sha1": "a9381712905201a89a9079d415568ac661135670",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2079-4991/13/14/2085/pdf?version=1689573845",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "030a61cb41d121d5b02b09f168b632c8e6b3f49c",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": []
} |
10851207 | pes2o/s2orc | v3-fos-license | Laparoscopic cholecystectomy in the elderly
Background
Few studies have examined the results of laparoscopic cholecystectomy (LC) in the elderly. We reviewed our experience with the procedure in 194 patients age 65 and older.
Methods
We conducted a retrospective study evaluating the medical records of 40 consecutive patients age 70 or older who underwent laparoscopic cholecystectomy at the Department of Surgical Sciences, Organ Transplantation and Advanced Technologies, Oncology Surgery Unit, University of Catania, Italy. Data included age and gender, American Society of Anaesthesiologists (ASA) score, comorbid illness, prior abdominal surgery, presentation, operative time, conversion rate and reasons for conversion, postoperative morbidity and mortality rates, pathologic diagnosis, and length of hospital stay. Patients were classified as having complicated (acute cholecystitis, biliary pancreatitis, obstructive jaundice, and cholangitis) or uncomplicated (biliary pain) gallstone disease. Ultrasonography evidence of a dilated common duct or presence of common duct stones, serum elevations in alkaline phosphatase, transaminase, or bilirubin were indications for preoperative magnetic resonance cholangiography. Preoperative endoscopic retrograde cholangiopancreatography (ERCP) with removal of possible common duct stones was performed in the patients suspected of having choledocholithiasis.
The timing for LC in patients with acute cholecystitis was 24-72 hours from admission. Patients with acute pancreatitis related to gallstone disease underwent surgery after resolution of clinical and biochemical symptoms. The supportive treatment during the acute phase consisted of intravenous infusion, antibiotics, and nasogastric suction when necessary. Early surgery was defined as laparoscopic cholecystectomy during the initial hospitalization usually within 5 days, whereas patients undergoing delayed surgery were treated conservatively, discharged, and readmitted for elective operation. Patients were included if surgery was performed primarily for symptomatic gallstone disease and excluded if cholecystectomy was performed incidentally or secondary to another procedure. Laparoscopic cholecystectomy was performed using a standard four-trocars technique. An "open technique" was used in all cases to introduce the subumbilical cannula as previously described by our group. Dissection of the gallbladder from the liver was accomplished using monopolar electrocautery. Intraoperative cholangiography was performed selectively to assist in defining the anatomy or intraoperative abnormalities. A closed suction drainage was used in all procedures.
Results
Forty patients with a mean age of 74.2 years (range 70 to 91 years) were evaluated. 13 (32.5%) were males and 27 (67.5%) were females. All 40 patients in this series of geriatric patients were symptomatic from their gallbladder disease. A variety of the classical symptoms of gallbladder disease consisting of epigastric pain, Murphy's sign, fatty food intolerance, nausea and emesis, right upper quad-rant pain radiating to the back, biliary colic, fever, dyspepsia, belching and bloating, were present in all patients in this series. The indications for surgery included biliary colic in 26 (65%) patients, acute cholecystitis in 9 (22.5%), acute cholecystitis with pancreatitis in 5 (12.5%). The patients were evaluated according to the American Society of Anaesthesiologists (ASA) classification. Twenty-two patients were classified as either class I or II, fifteen patients were classified as class III, and three were classified as class IV. Comorbid conditions included hypertension, cardiac disease, peripheral vascular disease, and diabetes mellitus. Ten patients had no comorbid disease (25%). ERCP was performed preoperatively in one patients who had common bile duct stones that required sphincterotomy and stone extraction. Operative time in this geriatric series of patients ranged from 46 minutes to 2 hours. Conversion to open cholecystectomy was required in three of 40 patients (7.5%) due to their distorted anatomy (intrahepatic gallbladder, severe acute inflammation) with inability to safely dissect the cystic duct and cystic artery or suspected cancer of the gallbladder. Postoperative complications occurred in four patients. Two of the four complications were related directly to the surgical procedure itself including one patient with a postoperative cystic duct leak, which was treated successfully with an ERCP and sphincterotomy. One patient had a postoperative bleeding from a trocarsite requiring reoperation. Medical complications that were not directly related to the procedure itself included one patient with postoperative myocardial infarction that was admitted to the cardiac intensive care unit and was discharged from the hospital after 11 days. The mean postoperative hospital stay was 3 days (range 2-11). A few patients required more than 48 hours postoperative hospitalization. The perioperative mortality rate was 0%.
Conclusion
The population of persons older than 80 years has increased by during the last century. Fifty percent of women and sixteen % of men in their 70s have been shown to have gallstone disease, and 20% of the abdominal procedures performed in those older than 80 years are hepatobiliary. Although the prevalence of gallstone formation increases with age, and many studies have examined the results of laparoscopic cholecystectomy in elderly patients, the treatment of gallstone disease in this age group is a challenging. This group of patients has in fact an incidence up to 55% of complicated gallstone disease, such as acute cholecystitis, jaundice, choledocholithiasis, cholangitis, and biliary pancreatitis. Elderly patients frequently suffer from significant comorbid diseases and limited cardiopulmonary reserves that may contribute to a complicated perioperative course and increase postoperative complications rate. Acute biliary disease in the elderly was associated with a considerable increase in operative morbidity and mortality, when compared with non-elderly patients.
Laparoscopic cholecystectomy is currently the procedure of choice for managing gallstone disease and were demonstrated the physiological benefits and positive socioeconomic effects over the open procedure. The many advantages of the laparoscopic procedure include less patient discomfort, early hospital discharge, early return to a normal lifestyle, and lower cost. Many studies have demonstrated the applicability and advantages of the laparoscopic cholecystectomy also in the geriatric population with low rates of morbidity and mortality. Despite the frequent presence of concomitant diseases involving the lung, heart, and frequent presence of diabetes, these medical diseases contributed minimally to the morbidity results for geriatric patients treated with laparoscopic cholecystectomy whereas open cholecystectomy have consistently demonstrated higher rates of morbidity and mortality and greater lengths of hospital stay than the general population. | 2018-05-08T18:20:13.784Z | 2009-04-01T00:00:00.000 | {
"year": 2009,
"sha1": "a5588f1c153db7b35df7f157f236362624dbfc92",
"oa_license": "CCBY",
"oa_url": "https://bmcgeriatr.biomedcentral.com/track/pdf/10.1186/1471-2318-9-S1-A11",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a5588f1c153db7b35df7f157f236362624dbfc92",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
234093142 | pes2o/s2orc | v3-fos-license | Hierarchical sparse recovery from hierarchically structured measurements with application to massive random access
A new family of operators, coined hierarchical measurement operators, is introduced and discussed within the well-known hierarchical sparse recovery framework. Such operator is a composition of block and mixing operations and notably contains the Kronecker product as a special case. Results on their hierarchical restricted isometry property (HiRIP) are derived, generalizing prior work on recovery of hierarchically sparse signals from Kronecker-structured linear measurements. Specifically, these results show that, very surprisingly, sparsity properties of the block and mixing part can be traded against each other. The measurement structure is well-motivated by a massive random access channel design in communication engineering. Numerical evaluation of user detection rates demonstrate the huge benefit of the theoretical framework.
The hierarchically sparse signal model
The paradigm of compressed sensing [1] can be boiled down to the notion that if a vector x ∈ K N (K refers to one of the two fields R or C) is sparse, it can be recovered from underdetermined and noisy linear measurements y = Ax + η ∈ K M , M N using a plethora of algorithms. The impact of compressed sensing on the signal processing cannot be understated. In many applications however, the signal x at hand is not only sparse, but enjoys a richer structure. In this work, we consider hierarchically sparse (hisparse) signals. Hi-sparse signals are structured into blocks that are sparse, or may even exhibit a block structure themselves. Sparsity is assumed on all levels of the block structure, i.e. in a hi-sparse signal x = (x 1 , . . . , x N ), only s blocks This work was partially funded by DFG SPP 1798 -Compressed Sensing in Information Processing. F. acknowledges support from the Chalmers AI Research center (CHAIR).
x i are non-zero, and each the non-zero blocks x i contains at most σ i non-vanishing entries. Prior work on the recovery of block-sparse signals includes [2,3], while a hi-sparse signal model has also been considered in [4][5][6][7]. Here, we will use the following definition. Definition 1. Let x = (x 1 , . . . , x N ) ∈ K n1 × · · · × K n N , where K is either R or C. For s and σ = (σ 1 , . . . , σ N ), we say that x is (s, σ)-sparse, if • at most s blocks x i are non-zero and Note that this definition can be readily generalized to more sparsity levels with a nested tree structure [8] by allowing the σ i to be multi-leveled themselves, i.e. of the form (σ i , ς i ), with ς i = (ς i,1 , . . . , ς i,ni ) etc.
Before formally introducing the new family of operators on hierarchically sparse signals, let us discuss a specific example in the context of massive random access in the IoT communication literature. For further application examples such as, e.g., massive MIMO please refer also to [9], [10].
Grouped Random Access Model
In random access, each user that wishes to communicate with a base station chooses a resource at random (out of ν available) and sends a pilot signal associated with that resource. Assuming equal power transmitted from the users, and letting b j (i) ∈ C m , m ≤ ν, be the pilot signal that user i chooses, the base station then receives the vector where x i = 0 if user i is inactive, and x i = 1 else, and B = (b 1 , ..., b ν ). As before, it is reasonable to assume that the user activity is sparse meaning that the vector x is sparse.
This protocol has a fundamental problem -if several users choose the same resource (a collision occurs), subsequent communication is impossible, since each resource can only serve one user at a time. The probability of a collision grows very fast with a growing amount of users -a phenomenon commonly referred to as the birthday paradox.
In order to reduce the probability of collisions, we propose to distribute the users in groups. That is, randomly subdivide the active users in N groups, and accordingly replace the x-vector replaced with N blocks x i . We then let the users in each group i choose one of n resources, with n ≤ ν. Obviously, there is then within each block a much lower probability of collision than before. Furthermore, each block is with high probability sparse, say σ i -sparse. Thus, the signal to be recovered, (x 1 , . . . , x N ) ∈ (C n ) N is no longer only sparse, but with high probability even (N, (σ 1 , . . . , σ N ))hierarchically sparse.
In order for the base station to be able to de-mix the individual block contributions, we may let them send their pilot signals during disjoint time intervals and recover each x i from B i x i , for i = 1, ..., N individually. This would however be very tedious compared to the original, one-shot scheme. Instead, we propose to mix the contributions B i x i over (much smaller) M incoherent slots (say in time, frequency or space), each time j with a different random modulation a j,i . The base station over the course of those time slots then receives To further increase efficiency, we may let the B i be subsampled, meaning that we have access to an M · m-dimensional measurement, instead of an N · n-dimensional. It is a priori not clear that recovery still is possible. However, the strict sparsity assumption of hierarchical sparsity still gives us some hope of recovery. We show that this intuition is indeed true and we bring a series of strong theoretical arguments why it is so.
Hierarchical measurement operators
The main aim of this paper is to analyze the properties of operators of the form (1) within the context of hierarchically sparse recovery. Let us give them a name.
Definition 2. We call a measurement operator H from
We will refer to A as the mixing matrix and the B i as the block operators.
In [8], a subset of the authors proposed to use an adapted version of the celebrated Hard Threshold Pursuit (HTP) [11] to recover hi-sparse signals -the Hierarchical HTP (HiHTP), which we will present in more detail in the experiment section. The main finding of [8] was that if a measurement operator M exhibits the so called hierarchically restricted isometry property (HiRIP), HiHTP recovers all hierarchically sparse signals in a stable and robust fashion from linear measurements y = Mx.
A matrix 'having the HiRIP' then refers to δ s,σ (M) being small enough for the correct parameters s and σ.
In this paper, we will analyse the HiRIP properties of hierarchical measurement operators. We will give bounds on the HiRIP constants in terms of the RIP constants of the mixing and block matrices. These generalize prior work on compressed sensing of hierarchically sparse signals from Kronecker-structured measurements [12].
In the following, we will refrain from presenting any mathematical proofs. These, along with additional results and discussions, can instead found be found in the journal version of this paper, of which a preprint [13] will soon be available on arXiv.
HIRIP-PROPERTIES OF HIERARCHICAL MEASUREMENT OPERATORS
As briefly mentioned in the introduction, in [12], some of the authors of this article proved that a measurement matrix being a Kronecker product A⊗B has the (s, σ)-RIP provided A has the s-RIP and B has the σ-RIP. The Kronecker product A⊗B is defined through its action on an element (x 1 , . . . , x N ) ∈ (K n ) N as follows where a i denotes the columns of A, and ⊗ the tensor product. It is evident that these operators are special cases of the hierarchical operators considered here.
The main result of [12] is that if A has the s-RIP and B has the σ-RIP, A ⊗ B has the (s, (σ, . . . , σ))-HiRIP. The first main result of this paper is a direct generalisation of this result.
Theorem 2.1. Let H be a hierarchical measurement operator, as in (2). Assume that the matrices B i all obey the σ i -RIP with constant δ σi (B i ) for all i. Assume further that A obeys the s-RIP with constant δ s (A). Then H obeys the HiRIP, with where again σ = (σ 1 , . . . , σ N ).
Remark 1.
The result applies to more levels of sparsity. To be concrete, the σ i in the result may be multileveled themselves.
This result already enables us to construct an abstract hierarchical measurement operator which has the (s, σ)-HiRIP with high probability. Remembering that a Gaussian matrix B ∈ K m,ni has σ-RIP with high probability already when the number o m σ log ni σi [14,Ch.9]. In the same manner, we can choose the matrix A ∈ K M ×N as a Gaussian matrix. It will have the s-RIP with high probability if M s log N s . The above result then implies that the hierarchical measurement operator formed by A and the (B i ) i∈N has the (s, σ)-HiRIP. We conclude that recovery of (s, σ)sparse vector is possible using asymptotically no more than M · m s max i σ i · polylog(N, n) total measurements, which is comparable to the results for unstructured measurement operators from [8].
Remark 2. These considerations could to some extend already be derived using the Kronecker result from [15]. The main difference is that we may use varying sparsity levels, ambient dimensions and measurement operators.
Refining the result through block operator incoherence
As for the Grouped Random Access model, Theorem 2.1 does not bring a particularly satisfactory guarantee. Indeed, in that model, we have to recover (N, σ)-sparse signals. In order for the above result to be relevant, we thus need A to have the N -RIP, which necessitates M ≥ N . This is discouraging.
Can the result however be strenghtened? In fact it can, if the block operators are incoherent in the following sense.
Definition 4. We say that the collection of operators
In fact, the B i in the Grouped Random Access is pairwise (δ, σ)-incoherent if they are independently subsampled and modulated with random signs, as the following result shows. Proposition 2.2. Let F ∈ C n,n be the unitary Fourier matrix. For each i, let the matrix B i ∈ C m,n be formed by uniformly and independently sampling m rows from F, multiplying each of them with a uniform random sign, and subsequently rescaling the rows by m −1/2 . Assuming that B i is independent of B j for i = j and m σδ −2 log(n) 4 log(N ), the collection is (δ, σ)-incoherent with probability higher than 1 − n − log n 3 .
is a pairwise incoherent collection, it can to some extent separate the contributions of the individual blocks of a hierarchically sparse vector from the measurement N i=1 B i g i . Therefore, the mixing matrix A intuitively does not need to have a full s-RIP in order for the hierarchical measurement operator H formed by A and (B i ) i∈[N ] to have the (s, σ)-HiRIP. Put differently, if A is Gaussian, the needed number of A does not scale as s. A formal result is as follows.
Assume further that (sδ * 2σ ) 2 ≤ N/ log(N ) and that A ∈ K M ×N is a Gaussian matrix. Let furthermore δ, > 0. Provided the hierarchical operator H defined by A and (B i ) i obeys δ (s,σ) (H) ≤ δ + δ * σ with a probability at least 1 − . The above result says that provided δ * 2σ is small, we need M to be of the order (sδ * 2σ ) 2 , rather than s, to allow for ssparse signals on the 'block level'. If δ * 2σ is very small, we may hence obtain the RIP already when M is less than s -a behaviour which cannot be explained by Theorem 2.1. Note that we can even reach the realm of the Grouped Random Access model, i.e. s = N .
Remark 3. The quadratic dependence on of sδ * 2σ is not sample optimal. We leave it to future work to determine whether this only is a proof artefact, or a fundamental limit.
NUMERICAL SIMULATIONS
In this section we empirically verify that the hierarchical sparsity framework, and in particular Theorem 2.3, can be used for user detection in the Grouped Random Access scenario.
HiHTP Algorithm
As mentioned in the introduction HiHTP is a low-complexity algorithm for solving hierarchically sparse compressed sensing problems of the form where H is a linear operator satisfying the HiRIP. Provided H has the HiRIP for an appropriate sparsity level, it will recover any (s, σ)-sparse vector. The algorithm consists of iteratively performing a gradient descent step, thresholding the new iterate onto the set of (s, σ)-sparse signals and then adjusting the non-zero values on the obtained support by linear least squares. The algorithm terminates once the support of two successive iterates does not change or a suitable stopping criterion is reached.
Importantly, the projection onto the set of (s, σ)-sparse signal (line 3 of Algorithm 1) can be performed in an efficient manner. We refer to [8] for details on this.
Grouped Random Access
We assume ν = n = 512 available resources and model the measurements at the base station as y = Fx, where F ∈ C n×n is a n × n-DFT matrix. Note that since we are aiming for user detection, we do not necessarily need to recover x exactly: we only need to determine which x i are non-zero. Our baseline method therefore consists of computing supp(F −1 y) to obtain the selected resources.
We compare this to the Grouped Random Access model with N subsampled B i , which each consist of m = 256 random rows from the n × n DFT matrix. The random modulation matrix A ∈ C M ×N , where M = 16 is chosen as complex Gaussian with variance 1 √ N . This results in the system C M m y = Hx, with measurement operator H formed by A and the B i as in (1). 8 10 12 14 16 18 20 22 24
Fig. 2: User detection with hierarchical measurements
The HiHTP Algorithm 1 is used to recover x from the measurements y. Figure 2 shows the average number of recovered users for varying sparsities σ = 16, . . . , 36 and number of blocks N = 8, . . . , 32 over 25 Monte-Carlo trials for each configuration. Note that from N > 8 on the system has more pre-image dimensions than measurements and the recovery results of our paper apply. Notably, since all the N blocks are filled, classical hierarchical CS does not apply and particularly Theorem 2.3 must be invoked. The baseline shown in Figure 2 is computed as the total number of users, N · σ, minus the expected number of collisions that occur, if these users choose randomly out of the n = 512 available resources. As can be seen, HiHTP is able to recover much more users compared to the baseline. For σ = 16 and σ = 32 the performance of the HiHTP algorithm for user detection is shown, when the σ · N users are distributed randomly over all available slots (i.e. the block sparsities are not uniformly fixed to σ). Even in this scenario, where the algorithm operates with the wrong sparsity parameters, reasonable performance is achieved.
CONCLUSION
We proved a number of results regarding the HiRIP for hierarchical measurement operators. The results generalize prior work on hierarchical sparse recovery with Kroneckerstructured matrices. Furthermore, we demonstrated that the HiHTP algorithm is capable of computing hi-sparse solutions under this measurement structure in a practical Grouped Random Access design for IoT communication scenarios exhibiting huge capacity gains. | 2021-05-10T01:15:41.994Z | 2021-05-07T00:00:00.000 | {
"year": 2021,
"sha1": "71a29ee9433b4d95d7747cb85d7081023607a874",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/2105.03169",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "71a29ee9433b4d95d7747cb85d7081023607a874",
"s2fieldsofstudy": [
"Computer Science",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science",
"Engineering",
"Mathematics"
]
} |
11409767 | pes2o/s2orc | v3-fos-license | Generating Approximate Geographic Descriptions
Georeferenced data sets are often large and complex. Natural Language Generation (NLG) systems are beginning to emerge that generate texts from such data. One of the challenges these systems face is the generation of geographic descriptions referring to the location of events or patterns in the data. Based on our studies in the domain of me-teorology we present a two staged approach to generating geographic descriptions. The first stage involves using domain knowledge based on the task context to select a frame of reference, and the second involves using constraints imposed by the end user to select values within a frame of reference. Because geographic concepts are inherently vague our approach does not guarantee a distinguishing description. Our evaluation studies show that NLG systems, because they can analyse input data exhaustively, can produce more fine-grained geographic descriptions that are more useful to end users than those generated by human experts.
Introduction
Disciplines such as environmental studies, geography, geology, planning and business marketing make extensive use of Geographical Information Systems (GIS); however, despite an explosion of available mapping software, GIS remains a specialist tool with specialist skills required to analyse and understand the information presented using map displays. Complementing such displays with textual summaries therefore provides an immediate niche for NLG systems.
Recently, research into NLG systems that generate text from georeferenced data has begun to emerge (Dale et al., 2005;Turner et al., 2006;Turner et al., 2008b;Thomas and Sripada, 2008). These systems are required to textually describe the geographic distribution of domain variables such as road surface temperature and unemployment rates. For example, descriptions such as 'road surface temperatures will fall below zero in some places in the southwest' and 'unemployment is highest in the rural areas' need to be generated by these systems. One of the main challenges such systems face is the generation of geographic descriptions such as 'in some places in the southwest' and 'in the rural areas'. Such a task is challenging for a number of reasons: • many geographic concepts are inherently vague (see for example (Varzi, 2001) for a discussion on this topic); • often the underlying data sets contain little explicit geographic information for a generation system to make use of (Turner et al., 2008b); • as input to a generation system, georeferenced data is often complex, constraints imposed on the output text (such as length) may make the traditional approach to the Referring Expression Generation (REG) problem in NLG of finding a distinguishing description implausible (Turner et al., 2008b).
This paper looks at the problem in the context of work the authors have carried out on summarising georeferenced data sets in the meteorology domain. The main feature of our approach is that geographic descriptions perform the dual function of referring to a specific geographic locations unambiguously (traditional function of REG) and also communicate the relationship between the domain information and the geography of the region (novel function of geographic descriptions).
We present a two staged approach to generating geographic descriptions that involve regions. The first stage involves using domain knowledge (meteorological knowledge in our case) to select a frame of reference and the second involves using constraints imposed by the end user to select values within a frame of reference. While generating geographic descriptions it is not always possible to produce a distinguishing description because of the inherent vagueness in geographic concepts. Therefore, in our case we aim to produce a distinguishing description wherever possible, but more often allow non-distinguishing descriptions in the output text, which approximate the location of the event being described as accurately as possible.
After a short overview of the background in §2, some empirical observations on geographic descrip-tions from knowledge acquisition (KA) studies we have carried out are discussed in §3. Taking these observations into account, in §4 we describe how this problem is approached using examples from RoadSafe (Turner et al., 2008b), which generates spatial references to events in georeferenced data in terms of regions that approximate their location. It pays particular attention to the use of different perspectives to describe the same situation and how factors that affect what makes a good reference in this domain are taken into account by the system. In §5 we present a qualitative discussion of aspects of geographic description from the evaluations of RoadSafe that were carried out, and how this relates to future possible work on this topic.
Background
Much work on generation of spatial descriptions has concentrated on smaller scale spaces that are immediately perceivable. For example, spatial descriptions have been studied from the perspective of robot communication (Kelleher and Kruijff, 2006), 3D animation (Towns et al., 1998) and basic visual scenes (Viethen and Dale, 2008;Ebert et al., 1996). In a more geographical context route description generation systems such as (Dale et al., 2005) and (Moulin and Kettani, 1999) have had wide appeal to NLG researchers. (Varges, 2005) also generate landmark based spatial descriptions using maps from the map task dialogue corpus.
RoadSafe is an NLG system that has been operationally deployed at Aerospace and Marine International (AMI) to produce weather forecast texts for winter road maintenance. It generates forecast texts describing various weather conditions on a road network as shown in Figure 1.
The input to the system is a data set consisting of numerical weather predictions (NWP) calculated over a large set of point locations across a road network. An example static snapshot of the input to RoadSafe for one parameter is shown in Figure 2. The complete input is a series of such snapshots for a number of parameters (see (Turner et al., 2008b) for details).
In applications such as RoadSafe, the same geographical situation can be expressed in a variety of different ways dependent upon the perspective employed, henceforth termed as a frame of reference. Space (geographic or otherwise) is inherently tied to a frame of reference that provides a framework for assigning different values to different locations in space. For example, locations on Earth's surface can be specified by latitude and longitude which provide an absolute frame of reference for geographic space. Cardinal directions such as {North, East, West and South} provide an alternative frame of reference for geographic space. As was noted in (Turner et al., 2008b), characterising the data in terms of frames of reference is important because often the only geographic information input data contains are coordinates (latitude and longitude), while the Overview: Road surface temperatures will fall below zero on all routes during the late evening until around midnight.
Weather: Snow will affect all routes at first, clearing at times then turning moderate during tonight and the early morning in all areas, and persisting until end of period. Ice will affect all routes from the late evening until early morning.
Hoar frost will affect some southwestern and central routes by early morning. Road surface temperatures will fall slowly during the evening and tonight, reaching zero in some far southern and southwestern places by 21:00. Fog will affect some northeastern and southwestern routes during tonight and the early morning, turning freezing in some places above 400M. output texts are required to employ a wider choice of frames of reference such as altitude, direction, coastal proximity and population. In RoadSafe the frames of reference employed are always absolute according to Levinson's terminology (Levinson, 2003).
Because the geographic descriptions in RoadSafe do not fit the traditional formulation of the REG problem as finding the most distinguishing description, the most pressing question to address is what makes an adequate reference strategy in this case? This is of course a difficult question and is reliant to a large extent on the communication goal of the system. This paper looks into this problem in the context of the RoadSafe application, that uses a simple spatial sublanguage to generate the types of descriptions required in this application domain.
Observations on geographic descriptions from the weather domain
In this section we summarise some empirical observations on how meteorologists use geographic descriptions in weather forecasts. It describes work carried out over the course of the RoadSafe project involving knowledge acquisition (KA) studies with experts on summarising georeferenced weather data, observations from data-text corpora (one aimed at the general public and one aimed at experts) and a small study with people from the general public. During RoadSafe we built two prototype georeferenced data-to-text systems that summarised georeferenced weather data: one that produces pollen forecasts based on very simple data (Turner et al., 2006), and the RoadSafe system, which generates road ice forecasts based on complex data. Small corpora consisting of forecast texts and their underlying NWP data were collected in both application domains. Using techniques described in (Reiter et al., 2005) these corpora have been analysed to understand the experts' strategies to describe georeferenced data. The major finding from our studies is the fact that experts tailor their geographic descriptions to the task context. Not only does the geographic knowledge of the end user have to be taken into account in their descriptions, but also how the geography of the region causes events and patterns in the data. The latter consideration has a large affect on the frame of reference experts employ to describe particular geographic situations. §3.1 looks at these observations from the point of view of end users of weather forecasts, while §3.2 looks at the descriptive strategies of experts.
End users' geographic knowledge
It is a well known and accepted fact that geographic knowledge varies greatly between individuals. To illustrate this point 24 students of a further education college in Scotland were asked a geography question, without reference to a map. Which of four major place names in Scotland (Ayr, Glasgow, Isle of Arran and Stirling) did they consider to be in the south west of the country? The responses showed a great variation in the subjects' geographic knowledge. Half of all sub-jects considered Glasgow and Ayr to be in the south west, one third considered Stirling to be in the south west and most surprisingly only four considered this to be true of the Isle of Arran. The results of this study are surprising because Stirling is the least south westerly place in the list while Isle of Arran is the most south westerly. This study actually agrees well with the studies in psychology on variation in individuals' mental representation of their geographic environment (Tversky, 1993).
Contrast this with the detailed knowledge of a road engineer who the RoadSafe texts are intended for. Road engineers rely upon a large amount of local geographic knowledge and experience when treating roads. Indeed, their spatial mental models are specified at a much finer detail. For example, they get to know where frost hollows tend to form and also come to learn of particular unexpected black spots, such as where garages allow hose water to cover part of a road during winter. This is an important point to be taken into account when communicating georeferenced data as geographic descriptions should be sensitive to that knowledge because it dictates how accurately they will be interpreted by the end user.
Both task context and structural features of data (e.g. number of observations, granularity of measurement), as well as functional features of data (how the entities being described function in space) influence how it is described geographically. Analysis of a small pollen forecast corpus (Turner et al., 2006) revealed that forecast texts, contain a rich variety of spatial descriptions for a location despite the data containing only six data points for the whole of Scotland. In general, the same region could be referred to by its proper name e.g. Sutherland and Caithness, by its relation to a well known geographical landmark e.g. North of the Great Glen, or simply by its geographical location on the map e.g. the far North and Northwest. In other words, experts characterise the limited geographic information contained within the data according to the task context. As the consumers of such forecasts are the general public, there is a greater onus on the expert to make the texts more interesting, unlike more restricted domains such as marine (see (Reiter et al., 2005)) or road ice forecasts that require consistent terminology.
Experts' descriptive strategy
Work in psychology has suggested that meteorologists use a dynamic mental model to arrive at an inference to predict and explain weather conditions (Trafton, 2007). Vital to this process is also their ability to take into account how the geography of a region influences the general weather conditions. Understanding the weathers interaction with the terrain enables them to make reliable meteorological inferences particularly when a certain pattern in the data may appear random. It is often unfeasible for a human forecaster to spend large amounts of time inspecting every data point in a detailed visual display. Using experience and expertise a forecaster can use her mental model to 'play out different hypothetical situations' (Trafton, 2007, p.2) and thus arrive at a plausible explanation for an apparently random weather pattern. Consider the following example description of a weather event by an expert taken from our road ice corpus: • 'exposed locations may have gales at times.' This is a good example of a forecaster using her meteorological expertise to make an inference about a random weather pattern. Clearly there is no way from inspection of a map one can ascertain with certainty where the exposed locations are in a region. However, an expert's knowledge of how the referent entity (the wind parameter) is affected by geographical features allow her to make such an inference. These pragmatic factors play a large part in determining an experts descriptive strategy, where certain frames of reference may be considered more appropriate to describe certain weather events (Turner et al., 2008a). This comes from weather forecasters' explicit knowledge of spatial dependence (the fact that observations points in georeferenced data at nearby locations are related, and the values of their non-spatial attributes will be influenced by certain geographical features). This is one of the most important and widely understood fact about spatial data from an analysis point of view, and one of the main reasons that it requires special treatment in comparison to other types of non-spatial data. This fact is most clearly outlined by an observation made in (Tobler, 1970, p.3) that 'everything is related to everything else, but near things are more related than distant things'. This is commonly known as the first law of geography and still resonates strongly today amongst geographers (Miller, 2004). The implication of Tobler's first law (TFL) is that samples in spatial data are not independent, and observations located at nearby locations are more likely to be similar. Recasting this into meteorological terms, exposed locations are more likely to be windier and elevated areas colder for example.
In fact, an analogy can be drawn between how meteorologists consider perspectives in their descriptive strategy and the preferred attribute list in the seminal work on REG by (Dale and Reiter, 1995). In their specification of an algorithm for generating referring expressions content selection is performed through the iteration over a pre-determined and task specific list of attributes. In our context, preferred attributes are replaced by preferred frames of reference. This means describing georeferenced data requires situational knowledge of when to apply a particular frame of reference given a particular geographic distribution to describe.
The most striking observation about the expert strategy is that the geographic descriptions in the corpora are approximations of the input (Turner et al., 2008a). The input is highly overspecified with 1000s of points for a small forecast region, sampled at sub hourly intervals during a forecast period. Meteorologists use vague descriptions in the texts to refer to weather events such as: • 'in some places in the south, temperatures will drop to around zero or just above zero.' There are a number of reasons they use this descriptive strategy: the forecasts are highly compressed summaries, as a few sentences describes megabytes of data; very specific descriptions are avoided unless the pattern in the data is very clear cut; experts try to avoid misinterpretation, road engineers often have detailed local geographic knowledge and experts may not be aware the more provincial terminology they use to refer to specific areas. The following section demonstrates how the problem of generating such descriptions is addressed in RoadSafe.
Generating Approximate Geographic Descriptions
In its current form, where summaries are meant to give a brief synopsis of conditions to the user, RoadSafe follows the approach taken by forecasters as discussed previously. This is unconventional in comparison to traditional REG approaches that aim to rule out all distractors in the domain (properties that are not true of the referent). In a description such as 'reaching zero in some places above 100M by 16:00' above, distractors can be defined as the set of points above 100M that do not satisfy the premise that temperatures will drop below zero. More succinctly, these can be defined as false positives. In fact, the problem can be formulated as a trade off between false positives and false negatives, where false negatives constitute points that are wrongly omitted from the description. For road gritting purposes, costs can be assigned to each type of error: road accidents in the case of false negatives and wasted salt in the case of false positives. As the task dictates, with the higher associated cost it is imperative that a referring expression eliminates all false negatives. Ideally a truly optimal description should then seek to minimise false positives as far as possible, thus reducing the overall cost for the reader. While reducing errors descriptions should also be meteorologically correct, as discussed in the previous section. Using certain frames of reference in certain contexts may result in a poor inference about a particular weather situation (Turner et al., 2008b). Given this domain knowledge, we can formulate constraints for what makes a good approximate geographic description in this task context:
Minimise false positives.
3. Complete coverage of the event being described (no false negatives).
These constraints have been realized in a two staged approach to generating geographic descriptions. The first stage involves using domain knowledge (meteorological knowledge in our case) to select a frame of reference, while the second accounts for end-user constraints to select values within that frame of reference. Before we describe the individual stages, two necessary pre-processing stages for generation are described.
Geographic characterisation
As noted in §2, observations in georeferenced data often contain little explicit geographic information apart from their coordinates. Geographic characterisation is responsible for assigning a set of qualitative descriptors to each observation based upon a set of reference frames, such that observations can be collectively distinguished from each other. This provides both a criterion for partitioning the data, and a set of properties to generate geographic descriptions. A frame of reference in this context consists of a set of descriptions based upon a common theme such as coastal proximity e.g. {inland,coastal} or population e.g. {urban,rural}. In RoadSafe four frames of reference have been implemented: altitude, coastal proximity, population and direction. Those that make use of human (population) and physical geographical features (altitude, coastal Proximity) can be represented by existing GIS data sets; therefore, in these cases geographic characterisation is simply responsible for mapping observation coordinates to areas of these data sets. In contrast, directions are abstract and require definition. In RoadSafe, geographic characterisation maps each observation to a set of directional areas with crisp boundaries, described in the following section.
Pattern formation
To generate descriptions, the geographic distribution of the event to be communicated has to be approximated using data analysis techniques such as clustering. While not new to data-to-text systems, the novel aspect here is that the data is partitioned based upon the frames of reference that make up the spatial sublanguage of the system. This process summarises the location of the event by measuring its density within each frame of reference's set of descriptions. An example of such a distribution is shown in Figure 3. Figure 3: Density of zero temperatures in Figure 2 While the descriptions within each frame of reference with human and geographical features are dictated by the granularity of available GIS data sets (altitude resolution for example), the boundaries of directional areas require definition. In RoadSafe, because some flexibility in the generated geographic descriptions is desirable, the system uses a four by four grid to split the domain into sixteen equally sized directional areas defined by their their latitude longitude extents. This configuration is shown below where T stands for true and C for central in this case: Using a simple set of adjacency matrices based on this grid, RoadSafe represents a set of descriptions depicting the traditional eight main points of the compass plus a further five that we term gradable (central, far south, far north, far east and far west). Alternative con-figurations using a greater number of gradable descriptions are possible. These matrices are used by the microplanner to choose attributes to refer to events using the direction frame of reference. One example matrix for each category of directional description are listed below. In each matrix a value of 1 indicates that the event has a non-zero density in that area. In what follows we describe how our two stage strategy is implemented in our system.
Frame of reference selection
The main content selection decision made by the document planner is the choice of which frame of reference to describe a specific weather event such as wind gusts increasing or road surface temperature falling below zero. This decision is based upon both the location of the event as discussed previously, and situational knowledge stored in the knowledge base of the system. Frames of reference where all descriptions have nonzero densities are not considered. Situational knowledge consists of the probability of using each frame of reference given the context (the weather parameter to describe), and is based on corpus frequencies. Rather than simply choosing the frame of reference with the highest density, weighting each frame of reference in this way ensures meteorological correctness as far as possible.
Attribute selection
Once a frame of reference has been selected the microplanner maps the descriptions to abstract syntax templates. As this is fairly trivial for most frames of reference in RoadSafe, because they contain a limited number of descriptions, we will provide an example how this is accomplished for directional descriptions. The input to the microplanner is a structure comprised of the density of the event within the containing area plus its associated adjacency matrix as shown in Figure 4. The attribute selection algorithm is based upon four constraints incorporating the first two principles of the descriptive strategy outlined at the beginning of this section. They are:
Location
1. Minimise false positives -The description describing the distribution should introduce the least number of distractors. For the above example distribution the set {South} ensures coverage but introduces three distractors: CSW, CSE and ESE. While the set of directions {Far South, South West} only introduces one: CSW. In general, a measure of how distinguishing a description x is of a distribution y is given by: Thus, for a distribution z and descriptions x and y, x is a more distinguishing description of z than y iff distinguishing(x,z) > distinguishing(y,z).
2. Coverage (no false negatives) -The description should completely describe the distribution. The set of directions {Far South,South West} completely describes the above example distribution while {Far South} does not. For the set of directions x and distribution y, the predicate covers(x, y) is true iff 3. Brevity -The set of directions should yield the shortest description of the distribution. For the above example distribution there is only one set of directions that ensures complete coverage. But when faced with a choice for example {South} and {South West, South East} brevity constraint favours {South}. In general,the set x should be chosen over y because it is a shorter description. For the distribution z and sets of directions x, y with equal coverage of z, x is a shorter description of z than y iff |x| < |y|.
4. Ordering: If two descriptions have equal coverage, cardinality and are equally distinguishing for a given distribution, a description is chosen based upon a predefined preference ordering. Each type of property is assigned a score: Cardinal = 3, Intercardinal = 2 and Gradeable = 1. Therefore, the set of directions {Far South, South West} would be assigned a value of 3.
In classification terms, the first constraint can be considered as precision and the second as recall. The algorithm firstly ranks each individual description in the set described in §4.2 according to the constraints outlined above. If a single directional term cannot be used to describe the distribution it then incrementally tries to find the highest ranking combination of directions that satisfy the coverage constraint and do not cover the whole region; otherwise, the algorithm terminates by returning the empty set. So, for the example input provided at the beginning of this section it would return the abstract syntax template shown in Figure 4. Quantifiers are selected by applying a simple threshold to the point ratio (which is recalculated should distractors be introduced): some = > 0, many = > 0.5, most = > 0.7. This would be realised as 'in some far southern and southwestern places'.
Evaluation and Discussion
RoadSafe has been evaluated in post-edit evaluations with meteorologists at AMI and by asking potential users to compare the quality of the summaries to corpus texts based on the same data. While evaluations have been intended to test the overall quality of the texts we have received much feedback on the geographic descriptions the system generates. We have also carried out some comparison of the direction descriptions to those in the corpus, by annotating the corpus descriptions with our adjacency matrices and running them through the system. Descriptions were compared by calculating the Jaccard coefficient between the two matrices. Overall the mean score was 0.53, with a fairly low perfect recall percentage of 30%. The low precision score is perhaps not surprising as the descriptions generated by RoadSafe are crisp and the corpus descriptions are not solely based on the input data we have available. However, the majority (67%) of partial alignments were the result of RoadSafe producing a subset of the human desciprition, e.g. northwest versus north, which indicates the system descriptions are more fine grained. In terms of the human descriptions, what was most apparent from this evaluation is the fact that they almost exclusively used the eight major points of the compass. In terms of feedback experts have commented that generally the location descriptions generated by the system are accurate but should be more general. Of 97 post edited texts generated by the system 20% of the geographic descriptions were edited.
Most notable was feedback from twenty one road maintenance personnel, who participated in an experiment asking them to compare expert written texts to RoadSafe generated texts based on the same five data sets. The details of this experiment are to be published elsewhere; however, one of the main reasons they gave for liking the style of the generated texts was because they contained more geographic descriptions than the corresponding human ones. The fact that a data-to-text system can analyse every data point is an advantage. In contrast experts have a huge amount of knowledge and experience to draw upon and this reflects in their more general and conservative approach in their geographic descriptions. Perhaps one of their biggest criticisms of the system as a whole is that it doesn't do a good job of generating geographic descriptions that involve motion, such as 'a band of rain works east across the area'. Indeed, this was the most edited type of generated phrase during the post-edit evaluation. There has been little work to our knowledge on describing motion in the NLG literature.
There are many aspects of the generation of geographic that haven't been addressed in this paper and warrant further exploration. Particularly at the content level, there is a need to consider how to account for semantic composition effects caused by overlaying frames of reference. Another question that arises is when is it best to use an intensional rather than extensional description. There is also the question of when to use descriptions that involve relations or gradable properties. These are all choices that a data-to-text system can make that will affect how the summary is interpreted.
Conclusions
This paper has described an approach for generating approximate geographic descriptions involving regions in the RoadSafe system, which is based on empirical work carried out in the weather domain. Our strategy takes into account constraints on what constitutes a good reference in the application domain described, by taking into account pragmatic factors imposed by both the task context and the end user. What is most apparent from our empirical studies is that geographic descriptions describing georeferenced data are influenced by not only by location but also task context. An important observation based on our evaluation studies is that NLG systems by virtue of their ability to analyse input data exhaustively can generate descriptions that are more useful to end users than those generated by human experts. | 2014-07-01T00:00:00.000Z | 2009-03-30T00:00:00.000 | {
"year": 2009,
"sha1": "8688d0d772df81c01e35488fcad55e991f1ce11f",
"oa_license": null,
"oa_url": null,
"oa_status": "CLOSED",
"pdf_src": "ACL",
"pdf_hash": "8688d0d772df81c01e35488fcad55e991f1ce11f",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.