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 |
|---|---|---|---|---|---|---|
253238078 | pes2o/s2orc | v3-fos-license | BIMRL: Brain Inspired Meta Reinforcement Learning
Sample efficiency has been a key issue in reinforcement learning (RL). An efficient agent must be able to leverage its prior experiences to quickly adapt to similar, but new tasks and situations. Meta-RL is one attempt at formalizing and addressing this issue. Inspired by recent progress in meta-RL, we introduce BIMRL, a novel multi-layer architecture along with a novel brain-inspired memory module that will help agents quickly adapt to new tasks within a few episodes. We also utilize this memory module to design a novel intrinsic reward that will guide the agent's exploration. Our architecture is inspired by findings in cognitive neuroscience and is compatible with the knowledge on connectivity and functionality of different regions in the brain. We empirically validate the effectiveness of our proposed method by competing with or surpassing the performance of some strong baselines on multiple MiniGrid environments.
I. INTRODUCTION
A major problem with most model-free RL methods is their sample inefficiency and poor generalizability. Even for simple tasks, these methods require thousands of interactions with the environment. Furthermore, even after learning a task, slight changes in the environment dynamics or the underlaying task will force us to basically retrain the agent from scratch. To combat these issues, a number of methods have been proposed that utilize ideas from multi-task and meta learning ( [1], [2]). Some of these methods treat the task-ID like an unobservable latent variable in the underlaying POMDP and condition their policy on a learned, task belief state [1]. However, these methods generally use some simplifying assumptions that limits their generalizability. Our proposed method, BIMRL, is similar in that it learns to identify a suitable task-specific latent and uses it to quickly adapt, but relaxes some of these assumptions. Similar to some prior works ( [1], [3]), we use variational methods to estimate this latent variable. However, by introducing a new factorization of the log-likelihood of trajectories, our method learns a different context embedding that is more consistent with POMDP assumptions and can also be interpreted in a more meaningful way.
Another central issue with RL is the well-known exploration-exploitation dilemma [1]. If an agent wants to quickly adapt to a new task, it must be able to explore the environment and obtain relevant experiences. Nonetheless, with complex tasks and environments, conventional exploration methods, such as epsilon-greedy exploration, are not enough. Intrinsic rewards that encourage suitable exploration 1 are currently one of the most successful approaches. Our novel memory module provides an intrinsic reward to guide the exploration; thus alleviating the problems that arise from exploration across different tasks. Aside from this reward our memory module, which consists of an episodic part that resets after each episode, and a Hebbian part ( [4]) which retains information across different episodes of a particular task, gives us a performance boost in memory-based tasks while helping deal with catastrophic forgetting that comes up in multi-task settings.
To summarize, our contributions are as follows: 1) Proposing a multi-layer architecture that bridges model-based and model-free approaches by utilizing predictive information on state-values. Each layer of our architecture corresponds to a different part of the Prefrontal Cortex (PFC) which is known to be responsible for important cognitive functionalities such as learning a world-model [5], planning [6], and shaping exploratory behaviours [5]. One of the objectives in this architecture is derived from our newly proposed factorization of the log-likelihood of trajectories. This factorization is more robust to violations of POMDP assumptions and also takes the predictive coding ability of the brain into consideration [7]. 2) Introducing a neuro-inspired memory module compatible with discoveries in cognitive neuroscience. This module will help in memory-based tasks and give the agent the ability to preserve information, both within an episode and across different episodes of some task. The design of this memory module is reminiscent of Hippocampus (HP) as the main region assigned to episodic memory in the brain [8]. 3) Proposing a new intrinsic reward based on our memory module. This reward does not disappear over time, is task-agnostic, and is very effective in a multi-task setting. It will encourage exploratory behaviour that leads to faster task identification.
A. Meta Reinforcement Learning
Meta reinforcement learning is a promising approach for tackling few-episode learning regimes. It aims to achieve quick adaptation by learning inductive biases in the form of meta-parameters. There are multiple approaches to meta-RL, two of the most popular being gradient based methods and metric based methods. Another popular family of approaches use a recurrent neural network (RNN) [2]. Some previous works try to extend this approach and use an RNN to extract some context that will be used to inform the policy. Recently, by observing the effectiveness of variational approaches for estimating the log-likelihood of trajectories, some methods proposed to use the estimated context (referred to as the belief state in the literature) to inform the policy about its current task ( [1], [3]). This belief state should contain information about environment dynamics as well as the reward function of the task. Ideally, conditioning on this belief would convert the POMDP into a regular MDP. However, these methods typically impose some constraints on the underlaying POMDP (for instance, [1], [3] consider Bayes-adaptive MDPs which are a special case of the general POMDPs). These restrictions may be violated in some scenarios. On the contrary, our proposed method is more robust to such violations.
B. Modular Architectures
Injecting a modular inductive bias into policies is a promising strategy for improving out-of-distribution generalization in a systematic way. [9] proposed a modular structure where each module can become an expert on a different part of the state space. These modules will be dynamically combined through attention mechanisms to form the overall policy. [10] introduced an extension called the bidirectional recurrent independent mechanism (BRIMs). BRIM is composed of multiple layers of recurrent neural networks where each module also receives information from the upper layer. A modified version of BRIM is used in our architecture as well.
C. Combining Model-Based & Model-Free
There has been efforts on trying to combine modelbased and model-free approaches. Dyna [11] and successor representation [12] are two such attempts. There has also been some investigations on whether our brains work in a model-based or model-free manner, which has resulted in some neuro-inspired computational models ( [13]). We propose a new way of combining these two approaches that is inspired by the predictive capabilities of PFC [14].
D. Exploration
Another central issue in RL is efficient exploration of the environment. Methods that encourage effective exploratory behaviour through the use of an intrinsic reward are among the most popular and effective ways of dealing with this problem. Intrinsic rewards can be obtained from a novelty or curiosity criterion ( [15]). It is worth noting that effective exploration becomes an even bigger issue in the metalearning setting ( [3]).
III. METHODS
In this chapter we introduce our proposed method. We will first derive a factorization for a trajectory's log-likelihood. Using this factorization, we will design a multi-layered architecture with multiple loss functions. We will then shift our focus and introduce our memory module. Finally, we will give some intuitive interpretations of our proposed method and mention some of the connections to different regions of the human nervous system.
A. Likelihood Factorization
Our method builds upon VariBAD [1] and we use the same terminology as is used there. Similar to VariBAD, we aim to optimize the following ELBO loss corresponding to loglikelihood of a trajectory τ: where ρ is the trajectory distribution induced by our policy, m is the belief state, and H + is the time horizon for a task, which is set to be four times the horizon for an episode. Notice that this objective is comprised of two parts. The first part, E q φ (m|τ :t ) [log p θ (τ :H + | m)], is known as the reconstruction loss and the second part is the KL-divergence between the variational posterior q φ and the prior over the belief state, p θ (m). Unlike VariBAD, we factorize the reconstruction loss as follows: It is worth mentioning that unlike previous methods which use the Bayes Adaptive Markov Decision Process (BAMDP) formulation and assume perfect observability given the belief state (which will not hold in extreme partial observability scenarios), our factorization uses the predictive information in m and asserts that our belief should contain enough information to predict the next n states and rewards, given the next n actions. Contrary to VariBAD that conditions the trajectory on the full history of actions, we condition a trajectory on only the first n actions (hence, the added fourth summation term in 2). This modification will result in an added action simulation network which will also help in learning better representations in the first level of the hierarchical structure. This predictive coding capability is compatible with those observed in the PFC ( [7]).
B. Brain Inspired Meta Reinforcement Learning (BIMRL)
Our proposed architecture is composed of a recurrent task inference module responsible for updating our estimate of belief state about the current task, followed by a hierarchical structure similar to BRIM [10], that conditions on the inferred belief state. As we will see, this hierarchical structure is similar to PFC in how they function. The last layer of this hierarchical structure is a controller whose hidden state will be fed into an actor-critic network. We also have a memory module that directly interacts with the controller. Figure 1 demonstrates how all these modules are put together. In what follows, we will briefly introduce each part 1 .
1) Hierarchical Structure (BRIM):
The task inference module is a recurrent network that processes the last (observation, action, reward) and produces a belief state that will be passed to each of the three layers in the hierarchical structure as well as the memory module.
The first level of BRIM (as shown in Fig. 1) is responsible for learning a world model. This layer uses the last observation, action and reward in addition to the current belief state to predict dynamics, inverse dynamics and the rewards. It does so using three simulation networks. This layer is trained to maximize the trajectories log-likelihood as is formulated in 2. Figure 2 illustrates this layer in more detail. The second level (the "Planning" box in Fig. 1) is responsible for predicting the value of the next n steps. It will take in the previous level's hidden state (h 1 ) as well as the current state and the next n actions, and predicts the value for the next n time-steps. It aims to minimize the following loss 1 More details about the architectures and hyperparameters that were used as well as the code can be found here function: where G t:t+k is the k-step TD return defined as G t:t+k = R t+1 + γR t+2 + · · · + γ k−1 R t+k + γ k V π θ (S t+k ) . (4) In the above equation, ψ denotes the parameters of n-step value decoder network, h 2 is the output of the second level and θ shows the parameters of the critic head in the actorcritic network. The existence of this layer between the upper layer, responsible for learning the model of the environment, and the lower layer, which directly affects decision making, allows the following interpretations: • Model-based approach: Model-based methods first obtain a model of the environment and then plan their next action using this model. In our proposed architecture, the first layer aims to learn as much as possible about the environment. The second layer receives this information and predicts the value for the next n steps.
To do so accurately, this layer must be able to perform some sort of planning. Therefore, in our model the two explicit stages in model-based approaches are performed implicitly by the networks in the first and second layer. • Context-based RL: Methods such as VariBAD inform the agent of its task by providing it with a context vector obtained from a dynamics prediction network. The context vector that we provide is even more informative as it also contains the necessary information for predicting the value of states in the next several steps.
Finally, the third level of BRIM, known as the controller, aggregates the information form the previous BRIM layer and the most recent observations form the environment. Additionally, the hidden state of this layer modulates access to the memory, which will be discussed in the next section. The aggregated information will then be used by shallow actor and critic networks to determine the next action and value. We use PPO [16], an on-policy learning algorithm, to train our policy networks.
This hierarchical structure is reminiscent of certain parts of the brain. The first layer is similar to the medial Prefrontal Cortex (mPFC) as they are both responsible for learning a predictive model of the environment [7]. Similarly, the second layer corresponds to the Obitofrontal Cortex (OFC) which, in some neuroscientific literature, is identified as the main region responsible for multi-step planning in the brain [6]. Lastly, a part of the nervous system that connects PFC to HP (episodic memory) and the Striatum (which corresponds to actor-critic networks [17]) is known as the Anterior Cingulate Cortex (ACC) [18]. This region manages the use of control or habitual behaviours based on the degree of uncertainty about the task. It is also known as one of the pathways for transferring information from PFC to HP to guide the transition between the working memory (corresponding to PFC [19]) and the episodic memory [8]. In our model this corresponds to the third layer which modulates the memory and aggregates information in time [18].
2) Memory: The memory module is itself composed of two parts: an explicit episodic memory and a Hebbian memory. At each time step, the output of these two parts are combined using an attention mechanism and is then sent to the controller (see Fig. 3). The hidden state of the controller is used as the query for this attention mechanism and weights the contribution of each module.
By concatenating the embedded observation of the agent along with the output of the task inference module at each time-step, we create the "event key" and the hidden state of the controller is used as the "event value". In what follows, we will reference these event keys and values. Explicit memory is a slot-based memory that stores the event keys and values. At each time-step, the hidden state of the last level of BRIM is used as the query and the explicit memory is accessed using a multi-head attention mechanism. At the end of each episode, we compute the normalized "reference time" for each slot, which indicates how often each stored event was called upon by the incoming queries, on average. Equation (5) shows how it is calculated.
In above equation r i is the reference time for the i-th slot, s i is the time when this slot was added to the memory, W t i is attention weight attributed to the i-th slot at time t and H is the episode length. As we will see, the reference time is needed for updating the Hebbian memory. Figure 4 demonstrates how the episodic memory works.
Fig. 4: Episodic memory
The Hebbian memory works on the time scale of tasks, i.e., it won't be reset after each episode. Once an episode is completed, the top k percent of slots with the largest reference times are transferred to the Hebbian memory through the process of memory consolidation [20]. These transferred slots will update the Hebbain memory using the Hebbian learning rule. Equation (6) shows a general form of this learning rule: Here, γ + is the correlation coefficient, γ − is the regularization coefficient, w max imposes a soft constraint on the maximum connection weight, and W assoc kl denotes the parameters of the Hebbian network. As can be seen in the above equation, this learning rule consists of multiple meta-parameters that control the plasticity of synaptic connections. Inspired by the phenomenon of meta-plasticity ( [21]), a meta-learning mechanism is used so that these meta-parameters can be learned. Using the Hebbian learning rule as learning mechanism in the inner loop will also circumvent the issue of facing a second order optimization problem.
This transfer of information from a dynamic memory that grows into a fixed-size memory can also help with the decontextualization of the event [22].
3) Intrinsic Reward: Our proposed intrinsic reward is added to the sparse extrinsic reward of the environment to encourage exploration, both on episodic and task-level time scales. This reward is obtained by multiplying two components. The first component, r curiosity t , is a curiosity based reward that encourages visiting surprising states. It is a convex combination of reward, state, and action prediction errors. This reward is scaled using a second component, α t . We can think of this second part as a factor that adjusts for the newness of the current observation. It is defined as the distance between the most recent observation and the k-th nearest event stored in the episodic memory. If the agent encounters an observation that is unlike what has been seen in the on-going episode, this coefficient will be large, resulting in a high intrinsic reward. The following equations fully define the intrinsic reward: where λ state + λ action + λ reward = 1.
Note that NN k,X is the k-th nearest event to the current one, among the records that are stored in the episodic memory, X.
A. Experimental Setup
For evaluation, we used MiniGrid [23], a set of procedurally-generated, partially-observable environments in which an agent can interact with multiple objects. At each time-step, the agent receives its observation in the form of a 7 × 7 × 3 tensor and can perform any of the available 7 actions (moving in different directions, picking up objects, etc.). In the meta-training, the agent will see four episodes of each task and has to learn to adapt and solve the task within the time-span of those four episodes. At test time, we report the performance on the last (fourth) episode of each task. We used four sets of tasks: MultiRoom-N4-S5, MultiRoom-N6, KeyCorridorS3R1, and KeyCorridorS3R2. In the KeyCorridor set of tasks, the agent has to pick-up an object that is behind a locked door. The key is hidden in another room and the agent has to find it. The MultiRoom set of tasks provide an environment with a series of connected rooms with doors between them. The goal is to reach a green square in the final room.
B. Results
Fig. 5 plots the average return of our model as well as those of our baselines. As can be seen, some of the baselines do not get any reward in the more challenging tasks. This is because those tasks have larger state-spaces and require performing several actions in a particular order. This makes them difficult to solve for methods without a suitable exploration strategy.
In all four sets of tasks, our method achieves the best performance while converging faster and observing fewer number of frames. In MultiRoom-N6, one of the more difficult tasks, our method outperforms VariBAD+BeBold by a significant margin. In this task, other baselines fail to get any reward even after training for more than 2M frames.
C. Ablation study
To investigate the significance of each part of our proposed architecture, we evaluated the performance of three ablated versions of our model on the MultiRoom-N4-S5 set of tasks. We examined a model without the memory module (BIMRL w/o Mem), one in which the second level of hierarchical structure (responsible for n-step value prediction) was removed (BIMRL w/o Value pred), and one with the vanilla VariBAD trajectory factorization instead of our proposed factorization (BIMRL w/o N step pred). The result of this ablation study is depicted in Fig. 6. It can be seen that all ablated versions are less sample-efficient, indicating the usefulness of different parts of our model. Most notably, the exclusion of the memory module significantly reduces the convergence rate. We introduced BIMRL, a novel, modular RL agent inspired by the human brain. BIMRL induces several useful inductive biases which helps it meta-learn patterns across different tasks and to adapt quickly to new ones. It emphasizes the significance of taking inspiration from biological systems in designing artificial agents. For future work, one could investigate adding another layer to the memory, in the form of a life-long generative memory module. Additionally, more extensive experiments on tasks other than those in MiniGrid, particularly memory-based tasks, would shed more light on the extent to which such multi-layered architectures can help with fast adaptation to new situations. Furthermore, extending our architecture with new modules that make use of textual instructions would allow the agent to test its adaptation capabilities on a much larger set of tasks. This provides yet another avenue for future research. | 2022-11-01T01:16:03.436Z | 2022-10-23T00:00:00.000 | {
"year": 2022,
"sha1": "2f262448d9df2751e0e7801faecdce59686b3f41",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "2f262448d9df2751e0e7801faecdce59686b3f41",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
3964882 | pes2o/s2orc | v3-fos-license | Activated aging dynamics and negative fluctuation-dissipation ratios
In glassy materials aging proceeds at large times via thermal activation. We show that this can lead to negative dynamical response functions and novel and well-defined violations of the fluctuation-dissipation theorem, in particular, negative fluctuation-dissipation ratios. Our analysis is based on detailed theoretical and numerical results for the activated aging regime of simple kinetically constrained models. The results are relevant to a variety of physical situations such as aging in glass-formers, thermally activated domain growth and granular compaction.
In glassy materials aging proceeds at large times via thermal activation. We show that this can lead to negative dynamical response functions and novel and well-defined violations of the fluctuationdissipation theorem, in particular, negative fluctuation-dissipation ratios. Our analysis is based on detailed theoretical and numerical results for the activated aging regime of simple kinetically constrained models. The results are relevant to a variety of physical situations such as aging in glass-formers, thermally activated domain growth and granular compaction. Glassy materials display increasingly slow dynamics when approaching their amorphous state. At the glass transition, where relaxation times exceed experimentally accessible timescales, they change from equilibrated fluids to non-equilibrium amorphous solids. In the glassy phase, physical properties are not stationary and the system ages [1]. A full understanding of the non-equilibrium glassy state remains a central theoretical challenge.
An important step forward was the mean-field description of aging dynamics for both structural and spin glasses [2]. In this context, thermal equilibrium is never reached and aging proceeds by downhill motion in an increasingly flat free energy landscape [3]. Two-time correlation and response functions depend explicitly on both their arguments, and while the fluctuation-dissipation theorem (FDT) does not hold it can be generalized using the concept of a fluctuation-dissipation ratio (FDR). This led in turn to the idea of effective temperatures [4], and a possible thermodynamic interpretation of aging [2,5].
However, in many systems of physical interest, such as liquids quenched below the glass transition or domain growth in disordered magnets, the dynamics is not of mean-field type, displaying both activated processes and spatial heterogeneity [6,7]. While some experiments and simulations [8] nonetheless seem to detect a mean-field aging regime, theoretical studies have found ill-defined FDRs [9], non-monotonic response functions [10,11,12,13], observable dependence [14,15], nontrivial FDRs without thermodynamic transitions [16] and a subtle interplay between growing dynamical correlation lengthscales and FDT violations [17,18]; experiments have also detected anomalously large FDT violations associated with intermittent dynamics [19]. It is thus an important task to delineate when the mean-field concept of an FDR-related effective temperature remains viable.
Independently of this interpretation, FDRs have additionally been recognized as universal amplitude ratios for non-equilibrium critical dynamics. This makes them important markers for distinguishing dynamic universality classes. This area has seen a recent surge of interest [20].
Here we study kinetically constrained models, which are simple models for glassy systems with heterogeneous dynamics [21]. We use them to study systematically the impact of activated, and therefore strongly non meanfield, dynamics on FDRs and associated effective temperatures. In addition, the dynamics of these models becomes critical at low temperatures, where dynamical lengthscales diverge. Our work therefore also pertains to the study of FDRs in non-equilibrium critical dynamics. We show that FDT violations retain a simple structure with well-defined FDRs in the activated regime, elucidate the physical origin of negative dynamical response functions, and predict the generic existence of negative FDRs for observables directly coupled to activated processes.
We focus mainly on the one-spin facilitated model of Fredrickson and Andersen [22] (FA model), defined in terms of a binary mobility field, n i ∈ {0, 1}, on a cubic lattice; n i = 1 indicates that a site is excited (or mobile). The system evolves through single-site dynamics obeying detailed balance with respect to the energy function H = i n i . The dynamics is subject to a kinetic constraint that permits changes at site i only if at least one nearest neighbor of i is in its excited state. In any spatial dimension d, relaxation times follow an Arrhenius law at low temperatures [21,23].
At low temperatures, T < 1, the dynamics of the FA model is effectively that of diffusing excitations, which can coalesce and branch [23,24]. Such a problem can be described in the continuum limit by a dynamical field theory action with complex fields ϕ(r, t),φ(r, t) [23,25], with n i → (1 +φ)ϕ [25]. In terms of the equilibrium concentration of excitations, c ≡ n i = 1/(1 + e 1/T ), the effective rates for diffusion, coalescence and branching are, respectively, D ∝ c, λ ∝ c, and γ ∝ c 2 [23,24]. The aging dynamics following a quench from T = ∞ to low T consists of two regimes. Initially, clusters of exci-tations coalesce on timescales of O(1), reaching a state made of isolated excitations. This process and the subsequent energy plateau are reminiscent of the descent to the threshold energy in mean-field models [2]. For times larger than 1/D, excitations diffuse via thermal activation and decrease the energy as they coalesce.
Following [26] we calculated the connected two-time correlation to all orders in λ at tree level, , and ϕ q (t),φ q (t) are those of ϕ(r, t),φ(r, t). Subscripts q indicate wavevector, N is the system size, and z ≡ Dq 2 t w . We have assumed that both waiting time t w and final time t are large (≫ 1/D), and set γ = 0 to get the leading contribution at low T . The function f (z) goes as f ≈ 1 − 1/z for z ≫ 1, and f ≈ (1 + z)/3 for z ≪ 1. At q = 0 we get the energy auto-correlation: These classical expressions should be accurate above the critical dimension d c where fluctuations are negligible. The limit γ = 0 corresponds to diffusion limited pair coalescence (DLPC) which has d c = 2 [25].
Consider now a perturbation δH = −h q A q at time t w , with A q = i cos(q · r i )n i . The action changes by [27] T δS δh q = p,s to leading order in c. For the two-time response function The energy response follows as R 0 (t, t w ) ≈ −n(t)/t. The corresponding susceptibility, χ 0 (t, t w ) = t tw dt ′ R 0 (t, t ′ ) is always negative, and proportional to the density of excitations at time t, At any waiting time t w , for inverse wavevectors q −1 smaller than the dynamical correlation length, ξ(t w ) = √ Dt w , FDT is recovered: X q ≈ 1. However, at small wavevectors, qξ ≪ 1, the FDR becomes negative. In the q → 0 limit, one gets the FDR for energy fluctuations, This simple form means that on large, non-equilibrated lengthscales a quasi-FDT holds with T replaced by T /X ∞ . Contrary to pure ferromagnets at criticality, however, X ∞ is negative. This feature is not predicted by earlier mean-field studies as it is a direct consequence of thermal activation: if temperature is perturbed upwards during aging, the dynamics is accelerated; the energy decay is then faster, giving a negative energy response.
The fluctuation-dissipation (FD) plots in Fig. 1 show that the tree-level calculations compare very well with numerical simulations of the FA model in d = 3. We have also confirmed the diffusive scaling with z = Dq 2 t w . Simulations were performed using a continuous time Monte Carlo algorithm. We measure C q (t, t w ) as the autocorrelation of the observable A q , on a cubic lattice with periodic boundary conditions, N = L 3 and a linear size L ≫ √ Dt w . The susceptibility χ q (t, t w ) is obtained by direct generalization of the "no-field" method of [28] to continuous time. We show data using the prescription of Ref. [29], plotting χ q (t, t w )/C q (t, t) as a function of 1 − C q (t, t w )/C q (t, t) for fixed observation time, t, and varying waiting time, 0 < t w ≤ t. The abscissa runs from 0 (t w = t) to ≈1 (t w ≪ t). Using t w as the running variable ensures that the slope of the plot is the FDR X q (t, t w ) [15]. Other procedures, e.g. keeping t w fixed, would give very different results [9,10,11,12,13] that could lead to erroneous conclusions.
In dimensions d < d c we need to take fluctuations into account. For d = 1 exact scaling results can be derived for the FA model by considering a DLPC process with diffusion rate D = c/2. Because the long-time behavior of DLPC dynamics is diffusion controlled [25] we are free to choose the coalescence rate as λ = D without affecting scaling results in the long-time regime (t, t w ≫ 1/c) [30]. We focus on the low temperature dynamics (c → 0) and times t ≪ 1/γ ∝ 1/c 2 where branching can be neglected; the system is then still far from equilibrium as it ages.
To analyze our DLPC process we use the standard quantum mechanical formalism [31]: probabil- [31]. If we interpret the n i in DLPA as domain walls in an Ising spin system, n i = 1 2 (1 − σ i σ i+1 ), then P i ≡ σ i1 σ i2 , and W A ≡ W σ becomes the T = 0 master operator for the Glauber-Ising spin chain. Substituting the two-spin propagator 1|σ i1 σ i2 e Wσ c∆t derived in [32] and mapping back to DLPC then gives where G (2) i,j (·) and H n (·) are given in [32]. Eq. (4) together with the identity n i = 1 − E (i,i+1) is the key ingredient for the calculation of the two-time energy correlation and response functions C 0 (t, t w ), R 0 (t, t w ) in the d = 1 FA model [30]. Qualitatively a picture similar to d > d c emerges, but with the FDR now showing a weak dependence on the ratio t/t w . At equal times, In d = 1 the data fall on the curved limit FD-plot [30] derived from (4). In dimensions d = 3, 4 the simulations are compatible with a constant FDR X ∞ = −3, Eq. (3). The data for d = d c = 2 suggest a slightly curved limit plot, possibly due to logarithmic corrections, but are still compatible with X ∞ ≈ −3.
Numerically, the no-field method of [28] becomes unreliable for small q. To get precise data as shown in Fig. 2 we used instead the exact relation 2R 0 (t, t w ) = (1 − 2c)∂ t n(t) + ∂ tw C 0 (t, t w ) + C * (t, t w ).
for the energy-temperature response of a general class of kinetically constrained spin models [30]. The quantity C * (t, t w ) is defined as N −1 H(t)U (t w ) c , with U = i f i (n i − c) and f i ∈ {0, 1} the kinetic constraint. The auto-correlation C(t, t w ) = n i (t)n i (t w ) c , and associated auto-response, R(t, t w ) = T δ n i (t) /δh i (t w ), to a local perturbation δH = −h i n i , also have a negative FDR regime. Previous studies [13,16] had suggested that the corresponding FD-plot has an equilibrium form, even during aging. A more careful analysis, however, reveals nonequilibrium contributions (Fig. 3). Exact long-time predictions for C(t, t w ) and R(t, t w ), and the resulting FDR, can be derived from (4) [30]. One finds that, as for the energy, the FDR has the aging scaling X(t/t w ). It crosses over from quasi-equilibrium behavior, X ≈ 1, at t/t w ≈ 1, to X ≈ X ∞ for t/t w ≫ 1; notably, X ∞ here is the same as for the energy, Eq. (5). However, the region in the FD-plot that reveals the nonequilibrium behavior shrinks as O(1/ √ t) as t grows. This explains previous observations of pseudo-equilibrium and makes a numerical measurement of X ∞ from the local observable n i very difficult. On the other hand, Fig. 2 shows that with the coherent counterpart, i.e. the energy, this would be straightforward, confirming a point made in [15]. Numerical simulations for d > 1 produce local FD-plots analogous to Fig. 3 but with X ∞ = −3. We emphasize that the non-monotonicity observed is not an artefact of using t as the curve parameter, as in e.g. [9,12].
We have shown that some important components of the mean-field picture survive in systems with activated dynamics, in particular the concepts of time sectors (initial relaxation versus activated aging on long timescales) with associated well-defined FDRs. However, activation effects make these FDRs negative. This effect has not previously been observed in non-equilibrium critical dynamics, and calls into question whether effective temperature descriptions are possible for activated dynamics. Our arguments show that negative non-equilibrium responses should occur generically for observables whose relaxation couples to activation effects; activation need not be thermal, but can be via macroscopic driving, e.g. tapping of granular materials [10,33]. To illustrate how quantitative effects may vary, we consider briefly the East model [21] of fragile glasses, leaving all details for a separate report. The behavior is richer due to the hierarchical nature of the relaxation, which leads to plateaux in the energy decay. Properly normalized energy FD-plots nevertheless have a simple structure, with three regimes (Fig. 4): for given t, equilibrium FDT is obeyed at small time differences t − t w , indicating quasi-equilibration within a plateau; a regime with a single negative FDR follows, coming from the activated relaxation process preceding the plateau at t; finally, the plot becomes horizontal, corresponding to all previous relaxation stages which decorrelate the system but do not contribute to the energy response. Interestingly, in the FD-plots for autocorrelation and auto-response (Fig. 4, inset), each relaxation stage of the hierarchy is associated with a welldefined effective temperature, a structure reminiscent of that found in mean-field spin glasses [2]. This work was supported by the Austrian Academy of Sciences and EPSRC grant 00800822 (PM); EPSRC grants GR/R83712/01, GR/S54074/01, and University of Nottingham grant FEF3024 (JPG). | 2017-04-20T13:56:25.362Z | 2005-08-29T00:00:00.000 | {
"year": 2005,
"sha1": "2067f41da0b8e280112d0dba6e2c553951645c98",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/cond-mat/0508686",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "e2d8e95162c1fe12d9384382c5cb37c6182e4bd2",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Medicine",
"Materials Science"
]
} |
266473626 | pes2o/s2orc | v3-fos-license | Navigating misinformation and political polarization of COVID-19: interviews with Milwaukee, Wisconsin county public health officials
Introduction The spread of misinformation combined with the political polarization of the COVID-19 vaccine created major challenges for public health officials responding to the COVID pandemic and vaccine roll-out. The challenges public health officials faced when making safety recommendations and promoting the vaccine only exacerbated the already exhausting work conditions they experienced since the start of the pandemic. Combating misinformation while receiving inadequate political support led to burnout for many public health officials. As such, they had to adapt and develop new strategies for increasing vaccine acceptance and decreasing vaccine hesitancies. Method This study was conducted through qualitative interviews with seven Milwaukee County public health officials. This study aimed to determine how public health officials perceived misinformation and political polarization during the pandemic. Additionally, the study aimed to learn more about strategies county health officials used to combat misinformation while increasing vaccine uptake in their communities. Results Thematic analysis of the interviews identified three major challenges faced by public health officials in promoting vaccination: dissemination of misinformation in media, political polarization of COVID and its contribution to vaccine acceptance and COVID fatigue, and assessment of the risks associated with disease severity versus vaccine safety considering limited public health resources. Discussion Learning from public health officials allows us to better understand their perceptions of the extent of local vaccine hesitancies and their advice on how to counteract fears and misinformation and to promote COVID vaccine uptake. Political polarization of COVID and misinformation affected community vaccine acceptance and challenged local public health leadership.
Introduction
The COVID vaccine has been proven to be safe and effective through rigorous efficacy trials.Vaccine effectiveness monitoring found the odds of hospitalization fell by about 70% after one or two doses, the chances of severe disease (having five or more symptoms in the first week of illness) dropped by about one-third, and the likelihood of having long COVID (symptoms for at least 28 days after infection) was halved (1).The addition of COVID boosters has increased overall vaccine effectiveness in protection against infection to nearly 70% among adults and 94% effectiveness in preventing death (2).However, despite considerable evidence of safety and effectiveness, vaccine uptake has been suboptimal.
Research findings indicate people have concerns about side effects and safety of the COVID vaccine, lack trust in the government, and are concerned that COVID vaccines were developed too quickly (12).However, unlike past vaccines, the decision to receive the COVID vaccine is also heavily affected by cultural norms, social and peer influences, and political views (6).Distrust of the government and health care systems has contributed to COVID vaccine hesitancies for many Americans (6).Opposition to the COVID vaccine by media outlets, political polarization of COVID, and the spread of misinformation has further reduced vaccine acceptance, especially in lower-income areas of the US (7).
Political polarization
The bi-partisan structuring of the US is believed to have contributed to political polarization of the COVID pandemic as people received information from polarizing, biased informational sources while having decreased cross-partisan social interactions and information sharing (13).Much of public health response during the pandemic, including safety recommendations, social distancing, mask wearing, and vaccine promotion, was disseminated through various media outlets.US media sources with differing political alignments portrayed COVID differently; certain politically charged media sources reported more negatively about COVID and recommendations made by health authorities (13).Throughout the pandemic, decision making authority was questioned as political parties were divided on how to respond to the pandemic while considering how the US economy would be affected, further influencing the polarization of public opinion (14).More research is needed to better understand how political polarization can be mitigated so that it does not affect public opinion to the degree it has throughout the COVID pandemic.
Public health leadership
COVID fatigue is a growing problem for the general population and the healthcare system as the pandemic lingers.A COVID-fatigued population in combination with health care provider burnout has exacerbated an already stressed health care system.Since 2020, 1 in 5 healthcare workers have quit their jobs, and over 50% of those who quit cited the COVID pandemic and burnout from work as main reasons (15).Burnout among healthcare providers increased as COVID related hospitalizations increased, many of which could have been prevented by increasing COVID vaccination rates, especially with the bivalent booster.More specifically, public health workers, compared to healthcare providers, saw even greater levels of burnout during the pandemic, accompanied with reports of exhaustion, anxiety, and depression (16).Over two-thirds of public health officials have reported experiencing increased burnout, many of whom also 10.3389/fpubh.2023.1215367Frontiers in Public Health 03 frontiersin.orgreported experiences of professional abuse, harassment, and personal threats which negatively impacted their jobs, further increasing burnout (16,17).For many public health workers, the burnout, harassment, stress, and depression stemming from the pandemic proved to be too much which has led to the resignation of hundreds of US public health officials since 2020 (17).Further research is needed to better understand why public health officials experienced such high levels of burnout, what was done to alleviate that burnout, and additional negative impacts they experienced while performing their duties to serve and protect their communities.
Study objectives
Factors affecting vaccine acceptance among Milwaukee County residents during the initial roll-out of the COVID-19 vaccine challenged public health officials who responded with new strategies.Public health officials provided their observations and experiences of factors that influence community beliefs about health interventions.Study findings can be used to counteract misinformation and to support public health officials during the next public health crisis.This study provides new insight, and a better understanding of how public health officials were constantly challenged by rapid, vast dissemination of misinformation and unsupported by decision makers.The challenges health officials faced led them to feel overwhelmingly burned-out and that they were no longer trusted as a key source for COVID prevention and safety information.The purpose of this study was to answer two major research questions using public health interviews: 1 How did misinformation and political polarization of COVID and the COVID-19 vaccine affect how Milwaukee County public health officials performed their duties and responsibilities during the pandemic? 2 What COVID vaccine confidence boosting strategies were used in Milwaukee County and what additional strategies were used by public health officials to counter misinformation and increase vaccine uptake among the different communities in Milwaukee County?
Study setting
Milwaukee County is one of the most racially and economically diverse and segregated counties in Wisconsin; it is home to approximately 920,000 adults with 28% Black or African American, 16% Hispanic/Latinx, 5% Asian, 1% American Indian and Alaskan Native, and 50% White/Non-Hispanic in 2022 (5).Low income and poverty are challenges faced by many Milwaukee County residents.The percent of persons in poverty in Milwaukee County (18%) is almost double the whole state of Wisconsin and the median household income is $55,000, compared to $67,000 for the state (US Census, 2021).Almost 10% of residents do not have health insurance, compared to almost 7% for the whole state (US Census, 2021).
Study design
The study was performed using an exploratory approach through qualitative interviews with seven public health officials in Milwaukee County.An explanatory design allowed for construction of interview questions that would obtain in depth and diverse responses from public health officials (18).No one health official's response to a question was the same as another's responses.Interviews with local public health officials allowed an analysis of unanticipated comments and to better understand responses in real-time, allowing the interviewer to ask additional follow-up questions (19).Interview questions were guided by findings from our literature review, the specific aims of the study, and the results of focus group interviews with Milwaukee County residents regarding COVID risks that were conducted earlier in the study (20).Interview questions asked about the vaccine trends public health officials witnessed, factors they noticed contributing to vaccine acceptance in their communities, vaccine promotion ideas, and interventions that they conducted to increase vaccinations among at risk and hesitant populations.Participants were provided the interview guide in advance so that they could prepare accordingly and share as much information as possible.Participants were given the opportunity to email study staff with any questions regarding the interview or study protocol.
Recruitment
Our goal was to establish a heterogeneous group of public health officials from various jurisdictions of Milwaukee County.The principal investigator emailed Milwaukee County local health department leaders to recruit them to the study.The Medical College of Wisconsin Human Rights Review Board reviewed and approved all study activities.Participants were sent an informed consent informational letter prior to the interview.Participants verbally provided informed consent upon their involvement in this study and were informed of additional research outcomes that may stem from their participation.Seven public health professionals were interviewed between March 30 and May 18, 2022.Five were women, two men.Their educational credentials included MPH (4), MS, MA, RN, and MD.Their titles included health officer (5), director (2), and nursing supervisor.They worked at city health departments (6) or a county health department.
Data collection
Interviews were conducted by Zoom and lasted approximately 30-45 min each.Interviews were professionally transcribed verbatim and deidentified for analysis.
Interview protocols
Public health officials responded to interview questions addressing three different survey constructs: social media activity, COVID and perception of risk, and public health employee burnout.An example of each construct is listed below.The interviewer addressed each area
Data analysis
An inductive analysis approach was used which included open coding, creation of categories, and abstraction (21).Our research team read the transcribed interviews multiple times to understand the shared information.A coding tree was created to capture specific terms or phrases using an inductive coding approach in which codes were generated as the transcribed interviews were read and analyzed.Once all text segments were coded, we then created categories and further synthesized into themes.Meaning was given to codes through the categorization process.For instance, specific codes were assigned to text segments that mentioned vaccine hesitancies, vaccination strategies, vaccine misinformation, etc., but these coded segments were all then categorized as "Contributing factors for vaccination." Intersecting codes and coded segments were identified which allowed for the recognition of relationships and theme development.Direct quotes and phrases from public health officials were analyzed for further meaning which led to the generation of possible themes, as part of the contextualization process.Themes were further developed through abstraction, using reoccurring codes and contextualizing quotes from interview participants.MAXQDA software was used for coding and generating reports with coded segments and quotes to be used for analysis.Noteworthy text segments were highlighted and used to support the credibility of our themes.Quotes were selected to be included in the results section to follow.Descriptions were developed from the reoccurring themes which provided further context to further support the created themes discussed throughout the results.
Results
Using thematic analysis, we identified the following three core themes: (1) misinformation in the media; (2) the role of political polarization in COVID fatigue; and (3) weighing the risks of COVID severity vs. vaccine resources.Descriptions of communication strategies public health officials used to counter-act misinformation and disseminate accurate information are included in a flowchart at the conclusion of the results (Figure 1).The flowchart additionally includes challenges public health officials had to persevere as they fulfilled their duties and responsibilities during the pandemic.
Misinformation in the media
When asked about how the population they served learned of the COVID pandemic and vaccine rollout updates, public health officials responded that they received information from various news outlets including television, radio, internet websites, and social media.Different media outlets delivered their messages in different ways and not all messages contained accurate information.Participants reported that it was difficult to monitor news delivered through social media for accuracy.
There were multiple instances where myths and facts regarding the COVID vaccine were mixed.All participants reported how social media allows for the dissemination of misinformation.One stated: "I think one of the themes throughout the pandemic given how politicized, for better or worse, that the topic of COVID became, was really coming to the reality that social media in particular is a platform that can promote misinformation, disinformation, or accurate information."
Another noted how they "felt and experienced the platform of social media providing both disinformation and misinformation. "
Multiple health officials mentioned vaccine safety was a concern among adult community members, even noting rumors circulating about the vaccine being unsafe or that it contained "trackers": "When the vaccine became available, especially early on, we had quite a few questions about the safety, rumors about trackers or those things are in vaccines." One participant noted they came across a social media post saying, "Vaccines Kill" followed by a several comments with some people in agreement.Social media was not the only source of COVID misinformation.Participants also mentioned that community members received information from radio or television news programs addressing the COVID vaccine in negative ways and that these sources also sometimes disseminated false claims about the vaccine and severity of COVID.With so many sources of information available to community members, health officials took on the responsibility to monitor the public's perception of the COVID vaccine.
Misinformation about disease severity and mortality caused people to question their need to be vaccinated.Another health official mentioned they had community members denying that deaths attributed to COVID were caused by COVID, claiming that the deaths were caused by other factors: "They were saying that we are exaggerating the seriousness of COVID, or the impact that it's having on some individuals, or the number of individuals that are dying.Thinking that the death data is being exaggerated or the -you know, we have those conversations with some stories, 'Oh, they died in a car accident, but they had COVID, so you said they died of COVID'." Claims that COVID mortality rates were much lower than those officially reported were common in those communities.No evidence was provided to support these claims, but they circulated, nonetheless.
Combatting misinformation is a challenge in the public health sector and necessary for increasing COVID vaccine acceptance.When asked how they responded to misinformation while increasing accurate information, our health officials provided several different strategies, such as working to obtain community leader support for their decision making on COVID recommendations and empowering grassroots community movements.When investigating strategies to connect with community members, one health official noted: "As we look over the span of the pandemic to date, we really just want to recognize when we empowered grassroots community neighborhood members to amplify messages on their own social media platforms, Facebook, TikTok, you name it, whatever, that is where we really were seeing some of the most direct influence to some of the most vulnerable populations and had the best sort of reach." Multiple public health officials noted the importance of connecting with their community members during the complicated times of the pandemic, the continued use of grassroot campaigns when developing effective messaging, and the collective efforts of public health and community centers when promoting accurate information.These public health officials took charge and noted how it was their responsibility to disseminate accurate information to their communities.One health official explained their unbranded, custom designed strategy for promoting truth about COVID information and testing through development of a dedicated, COVID informational website to be used by all community health organizations and unrestricted by a sole health entity: "As the pandemic marched forward and time passed, that then got converted into healthy MKE.And so our kind of constant narrative was, "Come here for a source of truth about all things COVID." And so that platform was created, again, with input from grassroots community members who were very clear about naming, "I want to see people from my neighborhood, who look like me, who are from my neighborhood and look like me that are behind the photographs that are on the websites.Who's behind the camera matters, who's on the webpage matters." I think what was really exceptional about the work that we continue to do, it's not branded.So, all of the health systems, all of the community health centers, other partners, contribute and used this.And we were able to, collectively, in the absence of a brand come together and say, "This will be our source of truth, to help streamline some of the narrative." Another public health official mentioned how they tried to establish networks throughout their community so their decisionmaking regarding COVID safety recommendations would be supported: Several participants also noted using social media platforms to disperse accurate information and vaccine updates.Health officials had to adapt with the times and navigate social media usage and messages to the point where they had to "outcompete" possible sources of misinformation: "I think I would say we have a couple of our social media posts that you know kind of were shared widely, went viral, whatever you want to say.So, I do think that when we took the time to kind of make a higher quality graphic that would illustrate, whether it was data or mitigation strategy that we were recommending.I think that that was probably more of the most effective, just given the number of people who viewed it and shared it." One public health department implemented a state funded survey that they disseminated to their community addressing health equity and barriers to vaccination with a focus on vaccine effectiveness and safety.They obtained over 500 responses.The survey participants had mixed opinions when asked what their trusted sources of COVID related information are.Some survey participants indicated public health and government officials as trusted sources, while other participants noted public health and government were not their first source for information, listing family and friends above government.The health department used results from the surveys, specifically the age range for participants who did note they use government and public health as sources of information, to tailor social media messaging and platforms.The tailored messages were perceived to be more useful for the targeted audience, residents in their 30s and 40s for the most part, who see the health department as a reliable source of information.Notably, the area this organization served had very high COVID vaccination rates.
Establishing trustworthy connections within the community was also recognized as a useful strategy for public health officials when addressing individuals' concerns about the vaccine: "We realized the one-on-one support was much more likely to lead to somebody then getting the vaccine if they were able to talk to a nurse or talk to somebody and answer those questions by somebody they trust, that was actually a medical person." One health official noted "using CDC and DHS wording" when recommending vaccines to their community members.Multiple participants stressed contacting older community members who may not have social media via phone calls or implementing "hotlines" (a 24/7 number that community members could call to ask vaccine related questions) to provide any vaccine updates and recommendations.Aforementioned strategies used by public health officials to navigate misinformation and promote accurate information are displayed in Figure 1 below.Ultimately their takeaways for effective strategies were developing vaccine intervention plans in different languages and tailored for different cultural groups and the re-direction and correction of misinformation public health officials came across on social media while trying to not be dragged into an argument or project any negativity.
Political polarization and burnout
Effective vaccine roll-out was dependent upon politicians and public health officials working together to develop a vaccine dissemination plan that would boost local vaccine acceptance.However, with the vaccine being rolled-out during the conclusion of an election year in the US, health departments began receiving backlash and negativity as it became a hot topic in the political forum with continued lower acceptance by the Republican Party versus the Democratic Party (22).Overt politicization of the public health response, including widespread misinformation related to COVID vaccination, was spread by various forms of media and politicians (11).The public health officials interviewed through our study reflected on the high degree of opposition to the COVID vaccine leading up to and during the roll-out.The lack of political support for public health officials exacerbated already exhausting jobs.One health official detailed their need for support in their decision making from political or administrative leaders, while trying not to politicize the vaccine: "So, having leaders of the community also express their support, I think was impactful, for sure.And we have seen that for other type of public health responses as well.But honestly, though, it was a little bit harder for something like this because people did not really want to always get involved in something that's controversial.Like, you know, "I support it, but I'm not gonna be public in supporting it." I think that also got to be a challenge because it was so politicized." With a country divided politically at the end of 2020, postelection, many Americans held fast to their beliefs.Participants in our study explained how more conservative media sources displayed more opposition to the vaccine than liberal media sources; thus, the more conservative community members who relied on those resources had more opposition to the vaccine.News outlets tend to be a source of information for many people; however, news stations do not always present a situation or event in the same way spreading contradictory information on a topic.FOX News and MSNBC, media sources traditionally on opposite ends of the political spectrum, were even discussed during one interview as being information resources for certain community members, creating further challenges for health officials attempting to combat misinformation.
"Fox News was probably giving a much different spin than MSNBC on a variety of topics. And so, I would say the place where that information originates, whether it's with a local health department, a state health department, is probably more important than what channel it was on or what source of information was out there. It was very interesting to see all of the negative sources of information and disinformation or misinformation that came out. "
Participants revealed how some community members and political leaders were in support of the vaccine, while others tended to downplay the severity of the pandemic and the need for a vaccine.One public health official explained their frustration by describing how their small team worked to provide services to their community while receiving pushback from "leaders": "Eleven of us are working together in a very small geographic area with very fluid borders and irregular borders -being on the same page, supporting each other, providing prospective and experiences is gonna be very, very important.But
that's what we can do. I cannot change the political leaders in the village next door to me that basically chastised the health officer in an open public forum and said that what they are doing is unnecessary and inappropriate. "
One public health official went so far as to say that vaccines were politicized to such a degree that for a Republican, getting vaccinated was tantamount to switching political loyalties."I do not think in the history of public health have we ever, ever predicted something would be so politicized to this level, where it wasn't really about really health, it was more political lines.You're betraying a certain thing if you were doing it, to be honest, that was a lot of it." Politicization of COVID not only affected vaccine acceptance, but also the day-to-day work of public health officials creating an already stressful work-life.Burnout due to COVID was encountered by all the public health officials involved in this study as they and their co-workers worked long hours, changed roles in the workplace frequently as co-workers retired or quit, and delivered difficult and important messages, recommendations, and restrictions to their communities.They fulfilled many different tasks and roles while receiving backlash from groups who opposed their guidance and working with a political system that sometimes failed to support them.When we asked our participating public health officials what exacerbated their feelings of burnout, they had many different responses, and all participants noted a lack of full support by a political system that they felt should be doing the most to unite and protect people:
"We run up against a situation where politicians, political leaders, school leaders have pulled away mask mandates and mask guidance or even mask recommendations. "
The same health official also discussed how the lack of support created additional burnout during an already overwhelming experience:
"People got dramatically burned out. They got frustrated with the political process. They got frustrated with the community members who continue to chastise them on social media, print media, at public meetings. Discouraging or discrediting their expertise. "
A second health official also noted the challenges politics created as they strived to fulfill their responsibilities and duties: "We've become a sort of lightning rod for threatening people's freedoms and having a negative impact on the economy, when, in fact, all we were trying to do was save people's lives.Like, at the end of the day that's all any of us wanted to do but because of the politics related to the pandemic it's become something very different." Half of our participants shared how they or their coworkers had been personally attacked on social media or through their organization's website.One health official stated they were sent a post with sheep wearing masks and were accused of committing crimes against humanity while trying to promote the COVID vaccine.Another participant shared how many health officials were threatened by various community members and elected officials while doing their job: "Burnout and exhaustion is probably the theme of it all amongst leaders.We were pretty fortunate here in our community where I can say I do not think our previous health officer and myself ever received death threats.I did not have to have police positioned outside my home.I did not have to be escorted to my car from board of health meetings or council meetings.But a lot of my peers did." Participants explained how they had co-workers experience fatigue and burnout to the extent that they quit their jobs, further exacerbating burnout as workloads of those remaining increased.Some health officials quit due to the arduous nature of the work.Others found it to be a good time to change careers as they were forced to provide guidance and restrictions to people who viewed it negatively.Others had to change from working on a public health task they enjoyed to something they did not enjoy or felt they lacked the experience to do and opted out of the position as a result.While factors contributing to burnout varied for our public health officials; all health officials reported some level of burnout.A summary of challenges contributing to public health official burnout is included in Figure 1.
Weight of COVID severity versus vaccine resources
Public health officials tried to help vaccine hesitant community members to weigh more appropriately the risks associated with the COVID vaccine with the risks of contracting COVID while health officials had to consider their own resource depth.Five public health officials reported that they encountered community members who were not concerned about contracting COVID because they perceived that they were at low risk for illness or minimized its severity.The severity of COVID proved to be a topic of debate along with the safety and effectiveness of the vaccine.One public health official explained how they had to weigh their limited resources and time when developing vaccine promotion messages.For some age groups, such as 65 and older, COVID complications can be more severe than in younger age groups.Additionally, parents who are already hesitant about childhood immunizations and fatigued from all the COVID information circulating already may be less receptive to public health messaging.Weighing resources for chronically underfunded public health organizations forced public health officials to make some difficult decisions.When deciding on how to use resources, health officials must address who can benefit most from public health messaging: Participants found promoting the COVID vaccine for children even more challenging, especially those with anti-vaccine parents.Children often have mild symptoms and parents assessed the vaccine as being a greater risk than COVID illness.While complications from the vaccine are extremely rare, they are not zero.One health official mentioned the risk of myocarditis in adolescent boys who received the vaccine.When this participant was asked if they had any concerns regarding the safety of the vaccine, their response was: "I mean, as far as safety of vaccines, no.Not really.I mean, the main risk that can come is myocarditis.Is there a small risk of myocarditis in adolescent boys?Yes.When you have that risk benefit discussion, if you actually look at the numbers, and there's a lot of great visualizations of the numbers, it's not even a comparison.The risk, if you get COVID, you are many more times likely to get myocarditis, and so you are preventing that.It's kinda of like, is there a risk of wearing a seatbelt?Yeah.I see people in the emergency department with broken ribs from a seatbelt, or liver injuries from a seatbelt, but for every one of those I see 1,000 more that these lives were saved by a seatbelt." To overcome child vaccine hesitancy, participants explained how children can expose family members or friends who are at higher risk for severe complications from COVID.Just because youths may not be as likely to develop severe complications from COVID, that does not mean they are any less likely to spread the virus to more vulnerable population if unvaccinated.It became evident for our health officials that everyone who can be vaccinated should have the chance to receive the vaccine.One health official even admitted they were unsure of the safety and effectiveness of the vaccine when it was initially rolled out, but they continuously saw the COVID death reports and hospitalizations statistics, so they trusted in science and promoted the vaccine.The challenges health officials encountered while weighing resources when promoting the vaccine among groups when vaccine safety concerns are included as the final branch in the flowchart below (Figure 1).This public health official referred to the vaccine as a gamble but viewed the long-term effects of COVID and the ability to contract COVID more than once as being a greater risk than the vaccine.With the vaccine being created at record speed and its safety being a topic of debate, public health officials had to strategically emphasize how the risks associated with the complications of COVID, ability to contract COVID multiple times, transmitting the virus to high risk loved ones, and the unknown long-term effects of COVID outweighed the risks associated with the COVID vaccine.
Effects of political polarization
This study investigated how Milwaukee County public health officials navigated political polarization of the COVID vaccine and misinformation in their communities.We identified three themes constructed from recurring observations and strategies.Factors similar to those found to influence vaccine uptake were mentioned in our participants' responses with particular emphasis on misinformation and political polarization of the vaccine creating challenges for promoting the vaccine in their communities.Rapid, vast dissemination of misinformation in media, political polarization of COVID and the vaccine, and risk assessment of disease severity vs. vaccine safety have received research attention as barriers to vaccination for COVID at both a collective and individual level.Research suggests populations across the world who believe misinformation about the vaccine and severity of COVID have increased vaccine hesitancy; for instance, voters affiliated with the Republican Party have higher rates of vaccine hesitancy than Democrats (22).Effects from political decisions regarding COVID prevention measures can be seen across the different states.States with Republican leadership saw fewer adoptions of COVID prevention recommendations with more delays and increased mortality across races than states with Democratic leadership (23).Wisconsin is one of the most divided states in the nation as noted by its election results.Milwaukee County is not quite as divided as the state with two-thirds of voters in favor of the Democratic presidential candidate, but when voting for their congressional representative, two-thirds of votes were for the Republican candidate (24).
Combatting misinformation
Social media is used more than ever for disseminating news; however, it can lead to rapid spread of misinformation and lead to increasing vaccine hesitancy among communities (11,25).This study provides evidence to suggest public health officials felt they were not trusted and lacked support when enacting pandemic prevention guidelines and promoting the COVID vaccine.The abundance of contradictory and misinformative messages, often through social media, challenged the actions of our public health officials making it more difficult for them to protect their communities from COVID.The high volume and reach of misinformative posts on social media networks has been explored through various studies (26).Public health officials described the different strategies they used to navigate the challenges that they faced when promoting the vaccine and safety recommendations all while struggling with increasing burnout and high employee turnover.Our public health officials faced a combination of challenges when trying to weigh their resources while attempting to promote the COVID vaccine.The limited supply of the COVID vaccine early-on during the pandemic, in combination with the limited personnel and funds of health departments made for difficult decisions when promoting the vaccine among certain populations who were considered to be less at risk for severe COVID complications who could still contract and spread COVID.These individuals were found to be less likely to accept the vaccine according to our health officials, which may have been due to the early on promotion of the vaccine for more at-risk groups.This helped create a false sense of security for less at-risk groups as they may have felt they did not need to be vaccinated.Once the US had a stable supply of the vaccine, the challenges for promoting the vaccine among less at-risk groups only grew.COVID fatigue began to set in for many people and some of those who had abstained from being vaccinated as they felt they did not need it as much as other more at-risk people, had no desire to receive it after it became available to everyone.Efforts to vaccinate everyone are still underway, but as more people weigh the decision to vaccinate and health departments are forced to weigh their resources, the trajectory for future booster vaccination coverage is ambiguous.It is important we learn from the early-on COVID vaccine promotion strategies and enforce the need for a highly vaccinated population to keep a virus from spreading and evolving.
Application of theory
Findings from this study can be used to guide interventions to promote vaccine uptake.Future research is needed to understand the perspectives of vaccine-hesitant individuals to learn more about the beliefs that drive the decision to vaccinate or not.Misconceptions and a general lack of trust in vaccines can be assessed, accounted for, and evaluated by using health communication strategies, such as the Health Belief Model (27).The Health Belief Model (HBM) serves as the framework for many public health campaigns.The HBM uses six constructs to predict health behavior: risk susceptibility, risk severity, benefits to action, barriers to action, self-efficacy, and cues to action (28).Components of this theory were intertwined in the vaccine promotion strategies our public health officials used as they focused on disseminating accurate information about the severity of COVID complications, often underestimated by the public, and the benefits of vaccination.The Theory of Planned Behavior (TPB) is the theoretical framework for the investigation of the influences on a person's decision to vaccinate as it allows us to better understand why something or someone else affects a person's decision making.The TPB states that behavioral intention is determined by more positive attitudes toward the behavior, approval of significant others for the behavior (subjective norms), and a sense of personal control over the behavior (perceived behavioral control) (29).Public health officials reported how they believed the rapid circulation of misinformation and political polarization of COVID influenced individuals' decisions to vaccinate.Strategies they implemented had to overcome these influences by targeting components of TPB.Public health officials incorporated TPB and targeted negative influences by developing grassroot campaigns, promoting community leadership and empowerment as they found community members were just as heavily influenced, if not more so by those around them in their own community who had their best interests in mind.Public health officials were not trusted and supported as well as they should have been due to the controversial, politically polarized misinformation regarding COVID circulating through communities who believed misinformation from sources they found to be more trusted or favorable than accurately informative public health officials.
Vaccine promotion recommendations
Health officials had to develop new plans to promote COVID safety recommendations and awareness among the public, all while counteracting circulating misinformation and politically polarizing media sources.To do this, many public health organizations turned to social media platforms.Research suggests social media campaigns can successfully inform the public on accurate COVID information to increase public awareness and education so that behavioral change can occur (30).Social media became a frontier for COVID information dissemination requiring health officials to learn more about effective social media campaign strategies and navigation of various social media platforms that they may have had little experience with prior to the pandemic.To combat misinformation spreading through communities who may not have viewed public health departments as the primary source of COVID prevention information, several public health officials enacted grassroots campaigns and tailored social media messages to increase vaccine acceptance and reduce misinformation that might be causing vaccine hesitancies.The public health officials created their strategies knowing that individuals were more perceptive to messages delivered by people they trust.The Theory of Planned Behavior also suggests that mass dissemination of accurate information can be more effective when tailored for a specific audience who will then reshare the information, as seen with multiple social media platforms used by public health officials.
Future implications
The climate of public health changed drastically once the pandemic began, but it also brought forth long standing issues about how the public health system is supported.Public health officials faced extreme challenges as a divided political system failed to properly fund health departments and support their evidence-based guidelines, restrictions, and recommendations (31).Even prior to 2019, governing parties had not properly supported public health systems in the US.The pandemic proved just how chronically underfunded and underserved the public health system was.Since 2008, the public health workforce has decreased by over 20% while 62% of local health departments have had no increases in funding (32).During the pandemic, public health officials across the nation, including several of our interview participants, were forced to change their roles and/or responsibilities to help with pandemic prevention, monitoring, and mitigation.Officials who had no experience with emerging infectious disease projects had to stop working on their projects to help with pandemic related projects.Opioid abuse prevention, blood lead investigations, health inspections, and countless other projects were halted as health departments did not have the funding or staff to keep up with the COVID response and these other areas at the same time (32).Health departments did not fare well when employees had to switch from their preferred projects to work on COVID related projects.Many of the health officials in our study noted how the reorganization of their department and changing of roles led to the resignation of many public health officials.The strenuous conditions public health officials faced began long before the pandemic.Moving forward, it is essential that our nation focuses on developing, funding, and supporting our public health system.Public health officials are experts in their field.Their expertise should be recognized and supported by political and administrative leaders who make decisions regarding mandates and guidelines created to better protect the health of public.
Additionally, the concerns expressed by public health officials regarding burnout and lack of support should be used to improve the emergency preparedness system in Milwaukee County and perhaps in other communities.This study found many public health officials were overwhelmed from the start of the pandemic, highlighting the lack of experienced personnel, support from the political system and administrative organizations, and communication with decision making entities.Experts predict future disease outbreaks will occur, and another pandemic is imminent.If another were to occur, it is essential we learn from the past 3 years by providing more support to our public health system, so they can better serve and protect members of the community without the backlash they faced during the COVID pandemic.
Limitations
This study could have benefited from asking public health officials additional interview questions addressing what theoretical models (if any) they used in crafting their messages, such as, using concepts from the Health Belief Model to understand vaccine decision making of community members.We were able to draw inferences from the interview results, but more precise questions would lead to more precise conclusions.Moving forward this study could expand outside of Milwaukee County to the greater Wisconsin area, adding more public health officials to our study who can provide insight into perhaps more conservative counties of Wisconsin.
Conclusions and implications
Through interviews with seven Milwaukee County public health officials and qualitative, thematic analysis, this study successfully identified factors contributing to COVID vaccine acceptance in Milwaukee County, factors affecting public health decision making during the pandemic, and the strategies public health officials used to promote the vaccine and enforce COVID safety precautions.Misinformation in the media, political polarization of COVID and its contribution to burnout among public health workers, and the weighing of COVID severity versus limited vaccine promotion resources created challenges for public health officials in Milwaukee throughout the pandemic.Public health officials guided much of the COVID pandemic response and the initial vaccine rollout.Many times, they received little support from political leaders as the vaccine became politically polarized and they were required to develop strategies to overcome an array of circulating myths and misinformation about the vaccine.By implementing tailored responses to the challenges that they faced, public health officials were able to create strategies for increasing vaccine acceptance and reducing hesitancies.Moving forward, public health officials need the support of all leaders (political, administrative, and community) to be able to best serve their community.
Disclosures
The project was approved by the Medical College of Wisconsin Institutional Review Board.
FIGURE 1
FIGURE 1Strategies used by public health officials to navigate the effects of misinformation, medical mistrusts, and political polarization during the pandemic are included in the flow chart above to the left.The resulting challenges health officials faced (burnout) during the pandemic are included on the right.Together the two columns depict the continuous efforts public health officials made while serving their communities throughout the pandemic. | 2023-12-23T16:12:05.644Z | 2023-12-21T00:00:00.000 | {
"year": 2023,
"sha1": "8ea7281885e462ed599281c6fcc8eafa7af80f22",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fpubh.2023.1215367/pdf?isPublishedV2=False",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3d13c3fd2f94724342c8e2a1819b3c0ba77bbeda",
"s2fieldsofstudy": [
"Political Science",
"Medicine",
"Sociology"
],
"extfieldsofstudy": []
} |
267249915 | pes2o/s2orc | v3-fos-license | Cognitive Performance of the Elderly People in an Urban Area of Bangladesh: A Sociodemographic Study
A cross-sectional, analytical study was conducted to see the sociodemographic characteristics of elderly people and their association with cognitive performance in an urban setting in Dhaka, Bangladesh, between January and December of 2016. A convenient sampling technique was adopted. A total of 189 elderly people (118 male and 71 female) were included in the study. A semi structured questionnaire was first developed in English and then translated into Bangla. The questionnaire contains questions related to sociodemographic characteristics of the respondents and their cognitive functions, as determined by using Mini Mental state Examination (MMSE). Sociodemographic data were analyzed with descriptive statistics. Pearson correlation test was applied to determine the relationship between quantitative variables, while Chi-square test was done to assess association of qualitative data. The age of the respondents ranged from 60 to 90 years. The mean age was 67.8±6.26 years. Higher proportion of male (37.3%) and female respondents (36.6%) were in 60-64 age group, while the lower proportion was in ≥75 years age group (male 12.7% and female 14.1%). Most of the male respondents were graduate and above education group (64.4%), while most female respondents belonged to SSC and below and informal education group (50.7%) group. The rest of them fell into HSC education group: 11% and 32.4% respectively. Most of the elderlies (57.67%) belonged to the extended family and the rest lived in the nuclear family (42.33%). Among the male respondents, 76.3% were retired. Rest of the respondents was service holder (9.3%), businessman (12.7%) and teacher (1.7%). Female respondents were mostly housewives (97.2%) and two were teachers (2.8%). Among 189 respondents, 35.4% had normal cognitive function. Half of the respondents (52.4%) had mild cognitive impairment and the rest had moderate cognitive impairment (12.2%). None had severe cognitive impairment. Significant relationships of the cognitive performance were observed with age, sex, marital status, educational status, and type of family they lived in. However, no significant association was found between occupation and cognitive performance.
Introduction
Cognition is about the processes behind human thinking and experiences.Cognition refers to "a process of identifying, selecting, interpreting, storing, and using information to make sense of and interact with the physical and social world, to conduct one's everyday activities, and to plan and enact the course of one's occupational life".
Address of Correspondence:
Email: irtifaazizoishee@gmail.com speed, logical thinking and spatial problem solving.A convenient sampling technique was adopted.
Based on the inclusion and exclusion criteria,
Results
The age of the respondents ranged from 60 to 90 years.The mean age was 67.8± 6.26 years.
Higher proportion of male (37.3%) and female respondents (36.6%) were in 60-64 age groups, while the lower proportion was in the age group of ≥75 years (male 12.7% and female 14.1%).
Most of the male respondents were graduate and No severe cognitive impairment was found (Fig. 1).A moderate negative association was revealed between age of the respondents and their cognitive performance, which was statistically significant (p<0.001) (Fig. 2).It was observed in the study that there was no association found between occupational status and cognitive performance; this may be due to our adoption of convenient sampling technique which could not show much variation in their pattern of earning livelihood and living standards.
Conclusion
Our data revealed significant relationships of the cognitive performance with age, sex, marital status, educational status, and type of family they lived in.However, no significant association was found between occupation and cognitive performance.
1
Scientists and researchers often refer to different cognitive domains such as perception, attention, memory, language, executive function (initiating, planning, organizing, controlling and evaluation of thinking and acting) and psychomotor speed.
2
Some of those cognitive functions decrease within normal aging; for example, short-term memory and the way we learn new skills, mental CBMJ 2024 January: Vol. 13 No. 01 finally 189 elderly people (118 male and 71 female) were included in the study.A semi structured questionnaire was first developed in English and then translated into Bangla.The questionnaire contains questions related to sociodemographic characteristics of the respondents and their cognitive functions, as determined by using Mini Mental state Examination (MMSE).The MMSE is one of the most commonly used tests worldwide for assessing cognitive function; it was developed by Folstein et al. in 1975. 12The MMSE assesses orientation in time and place, attention, memory, and language and visual construction.The MMSE has a maximum of 30 points where higher scores indicate better cognition.The cut-off levels used in this study were: ≥27 = no impairment; 21-26 = mild; 11-20 = moderate; and ≤10 = severe impairment.13 After data collection, each questionnaire form was checked to find out any incompleteness or missing data or outliers.Data sheets produced on the computer were double checked to ensure their accuracy and completeness.Statistical analysis was carried out using SPSS software version 23.0.Sociodemographic data were analyzed with descriptive statistics.Pearson correlation was done to determine the relationship between quantitative variables, while Chi-square test was carried out to assess association of qualitative data.To assess the strength of association, odds ratio (OR) and their corresponding 95% confidence interval (CI) were calculated.All tests were two tailed and a P value <0.05 was considered statistically significant.This research was approved by the Institutional Ethical Review Committee of National Institute of Preventive and Social Medicine (NIPSOM), Dhaka, Bangladesh.
Table -
II: Scores of mini mental state examination (MMSE) of the respondents (N=189) CBMJ 2024 January: Vol. 13 No. 01 Among the respondents, 35.4% had normal cognitive function.Half of the respondents (52.4%) had mild cognitive impairment and the rest had moderate cognitive impairment (12.2%). Table-III).
CBMJ 2024 January: Vol. 13 No. 01DiscussionThe present study was done on elderly people.14 In contrast, van Exel et al. reported that women were less prone to the development of cognitive impairment. | 2024-01-26T16:33:28.657Z | 2024-01-24T00:00:00.000 | {
"year": 2024,
"sha1": "c89ef49b10c817ea52d8137bfb783aced12d2d1a",
"oa_license": "CCBY",
"oa_url": "https://www.banglajol.info/index.php/CBMJ/article/download/71066/47620",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "5461a22fb3211b74fc286d84d8bd7dee5379c09d",
"s2fieldsofstudy": [
"Sociology"
],
"extfieldsofstudy": []
} |
46932160 | pes2o/s2orc | v3-fos-license | Forced differentiation in vitro leads to stress-induced activation of DNA damage response in hiPSC-derived chondrocyte-like cells
A human induced pluripotent stem cell line (GPCCi001-A) created by our group was differentiated towards chondrocyte-like cells (ChiPS) via monolayer culturing with growth factors. ChiPS are promising because they have the potential to be used in tissue engineering to regenerate articular cartilage. However, their safety must be confirmed before they can be routinely used in regenerative medicine. Using microarray analysis, we compared the ChiPS to both GPCCi001-A cells and chondrocytes. The analysis showed that, compared to both GPCCi001-A cells and chondrocytes, the expression of genes engaged in DNA damage and in the tumor protein p53 signalling pathways was significantly higher in the ChiPS. The significant amount of DNA double strand breaks and increased DNA damage response may lead to incomplete DNA repair and the accumulation of mutations and, ultimately, to genetic instability. These findings provide evidence indicating that the differentiation process in vitro places stress on human induced pluripotent stem cells (hiPSCs). The results of this study raise doubts about the use of stem cell-derived components given the negative effects of the differentiation process in vitro on hiPSCs.
Introduction
Stem cells (SCs), particularly human induced pluripotent SCs (hiPSCs), constitute a real hope to better understand the pathogenesis and improve the treatment of many disorders (e.g. neurodegenerative, neurovascular, and cardias diseases) that are unresponsive to current treatments [1]. HiPSCs hold great potential in regenerative medicine due to their potentially unlimited self-renewal capacity and ability to give rise to all of the somatic lineages in the body [2]. HiPSCs can be cultured in two main ways: feeder-dependent and feeder-free systems. In the feeder-dependent system, the cells are placed on a layer of inactivated murine embryonic a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 fibroblasts as feeder cells in a medium supplemented with fetal bovine serum or proprietary replacements such as KnockOut Serum Replacement [3]. However, this approach presents a risk of contamination with animal pathogens and is thus considered unsuitable for clinical applications. By contrast, feeder-free culture systems represent a significant improvement over feeder-dependent systems, making these cells and their derivatives more suitable for use in clinical practice [4]. However, it is essential that Good Manufacturing Practices (GMP)which provide strict conditions for production of these cells-be followed when generating hiPSCs in a feeder-free culture [5]. GMP validation-which involves genetic stability analyses, virus and pathogen testing and method of derivation-is crucial before hiPSCs-based cell therapies can move from bench to bedside. This extensive characterization and validation process helps to ensure patient safety. Conceivably, for some patients, bioproducts based on hiPSCs may offer a viable treatment in the future [6].
HiPSCs can be differentiated into specific lineages-cardiac, osteogenic, chondrogenic, and neural-that possess distinct features and characteristics. The differentiation process in vitro can be achieved with either two-or three-dimensional (3D) cell culture methods [7,8]. In 3D cultures, such as differentiation using scaffolds or bioreactors, cell-cell contacts and cell-matrix interactions mimicking the native environment play an important role [9]. For this reason, 3D cultures may be preferred because they improve and enhance the effectiveness of differentiation protocols [10]. The most common methods of hiPSC differentiation in vitro are embryoid body formation [11] and micromass [12] and pellet culture [13]. In these methods, hiPSCs are differentiated in the medium with the addition of exogenous growth factors (GFs). The most common GFs are epidermal growth factor (EGF), fibroblast growth factor (FGF), bone morphogenetic protein 4 (BMP-4), transforming growth factor 3 (TGF-β3), and Activin A [14].
According to the published data, differentiation in vitro leads to the activation of the tumor protein p53 signalling pathway that controls the course of the differentiation process in vitro throughout the inter alia downregulation of agents responsible for pluripotency (e.g., Nanog). The differentiation process is also accompanied by telomere shortening, an increase in the levels of reactive oxygen species (ROS) and deficient DNA repair activities [15,16]. However, there is notable lack of data on DNA damage and p53-associated signalling pathways in hiPSCs during chondrogenic differentiation. In this context, we conducted the present study. The main aim was to evaluate how the chondrogenic differentiation process influences essential genes engaged in DNA damage and p53 signalling pathways and to compare these aspects to both mature chondrocytes and hiPSCs created by our group. We performed a global gene expression analysis using the Affymetrix platform in order to perform this comparison.
Recently, our group generated a hiPS cell line (GPCCi001-A) from primary human dermal fibroblasts using a lentiviral system [17]; these cells were subsequently differentiated towards chondrocyte-like cells (ChiPS) [18]. Our protocol involves a 3-week monolayer culture with supplementation with a relatively small amount of pro-mesodermal (e.g., BMP-4) and chondrogenic (e.g., TGF-β3) GFs. The advantage of this approach over other protocols is that it does not require any additional steps (such as formation of embryoid bodies) and is thus a time-and cost-effective approach.
The present work provides a comprehensive understanding of the processes involved in the formation of DNA damage and stimulation of the p53 signalling pathway, which is activated during chondrogenesis in vitro of hiPSCs. Our study confirms that hiPSCs present decreased DNA damage-induced stress mechanisms, which are characteristics of pluripotent stem cells whose capacity to maintain a mutation-free cell population is high. By contrast, hiPSC-derived chondrocytes (ChiPS) are characterized by increased expression of genes involved in DNA damage and p53 mechanisms. These findings raise important questions concerning the genetic stability of cells derived from hiPSCs (including ChiPS), which show a notable downregulation of stress defense mechanisms after undergoing chondrogenic differentiation.
The global gene expression profile of ChiPS were compared to the hiPSCs (GPCCi001-A cell line) and to two cell lines of mature chondrocytes: HC-402-05a (Merck Millipore, Germany) (HC) and primary articular cartilage chondrocytes (ACC).
Ethics
Primary articular cartilage chondrocytes (ACC) used in this study were obtained from Prof. Kruczynski's at the Department Orthopedics and Traumatology, Poznan University of Medical Sciences. Department of Orthopedics and Traumatology obtained an informed consent from tissue donors.
Immunofluorescence analysis
The cells were transferred into a 0.1% gelatin-coated 48-well plate for 48 h and then washed with phosphate buffered saline (PBS) (Sigma Aldrich, MO, USA) and fixed for 20 min in 4% formaldehyde (CHEMPUR, POLAND) (400 μl of formaldehyde per well). Then, the cells were rinsed with PBS containing 1% bovine serum albumin (BSA) (Sigma Aldrich, MO, USA) and incubated for 30 min in PBS containing 1% BSA and 0.2% Triton X-100 (Sigma Aldrich, MO, USA). After 30 min, the cells were washed with PBS containing 1% BSA. The primary antibodies were diluted in PBS containing 1% BSA and 0.2% Triton X-100 and the cells were incubated overnight at 4˚C with the following primary antibodies (all from Abcam PLC, U. K.): type II collagen (1:100) (ab34712); SSEA4 (1:100) (ab16287). After conjugation with the primary antibodies, the cells were rinsed three times with PBS containing 1% BSA. The following secondary antibody was diluted with 1% BSA in PBS and incubated in the dark for 1 h at 37˚C: rabbit polyclonal (1:500) (711-546-152) and mouse monoclonal antibodies (1:500) (715-545-150) (Jackson ImmunoResearch, PA, USA). After the cells were washed three times with 1% BSA in PBS, they were stained for 5 min with diamidino-2-phenylindole dye (DAPI) (Sigma Aldrich, MO, USA) solution in water (1:10000) and then washed with PBS before undergoing microscope analysis.
RNA extraction
Total RNA was isolated according to the Chomczynski method using TRI Reagent (Sigma, St Louis, MO, USA) and the RNeasy Mini Elute cleanup Kit (Qiagen, Hilden, Germany) based on the manufacturer's guidelines. The concentration of isolated RNA was determined by spectrophotometric measurement at wavelengths of 260 nm (NanoDrop spectrophotometer, Thermo Scientific, MA, USA). The concentration of total RNA was assessed using the 260/280 nm absorption ratio; in all case the concentration was-as expected-at least 1.8. The quality and integrity of the RNA was verified in a Bioanalyzer 2100 (Agilent Technologies, Inc., Santa Clara, CA, USA). The RNA integrity numbers (RINs) ranged from 8.5 to 10, with a mean value of 9.2 (Agilent Technologies, Inc., Santa Clara, CA, USA). The RNA samples were diluted to a final concentration of 50 ng/μl. For the microarray study, 100 ng of good-quality, total RNA was used. The remaining RNA was exploited in the RT-qPCR validation.
Microarray expression studies
The microarray experiment was conducted according to previously published procedures [19][20][21][22]. The protocol involving preparation of RNA for further hybridization was carried out using GeneChip WT PLUS Reagent Kit (Affymetrix, Santa Clara, CA, USA). The first stage included performing a two-step cDNA synthesis reaction from 100 ng RNA with the use of random primers extended by the T7 TNA polymerase promoter. The cRNA was synthesized during in vitro transcription (16h, 40˚C) and then purified, re-transcribed into cDNA, and biotin labeled and fragmented using the Affymetrix GeneChip WT Terminal Labeling and Hybridization kit (Affymetrix, Santa Clara, CA, USA). Biotin-labeled fragments of cDNA were hybridized to the microarray probes included in the Affymetrix Human Gene 2.1 ST Array-Strip (20h, 48˚C). The microarrays were then washed and stained using the Affymetrix Gen-eAtlas Fluidics Station (Affymetrix, Santa Clara, CA, USA). The Imaging Station of GeneAtlas System (ThermoFisher Scientific, MA, USA) was used to scan the array strips. Preliminary analysis of the scanned chips was conducted with the Affymetrix GeneAtlas Operating Software (Affymetrix, Santa Clara, CA, USA). The software's quality criteria were applied to verify the quality of the gene expression data.
Microarray data analysis
The obtained CEL files were further analyzed using the R statistical language and Bioconductor package, including the relevant Bioconductor libraries. For the normalization, background correction, and calculation of the expression values of the examined genes, the Robust Multiarray Average (RMA) normalization algorithm implemented in the "Affy" library was applied [23]. A complete gene data table involving normalized gene expression values, gene symbols, gene names and Entrez IDs was generated on the basis of assigned biological annotations taken from the "pd.hugene.2.1.st" library. For the expression and statistical assessment, the linear models for microarray data included in the "limma" library was applied [24]. The established cut-off criteria were based on both differences in expression fold change (FC) greater than abs. 2 and an adjusted p value of 0.05. Genes fulfilling the aforementioned selection criteria were considered to be significantly different and therefore were subjected to the final analysis, described below in section 2.5.
Raw data files and a technical description were also deposited in the Gene Expression Omnibus (GEO) repository at the National Center for Biotechnology Information (http:// www.ncbi.nlm.nih.gov/geo/) under the GEO accession number: GSE108035.
Assignment of differentially expressed gene to relevant gene ontology (GO) terms
The set of significantly differentially expressed genes were subjected to functional annotation and clusterization using the Database for Annotation, Visualization and Integrated Discovery (DAVID) web-based bioinformatics tool, and they also underwent the Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis [25]. Gene symbols describing differentially expressed genes were uploaded to DAVID through the "RDAVIDWebService" BioConductor library [26], where classification of significantly enriched ontological terms associated with the biological processes involved in regulating DNA damage and p53 signalling pathways was performed. The p values of specific GO terms were corrected using the Benjamini-Hochberg false discovery rate method and are reported as adj p-values. All results were visualized using the GOplot library [27]. Differentially expressed genes linked to specific GO terms were subjected to a hierarchical clusterization algorithm leading to their visualization as heatmaps. Finally, differentially expressed genes were also mapped to the signalling pathways of KEGG using the Pathview library, where expression folds of differentially expressed genes belonging to cell cycle and p53 signalling pathways were assigned to a predetermined colour scale in which upregulated and down-regulated genes are shown, respectively, in green and red color for the relevant experimental groups.
RT-qPCR evaluation
Real Time-PCR reactions were performed using the PrimePCR TM Custom Plates (Bio-Rad, CA, USA) and the specific synthesized primers for each gene: BTG2, RGCC, CDK2, CDKN2A and PMAIP1. cDNA samples were analyzed for genes of interest and for the reference gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The expression level for each target gene was calculated as -2 ΔΔct . The reaction was performed in triplicate for the gene of interest.
Generation of chondrocyte-like cells from human induced pluripotent stem cells
GPCCi001-A cells were generated from primary human dermal fibroblasts through a reprogramming process using the Stemcca-tetO lentiviral vector coding transcription factors responsible for pluripotency (Oct 4, Sox2, c-Myc and Klf4). GPCCi001-A cells express markers characteristic of pluripotent SCs and possess the capacity to form embryoid bodies with the derivatives of three primary germ layers [17] (Fig 1). HiPSCs cells were subjected to chondrogenic differentiation via monolayer culture based on a previously established protocol (DIRECT) [18]. Differentiated chondrocyte-like cells revealed features characteristic of chondrocytes such as type II collagen (Fig 1). To assess global gene expression, a microarray analysis using bioinformatics tools was performed. The global gene expression profile of ChiPS were compared to the hiPSCs (GPCCi001-A cell line) and to two cell lines of mature chondrocytes: HC-402-05a (Merck Millipore, Germany) (HC) and primary articular cartilage chondrocytes (ACC).
Microarray gene expression profiling: Signalling pathways involving DNA damage and p53 are significantly upregulated in chondrocyte-like cells differentiated from hiPSCs
Affymetrix Human Gene 2.1 ST Array Strips were used to analyse whole gene expression and to perform a complete comparison of ChiPS, HC, ACC, and GPCCi001-A transcriptome profiles (40,716 different transcripts). A pairwise scatter plot analysis was performed to determine the general profiles of the whole gene expression in the ChiPS, HC, ACC and GPCCi001-A experimental groups; this pairwise analysis revealed the patterns of upregulated and downregulated ChiPS-specific genes compared to the control HC, ACC and GPCCi001-A cells (Panel A, B, and C in Fig 2). Each colour dot corresponds to one differentially-expressed transcript, with red and green dots representing, respectively, the downregulated and upregulated genes Differentiation leads to stress-induced activation of DDR in hiPSC-derived chondrocyte-like cells Differentially expressed gene sets from the comparisons (ChiPS vs HC, ACC, GPCCi001-A cells) were assigned to significantly enriched GO terms associated with biological processes engaged in DNA damage response (DDR), as follows: "signal transduction in response to DNA damage"; "mitotic DNA damage checkpoint"; "G1 DNA damage checkpoint"; "signal transduction involved in DNA integrity checkpoint"; "signal transduction involved in mitotic G1 DNA damage checkpoint"; "intracellular signal transduction involved in G1 DNA damage checkpoint"; "signal transduction involved in mitotic DNA damage checkpoint"; "DDR to signal transduction by p53 class mediator resulting in cell cycle arrest"; "DDR signal transduction by p53 class mediator"; "signal transduction by p53 class mediator"; and "regulation of signal transduction by p53 class mediator" ( Table 1). The GOplot library was used to visualize these results [27]. The general characteristics of enriched GO term are presented as bubble plots with the negative logarithms of the adj p-values on the y-axes and Z-score values on the x-axes. The Z-score, which indicates stimulated (positive value) or inhibited (negative value) processes, was calculated automatically using the GOplot library. In the ChiPS group, gene sets assigned to specific ontological groups demonstrated a notable increase in gene expression, as reflected by positive Z-score values (Panel A in Figs 3 and 4) The logarithms of the FC values for the differentially expressed genes corresponding to GO terms are shown in Panel B in Fig 4. Each blue circle presents an upregulated gene from the specific GO terms. In many cases, single genes are often assigned to many ontological terms and therefore the complex relationships between genes and GO terms were mapped using circos plots including logFC values and gene symbols (Panel C in Figs 3 and 4).
A pathway analysis was also performed for the differentially expressed genes based on the KEGG database. This analysis allowed us to determine the biological pathways "p53" and "cell cycle", which involve a significant enrichment of differentially expressed genes in the ChiPS group (p<0.05). Differentially expressed genes belonging to these pathways were assigned to a predetermined colour scale, which was subsequently imposed on the gene/protein symbol field (Fig 5).
Based on the circos plots that revealed the most upregulated genes in the ChiPS vs HC, ACC and GPCCi001-A groups, we selected the most upregulated genes (BTG2, RGCC, CDK2, CDKN2A and PMAIP1) for further RT-qPCR evaluation. All of these genes are involved in DDR signalling pathways (Panels A, B, and C in Fig 6). Expression of genes engaged in activating p53 and DNA damage signalling pathways was significantly higher in the ChiPS compared to the control groups (HC, ACC and GPCCi001-A cells).
Significantly differentially expressed genes from each GO term were also subjected to a hierarchical clusterization algorithm and visualized as heatmaps. Arbitrary signal intensities from the most up-or down-regulated genes are presented by colours (green = higher expression, red = lower expression) and described by their symbols. Log2 signal intensity values for any single gene were resized to the row Z-Score scale (Panels A, B, and C in S1 and S2 Figs).
Discussion
Our group previously established a novel protocol to obtain chondrocyte-like cells from hiPSCs derived from primary human dermal fibroblasts via a reprogramming process [17,18]. In the current study, analysed the global gene expression of ChiPS, HC, ACC and GPCCi001-A cells, with a focus on the stress-induced signalling pathways activated during the chondrogenic process in vitro. The main aim of this study was to assess how the differentiation process influences key genes engaged in DNA damage and p53 signalling pathways, comparing the results obtained for the ChiPS to those obtained for adult chondrocytes and hiPSCs. Our findings show that chondrogenic differentiation in vitro provokes an important stressinduced response for these cells, leading to DNA breaks and activation of major components of the p53 signalling pathway. Published data on chondrogenic differentiation of hiPSCs and the effect of this process on DDR mechanisms are scant. For this reason, we discuss our findings in the context of the available data regarding the regulation of p53 and DDR processes, Table 1. Significantly differentiated genes from all experimental groups assigned to specific GO terms involved in DNA damage and p53 signalling pathways. Lin et al. reported that rapid downregulation of Nanog-one of the key pluripotent transcription factors-during the differentiation of mouse embryonic stem cells (ESCs) correlates with the induction and Ser 315 phosphorylation of p53. The increased activity of p53 helps to maintain the genetic stability of mouse ESCs, which-as a result of the differentiation process in vitro-can give rise to other cell types. ESC-derived mouse cells, in contrast to undifferentiated mouse ESCs, have the capacity to undergo p53-dependent cell-cycle arrest and apoptosis [28]. In that study, murine MSCs with p53 knockout presented genomic instability involving increased expression of c-Myc and the time required by these cells to differentiate into adipocytes and osteocytes was significantly reduced [29].
ChiPS vs HC ChiPS vs ACC ChiPS vs GPCCi001-
P53 interacts with key regulators of neurogenesis to redirect mouse neural SCs (NSCs) to differentiate as an alternative to cell death during stress. When this occurs, the histone H3 lysine 27-specific demethylase JMJD3 induces p53 stabilization and nuclear accumulation through ARF-dependent mechanisms [30]. The available evidence [31] indicates that p53-a key factor in the regulation of neuronal differentiation of NSCs in organotypic cultures of Differentiation leads to stress-induced activation of DDR in hiPSC-derived chondrocyte-like cells mouse hippocampus and the p53 signalling pathway-are probably regulated by proteins associated with cell cycle regulation (e.g., Pim-1 and Phb1) [31]. Another group [32] reported that p53 is strongly involved in embryonic development by mediating the p38 mitogen-activated protein kinase (p38 MAPK) pathway during early differentiation of mouse ESCs into mesodermal and neural lineages. The p53 has the capacity to suppress expression of Nanog [32].
In the nucleus of hESCS, p53 expression is scarcely detectable due to HDM2 and TRIM24 negative regulators, which leads to p53 degradation. After differentiation has been induced, lysine 373 of p53 is acetylated by CBP/p300, thus disrupting the interaction between p53 and negative regulators, which ultimately stabilizes and activates p53. Activated p53 induces expression of p21-the cell cycle regulator-which is responsible for the accumulation of differentiated hESCs in the gap (G1) phase of the cell cycle. In addition, p53 promotes miR-34a and miR-145, which inhibit expression of the SC factors Oct4, Klf4, Lin28A and Sox2 in differentiated hESCs [33].
A comprehensive study involving the role of p53 in early hESC differentiation was carried out by Akdemir and collaborators [34]. Those authors performed genome-wide profiling of protein interactions with chromatin and the intersection with global gene expression analysis to investigate potential networks of transcription and epigenetic regulators of pluripotency and self-renewal of hESCs. Based on the genome-wide profiling, they showed that differentiation-activated p53 targets are bound by core pluripotency factors-that is, Oct4 and Nanog at chromatin enriched in H3K27me3 and H3K4me3. The conclusion of that study was that p53 mediates the expression of specific developmental genes [34].
Both p53 and p73 have been shown to play an important role in regulating apoptosis in mouse ESCs in response to DNA damage or differentiation. Importantly, neither p53 nor its target p73 is required for G2/M arrest. This indicates that the major role of p53 in maintaining genetic stability of mouse ESCs is achieved by eliminating damaged cells [35]. In our study, we observed a notable regulation of genes engaged in the p53 signalling pathway and the cell cycle in ChiPS, in which cell cycle phases (such as mitotic and G1 DNA damage checkpoints) were activated, with cell cycle arrest in response to chondrogenic differentiation in vitro.
The p53, which has been described as "a guardian of the genome", is a key factor responsible for the safety of adult cells because it induces cell-cycle arrest and apoptosis during stress. Moreover, p53 plays an important role in pluripotent SC differentiation and in reprogramming, leading to the generation of hiPSCs. DNA damage strongly activates the p53 signalling pathway, thereby inhibiting pluripotent transcription factors while promoting differentiation. Pluripotent SCs are capable of transforming into malignant teratomas, carcinomas or cancer SCs when p53 is not controlled. On the other hand, p53 has the ability to effective inhibit cancer by suppressing pluripotent SC-mechanisms [36,37]. For these reasons, it is essential to carefully assess p53-related activities in pluripotent SC-based research and applications. Although the aforementioned studies concerning activation of p53 are highly informative, they do not, unfortunately, provide unambiguous data that would allow us to reach definitive conclusions about stress-induced DDR mechanisms in SCs undergoing differentiation process in vitro. Saretzki et al. (2008) provided valuable data on the repression of stress defence mechanisms during spontaneous differentiation of hESCs. Those authors discovered that the differentiation For the evaluation, we selected the most highly expressed genes involved in both DNA damage and p53 signalling pathways based on the previously created circos plot GO terms. The top panel represents normalized ChiPS fold changes (FC) of selected genes based on microarray data (A). The bottom panel represents microarray data assessed by RT-qPCR technique. The graph represents means ± SD from three replicates per group (B). The table shows normalized FC and p-values of selected genes from the microarray and RT-qPCR analyses (C). https://doi.org/10.1371/journal.pone.0198079.g006 Differentiation leads to stress-induced activation of DDR in hiPSC-derived chondrocyte-like cells process in vitro leads to a notable downregulation of telomerase activity due to deacetylation of histones H3 and H4 at the hTERT promoter as well as deacetylation of histone H3 at the hTR promoter. Those authors also reported a higher level of mitochondrial superoxide production and cellular levels of ROS during differentiation. Nevertheless, in that study they did not observe an increased expression of major antioxidant genes. They did, however, find evidence indicating an accumulation of DNA damage in hESCs undergoing differentiation in vitro and a significant decrease in the expression of genes involved in the main DNA repair mechanisms. Those findings suggest that hESC-derived cells are less proficient than hESCs in protecting their genome [15], an observation that is partially confirmed by our results, particularly the strong correlation between DNA damage and p53 activation in hiPSCs during chondrogenic differentiation in vitro.
Conclusions
This study evaluated chondrocyte-like cells (ChiPS) derived from the GPCCi001-A cell line via a chondrogenic process. We found that the chondrogenic process in vitro is controlled by p53, which is mainly responsible for diminishing pluripotency factors and for preventing hiPSCderived cells to backslide into the pluripotency state. Our findings also demonstrate that hiPSCs undergoing chondrogenic differentiation present significantly higher levels of DNA damage, which may be ascribed to differentiation-associated stress. We found that numerous genes engaged in cell cycle arrest and checkpoint were significantly up-regulated in the ChiPS as a response to accumulated DNA damage and activation of the p53 signalling pathway.
Although the precise nature of the cells obtained through chondrogenic differentiation in vitro of hiPSCs remains an open question, the ChiPS assessed in this study present features that are characteristic of cells involved in chondrogenesis. However, the use of such cells in regenerative medicine could present a risk given that ChiPS may be less proficient in maintaining their genetic stability; this characteristic, if confirmed, may constitute a serious drawback for the future use of these cells in clinical practice. , and GPCCi001-A (C) from the specific GO terms. The significant GO terms were as follows: "signal transduction in response to DNA damage"; "mitotic DNA damage checkpoint"; "G1 DNA damage checkpoint"; "signal transduction involved in DNA integrity checkpoint"; "signal transduction involved in mitotic G1 DNA damage checkpoint"; "intracellular signal transduction involved in G1 DNA damage checkpoint"; and "signal transduction involved in DNA damage checkpoint". Arbitrary signal intensity obtained from the microarray analysis is represented by the appropriate colours (green = higher expression; red = lower expression). Log2 signal intensity values for each gene were resized to row Z-score scales. Genes belonging to the relevant GO term are described by their symbols (A,B,C). , and GPCCi001-A (C) from the specific GO terms. The significant GO terms were as follows: "DNA damage response signal transduction by p53 class mediator resulting in cell cycle arrest"; "DNA damage response signal transduction by p53 class mediator"; "signal transduction by p53 class mediator"; and "regulation of signal transduction by p53 class mediator". Arbitrary signal intensity obtained from the microarray analysis is represented by the appropriate colours (green = higher expression; red = lower expression). Log2 signal intensity values for each gene were resized to row Z-score scales. Genes belonging to the relevant GO term are described by their symbols (A,B,C). | 2018-06-18T01:04:47.679Z | 2018-06-04T00:00:00.000 | {
"year": 2018,
"sha1": "83356083ecc9449297db8a08da4e16f3b63ea234",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1371/journal.pone.0198079",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "83356083ecc9449297db8a08da4e16f3b63ea234",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Chemistry"
]
} |
271479003 | pes2o/s2orc | v3-fos-license | Evaluation of soil nutrients and berry quality characteristics of Cabernet Gernischet (Vitis vinifera L.) vineyards in the eastern foothills of the Helan Mountains, China
Soil is the basis of the existence of fruit tree and soil nutrients plays a crucial role in plant growth and berry quality. To investigate the characteristics and interrelationships between soil nutrients and berry quality in Cabernet Gernischet vineyards, this study focused on seven representative vineyards in the eastern foothills of the Helan Mountains. Fifteen soil physicochemical factors and 10 berry quality factors were measured, followed by variation analysis, correlation analysis, multiple linear regression (MLR), partial-least squares regression (PLSR), principal component analysis (PCA), and systematic cluster analysis. We identified the main soil nutrient indicators influencing berry quality and developed linear regression equations. Utilizing PCA, a comprehensive evaluation model for berry quality was constructed, which enabled the calculation and ranking of integrated berry quality scores. The results indicated that soil nutrients in the vineyards of the eastern foothills of the Helan Mountains are relatively deficient and alkaline. The coefficient of variation for soil nutrient factors ranged from 3.19 to 118.08% and for berry quality factors 2.41–26.37%. Correlation analysis revealed varying degrees of correlation between soil nutrient indicators and fruit quality indicators. PCA extracted four principal components with a cumulative contribution rate of 91.506%. Based on the scores of these components and their corresponding weights, a comprehensive model for evaluating the quality of Cabernet Gernischet berries was established. The vineyards were ranked from the highest to the lowest combined scores as Zhenbeibu (ZBB), Yuquanying (YQY), Dawukou (DWK), Beihaizi (BHZ), Shuxin (SX), Huangyangtan (HYT), and Hongde (HD). These findings provide insights into soil nutrient management and comprehensive quality assessment of vineyards in the eastern foothills of the Helan Mountains. In conclusion, this study offers a theoretical foundation for vineyard managers to enhance grape berries quality through soil nutrient management. This will aid in the diagnosis of vineyard soil nutrition and the efficient use of fertilizers, with critical practical and theoretical implications for the meticulous management of vineyards and the production of high-quality wines.
Introduction
The eastern foothills of the Helan Mountains in Ningxia, situated between 37°43′N -39°23′N latitude, 105°45′E -106°47′E longitude, and at an altitude of 1100 -1120 m, form an open area flanked by the Helan Mountain alluvial fan and the Yellow River alluvial plain.This area spans 2410.7 km 2 , approximately 200 km in length from north to south and 5 -30 km in width from east to west (Qi et al., 2019).This region experiences an arid climate within the temperate zone, as characterized by drought conditions, scant rainfall, abundant sunlight, significant day-night temperature variations, deep soil, and favorable conditions for irrigation from the Yellow River.Such distinctive geographical and natural attributes render it one of the premier ecological locales for wine grape cultivation in China (Wang et al., 2015).Cabernet Gernischt, a red wine grape variety renowned for its distinctive herbaceous flavor, thrives in the sandy soils of this region and exhibits drought tolerance, making it a prominent cultivated variety in the research region.
Soil serves as the foundation for fruit tree growth, and its nutrient status directly influences tree development, fruit yield, quality, and, ultimately, the economic profitability of orchards (Reeve et al., 2005).Extensive research conducted both domestically and internationally has investigated the relationship between vineyard soil nutrients and berry quality, revealing the pivotal role of soil nutrients in shaping and enhancing berry quality (Qi et al., 2019;Li Q. J. et al., 2024).However, many of these studies have primarily focused on simple correlation analyses between berry quality and soil nutrient factors, failing to adequately capture the complex relationship.Therefore, there is a need to delve deeper into this complex relationship through multivariate statistical analysis.
Partial least-squares regression (PLSR) emerges as a robust method for modeling multiple linear regression, which is capable of analyzing datasets with numerous, noisy, collinear, or incomplete variables across both dependent and independent variables (Wold et al., 2001).Thus, PLSR finds widespread application in analyzing the correlation between soil nutrients and berry quality (Zhang et al., 2018).Using the insights from such analyses, vineyard managers can identify the primary soil nutrient factors influencing berry quality, thereby facilitating improvements in the nutrient status and improving berry quality.
Various methods exist for assessing berry quality, including factor analysis (Feng et al., 2016), The technique for order preference by similarity to the ideal solution (TOPSIS) model (Hao et al., 2022), and principal component analysis (PCA) (Cao et al., 2014;Shi et al., 2015;Zheng et al., 2022), among others.As a commonly employed multidimensional data analysis method, PCA is widely applied in practical settings.In this study, seven representative vineyards served as the study sample.Fifteen soil nutrient indices and ten berry quality indices were assessed, followed by correlational analysis, PLSR, and MLR to elucidate the intrinsic relationship between soil nutrients and berry quality.In addition, PCA facilitated a comprehensive analysis and ranking of berry quality, offering valuable insights and guidance for local vineyard management.
Experimental materials
We investigated the vineyards of Cabernet Gernischet situated in the wine-producing region of the eastern foothills of the Helan Mountains.Seven representative Cabernet Sauvignon vineyards were chosen from five sub-regions, including Shizuishan (Dawukou; DWK), Yinchuan (Zhenbeibu; ZBB), Yongning (Yuquanying [YQY] and Huangyangtan [HYT]), Qingtongxia (Shuxin; SX), and Hongsibu (Beihaizi [BHZ] and Hongde [HD]) (Figure 1).These vineyards are equipped with water and fertilizer systems for irrigation and fertilization.Drip irrigation is supplied via water sources from the Yellow River, which traverses the eastern part of the production area, as well as meltwater from the Helan Mountains.Fertilization practices aligned with the annual growth requirements of the grapes, whereas natural grass management models were implemented for field soil management.Grapevines were buried in the soil to protect them from severe winter cold.Refer to Table 1 for detailed information on each vineyard.
Sampling methods
Three soil sampling units measuring 10 m × 10 m were randomly selected from within each vineyard.Within each unit, an S-type soil sampling approach was employed, avoiding areas with fertilized ditches and drip irrigation zones.Fifteen soil sampling points were designated approximately 30-40 cm from the grape lines, and soil samples were collected using a tubular soil drill up to a depth of 40 cm from the soil surface.These samples were thoroughly mixed, and approximately 1 kg of soil was obtained by using the quartering method.Each soil sample was then placed into a self-sealing bag, labeled, and marked with the sampling site using a pencil, which served as the final soil sample for each unit.Three replicates were performed in each vineyard.All soil samples were subsequently placed in a freezer and transported to the laboratory.Approximately 100 g of fresh soil was selected from each sample for immediate identification of nitrate and ammonium nitrogen contents.The remaining soil samples were air-dried naturally in a shady and cool environment.Once dried, the soil samples were finely ground with a wooden stick and sifted through a sieve with a 1-mm aperture.After thorough mixing, two samples were selected using the quartering method for the determination of remaining soil chemical indicators.
Berry sampling was conducted during the commercial ripening period of the grapes within the same sampling area as the soil sampling.Fifty healthy clusters with consistent growth on both sides of the vine were randomly selected in a sampling unit.From each cluster, six grapes were harvested from different parts using miniature scissors.A total of 300 grape berries were collected from each sampling area and placed in individual self-sealing bags.Three bags were collected from each vineyard and transported to the laboratory freezer.From each bag, 100 grape berries were randomly selected for weighing, whereas another 100 berries were stored in a −80°C ultra-low temperature refrigerator for The location of the sampling sites.the measurement of the amount of anthocyanins, flavonols, and tannins.The remaining 100 grape berries were manually juiced, and the physicochemical factors of the grape juice were determined.Mature grapes from the same sampling unit were hand-picked at the harvest date.Approximately 20 kg of healthy grapes per unit were obtained for small-scale wine production.The vinification was performed with reference to a previous method (Duan et al., 2022a).The winemaking was performed in triplicate for each vineyard.
Berry weight was measured with an electronic balance (JY10002, Shanghai Hengping Instrument Co., Ltd, Shanghai, China).Longitudinal and transverse diameters were assessed with an electronic digital display vernier caliper (IP54, Shanghai Meinaite Industrial Co., LTD, Shanghai, China).Soluble solids of grape juice were measured using a handheld digital refractometer (PAL-1, Atago, Japan).Reducing sugar content was determined through Fehling's solution-titration (Ma et al., 2017).Titratable acidity and grape juice pH were measured by using the Easyplus Titration System (Mettler Toledo, Columbus, OH, USA) and PHS-3C pH meter (Precision Science Instruments Co., Ltd., Shanghai, China), respectively.Tannins, anthocyanins, and flavonols were extracted and determined based on previous literature (Li Y. S. et al., 2024).Anthocyanins, flavonols and tannins are classified according to their substituents at the 3' and 5' positions of the B-ring of flavonoids, and the types of acylation at the 3' position of the C-ring (Duan et al., 2022a).The basic physicochemical indicators of wine were determined by a Fourier-transformed infrared spectroscopy automatic wine analyzer (Lyza 5000 Wine, Anton Paar GMBH, Austria).The sensory evaluation of biological quality was conducted in a single day in a wine sensory analysis laboratory located in the College of Enology of Northwestern A&F University in Yangling, China.A 13-member panel of experienced professionals was appointed.The panelists were asked to taste the wines and perform descriptive analyses on each sample in a blind tasting.Each wine sample was scored on a 100-point scale based on a wine sensory rating scale.
Screening of main soil nutrient indices affecting grape berry quality and constructing regression equations
The PLSR method was primarily employed to screen soil nutrient factors, following the approach outlined by Wang et al. (2017).Using SPSS 25.0 software, the soil nutrient indices were designated as independent variables and each berry quality index was the dependent variable for PLSR analysis.Standardized regression equations were derived from the output results of the software by using the weight results of potential factors.Subsequently, based on the magnitude of standardized regression coefficients and professional expertise, the indices were screened.Multiple linear regression (input method) was then conducted using SPSS 25.0 to establish regression equations and to perform significance tests.
Statistical analyses
Original data were processed using Excel 2016 software (Microsoft, WA, America).One-way analysis of variance, PLSR, and multiple linear regression were conducted using SPSS 25.0 software (IBM, Chicago, IL, USA).Tukey's multiple-range tests at a significance level of P< 0.05 were performed using SPSS 25.0.Origin 2023 software (OriginLab Corporation, Massachusetts, America) was utilized for correlation mapping and cluster analysis.
Soil nutrients profile
The statistical analysis results of soil nutrients in Cabernet Gernischet vineyards are summarized in Table 2. Macroelement analysis revealed significant disparities among vineyards.In the DWK vineyard, total nitrogen, total potassium, total phosphorus, and nitrate nitrogen levels were notably higher compared to those in other vineyards.YQY and HD vineyards exhibited significantly higher ammonium nitrogen levels, whereas ZBB demonstrated significantly elevated available potassium levels.Exchangeable magnesium levels of ZBB were significantly higher than for the other vineyards.Trace element distribution also showcased disparities, with DWK showing significantly higher available iron levels than other vineyards, and ZBB exhibiting notably higher available manganese levels.Conversely, the HD vineyard displayed significantly lower available copper and zinc contents when compared to the other vineyards.In addition, the organic matter content of different vineyards varied significantly.DWK had the highest organic matter content, that is, significantly higher than that for the other vineyards and 11.61-times greater than HD.Across all vineyards, soil exhibited alkalinity, with pH levels reaching 8.54 in the HYT vineyard, which was significantly higher than for the other vineyards.The DWK vineyard exhibited relatively favorable soil nutrient content, whereas SX, BHZ, and HD vineyards displayed comparatively impoverished soil nutrient content.When compared with the grading standard of soil nutrients in vineyards (Wang, 2016), the vineyard soil in the eastern foothills of the Helan Mountains was relatively poor, with low availability of most of the soil nutrient elements.The elevated pH may be one of the main factors limiting soil nutrient activation and nutrient retention in this region.
The variation of soil nutrient factors across Cabernet Gernischet vineyards is depicted in Table 3. Variations in different soil nutrient factors were found to vary significantly, with organic matter and nitrate exhibiting coefficients of variation > 100% across different vineyards, indicating a strong variability intensity.In contrast, total potassium, exchangeable magnesium, and soil pH have lowest coefficients of variation (< 10%), indicating weak variability.The residual soil nutrient index showed a coefficient of variation of 10% -100%, indicating a middle range of intensity.
Correlation of soil nutrient indices
The correlations among soil nutrients in the Cabernet Gernischet vineyards are depicted in Figure 2. The correlations between total nitrogen and total potassium, organic matter and available iron, respectively, were at a significant level, and, for the latter two, it was an extremely significant level.There was a significant correlation between total phosphorus and available phosphorus, exchangeable calcium, and exchangeable magnesium.Total potassium level was significantly correlated with the organic matter and available iron.There was a significant correlation between organic matter and available iron.The correlation between available phosphorus and exchangeable calcium and exchangeable magnesium was at a significant and extremely significant level, respectively.The correlation between exchangeable calcium and exchangeable magnesium also reached a highly significant level.All correlations among the above mentioned nutrients were positively correlated.The correlations between other soil nutrients did not reach significant levels.Overall, soil nutrient indicators were mainly positively correlated, with the synergistic effects being greater than the antagonistic ones.
Berry quality profile
The results of grape quality index determination for Cabernet Gernischet are presented in Table 4. ZBB showed significantly higher berry weights and transverse diameters than did the other vineyards.HYT, SX, and BHZ showed lower longitudinal and transverse diameters than those shown by the remaining vineyards.The contents of reducing sugars and soluble solids in HD were significantly higher than those in the other vineyards, the contents of soluble solids and reducing sugars in DWK were significantly lower than those in the other vineyards, and the content of titratable acids in YQY was significantly lower than those in the other vineyards.Grape juice from berries in YQY showed the highest pH and that from berries in HD showed the lowest pH.In terms of the flavonoid content, ZBB and SX showed significantly higher tannin content than the other vineyards, while BHZ showed the highest anthocyanin and flavonol levels.
Table 5 displays variations in the berry quality index for the seven vineyards.Different indicators varied widely among the vineyards, with higher degrees of variations observed in the tannin levels and berry weight, both of which were > 20% (Figure 3), followed by flavonoid levels, titratable acidity, and anthocyanin levels, all exhibiting variation coefficients > 10%.However, their coefficient of variation is< 100%, hence they belong to the medium level of variation.The coefficient of variation of longitudinal diameter, transverse diameter, soluble solid content, reduced sugar content, and berry juice pH was< 10%, which belonged to weak variation.
Flavonoid (anthocyanins, flavonols, and tannins) profiles of grape berries
The flavonoids in grape berries mainly contain three types of substances, anthocyanins, flavonols, and condensed tannins.In this study, 19 anthocyanins, nine flavonols, and 4 tannin units were Of the nine flavonol monomers, myricetin-glucoside (My-glc), quercetin-glucuronide (Qu-glcU), and quercetin-glucoside (Quglc) had higher levels.Kaempferol-glucoside (Ka-glc) in BHZ was significantly higher than that in the other vineyards.There was no significant variation in Isorhamnetin-3-O-glucoside (Isorh-glc) content across different vineyards.The (-)-epicatechin (EC unit) showed the relatively highest content of the four monomers of tannins, but it was significantly lower in DWK than in the other vineyards.Conversely, the (-)-epicatechin-3-O-gallate (ECG unit) in DWK was significantly higher than the others.
Correlation of berry quality indicators
The results of the correlation analysis of berry quality indicators are presented in Figure 4.The berry weight was significantly positively correlated with the longitudinal and transverse diameters.The longitudinal diameter was significantly positively correlated with the transverse diameter.
In addition, soluble solids and reduced sugar contents with correlation coefficients >0.90 were significantly positively correlated with each other.
Assessment of the enological quality of each berry sample
The berry weight and volume of DWK were moderate.Moderate to strong acidity was attributable to northerly location, lack of heat, fewer soluble solids and reduced sugar content.There was also minor flavonoid, especially tannin, content.These properties make DWK's grape berries suitable for the production of low-alcohol fresh wines.The vineyards of ZBB and YQY are located in the heart of the wine-producing region at the eastern foothills of the Helan Mountains, close to the mountain's abundant sources of heat.Not only are the berries large and heavy, but they also contain more amount of soluble solids and reducing sugars, have appropriate acidity, tannins, anthocyanins, flavonols and other flavoring substances, and have the potential to make superiorquality wines.Although the vineyard is rich in calories, the grape berries of HYT obtain higher soluble solids and reduced sugars, and the resulting wine has a higher alcohol content.However, it has a relatively low acid content and tannin content, which easily creates an imbalance in the balance triangle of alcohol, acid, and tannin.SX and BHZ berries are medium-sized, moderately soluble solids, low in sugar and tannins, and have a degree of acidity to produce balanced wines.HD's grape berries are rich in sugar, and the resulting wines are elevated in alcohol content.However, the lack of anthocyanins and tannins can produce wines that are dull and unbalanced in flavor.
Correlational analysis of soil nutrient and berry quality indicators
To determine the correlation between soil nutrient indicators and the berry quality indicators, a correlation analysis was performed (Figure 3).Berry weight, longitudinal diameter, and transverse diameter were positively correlated with all soil nutrient indicators, except for pH, and the correlation with available manganese reached significant or extremely significant levels.Soluble solids and reduced sugar displayed a high degree of agreement in relation to soil nutrient indicators.They were only positively correlated with ammonium nitrogen, exchangeable calcium, and soil pH, but negatively correlated with other soil nutrient indicators.The relationship between soluble solids and total nitrogen, organic matter, and available iron was significant or highly significant.The same was true for reducing sugar.Flavonol was negatively correlated with most of the soil nutrient indicators and significantly negatively correlated with total nitrogen and available iron.In addition, there was a significant positive correlation between the berry juice pH and available zinc.None of the other interactions reached significant levels.
Correlation analysis of soil nutrient and grape berry flavonoids
The results of the correlation analysis between soil nutrient indicators and grape berry flavonoids are shown in Figure 3.Among the anthocyanin monomers, cyanidin-3-O-glucoside (Cy) and TP, Cy and ECa, Cy and EMg, petunidin-3-O-glucoside (Pt) and TP, Pt and AP, peonidin-3-O-glucoside (Pn) and TP, delphinidin-3-(6-Ocoumaroyl) glucoside (cis isomer) (cDp) and ACu, and cyanidin-(6-O-coumaryoyl) glucoside (trans isomer) (tCy) and soil pH were significantly negatively correlated, while the correlation between Cy and ECa reached a highly significant level.Delphinidin-3-Oglucoside (Dp) was positively correlated with TK, peonidin-(6-Ocaffeoyl) glucoside (Pn-caf), and AZn, with the latter reaching extremely significant levels of correlation.Among the flavonoids monomers, quercetin-glucoside (Qu-glc) was negatively correlated with TN and AFe and laricitrin-3-O-rhamnose-7-Otrihydroxycinnamic (La) with TN and AMn, respectively.In tannin monomers, the EC unit was negatively correlated with OM, while the ECG unit was positively correlated with TK and OM, respectively.The correlation between other flavonoids monomer components and soil nutrient index was not significant.
Screening of main soil nutrient factors affecting berry quality of Cabernet Gernischet and the establishment of regression equations
Soil nutrient and berry quality indices are two different normal distribution aggregates.While soil nutrients affect berry quality, internal relationships between the two are complex.Twelve terms exhibited highly significant correlations among soil nutrient indicators, with correlation coefficients > 0.75 (Figure 2), thereby indicating multiple collinearities among the indicators.Thus, a simple linear analysis cannot reveal the true relationship between soil nutrients and berry quality.Hence, PLSR analysis was performed by considering the vineyard soil nutrient indicators as an independent variable and the berry quality indicators as a dependent variable to determine the multiple-linear relationship between soil nutrients and berry quality (Wold et al., 2001).The variable selection method in PLSR was conducted in accordance with the regression coefficients (Mehmood et al., 2012).Additionally, considering screened soil nutrient indicators as an independent variable and the corresponding berry quality index as a dependent variable, MLR was performed sequentially to establish the linear regression equations of berry quality and soil nutrient indicators (Table 7).The regression coefficients and symbols of the equations could reflect the important degrees and positive or negative impacts of the effects of different soil nutrient indicators on the berry quality factors (Zhang et al., 2018).Furthermore, a significance test was conducted on the established regression equations.The results were significant (Table 7), which suggested that the constructed regression equations were reliable.
As shown in Table 7, each indicator of berry quality was affected mainly by the various nutrients (except for soluble solids and reducing sugar).For example, berry weight was primarily affected by available potassium, available zinc, available iron, and available manganese.Available potassium, available zinc and available manganese had positive effects on the berry weight, available zinc had the largest effect on the berry weight, and iron had a negative effect.Notably, soluble solids and reducing sugars were affected by total nitrogen, ammonium nitrogen and available iron.Total nitrogen and ammonium nitrogen had a negative effect on both, while ammonium nitrogen had a positive effect.Among them, total nitrogen had the greatest effect on both.This effect can be attributed to the high correlation between soluble solids and reduced sugar content (Figure 4).
PCA of berry quality indices
The berry quality indices of the seven vineyards were analyzed by performing PCA (Tables 8, 9).Following the principle that the eigenvalues are > 1.0, four principal components were identified, and the cumulative contribution rate reached 91.506%.The first principal component accounted for 38.493% of the original information content, and berry weight and longitudinal and transverse diameters displayed high positive coefficients (0.794, 0.695, and 0.818, respectively).Soluble solids, reducing sugar, and flavonol contents showed high negative coefficients (−0.691, −0.730, and −0.707, respectively).The results indicated that the greater the first principal component, the higher the berry weight and longitudinal and transverse diameters and the lower the soluble solid, reducing sugar, and flavonoid contents.The second principal component accounted for 22.229% of the original information content, with high positive coefficients for berry weight, longitudinal and transverse diameters, and soluble solid, reduced sugar, and flavonol contents, with high loading values of 0.530, 0.644, 0.556, 0.639, 0.598, and 0.565, respectively.The third principal component accounted for 17.144% of the original information, and the loading value was 0.925.Tannin content had a larger negative coefficient, and the loading value was −0.612, indicating that the larger is the third principal component, the higher is the berry juice pH and the lower is the tannin content.The fourth principal component accounted for 13.650% of the original information, with large positive coefficients and loading values of 0.529 and 0.692 for tannin and anthocyanin contents, respectively, indicating that the greater the fourth principal component, the higher the is tannin and anthocyanin contents.
where F 1 -F 4 represent the fractions of the principal components of fruit quality indices for different vineyards, and Z 1 -Z 10 represents ten quality indices, including berry weight, longitudinal diameter, transverse diameter, soluble solid, reducing sugar, titratable acidity, berry juice pH, tannin, anthocyanin, and flavonol contents.
Using the variance contribution rates corresponding to the four principal components as weight coefficients, an integrated evaluation model Equation for berry quality was developed from the weighted sum of the principal component scores and the corresponding weight coefficients (Nie et al., 2019;Zhao et al., 2023): (5) The model was used to obtain comprehensive berry quality scores for the seven vineyards (Table 10).The higher the overall score, the better the overall berry quality (Wu et al., 2022).As shown in Table 10, the overall ranking of fruit quality from highest to lowest was as follows: ZBB, YQY, DWK, BHZ, SX, HYT, and HD.
In order to verify the reliability of the results of the comprehensive evaluation of grape berry quality, the wines from different vineyards were tested for physical and chemical indexes (Supplementary Table 1) and sensory evaluation (Supplementary Table 2).As shown in Supplementary Table 2, wines from YQY and ZBB were rated excellent by sensory ratings, wines from HD vineyards were rated acceptable by sensory ratings, and wines from other vineyards were rated good by sensory ratings.It can be seen that the results of the sensory assessment are consistent with the results of the integrated evaluation of the quality of the berries.
Cluster analysis
Cluster analysis was performed on the normalized berry quality indices (Figure 5).When the Euclidean distance was 4, the seven Cabernet Gernischet vineyards were divided into three categories.The first category comprised DWK and ZBB showing longer longitudinal and transverse diameters and lower soluble solid and reducing sugar contents.The second category comprised only YQY showing longer longitudinal diameter, higher berry juice pH, higher anthocyanin and flavonol content, and lower titratable acidity.The third category comprised the remaining vineyards displaying lower berry weights, shorter longitudinal and transverse diameters, and higher anthocyanin content.The results of the systematic clustering analysis were consistent with those of the integrated score ranking of PCA, which indicated the reliability and consistency of the present results.
Discussion
The unique terroir characteristics are the primary reason why the eastern foothills of the Helan Mountains wine region has become a world-renowned wine region (Yang et al., 2024).The study results revealed significant variations in soil nutrient levels among different vineyards within this wine region.It is wellestablished that diverse soil nutrient conditions can significantly impact the quality of grape berries.Therefore, this study's findings offer valuable insights into how grape berry quality responds to soil nutrient properties in the eastern foothills of the Helan Mountains, and provide a solid foundation for vineyard soil management.Soil properties play a significant role in influencing grape berry composition and the resulting wine quality (Cheng et al., 2014;Zerihun et al., 2015).Among all soil properties, soil nutrients are the most easily regulated and managed (Sun, 2022).It is crucial to investigate the relationship between soil nutrients and grape berry quality in order to provide guidance for vineyard production and management.The impact of nitrogen on grape berry quality primarily occurs through its direct influence on grape metabolism and indirect effects on vigor and yield (Poni et al., 2018).Numerous studies have indicated that high levels of nitrogen fertilization can decrease soluble solids content in berries while increasing titratable acidity content (Delgado et al., 2004;Thomidis et al., 2016).In this study, the correlation between different forms of nitrogen and fruit quality indexes was inconsistent (Figure 3), mainly due to the varying functions performed by different forms of nitrogen.Despite its significant role in plant growth and development, there is relatively limited literature regarding the effects of phosphorus on grape fruit quality.Amongst the few available experiments, only a minimal effect of phosphorus on grape berry quality has been observed (Topalovićet al., 2011;Schreiner et al., 2013;Brunetto et al., 2015).The results of this study also align with these findings as evidenced by small correlation coefficients between both total and available phosphorus and grape berry quality indicators (Figure 3).However, Duan et al. (2022b) demonstrated a positive correlation between available phosphorus and the content of reducing sugars and anthocyanins in grape berries.However, the results of this study are contradictory, which may be attributed to differences in species and regions.It is well known that grapevines require a large amount of potassium, which plays an essential role in sugar accumulation in berries (Brunetto et al., 2015).In most cases, potassium is positively correlated with soluble solids (Ramos and Romero, 2017;Tassinari et al., 2022).Although the results of this study contradict this general trend, it is worth noting that the correlation coefficient is small, similar to the findings of Ciotta et al. (2021).This discrepancy may be due to the abundance of total and available potassium in all vineyards according to vineyard soil nutrient grading standards (Wang, 2016).
In addition to nitrogen, phosphorus, and potassium, grapevines are highly in demand for calcium and magnesium.The optimal levels of exchangeable calcium were found to be 500-700 mg•kg −1 (Wang, 2016), which all the included vineyards did not meet.Therefore, timely calcium supplementation is especially necessary.According to the research results of Wang et al. (2019), the optimal concentration of calcium fertilizer was determined to be 30 kg•ha -1 in the vineyard of the eastern foothills of the Helan Mountains.Magnesium plays a crucial role as an element constituent in chlorophyll molecules that regulate photosynthesis processes (Abo El-Ezz et al., 2022).Other studies have indicated that magnesium treatment can inhibit the degradation of anthocyanins and alter the proportion of different anthocyanin monomers (Sinilal et al., 2011).In this study's results, a significant correlation between magnesium and different anthocyanin monomers was observed (Figure 6), indicating that varying magnesium content would have distinct effects on different monomers, ultimately impacting the final color of the fruit.
The presence of trace elements does indeed exert an influence on the composition of wine grapes (Mackenzie and Christy, 2005).The findings of this study corroborate this assertion.Notably, a significant correlation was observed between iron and manganese levels and certain indicators of berry quality (Figure 3).Furthermore, it was determined that iron and manganese play pivotal roles in influencing Cluster analysis of grape berry quality.various aspects of fruit quality (Table 7).However, based on the classification criteria for soil nutrients in vineyards (Wang, 2016), it was found that the vineyards located in the eastern foothills of the Helan Mountains exhibited a deficiency in iron content.This deficiency may be attributed to high lime content and pH levels within the soil (Dell'Orto et al., 2000).It is important to note that insufficient iron levels can lead to leaf chlorosis and stunted growth in grapevines, ultimately impacting grape berry quality (Bertamini and Neduchezhian, 2005).Therefore, it is imperative to adequately replenish these nutrients in order to regulate grapevine growth and enhance berry quality (Wang et al., 2019;Jiang et al., 2022;Ma et al., 2022).Soil organic matter plays a crucial role in supplying plant nutrients, improving soil fertility and buffering capacity, as well as enhancing soil physical properties (Fageria, 2012;Waibel et al., 2023).Several studies have indicated a positive relationship between soil organic matter content and soluble solids and reduced sugar contents (Wang, 2016;Li Y. S. et al., 2024), while others have shown a negative correlation between soil organic matter content and soluble solid content (Xu et al., 2013;Qi et al., 2019).In this study, a significant negative correlation was observed between soil organic matter and soluble solids as well as reduced sugar contents, which aligns with the findings of previous research (Qi et al., 2019).This may be attributed to the increasing berry volume associated with higher organic matter levels, resulting in lower soluble solids and reduced sugar contents.Similar trends have been reported in apple and orange production (Liu et al., 2023;An et al., 2024).Research has suggested that the optimal organic matter content falls within the range of 1 -3% (Aguirre et al., 2023), yet most vineyards in this region do not meet this standard.
The analysis of the correlation between soil nutrient indicators and monomers of the three main flavonoids in grape berries provides a more comprehensive research perspective.Previous studies have primarily focused on the relationship between soil properties and total anthocyanins, total flavonols, and total tannins (Zerihun et al., 2015;Qi et al., 2019;Barbagallo et al., 2021;Wang R. et al., 2021).However, limited research has been conducted on the relationship between soil properties and the individual monomers.This is partly due to the technical challenges associated with detecting monomers, as well as the complexity of data analysis.The flavonoids examined in this study share a common upstream synthesis pathway (phenylpropanoid pathway) (Gouot et al., 2019), resulting in complex internal connections among them.It is worth noting that there have been studies exploring the relationship between soil characteristics and genes related to anthocyanin synthesis, offering insights into understanding how soil nutrients affect anthocyanin monomers (Jiang et al., 2024).This approach can serve as a reference for future analyses of the relationship between soil properties and flavonol and tannin monomers.
Of course, there are certain limitations in this study.Firstly, it is important to note that the quality of wine grape berries is influenced by a combination of factors including local climate, soil composition, vineyard management techniques, and other terroir elements (Meinert, 2018;Qiao et al., 2023).Therefore, while soil nutrients play a role in grape berry quality, they are just one part of the equation.A Correlation analysis between soil nutrients and berry flavonoids.TN, TK, TP, OM, NN, AN, AP, AK, ACu, AZn, AFe, AMn, ECa, and EMg are abbreviations for total nitrogen, total potassium, total phosphorus, organic matter, nitrate nitrogen, ammonium nitrogen, available phosphorus, available potassium, available copper, available zinc, available iron, available manganese, exchangeable calcium, and exchangeable magnesium.* and ** indicated that the significance level reached P < 0.05 and P < 0.01, respectively.
comprehensive understanding of grape berry quality requires consideration of climatic conditions, management practices, and other contributing factors.Secondly, this study specifically focused on analyzing the relationship between soil nutrient indicators and fruit quality indicators without addressing additional soil characteristics such as physical, chemical and microbial properties.However, it is essential to recognize that these properties and their interactions can impact the availability of soil nutrients.As such, future studies should explore these aspects further.Thirdly, given that both soil nutrients and fruit quality can vary from year to year, it is necessary to conduct continuous studies over multiple years at fixed points in order to verify the stability and reliability of results.Finally, the uptake of soil nutrients by grapevines largely depends on water transport.How to effectively couple water and fertilizer, and enhance their utilization efficiency for different types of soils, is an important scientific question that warrants further exploration by researchers.
Conclusions
1.The soil of the vineyards in the eastern foothills of the Helan Mountains was alkaline, and most vineyards had poor soil nutrients.With the exception of the DWK vineyard, total nitrogen, total phosphorus and organic matter replenishment were required in all vineyards.In addition, all vineyards required increased application of nitrate nitrogen, ammonium nitrogen, available phosphorus, available iron, available calcium, and available magnesium, as well as reduction of the soil alkalinity.For the vineyard of DWK, ZBB, and YQY, it was needed to reduce the weight and volume of the berries so as to increase in the specific surface area of the berries, which promotes the accumulation of flavor substances in the berries and, thus, the quality of the berries.2. Multilinearity and strong coordination were observed in the soil nutrient indicators, and there was a close correlation between the soil nutrient indicators and the berry quality indicators.Ammonium nitrogen, available potassium, available copper, available zinc, available iron, and available manganese were the key soil nutrient indicators that affected the highest number of grape berry quality indicators.3. ZBB and YQY vineyards showed better fruit quality, while HD vineyards showed worse fruit quality.HD vineyards should determine earlier harvest time to reduce the reducing sugar content of the berry, increase the titratable acid content of the berry, and improve the balance of the wine's mouthfeel.4. In this study, the internal relationship between soil nutrients and fruit quality was investigated using quantitative methods.The fruit quality of different vineyards was evaluated and ranked.The findings not only provide guidance for vineyard managers to enhance fruit quality through soil improvement, but are also of significant importance in exploring distinct vineyard terroir characteristics and developing differentiated products.In addition, the study provides new insights and ideas for a comprehensive assessment of grape beery quality, efficient fertilizer utilization, classification criteria for wineproducing sub-regions, and contributes to the development of refined vineyard management schemes.
FIGURE 2
FIGURE 2 Correlation of the soil nutrient indices.* and ** indicated that the significance level reached P< 0.05 and P< 0.01, respectively.
FIGURE 3
FIGURE 3Correlation analysis of soil nutrients and berry quality indices.* and ** indicated that the significance level reached P < 0.05 and P < 0.01, respectively.
FIGURE 4 −
FIGURE 4Correlation of berry quality indices.** indicated that the significance level reached P < 0.01.
TABLE 1
Basic information about Cabernet Gernischt vineyards from seven representative vineyards.
TABLE 2
Soil nutrient content across different vineyards.
TABLE 3
Variations of soil nutrient indexes across seven vineyards of Cabernet Gernischet.
SD stands for standard deviation, CV means coefficient of variation.The same as below.
TABLE 5
Variations of berry quality indices among seven vineyards.
TABLE 4
Quality characteristics of grape berries of Cabernet Gernischet in seven vineyards.
Different small letters in the same column mean a significant difference at P < 0.05 among vineyards according to Tukey's test.
TABLE 6
Flavonoid components content of grape berries of Cabernet Gernischet across seven vineyards.
TABLE 7
Selection of soil nutrient indices and construction of regression equations affecting berry qualities.
TABLE 9
Loading matrix of the principle components of fruit quality indexes.
TABLE 8
Initial eigenvalues and contribution rates of principle.
TABLE 10
Principal component score and comprehensive evaluation ranking. | 2024-07-27T15:13:35.810Z | 2024-07-25T00:00:00.000 | {
"year": 2024,
"sha1": "3dcd3a2666a3c9ad310694957c7f5b0e4df8da71",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3389/fpls.2024.1418197",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "9020c36e60ad5c8cb13555307975b217e2127ea2",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": []
} |
232148007 | pes2o/s2orc | v3-fos-license | Deep Model Intellectual Property Protection via Deep Watermarking
Despite the tremendous success, deep neural networks are exposed to serious IP infringement risks. Given a target deep model, if the attacker knows its full information, it can be easily stolen by fine-tuning. Even if only its output is accessible, a surrogate model can be trained through student-teacher learning by generating many input-output training pairs. Therefore, deep model IP protection is important and necessary. However, it is still seriously under-researched. In this work, we propose a new model watermarking framework for protecting deep networks trained for low-level computer vision or image processing tasks. Specifically, a special task-agnostic barrier is added after the target model, which embeds a unified and invisible watermark into its outputs. When the attacker trains one surrogate model by using the input-output pairs of the barrier target model, the hidden watermark will be learned and extracted afterwards. To enable watermarks from binary bits to high-resolution images, a deep invisible watermarking mechanism is designed. By jointly training the target model and watermark embedding, the extra barrier can even be absorbed into the target model. Through extensive experiments, we demonstrate the robustness of the proposed framework, which can resist attacks with different network structures and objective functions.
INTRODUCTION
I N recent years, deep learning has revolutionized a wide variety of artificial intelligence fields such as image recognition [1], [2], medical image processing [3], [4], [5], [6], speech recognition [7], [8] and natural language processing [9], and outperforms traditional state-of-the-art methods by a large margin. Despite its success, training a good deep model is not a trivial task and often requires the dedicated design of network structures and learning strategies, a large scale high-quality labeled data and massive amount of computation resources, all of which are expensive and full of great business value. For some companies, these trained models are indeed their core competitiveness. Therefore, IP protection for deep models is not only important but also essential for their success.
However, compared to media IP protection, deep model IP protection is much more challenging because of the powerful learning capacity of deep models [10], [11]. In other words, given one specific task, numerous structures and weight combinations can obtain similar performance. This actually makes the IP infringement rather convenient. In the white-box case, where the full information including the detailed network structure and weights of the target model is known, one typical and effective attack way would be fine-tuning or pruning based on the target model on new datasets. Even in the black-box case where only the output of the target model can be accessed, the target model can still be stolen by training another surrogate model to imitate its behavior through student-teacher learning [12], [13]. Specifically, we can first generate a large scale of input-output training pairs based on the target model, and then directly train the surrogate model in a supervised manner by regarding the outputs of the target model as ground-truth labels.
Very recently, some preliminary research works [10], [11], [14], [15] emerge for deep model IP protection. They often either add a weight regularizer into the loss function to make the learned weight have some special patterns or use the predictions of a special set of indicator images as the watermarks. Though these methods work pretty well to some extent, they only consider the classification task and the white-box attacks like fine-tuning or pruning. But in real scenarios, labeling the training data for low-level computer vision or image processing tasks is much more complex and expensive than classification tasks, because their ground-truth labels should be pixel-wise precise. Examples include removing all the ribs in Chest X-ray images and the rain streaks in real rainy images. In this sense, protecting such image processing models is more valuable. Moreover, because the original raw model does not need to be provided in most application scenarios, it can be easily encrypted with traditional algorithms to resist the white-box attack. Therefore, more attention should be paid to the black-box surrogate model attack.
Motivated by this, this paper studies the IP protection problem for image processing networks that aims to resist the challenging surrogate model attack [16]. Specifically, a new model watermarking framework is proposed to add watermarks into the target model. When the attackers use a surrogate model to imitate the behavior of the watermarked model, the designed watermarking mechanism should be able to extract pre-defined watermarks out arXiv:2103.04980v1 [cs.CV] 8 Mar 2021 Input: Output: IEEEIEEEIEEEIEEE Fig. 1. The simplest watermarking mechanism by adding unified visible watermarks onto output of the target model, which will sacrifice the visual quality and usability.
from outputs of the learned surrogate model.
For better understanding, before diving into the model watermarking details, let us first discuss the simplest spatial visible watermarking mechanism shown in Figure 1. Suppose that we have a lot of input-output training pairs and we manually add a unified visible watermark template to all the outputs. Intuitively, if a surrogate model is trained on such pairs with the simple L2 loss, the learned model will learn this visible watermark into its output to get lower loss. That is to say, given one target model, if we forcibly add one unified visible watermark into all its output, it can resist the plagiarism from other surrogate models to some extent. However, the biggest limitation of this method is that the added visible watermarks will seriously degrade the visual quality and usability of the target model. Another potential threat is that attackers may use image editing tools like Photoshops to manually remove all the visible watermarks.
Inspired by the above observation, we propose a general deep invisible model watermarking framework as shown in Figure 2. Given a target model M to be protected, we denote its original input and output images as domain A and B respectively. Then a spatial invisible watermark embedder H is used to hide a unified target watermark δ into all the output images in the domain B and generate a new domain B . Different from the aforementioned simple visible watermarks, all the images in the domain B should be visually consistent to domain B. Symmetrically, given the images in domain B , another watermark extractor R will extract the watermark δ out, which is consistent to δ. To protect the IP of M, we will pack H and M into a whole for deployment. The key hypothesis here is that when the attacker uses A and B to learn a surrogate model SM, R can still extract the target watermark out from the output B of SM.
We first test the effectiveness of our framework by using traditional spatial invisible watermarking algorithms such as [17], [18] and [19], which work well for some special surrogate models but fail for most other ones. Another common limitation is that the information capacity they can hide is relatively low, e.g., tens of bits. In order to hide high capacity watermarks like logo images and achieve better robustness, a novel deep invisible watermarking system is proposed. As shown in Figure 3, it consists of two main parts: one embedding sub-network H to learn how to hide invisible watermarks into the image, and another extractor sub-network R to learn how to extract the invisible watermark out. To avoid R generating watermark for all the images no matter whether they have invisible watermarks or not, we also constrain R not to extract any watermark out if its input is a clean image. Moreover, to make R generalize better for outputs of any surrogate model, an extra adversarial training stage is incorporated. In this stage, we pick one example surrogate model and add its outputs into R s training.
For forensics, we leverage both the classic normalized correlation metric and a simple classifier to judge whether the extracted watermark is valid or not. Through extensive experiments, the proposed framework shows its strong ability in resisting the attack from surrogate models trained with different network structures like Resnet and UNet and different loss functions like L1, L2, perceptual loss, and adversarial loss. By jointly training the target model and watermark embedding, we find the functionality of H can be even absorbed into M, making M a watermarked model itself.
The contributions of this paper can be summarized in the following five aspects: • We introduce the IP protection problem for image processing networks. We hope it will help draw more attention to this seriously under-explored field and inspire more great works. • Motivated by the loss minimization property of deep networks, we innovatively propose to leverage the spatial invisible watermarking mechanism for deep model watermarking. • A novel deep invisible watermarking algorithm along with the two-stage training strategy is designed to improve the robustness and capacity of traditional spatial invisible watermarking methods. • We extend the proposed framework to multiple-watermark and self-watermark case, which further shows its strong generalization ability. • Both the classic normalized correlation metric and a watermark classifier demonstrate the strong robustness to surrogate model attack with different network structures and loss functions. Compared with the preliminary conference version [16], we have made significant improvements and extension in this manuscript. The main difference are from four aspects: 1) We rewrite the theoretical analysis Section 3.2 to make it more clear and solid; 2) We design a new classifier based watermark verification mechanism in Section 3.4.4; 3) We extend our framework to multiple watermarks and self-watermarked models in Section 3.4.5 and Section 3.4.6; 4) More corresponding experimental results and analysis are given in Section 4 and Section 5.
Media Copyright Protection
Compared to model IP protection, media copyright protection [20], [21], [22], [23], [24], [25] is a classic research field that has been studied for several decades. The most popular media copyright protection method is watermarking. For image watermarking, many different algorithms have been proposed in the past, which can be broadly categorized into two types: visible watermarks like logos, and invisible watermarks. Compared to visible watermarks, invisible watermarks are more secure and robust. They are often embedded in the original spatial domain [17], [18], [19], [26], or other image transform domains such as discrete cosine transform (DCT) domain [27], [28], discrete wavelet transform (DWT) domain [29], and discrete Fourier transform (DFT) domain [30]. However, all these traditional watermarking algorithms are often only able to hide several or tens of bits, let alone real logo images. More importantly, we find only spatial domain watermarking work to some extent for this task and all other transform domain watermarking algorithms totally fail.
In recent years, some DNN-based watermarking schemes have been proposed. For example, Zhu et al. [31] propose an autoencoder-based network architecture to realize the embedding and extracting process of watermarks. Based on it, Tancik et al. [32] further realize a camera shooting resilient watermarking scheme by adding a simulated camera shooting distortion to the noise layer. In [33], a full-size image is proposed to be used as watermark under the hiding image in image framework. Compared to these media watermarking algorithms, model watermarking is much more challenging because of the exponential search space of deep models. But we innovatively find it possible to leverage spatial invisible watermarking techniques for model protection.
Deep Model IP protection
Though IP protection for deep neural networks is still seriously under-studied, there are some recent works [10], [11], [14], [34], [35], [36], [37] that start paying attention to it. For example, based on the over-parameterized property of deep neural networks, Uchida et al. [10] propose a special weight regularizer to the objective function so that the distribution of weights can be resilient to attacks such as fine-tuning and pruning. One big limitation of this method is not task-agnostic and need to know the original network structure and parameters for retraining. Adi et al. [11] use a particular set of inputs as the indicators and let the model deliberately output specific incorrect labels, which is also known as "backdoor attack". Similar scheme is adopted in [14] and [38] for DNN watermarking on embedded devices. Recently, Fan et al. [39] find all the above backdoor based methods are fragile to ambiguity attack. To overcome this limitation, a special digital passport layer is embedded into the target model and trained in an alternative manner. But this special passport layer often needs to make some network structure changes and causes performance degradation. So in the recent work [40], Zhang et al. further introduce a new passport-aware normalization layer for model IP protection.
Despite the effectiveness of the above methods, they all focus on the classification tasks, which is different from the purpose of this paper, i.e., protecting higher commercial valued image processing models. And for the challenging surrogate model attack, they will totally fail. In this paper, we consider the surrogate model attack and study the IP protection for image processing networks. And different from the above methods, we innovatively propose to leverage spatial invisible watermarking algorithms for deep model IP protection.
Image-to-image Translation
In the deep learning era, most image processing tasks such as style transfer [41], [42], [43], deraining [44], [45], [46], and X-ray image debone [47], can be modeled as an image-to-image translation problem where both the input and output are images. Thanks to the emergence of the generative adversarial network (GAN) [48], this field has achieved significant progress recently. Isola et al. propose a general image-to-image translation framework Pix2Pix by combining adversarial training in [49], which is further improved by many following works [50], [51], [52] in terms of quality, diversity and functionality. However, all these methods need a lot of pairwise training data, which is often difficult to be collected. By introducing the cycle consistency, Zhu unpaired image-to-image translation framework CycleGAN [53].
A similar idea is also adopted in concurrent works DualGAN [54] and DiscoGAN [55]. In this paper, we mainly focus on the deep models of paired image-to-image translation, because the paired training data is much more expensive to be obtained than unpaired datasets. More importantly, there is no prior work that has ever considered the watermarking issue for such models.
METHOD
In this section, we will first define the target problem formally in Section 3.1, then provide the theoretical analysis why the spatial invisible watermarking can be utilized for model watermarking in Section 3.2. Based on the analysis, traditional spatial invisible watermarking and a new deep invisible watermarking framework will be elaborated in Section 3.3 and Section 3.4 respectively. Next, two different watermark verification methods will be considered in Section 3.4.4. Finally, we will show how to extend the proposed framework to enable multiple watermarks within one single sub-network in Section 3.4.5 and make the target model self-watermarked in Section 3.4.6.
Problem Definition
Given a target model M to protect, this paper mainly considers the surrogate model attack, where the attacker does not know the detailed network structure or weights of M but is able to access its output. This is a very common and sensible setting for real systems because most existing commercial deep models are deployed as cloud API service or encrypted executable program. In these scenarios, though the attacker cannot use common whitebox attack methods like fine-tuning and pruning, a surrogate model can be trained to imitate the target model's behavior in a teacher-student way. Specifically, the attacker collects a lot of input images {a 1 , a 2 , ..., a m } (domain A) and feeds them into M to get corresponding output images , then a surrogate model SM is trained in a supervised way by regarding M's outputs as ground truth. For effective forensics, the goal of resisting surrogate model attack is to design an effective way that is able to identify SM once it is trained with the data generated by M. And in real scenarios, it is highly possible that we cannot access the detailed information of model SM (e.g., network structure and weights) like the target model neither. This means that, even when we suspect one model is a pirated model, the only indicator we can utilize is its output unless we start the costly legal proceeding. Therefore, we need to figure out one way to extract some kinds of forensics hints from the output of SM. In this paper, we propose a new deep model watermarking mechanism so that the pre-defined watermark pattern can be extracted from SM's output.
Theoretical Pre-analysis.
In traditional watermarking algorithms, given an image I and a target watermark δ to be embed, they will first use a watermark embedder H to generate an image I which contains δ. Symmetrically, the target watermark δ can be further extracted out by the corresponding watermark extractor R. As shown in Figure 1, if each image b i ∈ B is embedded with a unified watermark δ, forming another domain B , then the objective of the surrogate model SM is to minimize the distance L between SM(a i ) and b i : Below we will first show SM can learn the watermark δ into its output, then show the inherent loss minimization property of deep network will theoretically guarantee δ to be learned by a good surrogate model SM.
In detail, for each a i we have b i = M (a i ), thus there must exist a model SM that can learn good transformation between domain A and domain B because of the existence of below equivalence.
In other words, one simplest solution of SM is to directly add δ to the output of M with a skip connection. On the other hand, because the objective of SM is calculated based on b i , the loss minimization property of deep network theoretically guarantees a good surrogate model SM should learn the unified watermark δ into its output. Otherwise, its objective loss function L cannot achieve a lower value than the above simple solution.
Based on this observation, we propose a general deep watermarking framework for image processing models shown in Figure 2. Given a target model M to protect, we add a barrier H after it and embed a unified watermark δ into its output before showing them to the end-users. In this way, the final output obtained by the attacker is actually the watermarked version, so the surrogate model SM has to be trained with the image pair After SM's training, we can leverage the corresponding watermark extractor R to extract the watermark out from the output of SM.
To ensure the watermarked output image b i is visually consistent with the original one b i , only spatial invisible watermarking algorithms are considered in this paper. Below we will introduce both traditional spatial invisible watermarking algorithm and a novel deep invisible watermarking algorithm.
Traditional Spatial Invisible Watermarking.
Additive-based embedding is the most common method used in the traditional spatial invisible watermarking scheme. Specifically, the watermark information is first spread into a sequence or block which satisfies a certain distribution, then embedded into the corresponding coefficients of the host image. This embedding procedure can be formulated by where I and I indicate the original image and embedded image respectively. α indicates the embedding intensity and C i denote the spread image block that represents bit "w i "(w i ∈ [0, 1]). In the extraction side, the watermark is determined by detecting the distribution of the corresponding coefficients. The robustness of such an algorithm is guaranteed by the spread spectrum operation. The redundancy brought by the spread spectrum makes a strong error correction ability of the watermark so that the distribution of the block will not change a lot even after image processing. However, such algorithms often have very limited embedding capacity because many extra redundant bits are needed to ensure robustness. In fact, in many application scenarios, the IP owners may want to embed some special images (e.g., logos) explicitly for convenient visual forensics. This is nearly infeasible for these algorithms. More importantly, the following experiments show that these traditional algorithms can only resist the attack from some special types of surrogate models. To enable more highcapacity watermarks and more robust resistance ability, we will propose a new deep invisible watermarking algorithm below.
Overview
As shown in Figure 3, to embed an image watermark into host images of the domain B and extract it out afterward, one embedding sub-network H and one extractor sub-network R are adopted respectively. Without sacrificing the original image quality of domain B, we require watermarked images in domain B to be visually consistent to the original images in the domain B. As adversarial networks have demonstrated their power in reducing the domain gap in many image translation tasks, we append one discriminator network D after H to further improve the image quality of domain B .
During training, we find if the extractor network R is only trained with the images of domain B , it is very easy to overfit and output the target watermark no matter whether the input images contain watermarks or not. To circumvent this problem, we also feed the images of domain A and domain B that do not contain watermarks into R and force it to output a constant blank image. In this way, R will have the real ability to extract watermarks out only when the input image has the watermark in it.
Based on the pre-analysis, when the attacker uses a surrogate model SM to imitate the target model M based on the input domain A and watermarked domain B , SM will learn the embedded watermark δ into its output. Despite higher embedding capacity, we find that the extractor sub-network R cannot extract the watermarks out from the output of the surrogate model SM neither like traditional watermarking algorithms if only trained with this initial training stage. This is because R has only observed clean watermarked images but not the watermarked images from surrogate models which may contain some unpleasant noises. To further enhance the extracting ability of R, we choose one simple surrogate network to imitate the attackers' behavior and fine-tune R on the mixed dataset of domain A, B, B , B in an extra adversarial training stage. The following experiments will show this can significantly boost the extracting ability of R. More importantly, this ability generalizes well to other types of surrogate models.
Network Structures
By default, we adopt UNet [56] as the network structure of H and SM, which has been widely used by many translation based tasks like Pix2Pix [49] and CycleGAN [53]. It performs especially well for tasks where the output image shares some common properties of input image by using multi-scale skip connections. But for the extractor sub-network R whose output is different from the input, we find CEILNet [57] works much better. It also follows an autoencoder like network structure. In details, the encoder consists of three convolutional layers, and the decoder consists of one deconvolutional layer and two convolutional layers symmetrically. In order to enhance the learning capacity, nine residual blocks are inserted between the encoder and decoder. For the discriminator D, we adopt the widely-used PatchGAN [49]. Note that except for the extractor sub-network R, we find other types of translation networks also work well in our framework, which demonstrates the strong generalization ability of our framework.
Loss Functions
The objective loss function of our method consists of two parts: the embedding loss L emd and the extracting loss L ext , i.e., where λ is the hyper parameter to balance these two loss terms. Below we will introduce the detailed formulation of L emd and L ext respectively.
Embedding Loss. To embed the watermark image into a cover image while guaranteeing the original visual quality, three different types of visual consistency loss are considered: the basic L2 loss bs , perceptual loss vgg and adversarial loss adv , i.e., Here the basic L2 loss bs is simply the pixel value difference between the input host image b i and the watermarked output image b i , N c is the total pixel number, i.e., And the perceptual loss vgg [58] is defined as the difference between the VGG feature of b i and b i : where V GG k (·) denotes the features extracted at layer k ("conv2 2" by default), and N f denotes the total feature neuron number. To further improve the visual quality and minimize the domain gap between B and B, the adversarial loss adv will let the embedding sub-network H embed watermarks better so that the discriminator D cannot differentiate its output from real watermark-free images in B.
Extracting Loss. The responsibility of the extractor sub-network R has two aspects: it should be able to extract the target watermark out for watermarked images from B and instead output a constant blank image for watermark-free images from A, B. So the first two terms of L ext are the reconstruction loss wm and the clean loss clean for these two types of images respectively, i.e., where δ is the target watermark image and δ 0 is the constant blank watermark image. Besides the reconstruction loss, we also want the watermarks extracted from different watermarked images to be consistent, thus another consistent loss cst is added: The final extracting loss L ext is defined as the weighted sum of these three terms, i.e., L ext = λ 4 * wm + λ 5 * clean + λ 6 * cst . A, B, B , B in this stage. Since the embedding sub-network H is fixed in this stage, only the extracting loss L ext is considered to refine the capability of R. Specifically, the clean loss clean is kept unchanged, while wm , cst are updated as below respectively:
Watermark Verification
For effective forensics, if we want to know whether a surrogate model SM is trained by stealing the IP of one target model, we first feed SM's output into the extractor sub-network R, then verify whether R's output matches the pre-defined watermark.
In this paper, we consider two different verification methods: the classic normalized correlation (NC) metric and an extra classifier C. For the former one, it is simply defined as: where < ·, · > and · denote the inner product and L2 norm respectively. Compared to the NC metric, the latter classifier is more straightforward and robust to noisy parts of R(b i ) potentially. Because it is just a binary classification problem (yes/no), we find a simple network consisting of three convolutional layers is enough. In our method, C is joint trained with R during the adversarial training stage, and label "1" and "0" represent watermarked and watermark-free images respectively. By default, the simple cross entropy loss is used:
Extension to Multiple Watermarks
In our default setting, for one specific watermark, an embedding sub-network and a corresponding extractor sub-network will be trained. This makes sense for the scenarios where only one specific deep model needs to be protected. However, it is unfriendly to some other scenarios where one deep model has multiple release versions and different versions are represented by different watermarks, because training multiple embedding and extractor subnetworks is both storage-and computation-consuming. Thanks to the flexibility of the proposed framework, enabling multiple different watermarks within one single embedding and extractor sub-network is easy to be supported. The overall training strategy is similar, and the biggest difference is that different watermarks will be randomly selected for H to embed and R to exact correspondingly. For the verification classifier C, it will be changed from the binary classification to multi-class classification.
Extension to Self-watermarked Models
As mentioned before, in order to protect one specific target model, the embedding sub-network acts as an extra barrier and will be appended after the target model as a whole. The main benefit of this pipeline is task-agnostic, which means that it is a very general solution and can work independently without the need of knowing the information of the target model. However, requiring an extra embedding sub-network may sometimew be a limitation. So we propose an extended solution to absorb the watermarking functionality into the target model itself if we can control its training process. In this sense, the target model is self-watermarked without the extra embedding sub-network. In other words, the target model itself will automatically embed a watermark into its output. The training difference between the original target model and the self-watermarked target model is shown in Figure 4. Assuming the original target model M is trained with the task-special loss L task , we add an extractor sub-network R after M and jointly train them by combining the extracting loss L ext mentioned above with L task . To ensure the robustness and capability of R, R still needs to be fine-tuned with the extra adversarial training stage.
EXPERIMENTS
To demonstrate the effectiveness of the proposed system, we use two example image processing tasks in this paper: image deraining and Chest X-ray image debone. The goal of these two tasks is to remove the rain streak and rib components from the input images respectively. We will first introduce the implementation details and evaluation metric in Section 4.1, then evaluate the proposed deep invisible watermarking algorithm quantitatively and qualitatively in Section 4.2. Next, we will demonstrate the robustness of the proposed deep watermarking framework to surrogate models with different network structures and loss functions in Section 4.3. Furthermore, we compare our method with traditional and DNNbased watermarking algorithm in Section 4.4 and show the robustness of our method to watermark overwriting in Section 4.5. Finally, some ablation analysis are provided in Section 4.6.
Implementation Details
Dataset setup. For image deraining, we use 6100 images from the PASCAL VOC dataset [59] as target domain B, and use the synthesis algorithm in [60] to generate rainy images as domain A. These images are split into two parts: 6000 are for both the initial and adversarial training stage and 100 for testing. Since the images used by the attacker may be different from that used by the IP owner to train H, R, we simulate it by choosing 6000 images from the COCO dataset [61] for the surrogate model training. Similarly, for X-ray image debone, we select 6100 high-quality chest X-ray images from the open dataset chestx-ray8 [62] and use the rib suppression algorithm proposed by [47] to generate the training pair. They are divided into three parts: 3000 for both the initial and adversarial training, 3000 for the surrogate model training, and 100 for testing. All the images are resized to 256 × 256 by default.
Training details. By default, we train sub-network H, R and discriminator D for 200 epochs with a batchsize of 8. Adam optimizer is adopted with the initial learning rate of 0.0002. We decay the learning rate by 0.2 if the loss does not decrease within 5 epochs. For surrogate model training, batchsize of 16 and longer epochs of 300 are used to ensure better performance. And the initial learning rate is set to be 0.0001, which stays unchanged in the first 150 epochs and is linearly decayed to zero in the remaining 150 epochs. In the adversarial training stage, subnetwork R is trained with the initial learning rate of 0.0001. For classifier C, it is also trained in a similar way. By default, all hyperparameters λ i equal to 1 except λ 3 = 0.01.
Evaluation Metric. To evaluate the visual quality, PSNR and SSIM are used by default. When using the NC metric for watermark verification, the watermark is regarded as successfully extracted if its NC value is bigger than 0.95. Based on it, the success rate (SR N C ) is further defined as the ratio of watermarked images whose hidden watermark is successfully extracted in an image set. When using the classifier for verification, we use the threshold 0.5 for the binary case and adopt the label with maxpossibility for the multi-class case. Similarly, the success rates based on the classifier are denoted as SR C .
Deep Image Based Invisible Watermarking
In this experiment, we give both qualitative and quantitative results about the proposed deep image based invisible watermarking algorithm. To demonstrate its generalization ability to various types of watermark images, both simple grayscale logo images and complex color images are considered. For the grayscale case, we even consider a QR code image. Several example images used for the debone and deraining task are shown in Figure 5. It can be easily seen that the proposed deep watermarking algorithm can not only embed the watermark images into the host images in an imperceptible way but also successfully extract the embedded watermarks out. Especially for the challenging "Peppers" and "Lena" watermarks that have rich textures, the embedding subnetwork and extractor sub-network still collaborate very well and guarantee the visual quality of the watermarked image and extracted watermark simultaneously.
To investigate whether the extracting sub-network embeds watermark in a naive additive way, we further show the residuals (even enhanced 10 ×) between watermarked image b i and watermark-free image b i in Figure 5. Obviously, we cannot observe any watermark hint from the residuals, which means the watermark is indeed embedded in an advanced and smart way. Taking one more step forward, we use the "Flower" watermark image as an example and show the color histogram distribution difference between the watermarked image and original watermark-free image in Figure 6. It shows that, even though the colorful "Flower" watermark image is embedded, the gray histogram distribution is almost identical to that of the original watermark-free image.
Quantitative results with respect to the visual quality and extracting ability are provided in Table 1. We can find that the average PSNR and SSIM between watermarked and watermarkfree images are very high, which double confirms the original image quality is well kept. On the other hand, though high extracting ability is contradictory to high visual quality to some extent, our extractor sub-network R is still able to achieve very high extracting ability with an average NC value over 0.99 and 100% success rate both based on the NC metric and the classifier C.
Robustness to Surrogate Model Attack
Besides the visual quality consistency, another more important goal is to guarantee IP protection ability. Considering the attacker may use different network structures trained with different loss functions to imitate target model's behavior, we simulate this case by using a lot of surrogate models to evaluate the robustness of the proposed deep watermarking framework. In details, we consider four different types of network structures: vanilla convolutional networks only consisting of several convolutional layers ("CNet"), an auto-encoder like networks with 9 and 16 residual blocks ("Res9", "Res16"), and the aforementioned UNet network ("UNet"). For the objective loss function, popular loss functions like L1, L2, perceptual loss L perc , adversarial loss L adv and their combination are considered. But we discard the case that only utilizes perceptual loss L perc for surrogate model learning because it will generate very terrible image quality (PSNR:19.73; SSIM:0.85) and make the attack meaningless. As the surrogate model with "UNet" and L2 loss function is utilized in the adversarial training stage, this configuration can be viewed as a white-box attack and all other configurations are black-box attacks. Without losing generality, we use grayscale "IEEE" and color "Flower" image as default target watermark for debone and deraining respectively in this experiment.
Because of the limited computation resource, we do not consider all the combinations of network structures and loss functions. Instead, we choose to conduct the control experiments to demonstrate the robustness to the network structures and loss functions respectively. In Table 2 (Column 2 ∼ 5), it shows that, though only UNet based surrogate model trained with L2 loss is leveraged in the adversarial training stage, the proposed deep model watermarking framework can resist both white-box and black-box attacks when equipped with the newly proposed deep image-based invisible watermarking technique.
To further demonstrate the robustness to different loss functions, we use the UNet as the default network structure and train surrogate models with different combinations of loss functions. As shown in Table 2 (Column 6 ∼ 10), the proposed deep watermarking framework has a very strong generalization ability and can resist different loss combinations with very high success rates. In Figure 7, we provide one example watermark extracted from different surrogate models.
Comparison with other methods
We further compare our method with one representative traditional watermarking algorithm [19] and one DNN-based watermarking algorithm HiDDeN [31] by hiding 64-bit as the watermark. For HiDDeN, two variants proposed in their paper are considered: with/without noise layer between the encoder and the decoder. Compared to the "without noise layer" version, "with noise layer" version is shown to be more robust.
As shown in Table 3, our method not only has better embedding and extracting ability, but also has better robustness to surrogate model attack. For the traditional watermarking algorithm [19], it can only resist the attack of some special surrogate models (only UNet in our experiment) because its extracting algorithms cannot handle the degraded watermarked images from different surrogate models (shown in Figure 8). More importantly, they cannot embed high-capacity watermarks like logo images. Although the visual quality is acceptable, the remnants of the original watermark can be discovered from the residual image ( Figure 5), which may be utilized by the attacker. In terms of the embedding time and extracting time, our method is about 0.0051s and 0.007s for a 256×256 image on GPU, while traditional spatial invisible watermarking algorithm [19] needs 0.0038s TABLE 2 The success rate (SR N C /SR C ) of our method resisting the attack from surrogate models. Column 2 ∼ 5 are trained with L2 loss but different network structures and Column 6 ∼ 10 are trained with UNet network structure but different loss combinations. We take Debone-IEEE task and Derain-Flower task as example, and † denotes the results without adversarial training. 4) and robustness to surrogate model attack (Column 5 ∼ 8) among our method and two typical methods in debone task. TSIW-64 denotes using traditional spatial invisible watermarking algorithms [19] to hide 64-bit, while HiDDeN is a typical DNN-based watermarking method. We use HiDDeN to hide 64-bit as well and ‡ denotes training HiDDeN without noise layer.
The surrogate models are trained with L2 loss but different network structures. For TSIW-64 and HiDDeN, only SR N C is reported. and 0.0047s on CPU. We have also tried some other traditional transform domain watermarking algorithms like DCT-based [63], TABLE 4 Quantitative results with different watermark images for watermark overwriting. The 2nd and 3rd rows are the extracting success rate of our extractor, and ‡ denotes the results with the enhanced training strategy. We also show the visual quality of surrogate model trained with corresponding re-watermarked images in the last row.
Robustness to Watermark Overwriting
In real applications, the attackers may follow a similar watermarking strategy and add another watermark upon the model outputs before training the surrogate model, which can potentially destroy the original watermark or cause forensics ambiguity. This type of attack is often called "watermark overwriting". By using our default setting, we find it cannot resist this attack very well. But following a similar idea as the adversarial training, we find a simple enhanced training strategy can help resist it. Specifically, we first train a single embedding network (simulate the potential overwriting network) to embed ninety diverse watermark images as described in Section 3.4.5, and then add a noise layer between the embedding sub-network and the extractor sub-network in the initial training stage by using the trained embedding network to re-watermark the watermarked images. We require the extractor to be able to extract the original watermark out from re-watermarked images but blank watermarks from clean images or images only watermarked by the overwriting network. In this way, the extracting ability of the extractor sub-network can be significantly enhanced.
To simulate the attack scenario, we utilize "IEEE" as the original watermark and use other different watermarks (no overlap with the ninety watermarks) to perform overwriting, which can be regarded as a black-box setting. As shown in Fig. 9 and Table 4, overwriting with different watermark images makes the extractor fail to some extent, but the extractor sub-network will work well with the above enhanced training strategy. In addition, Table 4 shows a degradation of the performance of surrogate model trained with re-watermarked images, which reveals watermark overwriting is a two-sided sword for the attacker. Need to note that, although our algorithm can still extract our original watermark out, it still suffers from the ambiguity issue (the extractor sub-network from the attacker can also extract their watermark out). And we still need the watermarking protocol to resolve this ambiguity issue.
Ablation Study
Importance of Clean Loss and Consistent Loss. At the watermark extracting end, besides the watermark reconstruction loss, we also incorporate the extra clean loss for watermark-free images and watermark consistent loss among different watermarked images. To demonstrate their importance, two control experiments are conducted by removing them from the extracting loss. As shown in Figure 11, without clean loss, the extractor will always extract meaningless watermark from watermark-free images of domain A, B with high NC values, which will make the forensics invalid.
Similarly in Figure 12, when training without the consistent loss, we find the extractor sub-network R can only extract very weak watermarks or even cannot extract any watermark out from the outputs of the surrogate mode. This is because the watermark consistency hidden in watermarked image b i is destroyed to some extent and makes it more difficult to learn unified watermarks into the surrogate model. By contrast, our method can always extract very clear watermarks out. Importance of Adversarial Training. As described above, to enhance the extracting ability of R, another adversarial training stage is used. To demonstrate its necessity, we also conduct the control experiments without adversarial training, and attach the corresponding results in Table 2 (labelled with " †"). It can be seen that, with the default L2 loss, its extracting success rate is all about 0% for surrogate models of different network structures. When using UNet as the network structure but trained with different losses, only the embedded watermarks of some special surrogate models can be partially extracted, which demonstrates the significant importance of the adversarial training. In our understanding, the reason why adversarial training with UNet can generalize well to other networks may come from two different aspects: 1) Different surrogate models are trained with the similar task-specific loss functions, so their outputs are similar; 2) As shown in the recent work [66], different CNN-based image generator models share some common artifacts during the generation process. So training with the degradation brought by UNet makes the extracting network robust to other networks.
Extracting Ability vs Surrogate Model Performance. In this section, we further analyze the relationship between the extracting ability and the surrogate model performance. Intuitively, the extracting ability will be high when the surrogate model's performance is good, and vice versa. To simulate this case, we take CNet with L2 loss as the baseline and adjust its learning ability by changing the feature channel number from 8 to 256. It can be observed from Figure 10 that, as the feature channels increase, the performance of surrogate model (PSNR from 20.77 to 23.55) and the extracting ability of extractor R ( SR N C : from 3% to 99% and SR C : from 18% to 100%) both increase. From the model protection perspective, when the surrogate model does not perform well, we can view it as successful IP protection. For the visual illustration, we give one example for CNet 8 and CNet 256 on the right of Figure 10. It can be seen that our model fails to extract the watermark only when the surrogate model completely fails to conduct the debone task.
Influence of Hyper-parameter and Watermark Size. For the hyper-parameter setting, we conduct control experiments with different λ, which is used to balance the ability of embedding and extracting. As shown in Table 5, although the visual quality of watermarked images and the NC value will be influenced by different λ, the final results are all pretty good, which means our algorithm is not sensitive to λ. We further try different sizes of watermark to test the generalization ability. Since we require the size of watermark to be same as the cover image in our framework, we pad the watermark with 255 if its size is smaller than 256. As shown in Table 6, our method generalizes well to watermarks of different sizes.
EXTENSIONS
In this section, we will provide the experiment results for the multiple watermarks and self-watermarked extensions in Section 5.1 and Section 5.2 respectively, then discuss how to leverage the proposed framework to protect private data and traditional non-CNN models in Section 5.3.
Multiple watermarks within one network.
As mentioned in Section 3.4.5, the proposed framework is flexible to embed multiple different watermarks with just a single embedding and extractor sub-network. To showcase it, we take the debone task for example and select 10 different logo images from the Internet as watermarks for training. For comparison, we use the logo "IEEE" and "Flower" as two representatives and compare them with the results of the default per-watermark-per-network setting. In terms of the original embedding and extracting ability, as shown in Table 7, the visual quality of this multiple-watermark setting measured by PSNR degrades from 47.76 to 41.87 compared to the default setting but is still larger than 40, which is reasonably good. Some visual examples are further displayed in Figure 13. In terms of the robustness to the surrogate model attack, we show the comparison results of different networks and loss functions in Table 8. It can be seen that the extracting success rates of the multiple-watermark setting are comparable to those of the default setting, thus preserving a similar level of robustness.
Self-Watermarked Model
Since the embedding sub-network H itself is an image processing network and can have different network structures, given a target image processing network M, it is possible to absorb the watermark embedding functionality of H into M as shown in Figure 4. In this way, M is self-watermarked without the need of an extra barrier H. To demonstrate this possibility, we use the derain task with the "Flower" logo image as an example and jointly learn the deraining and watermark embedding process within one single network. In Table 9 Figure 14. Besides the deraining functionality, the self-watermarked model can also hide the watermarks well with 0.9999 NC values. In Table 11, we further demonstrate the robustness to the surrogate model attacks of different network structures and loss functions. It shows that the self-watermarked model is very robust with near 100% extracting success rates.
Protect Valuable Data and Traditional Algorithms
In this paper, though most experiments are conducted to simulate CNN model protection, the proposed idea is very general and can be easily applied to data protection and traditional non-CNN algorithm protection. In details, we can follow the default task-agnostic watermarking setting and embed watermarks into labeled target data or add a watermarking sub-network barrier after traditional algorithms' output. To simulate data protection, we adopt the DPED ("DSLR Photo Enhancement Dataset") dataset [67] as an example. It is collected by using high-end cameras and careful alignment to train a high-quality image enhancement network. For traditional algorithm protection, we consider the famous structure-aware texture smoothing method RTV [68] and choose 6000 images from PASCAL VOC dataset [59] and COCO dataset [61] as its inputs. As our default setting, we split the dataset into different parts for embedding sub-network training and surrogate model attack. The detailed evaluation results are given In Table 10. On the one hand, the learned embedding subnetwork can keep the original visual quality very well with high PSNR/SSIM values and the extracting sub-network has a 100% extracting success rate. On the other hand, we take the Res9 with L1 + L adv as an example surrogate model, and the proposed framework is able to extract the watermark out when the surrogate model is trained with the target dataset or imitates the behavior of the traditional algorithm. Some visual results are shown in Figure 15.
CONCLUSION AND DISCUSSION
IP protection for deep models is an important but seriously under-researched problem. Inspired by traditional spatial invisible media watermarking and the powerful learning capacity of deep neural networks, we innovatively propose a novel deep spatial watermarking framework for deep model watermarking. To make it robust to different surrogate model attacks and support imagebased watermarks, we dedicatedly design the loss functions and training strategy. Experiments demonstrate that our framework can resist the attack from surrogate models trained with different network structures and loss functions. By jointly training the target model and watermark embedding together, we can even make the target model self-watermarked without the need of an extra watermark embedding sub-network. Moreover, though our motivation is to protect deep models, it is general to data protection and traditional algorithms protection. We hope our work can inspire more explorations of deep model IP protection, such as more types of models (detection, semantic segmentation) and protection strategies. There are still some interesting questions to explore in the future. For example, though the joint training of the embedding and extracting sub-networks in an adversarial way makes it difficult to find some explicit pattern in the watermarked image, it is worthy to study what implicit watermark is hidden. Besides, our method is not robust enough to some pre-processing techniques for surrogate model attack, such as random cropping and resizing, because the consistency we rely on will be destroyed. So it is necessary to design some new kind of consistency which is intrinsically robust to such pre-processing. 10 Quantitative results of applying the proposed framework to data (DPED [67]) and traditional algorithms (RTV [68]) protection. The extracting success rate marked with ‡ denotes the results of resisting an example surrogate model attack (L1 + L adv ). The second to fourth columns represent the results of the learned embedding and extracting sub-network. TABLE 11 The comparison between task-agnostic watermarking model and self-watermarked model about the success rate (SR N C /SR C ) of resisting the attack from surrogate models trained with different loss combinations.. | 2021-03-09T02:16:23.638Z | 2021-03-08T00:00:00.000 | {
"year": 2021,
"sha1": "47d240d5c70da5c2995ec6e591de8fc843fba01c",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/2103.04980",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "47d240d5c70da5c2995ec6e591de8fc843fba01c",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
268986050 | pes2o/s2orc | v3-fos-license | Analysis of heavy metals and pathogen levels in vegetables cultivated using selected water bodies in urban areas of the Greater Accra Metropolis of Ghana
Open-surfaced water sources have been used to irrigate vegetable farms in cities. Open-surface water often contains unmonitored concentrations of health-threatening contaminants that pose health risks, especially when used to produce vegetables for human consumption. However, information on levels of heavy metals and faecal coliform bacteria in such vegetables in selected sites, especially in Greater Accra Metropolitan Area (GAMA) of Ghana is rare. This study examines the levels of heavy metals and faecal coliform in two vegetables-lettuce and bell pepper - that were cultivated using open-surface wastewater from drains and constructed reservoirs at different locations of the GAMA. Using concurrent mixed methods, questionnaires were administered to 67 vegetable farmers, followed by the collection of vegetable samples from three urban farm sites, Haatso and Dzorwulu and Weija irrigation scheme site (WISS) for laboratory analysis. The concentrations of Lead (Pb), Mercury (Hg) and Cadmium (Cd) were determined using atomic absorption spectroscopy after microwave digestion of the vegetables while total faecal coliform was quantified using MacConkey-Endo broth method. The results from all three sites showed that the concentrations of Cd (=0.001 μg/mg) and Pb (=0.005 μg/mg) in lettuce were within the World Health Organization's (WHO) permissible levels. However, the Hg (≥0.309 μg/mg) and faecal coliform (>5 count/100 ml) in the vegetables from all three sites exceeded the WHO permissible limits. Therefore, consumers of vegetables from such urban farms are exposed to health risks associated with Hg and faecal coliforms. There is the need to intensify education on the health risks of consuming vegetables produced from open-surface water sources from the observed sites. The enforcement of existing phytosanitary standards to enhance food safety and the quality of urban vegetables is also recommended.
Open-surfaced water sources have been used to irrigate vegetable farms in cities. Open-surface water often contains unmonitored concentrations of health-threatening contaminants that pose health risks, especially when used to produce vegetables for human consumption.However, information on levels of heavy metals and faecal coliform bacteria in such vegetables in selected sites, especially in Greater Accra Metropolitan Area (GAMA) of Ghana is rare.This study examines the levels of heavy metals and faecal coliform in two vegetables-lettuce and bell pepper -that were cultivated using open-surface wastewater from drains and constructed reservoirs at different locations of the GAMA.Using concurrent mixed methods, questionnaires were administered to 67 vegetable farmers, followed by the collection of vegetable samples from three urban farm sites, Haatso and Dzorwulu and Weija irrigation scheme site (WISS) for laboratory analysis.The concentrations of Lead (Pb), Mercury (Hg) and Cadmium (Cd) were determined using atomic absorption spectroscopy after microwave digestion of the vegetables while total faecal coliform was quantified using MacConkey-Endo broth method.The results from all three sites showed that the concentrations of Cd (=0.001 μg/mg) and Pb (=0.005 μg/mg) in lettuce were within the World Health Organization's (WHO) permissible levels.However, the Hg (≥0.309 μg/mg) and faecal coliform (>5 count/100 ml) in the vegetables from all three sites exceeded the WHO permissible limits.Therefore, consumers of vegetables from such urban farms are exposed to health risks associated with Hg and faecal coliforms.There is the need to intensify education on the health risks of consuming vegetables produced from open-surface water sources from the observed sites.The enforcement of existing phytosanitary standards to enhance food safety and the quality of urban vegetables is also recommended.
Introduction
The world is experiencing rapid urbanization, with a growing dependence on urban agriculture to meet increasing human demand for healthy vegetables to augment dietary needs [1][2][3].Urban and peri-urban agriculture encompasses production, processing, distribution and sale of food particularly vegetables within cities and city peripheries [4,5].In this context, "urban vegetables" are plants like lettuce (Lactuca sativa) and bell pepper (Capsicum annuum) grown for their edible parts in different parts of an urban area.
In Ghana, Mordor Intelligence report [6] suggests that the vegetable market share was around USD 0.95 billion in 2024 and is projected to reach USD 1.2 billion in 2029.Apart from export of vegetables from Ghana to other countries, estimates point out that, total vegetable produced in the country was respectively 786,457, 787,160, 788,010 and 788,396 metric tons in 2017, 2018, 2019 and 2020 [7].
Unfortunately, rapid rate of urbanization (56.7%) in Ghana and unpredictable rainfall patterns resulting in scarcity of freshwater, have forced many small-scale farmers to use polluted or contaminated wastewater sources for urban vegetable farming [8,9].This practice does not only raise concerns about quality of the agricultural produce but also gives rise to potential food safety and health risks for urban families [10,11] who rely on these vegetables in the Greater Accra Metropolitan Area (GAMA) of Ghana.
Wastewater, originating from urban households, industrial, and commercial activities [12,13], often contains unwholesome physical, chemical, and microbial substances, making it unsuitable for irrigating crops intended for human consumption [14,15].The contamination of wastewater sources by toxic faecal coliform bacteria (E. coli which indicates faecal contamination) and heavy metals (HMs) like lead (Pb), nickel (Ni), cadmium (Cd), chromium (Cr), and mercury (Hg) render many urban water systems unwholesome.Escherichia coli (E.coli) bacteria and heavy metals compromise water systems including dams, reservoirs, lakes rivers and streams making them unsuitable for drinking, irrigation and sustaining aquatic life [16][17][18][19].However, in many urban areas like GAMA where vegetable cultivation is a notable farming practice, these polluted water systems continue to serve as the primary source of irrigation water for vegetable farming [19,20].
Although, recycling or treating wastewater for urban agriculture may offer a viable solution [17], persistent lack of institutional oversight in monitoring urban agriculture phytosanitary conditions exacerbates concerns about the use of wastewater and contaminated streams.As a result, several studies have already established strong links between irrigation water quality and the resulting agricultural produce [10,11].Some of these studies indicate that urban families are exposed to contaminated farm produce due to the use of polluted surface water sources for irrigation [10].Notably, the recent study by Onalenna and Rahube [21] identified presence of blaTEM gene in soil and vegetables irrigated with wastewater, raising legitimate concerns for urban vegetable production for consumption both locally and for export.
Ghana's efforts to diversify its export base by promoting non-traditional crops like tomatoes, chili peppers, eggplants and bottle gourds, among others like sweet peppers, onions/shallots, okra, carrots, cauliflower, cabbage, lettuce and cucumber, faced challenges [22,6].According to GIRSAL [22], the European Union (EU) imposed a ban on vegetable imports from Ghana in 2015 due to the discovery of banned substances in some of the vegetables.As a result, the export of fresh chili to the UK, which was valued at about US$ 5.473 million in 2014 (Fig. 1), reduced to US$ 1.312 in 2015 and continued to decline.
The ban had a significant impact on the livelihoods of farmers across both rural and urban areas, many of whom relied on income from vegetable production for their daily sustenance.In response to a sustainable urban vegetable production drive to boost the Fig. 1.Chili export from Ghana to the UK from 2012 to 2021: Source [22].
S. Francis Gbedemah et al.
international competitiveness of Ghana, the GhanaVeg program was established.The program aims to enhance vegetable production throughout the country.Hence, it is worth investigating the current quality of urban vegetable produce to ascertain the programme's impact on GAMA.
Although, several studies have investigated the presence of heavy metals [23][24][25][26][27] and bacteriological substances [28,29] in urban vegetables, little knowledge exists in terms of spatial differences in the locations of urban vegetable production sites and water sources in GAMA.In addition, most studies focused on well-known repositories of heavy metal-contaminated wastewater sources like Korle Lagoon.For instance, Osae et al. [23] evaluated the concentration of heavy metals in vegetables in the catchment area of the Korle Lagoon and found that the heavy metal concentration in lettuce was far above the recommended World Health Organization's (WHO) guideline level.However, studies comparing the level of heavy metals and faecal coliform contamination of vegetables irrigated using presumably wastewater and water from official irrigation schemes are limited.
The current study takes a different approach by employing a unique combination of experimental and social research incorporating interviews to gain an in-depth understanding from the farmers' perspectives.To the best of our knowledge, no previous study has employed this hybrid exploratory concurrent mixed methods research design of combining laboratory experiments on 'potentially contaminated vegetables' and farmer surveys to understand farmers' choices of water sources used in vegetable farming in GAMA.Our study, therefore seeks to fill the above research gaps by analysing the levels of heavy metals and faecal contamination in two urban vegetables from three different locations in GAMA.Specifically, our study objective aims to examine the spatial differences in heavy metal and faecal contaminants in the lettuce and bell pepper cultivated by farmers in Dzorwulu, Haatso and Weija irrigation scheme site (WISS), with the first two sites using wastewater from drains, while the latter uses water reservoir under official irrigation scheme.The study also examines the socioeconomic attributes of the farmers and their motivation for the choice of irrigation water source used in producing the vegetables.This research thus contributes to the broader field of sustainable urban agriculture by addressing differences in locations of farms under diverse irrigation schemes, offering insights into the complex spatial ecology of urban vegetable farming, water quality and food safety.
Study area
The study was conducted on farm sites in three different Municipalities in the Greater Accra Metropolitan Area (GAMA) as shown in Fig. 2. Map of the study area and sites.
S. Francis Gbedemah et al.
Fig. 2. The locations of the farm sites were Haatso in Ga East Municipality (GEM), Dzorwulu in the Ayawaso West Municipality (AWM), and Weija (WISS) in Weija-Gbawe Municipality (WGM).Dzorwulu and Haatso are 17 km and 19 km respectively away from Weija which is in the eastern part of GAMA (see Fig. 2).The Weija irrigation scheme site (WISS) is about 21 km from Accra central.The farmers take their irrigation water from the Weija reservoir which is one of the two reservoirs supplying drinking water to many homes in the GAMA.The WISS is a formal irrigation scheme recently renovated by the Ghana Irrigation Development Authority (GIDA) with support from the Food and Agricultural Organization (FAO) and the Ministry of Food and Agriculture (MoFA) of Ghana.The Weija reservoir takes its source from Densu which is a major river that flows towards the south-eastern portion of the country (please refer to Fig. 2 for the locations of the vegetable production sites).
The Densu River traverses three regions in Ghana.It takes its source from the Atiwa Mountains in the Eastern Region and passes through the Central Region to GAMA.In the GAMA, the river passes through the Ga South Municipality and enters the sea at Bortianor.Increasing anthropogenic activities along the path of Densu River affect the water quality in the reservoir which has the potential to also affect the quality of water for domestic and agricultural purposes [30,31].These activities include mining (both legal and illegal), industrialisation, urbanization and commercial activities.Due to these activities, the river was labelled among the highly polluted rivers in Ghana [32].Most farms in the Haatso and Dzorwulu areas get their water supply sources from drains or streams such as the Onyasia Stream [27], which is one of the tributaries of the Odaw River in GAMA [28].Other sources include other natural and human-induced drains, raw wastewater, storms, and grey water.These water sources often contain high faecal contamination with dilution changes between locations and seasons.These polluted streams are sometimes supplemented by farmers with pipe-borne water [33,34].A few farmers also use partially treated effluents from wastewater treatment system ponds that can be found around the Burma Camp military facility by puncturing the sewage pipes [27].The three farm sites host the cultivation of major vegetables such as okra, pepper, carrot, cabbage, and lettuce farms for local consumption in the GAMA and Kasoa in the Central Region.Vegetable farming is the major economic activity for about 55% of these municipalities' economically active farming population [28].
Survey of vegetable farmers
The study generally adopted the concurrent mixed methods approach, where experimental laboratory research and a semistructured survey were designed and implemented together.The survey was a stratified convenient farm habitat assessment [35,36] involving sampling of vegetable (lettuce and bell pepper) farmers across the three sites.The initial intention of the study was to use quota sampling to sample 90 vegetable farmers, 30 from each site.However, a total of 67 vegetable farmers were conveniently sampled, 22, 21, and 24 from Haatso, Dzorwulu, and WISS respectively.The shortfall was because some of the identified vegetable farmers were unwilling to participate in the survey.
A semi-structured survey questionnaire was used.The questionnaire aimed at soliciting farmers' socio-economic attributes (age, education, crops, and inputs) and their opinions on sources of water for farming.The study was interested in the perceived impact of the water and chemicals being used on the vegetables and experiences of vegetable farming, the sustainability and health risks of their activities.The questionnaire was pre-tested to ensure its reliability.The pre-testing was conducted using 15 non-selected farmers, 5 each from the three study sites.After a successful pre-test of the questionnaire, a one-on-one individual interview of the selected farmers was conducted, following ethical standards.
Ethical compliance was observed by obtaining informed consent from all respondents (orally or written).This assures anonymity of their identity and confidentiality of their information.Confidentiality was assured by using codes for each respondent instead of using their names.Also, farmers who were not comfortable with the survey were exempted.The identity of those who willingly participated were protected throughout the study.
Vegetable sample collection, preparation and analyses
The experimental laboratory analysis was conducted to measure the levels of bacteria pathogens and heavy metals in samples of ready-to-sell fresh lettuce (lactuca sativa) and bell pepper (capsicum annuum) collected from farms in all the three sites.The vegetable samples from Haatso and Dzorwulu served as the primary experimental samples because the study's major objective was to assess contaminants in lettuce and pepper caused by wastewater usage.The vegetable samples from the WISS were used as a control sample because the farms were irrigated using fresh water from the Weija reservoir, which is not wastewater.The WISS was usually monitored by the Ghana Irrigation Development Authority (GIDA) to ensure that the water used for irrigating the farms within the irrigation scheme is good.The samples were collected into transparent plastic bags with zip-lock closures that were labelled with the site's information.The samples were transferred to the University's laboratory at a temperature of 4 • C in an ice chest for analysis.To prevent microbial contamination, the laboratory apparatus used for the microbiological analysis were all sterilized using accepted practices.The Petri dishes and test tubes were washed with soap, followed by a water rinse, and then allowed to air dry.To prevent fungus and other cross-contaminations, the inoculation room was sprayed with 3% thymol.Before and following use, the workbenches were cleaned with ethanol at a concentration of about 80%.To guarantee there were no residues of thymol before the room was used, the room was made to dry for about an hour.All inoculum transfers were carried out in the inoculation chamber to prevent medium contamination.Before and after the transfer of samples, inoculation loops or bent wires were sterilized until they were red hot under a naked flame.Additionally, the media-and sample-containing test tubes' opening ends were flamed before and after transfers.Before measuring 100 ml of De-Ionized (DI) water into a beaker, the pH meter was calibrated.The pH of the mixed puree was measured using the pH meter after 100 g of the vegetables and 100 ml of the DI water were added to a clean blender and blended for 3 min.
Media preparation and faecal coliform tests
We used specific media for coliform bacteria detection in the lettuce and bell pepper purees.The medium [37,38] for total coliform detection and enumeration was MacConkey(M)-Endo broth [39] and that for Faecal Coliform (FC) was Modified(M)-FC broth base [40].For Escherichia coli (E.coli), tryptone water was used.
M-Endo broth [39]: 4.8 g of dehydrated endo medium was suspended in 100 ml of distilled water containing 2 ml of 95% ethanol.The solution was covered with aluminium foil and heated on a hot plate while stirring with a magnetic stirrer.The solution was removed from the heat source immediately after boiling began.It was allowed to cool at room temperature.It was dispensed into Petri dishes.
M − FC broth [40]: 3.7 g of dehydrated FC medium was suspended in 100 ml of distilled water.About 1 ml of a 1% solution of rosolic acid in 0.2 N NaOH was added, and the solution was covered with aluminium foil.It was heated whilst stirring.The solution was taken off the hotplate immediately after boiling began.Cooling was done at room temperature and the broth was dispensed into Petri dishes.Rosolic Acid: 0.08 g of NaOH was dissolved in 10 ml distilled water.0.1 g of rosolic acid was dissolved in the NaOH solution.
Tryptone water [39]: 20 g of tryptone and 5 g of calcium chloride were weighed and dissolved in distilled water and made up to 1000 ml using distilled water.A pH of 7.5 was recorded.About 6 mls each of the solutions were dispensed into MacConkey bottles and autoclaved at 115 • C for 10 min.
Preparation of Petri dishes: About 47 mm microbiological grade Petri dishes were labelled on the underside.Absorbent pads were deposited in each dish using a pair of sterile forceps.A sterile pipette was used to dispense 2 ml of the appropriate medium onto the absorbent paper.
Incubation: Incubation of total coliforms was done at 37 • C for 24 h in a Millipore incubator.Incubation for faecal coliform was done at 44.5 • C for 24 h.
Faecal coliform bacteria identification and counting: After leaving the solution for over 24 h, developed total coliform in the vegetables appeared pinkish and grey in colour with a metallic sheen.Faecal coliforms were yellow or blue in colour.
Heavy metal concentration analysis
This section explains the procedures used to prepare the samples to determine the concentration (mg/kg) of the heavy metals (Pb, Cd, Fe, Hg, As, Se) in the vegetables.
Determination of heavy metals in vegetables (Pb, Cd, Fe)
Sample preparation: To remove all surface and tissue water, vegetables were dried in an oven at 105 • C. The dried vegetables were ground into a powder, and 1 g of that powder was placed in a clean porcelain crucible together with 1 ml of strong nitric acid.To allow the contents to char, the crucible was put on a hot plate set at 200 • C for 2 h.The crucible was removed from the hot plate and allowed to cool for 30 min 1 ml of concentrated nitric was added and the crucible was placed on the hot plate again for 30 min.The crucible was allowed to cool.The crucible and its content were placed in a furnace at 400 • C for 3 h after which it was allowed to cool for 1 h.
The ash was dissolved with 10% 7 M nitric acid and the sample was filtered into a 100 ml volumetric flask.DI water was added to the 100 ml mark, and the solution was sent for metal concentration measurement using atomic absorption spectrophotometry (PINAAcle 900T PerkinElmer Inc., Massachusetts, United States).
Determination of heavy metals in vegetables (Hg, As, Se)
Sample preparation: Five grams (5 g) of dried sample was placed in an Erlenmeyer flask and 10 ml of concentration.HNO 3 was added to it.The sample was then warmed on a hot plate until solubilization.The temperature was increased to near boiling until the solution began to turn brown.After allowing the sample to cool down, 10 ml of conc.HNO 3 was again added and thereafter returned to the hotplate to reduce the volume to 5-10 ml.About 30% of H 2 0 2 was added to the sample when it was allowed to cool.The sample was again heated, and the process was repeated until the solution was clear.About 2 ml of concentrated HCl was added after the sample was cooled and then it was heated again.After cooling, the sample was transferred into a 100 ml volumetric flask.Dilution was done with the American Society for Testing and Materials (ASTM) type 1 water [41].According to ASTM [41], this water can be described as ultrapure based on their standard.The insoluble material was allowed to separate after which the sample was analysed using the hydride generation/atomic absorption spectrometric method for As and Se.The cold vapour atomic absorption spectrometric method was used for Hg analysis.
Statistical analysis
The farmers' socio-economic attributes were described using the frequency (n) and percentages (%) while the levels of heavy metals (mg/kg) and coliform (count/100 ml) were estimated using the mean.The results and discussion section are discussed next.
S. Francis Gbedemah et al.
Background profile of vegetable farmers
The basic profile of the vegetable farmers at the three study sites is represented in Table 1.The findings revealed that 60 (89.6%) of the respondents were males and 7 (10.4%)were females.This finding is not new to this study site but has been observed in other similar studies conducted in Ghana (see Refs. [10,27]).Women do not dominate in farming but rather are involved in the vending of the produce [42].In Table 1, majority of the farmers are in the age range of 41-50 years (53.7%)followed by the 51-60 years age group (19.4%), and 7.5% are those who are 61 years and above.The lowest age group is 18-30 years, which is the most youthful.In terms of education, majority had Senior High School/'O' level education (41.7%), followed by Middle/Junior High School (29.9%).Those educated up to primary school level were 17.9% and the least educational attainment was 4.5%, and this is for respondents that had non-formal education.As far as crop varieties that were planted are concerned, 32.8% grew more than four crops, 29.9% grew three varieties of crops, 19.4% grew two varieties and 17.9% grew only one variety.The most common fertilizer used was organic fertilizer according to 53.7% of the farmers.Those who used NPK and Potassium fertilizer were 22.4% for each fertilizer type.Only a few (1.5%) did not apply fertilizer.These basic profiles of the respondents can be seen in Table 1.
Main sources of water for vegetable farms and reasons for using them
The Onyasia stream was found to be the primary source of water for the vegetable farmers at the Haatso and Dzorwulu farm sites (see Table 2).Approximately 26.8% of the farmers in Haatso and about 25.4% of the farmers in Dzorwulu used this water as their primary irrigation water source.A total of 52.2% of the farmers stated that they used the stream as their main irrigation water source, which receives effluents from numerous drains and dwellings in the city.
The finding of the main sources of water for vegetable farming in Accra confirms the claim by [42] that vegetable farmers in urban and peri-urban areas mostly rely on wastewater flowing in streams.While environmentalists and water conservation experts have argued for the reuse of wastewater in aquacultural activities to ensure sustainable water resources management which is in line with the United Nations Sustainable Development Goal 6, wastewater has to be treated to meet some quality standards as specified by [43] for it to be safe to be used to water crops for human consumption.This is because using untreated wastewater to water crops can pose serious health risks to humans.An example of such health risk is salmonella bacteria outbreaks which are common in fruits and vegetables.
Table 3 shows the responses of respondents (farmers) from the three study sites, where closeness of the water source to the farm was the main reason why most farmers used their current water source for irrigating their vegetable farms.For instance, 41%, 38%, and 54% of the respondents at Haatso, Dzorwulu and WISS attributed their use of the current water sources to closeness to their farms.At Haatso, the other two main reasons were that they could access it for free (16.2%) and that it is reliable to use (18.2%).At Dzorwulu the Note: All figures in parentheses are the percentages of the total sample size (N = 67).
S. Francis Gbedemah et al.
two other main reasons were that they are available for free (23.8%).According to the farmers, that water is what is available to them (14.2%) to use on their farms.Similar findings were reported for Weija.On the whole, the closeness of the water source to the farm, availability, and reliability are very important when it comes to consideration in the use of water for irrigation [44].
Concentration of selected heavy metals in lettuce from the 3 farm sites
Heavy metal concentrations in the lettuce samples from Haatso, Dzorwulu, and Weija are presented in Table 4.The results show a mean Hg concentration of 0.036 × 10 4 mg/kg in lettuce from Haatso, 0.116 × 10 4 mg/kg for Dzorwulu, and 0.4860 × 10 4 mg/kg for Weija irrigation scheme site (WISS).In other words, the Hg concentration in lettuce from Haatso (360 mg/kg) was lower than those from Dzorwulu (1160 mg/kg) and Weija (4860 mg/kg).The results also show that lettuce from Dzorwulu and Weija had a Hg concentration above the maximum WHO permissible level of 1000 mg/kg for human consumption.The concentrations of Pb and Cd from all the irrigation sites were within WHO [43] recommended limits for human consumption.However, the sample collected from the WISS, which does not use wastewater but river/dam, recorded high levels of mercury.This may be due to the high levels of mercury in the irrigation water since some studies have already reported high levels of Hg in the Densu River which is used for irrigation by the farmers at the Weija irrigation scheme site [45].It should be noted that the high concentration of Hg in the vegetables can also come from the soil or from the fertilizer the farmers use on their farms.
It should be indicated that although mercury can be released into the environment from natural sources, from the soil, or the fertilizer the farmers used, anthropogenic factors along the Densu River could be the main source of heavy metals in the vegetables [45].For instance, mercury is used in mining along the Densu River, especially those engaged in illegal mining [46] and indiscriminate dumping of solid and electronic waste [47].
Concentration of selected heavy metals in pepper from the 3 farm sites
The concentrations of Hg, Pb, and Cd in pepper samples from Haatso, Dzorwulu, and WISS are shown in Table 5.Except for Hg, all the concentrations were within the maximum permissible limit set by WHO [43].The results show low Hg concentration in vegetable samples from Haatso (0.0150 * 10 4 mg/kg) and Dzorwulu (0.0280 * 10 4 mg/kg) compared to that of the Weija irrigation scheme site (0.309 * 10 4 mg/kg) which is on the high side.The amount of Hg obtained in this study is significantly elevated compared with the maximum acceptable limit set by WHO.The Onyasia Stream serves as the main water source for the farms in Haatso and Dzorwulu, and this may be the source of pollution of the vegetables, or it may be due to uptake from the soil which eventually accumulates in the sampled vegetables.Note: All figures not in parentheses are the frequency responses (N = 67).
Table 3
Reasons for using a particular water source.
Study Sites
Reasons for using a particular water source to water vegetables N (%) Note: All figures not in parenthesis are the frequency responses (N = 67).
Table 4
The concentration of heavy metals in lettuce.Wastewater can considerably contribute to the accumulation of heavy metals in soils and crops [48,49].Apart from airborne contamination from traffic, the application of manures, fertilizers, and pesticides are additional sources of heavy metals in irrigated agriculture farms [50,51].Despite Ghana's lower level of industrialisation, untreated but diluted wastewater (wastewater mixed with stream/stormwater) is often used for vegetable cultivation in urban areas [33].For instance, Accra is home to roughly 1000 active vegetable producers and these vegetables are consumed by about 200,000 urban residents daily [29].Heavy metals, often known as trace metals, are one of the most enduring contaminants in wastewater.The environment and human health are negatively impacted by high concentrations of heavy metals released into water bodies in several ways [52].Wastewater that is released into the Onyasia River may contain heavy metal pollutants, which may also end up being absorbed via the soil and bioaccumulating by the vegetables that are irrigated with the stream water.This study shows that three common heavy metals-lead, mercury, and cadmium-pose severe risks to both humans and the environment [53].
The finding in this paper is similar to that of [23] who also assessed heavy metal concentrations in some vegetables and their corresponding soil from a farm in the Korle Lagoon's catchment.They found that heavy metals in lettuce were more than the recommended guideline level.Specifically, Fe and Zn in the vegetables were above the recommended guideline level.Due to these findings, they posit that vegetables grown around the Korle Lagoon are not good for human consumption because they are carcinogenic while few are noncarcinogenic to both adults and children.Accumulation of heavy metals in tissues, organs and cells of the human body can produce health effects such as reduced growth, short and long-term morbidities, organ and nervous system damage, and in extreme cases, death [54].
Comparison of faecal coliform counts in lettuce and pepper in the three sites
Table 6 lists the faecal coliform and total coliform counts in the lettuce and pepper samples taken from the three farm sites.The findings show that coliforms and faecal coliforms were present in the lettuce and pepper samples taken from the irrigation sites, and their levels were higher than the geometric mean count recommended by the WHO [43].This suggests that eating the lettuce and pepper from the three irrigated sites is unsafe for the general population's health.The level of faecal coliforms suggests that water used for irrigation of the vegetables is the most likely source of the faecal contamination even though the water has not been tested to confirm this assertion.The good thing about this finding is that, faecal coliform and total coliform counts in the samples from the WISS are lower than those of Haatso and Dzorwulu.
This study supports the assertions made by [27,33] that the use of contaminated water for the cultivation of vegetables poses risks to people's health.Since fresh produce and vegetable farmers in urban and peri-urban areas of Ghana and other West African countries depend on water from various wastewater sources to irrigate their crops, the consumers of these crops are likely to be exposed to diseases like typhoid fever.This is because, vegetables bioaccumulate the bacterial, biological, and chemical pollutants found in most streams and wastewater sources [29].
Changing risky behaviour seems to be motivated in most cases by the knowledge of a health threat associated with the behaviour [55].Hence, a lack of awareness of the outcome of risky behaviour may not elicit behavioural change in people or society.For most farmers, the lack of understanding of the potential health risks or hazardous nature and effect of wastewater use on their produce and consumers makes it difficult to change their behaviour.In situations where they want to change the source of water for irrigation but there are no alternative sources, they have no option but to continue to use the contaminated source.Risk perception relating to hazards is usually based on group or individual's experiences and the prevailing environmental conditions [56].This may mean that vegetable farmers may see no problem with the use of their sources of water for farming if the community accepts, does not complain about it and have not experienced any negative health outcomes attributed to the use of their farming products.Again, vegetable farmers may continue with risky farming practices like using wastewater sources probably because they are not aware of the level of pollution and health consequences of their actions.It will also continue when authorities are not enforcing regulations on best farming practices, or simply because consumers are not perturbed with the harm that they are ingesting.They will therefore continue to use these water sources for their urban farming.This study together with others like [57] points out that, these water sources may be the cause of heavy metal contaminants.Among agricultural sources of heavy metals in plants, it can be said that the most common sources are pesticides, fertilizers, and sewage sludge [58] to mention only a few.Fertilisers, especially the organic and inorganic types are accountable for producing a lot of heavy metals in the soil [59].Pesticides are very important for plant growth and for a growing population as those found in urban areas of West Africa including Ghana.Some crops cannot grow in the urban environment without the application of pesticides.This is because pesticides protect most crops in tropical countries like Ghana in the wake of climate change.It should be noted that the farmers interviewed said they mainly use organic fertilizer on their farms.The point is that unless the soil is contaminated with heavy metals from a particular source, natural sources of heavy metals in soils may not reach the levels found in the vegetables in the sites under investigation.The obvious source of contamination may be the water from the river used to irrigate the crops.
Research conducted by [34] through the examination of both conventional and unconventional vegetables utilizing wastewater sourced from specific regions in Accra shows that the irrigation water derived from wastewater exhibited the highest levels of manganese (Mn), followed by iron (Fe), zinc (Zn), lead (Pb), copper (Cu), and nickel (Ni).However, chromium (Cr), cadmium (Cd), and cobalt (Co) concentrations remained within the thresholds established by the World Health Organization (WHO) and the Food and Agriculture Organization (FAO).They also found that iron (Fe), manganese (Mn), copper (Cu), lead (Pb) and nickel (Ni) levels were higher in wastewater-irrigated soils than in groundwater-irrigated soils respectively.These are interesting studies that show that using untreated wastewater on crops has a high concentration of heavy metals in them.However, evidence from studies like [60,61] suggests that mining and municipal solid wastes are the key sources of heavy metal contamination in the Weija reservoir.
Conclusions, limitations, and future works
Conclusions: The fruit and vegetable industry has become a very important part of Ghanaian agriculture because it employs a large segment of the unemployed, provides nutrition to a large section of the population, serves as a source of income to people and provides foreign exchange to the country.Over the past years however, the export of some vegetables has been banned due to the presence of pests and other substances in the produce.Also, more vegetables are being cultivated in urban areas calling for this study to ascertain the presence of heavy metals and faecal coliform in two species of vegetables cultivated in selected locations in the Accra metropolitan area to determine their wholesomeness for human consumption.This paper is among the few to use exploratory concurrent mixed methods research design to examine the spatial differentiation in levels of heavy metals and coliform bacteria in two vegetable crops that were cultivated under different sources of irrigation water.Interestingly, the study found no spatial variation concerning total coliform and faecal coliform count in the vegetable samples at the three sites studied.The total coliform and faecal coliform count levels were higher than what is recommended by the WHO.However, heavy metal concentrations for Pb and Cd in the vegetables were below permissible levels at the sites, except for Hg, which was higher than the permissible levels set by the World Health Organization at the WISS.This paper contributes to the literature on vegetable cultivation in urban areas of Ghana using streams by pointing out that mercury (Hg) concentration in vegetable samples at the WISS, which was thought to be good was much higher compared to the other two farming sites which, on the surface does not look good.The conclusion is that the consumption of lettuce and pepper from these three sites can pose potential health risks to consumers.In particular, this study has demonstrated that the location of vegetable farms can contribute to the risk of ingestion of Hg and toxic faecal coliform bacteria from contaminated vegetables irrespective of the irrigation scheme type (formal or informal) being used by farmers.This is evident when comparing Hg concentrations in the vegetables grown at the control site (WISS) where higher values were recorded in comparison with vegetables cultivated using wastewater sources such as those in Dzorwulu and Haatso, which recorded lower values.It should however be noted that farmers in all the study sites said they access water from streams and dams that are close to their farms to irrigate their crops.Thus, it is concluded in this paper that, the vegetables produced at the study sites, especially those at the WISS have high Hg and microbial contaminants as such may not be safe for human consumption.
Limitations of the study: One of the limitations of this present study is that vegetable samples were not taken from all important hotspots of vegetable production in the Accra-Tema metropolitan area, or farming areas in the country where vegetables are cultivated for exports.Doing this would have given credence to the conclusions on the country's applicability of the findings.Another limitation of the study is that the heavy metal concentrations were not assessed for the soils and water samples of the farms where the vegetables are being grown but only on the vegetable sources.This would have helped to pinpoint the specific source of the contamination.Another limitation of this study is the small number of vegetable samples ( lettuce and bell pepper) that were used.Related to this limitation is the number of times the samples were taken.Due to financial limitations, only one sample was collected and analysed.This can call into question the extent to which the results of heavy metals and faecal contamination of vegetables in the study can be generalized.These limitations aside, we feel this is a good starting point to conduct more studies on the fruit and vegetable industry in Ghana and other countries, especially vegetables produced from urban agriculture.This is because the fruits and vegetables industry has come to stay and should be made safe and sustainable.
Recommendations for future works: Ghana should also come out with a good framework for quality standards.Enhanced sensitization of the populace and the farmers in urban areas as a whole on quality issues and vegetable production and sources should be embarked upon.The general public should be sensitised to the health implications of consuming crops laden with heavy metals and faecal coliform.This education drive may help reduce the risk of consumers patronising urban vegetables cultivated using polluted water.Health risk assessment for consuming faecal coliform and Hg contaminated vegetables should be evaluated in future studies as well as S. Francis Gbedemah et al.
assessing the quality of water used to irrigate urban farms.
Research should be carried out on the vegetable farms that are being assisted in the various locations of the country.Also, research should be conducted on other vegetables like okra, onions, turia, tomatoes, herbs, cucumber and tinda to ascertain the status of heavy metals and faecal contamination in them.Further, comparative studies should be done on organic vegetables sourced from big retail stores, markets and organic home-based gardens to compare the levels of pollution in these crops in the country.In terms of the spatial location of vegetable farms, more studies should be embarked upon to compare the crops that are produced in urban areas and those cultivated in rural locations.
Table 1
Basic socioeconomic profile of the respondents.
Table 2
Percent distribution of the main sources of water used by vegetable farmers.
Table 5
Mean concentrations of heavy metals in pepper.
Table 6
Faecal and total coliform contamination levels of lettuce and pepper from the vegetable farms. | 2024-04-08T05:07:06.628Z | 2024-03-20T00:00:00.000 | {
"year": 2024,
"sha1": "c76a6e9de5c763d94c60e504b91e48be334ee20b",
"oa_license": "CCBYNC",
"oa_url": "http://www.cell.com/article/S2405844024039550/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c76a6e9de5c763d94c60e504b91e48be334ee20b",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
150056 | pes2o/s2orc | v3-fos-license | Control of traveling-wave oscillations and bifurcation behavior in central pattern generators
Understanding synchronous and traveling-wave oscillations, particularly as they relate to transitions between different types of behavior, is a central problem in modeling biological systems. Here, we address this problem in the context of central pattern generators (CPGs). We use contraction theory to establish the global stability of a traveling-wave or synchronous oscillation, determined by the type of coupling. This opens the door to better design of coupling architectures to create the desired type of stable oscillations. We then use coupling that is both amplitude and phase dependent to create either globally stable synchronous or traveling-wave solutions. Using the CPG motor neuron network of a leech as an example, we show that while both traveling and synchronous oscillations can be achieved by several types of coupling, the transition between different types of behavior is dictated by a specific coupling architecture. In particular, it is only the “repulsive” but not the commonly used phase or rotational coupling that can explain the transition to high-frequency synchronous oscillations that have been observed in the heartbeat pattern generator of a leech. This shows that the overall dynamics of a CPG can be highly sensitive to the type of coupling used, even for coupling architectures that are widely believed to produce the same qualitative behavior.
I. INTRODUCTION
Coupled limit cycle oscillators are used to model many physical phenomena in biology and engineering. While most prior studies have focused on synchronous oscillations, there has been a growing interest in traveling-wave oscillations, particularly in the context of central pattern generators (CPGs). These are neural circuits found in both vertebrate and invertebrate animals that can produce rhythmic patterns of activity without an external periodic drive. CPGs are fundamental to many periodic activities such as chewing, moving, and breathing and have inspired many robotic models of animal and even human locomotion [1][2][3][4]. The advantages of using CPGs to model locomotion in robots include [5] (i) the use of limit cycle behavior to produce stable periodic patterns, (ii) suitability for distributed implementation [6], and (iii) few control parameters, which allows for the modulation of locomotion [1].
A central problem in modeling CPGs using coupled limit cycle oscillators is finding and proving the stability of solutions that correspond to relevant phenomena, such as, for instance, different types of gait in animal locomotion. While it has been shown that it is possible to find solutions using spatiotemporal symmetries of the model [7], proving global stability of these solutions remains one of the central problems in CPG design and implementation [5]. Much work has been done showing stability of synchronous solutions (see, for example, Ref. [8] for a review). In this case, it is possible to prove local stability by linearizing around the synchronous state [9]. A particular challenge is proving global stability of out-of-phase oscillations. These oscillations model the traveling wave of electrical activity that is behind undulatory swimming.
In this work, we show that contraction theory [10] can be used to establish the global stability of a traveling wave, allowing more effective design of traveling waves of controlled amplitude, as well as synchronous oscillations. In a recent paper, a traveling wave generated by a double chain of oscillatory centers (representing groups of neurons on the two sides of the spinal cord) has been shown to be involved in both the swimming and walking behavior of a salamander [1]. Some work has been done in proving the stability of a traveling-wave solution in phase-coupled oscillators [3,11]. Phase coupling (which has no dependence on the amplitude of the oscillations) is frequently used to model weakly coupled limit cycle oscillators [12], where the dynamics is assumed to remain on a limit cycle. However, it has been shown that in weakly coupled cortical neurons off-limit-cycle dynamics contributes to synchronization by modulating the period of oscillation [13].
In the lamprey and salamander, both phase and amplitude dependence in coupling is needed to produce three different phases of the dynamics [1,14] corresponding to (i) a subthreshold phase without bursts, (ii) a bursting phase, with the frequency and amplitude of the bursts dependent on the drive, and (iii) a saturation phase, where no oscillations occur. Here, we give an example of three analogous types of dynamics in a leech, demonstrating that in this case as well, both amplitude and phase dependence in coupling are needed to account for the dynamics. The combined results suggest that the commonly used phase coupling may not be sufficient to model many CPG systems of interest.
Rotational coupling, which has both a phase and an amplitude dependence, has been previously shown to generate a stable traveling wave in nearest-neighbor coupled oscillators [15,16]. Here, we use phase-and amplitude-dependent inhibitory (repulsive) coupling to achieve a similar effect in a more physical way, in essence allowing the system itself to compute the couplings achieving the out-of-phase behavior.
Our basic model consists of a chain of nearest-neighbor coupled Andronov-Hopf limit cycle oscillators, whereby the nth oscillator is connected to the first oscillator in a ring structure. The system is given bẏ where x j = {x j ,y j } is a two-dimensional vector describing the dynamics of the j th oscillator, with F(x) given by Due to symmetry considerations [7], the synchronous state is one possible solution to the above equation, resulting in an amplitude of oscillation given by |x j | = √ α + |k|. This solution will be shown in the next section to be globally stable for diffusive types of coupling, given by k < 0. For repulsive coupling, given by k > 0, the system tends to an out-of-phase state, whereby the neighboring oscillators are maximally out of phase with each other [17]. This results in two different types of dynamics, depending on whether the number of oscillators, N, is even or odd. When N is even, the array oscillates with a difference of π between nearest neighbors, splitting into two equal synchronous groups that are 180 • out of phase with each other (see Fig. 1, plotted for N = 4). The phase difference between nearest neighbors is thereby given by A traveling-wave solution is possible when the number of oscillators is odd. Requiring the neighboring oscillators to be maximally out of phase while preserving the symmetry of the system leads to two possible degenerate solutions given by The smallest phase difference between the two oscillators is given by the next-to-nearest-neighbor phase difference: φ odd j,j +2 ± 2π/N. The vectors x j therefore fall on a circle where they are spaced with an equal phase difference of 2π/N, forming a symmetric out-of-phase solution. Figure 2 illustrates the stable out-of-phase dynamics for N = 5, where each oscillator is shifted in phase from its neighbor by π + π/5.
To solve for the amplitude of the out-of-phase solution for N even and N odd, we use Eqs. (3) and (4), respectively, and substitute these solutions for the nearest-neighbor terms into Eq. (1), resulting in Note that in Eq. (6), the amplitude of the oscillation is independent of the total number of oscillators, while in the traveling-wave amplitude, given by Eq. (5), it increases with increasing N , asymptotically approaching (α + 3k) 1/2 as n → ∞. We now prove the global stability of the synchronous and traveling-wave solutions when k < 0 and k > 0, respectively, for the N = 3 case (it is straightforward to extend our approach to higher N ). While our results are limited to nominally identical oscillators, they are robust to variations in individual parameters and the associated deviations from the ideal case can be explicitly quantified [15,18]. For nonidentical frequencies, the stability of phase-locked solutions is determined by Arnold tongues [8] that set an upper bound on the possible variation in the frequency of individual oscillators as a function of the coupling strength. Therefore, our analysis should be robust to variation of the parameters, as long as the coupling is sufficiently strong to overcome the spread in frequencies of nonidentical oscillators.
A sufficient condition for global exponential stability on a flow-invariant linear subspace M [i.e., a linear subspace M such that ∀ t : F(M,t) ⊂ M] as given by [15] 041914-2 where V forms a basis of the linear subspace M ⊥ (orthogonal to M), ∂F/∂x is the Jacobian of the uncoupled system, with F given by Eq. (2), and L is the coupling matrix, to be given below. The terms λ min and λ max indicate the minimum and maximum eigenvalues of the symmetric parts of the matrices L and ∂F/∂x, respectively. Intuitively, the above condition ensures that the system is contracting in the orthogonal subspace M ⊥ , thereby ensuring that the dynamics converges exponentially to M. From Eq. (1), the coupling matrix L for the N = 3 case is given by where I above is a 2 × 2 identity matrix. The entire phase space M ⊕ M ⊥ is spanned by a total of six vectors: the two vectors spanning the synchronous solution, plus the four vectors spanning a linear vector subspace formed by the two degenerate out-of-phase solutions. For convenience, let us define the subspace corresponding to the synchronous case as and the subspace corresponding to the out-of-phase states as where R is a 2 × 2 rotation matrix, with the rotation angles of {2π/3,4π/3} obtained from Eq. Having obtained all the eigenvectors, we can now use Eq. (7) to analyze the stability of either the synchronous or the out-of-phase states, represented by M sync and M phase , respectively. As will be shown shortly, this stability depends on the value of the coupling constant k, with the synchronous solution being stable for negative k (or a diffusive type of coupling) and the out-of-phase subspace being stable for positive values of k (representing repulsive coupling).
To analyze the stability of the synchronous state, we equate M sync ≡ M and M phase ≡ M ⊥ , thereby obtaining the 6 × 4 matrix V from Eq. (10). The four eigenvalues of VLV T , with L given in Eq. (8), are all identical and given by λ 1,2,3,4 = 2k. The eigenvalues of the symmetric part of the Jacobian, ∂F/∂x, are given by α − |x| 2 and α − 3|x| 2 , which are upper-bounded by α. It follows from Eq. (7) that the solution converges exponentially to M sync when k < −α/2.
Next, doing the reverse by equating M phase ≡ M and M sync ≡ M ⊥ , we again obtain the eigenvalues of VLV T [with V now given by Eq. (9)]: λ 5,6 = −k. Combining the above results and again using Eq. (7), we now get the condition for the global stability of M phase : Equations (11) and (12) give conditions for the global stability of synchronous and traveling-wave solutions, respectively. We now compare the model in Eq. (1), which has repulsive coupling, given by k > 0, to an alternative way of generating a globally stable traveling wave using either (a) rotational or (b) phase coupling. For rotational coupling, Eq. (1) becomeṡ While the rotational coupling given in Eq. (13) also produces a globally stable traveling wave [15,16], there are important differences in dynamics, including the following: (I) The existence of an out-of-phase solution for any number N of oscillators. This is in contrast to the system analyzed here, where an odd N is needed for an out-of-phase state. (II) The existence of a single (nondegenerate) out-of-phase solution, unlike the two solutions given in Eq. (10). (III) Different phase differences between nearest neighbors. Thus, the phase difference given by Eq. (4) is such that the nearest neighbors are maximally out of phase, while the phase difference from rotational coupling is such that oscillator j is advanced from j − 1 by a phase of 2π/N. (IV) With rotational coupling, the amplitude of the oscillation is preserved, as compared to the uncoupled case. This can be seen by directly substituting the out-of-phase solution x j = R 2π/N x j −1 into Eq. (13), whereby the coupling term drops out. In contrast, in the repulsive coupling, the amplitude of the traveling wave increases with the coupling constant, as can be seen from Eq. (5). Similarly, for phase coupling, the amplitude of oscillation is independent of the coupling, since the dynamics is assumed to remain on the limit cycle.
Here, we present an example where, due to property IV (i.e., the amplitude increase of a traveling wave), only repulsive but not phase or rotational coupling can explain an important qualitative change in behavior.
Example. The heartbeat pattern generator of a medicinal leech consists of two groups of neurons coupled together by a common element (namely, interneurons located in ganglion 5 [19]). The first group are called oscillator interneurons; they form mutually inhibitory synapses with each other. The second group, called premotor heart interneurons, have excitatory synapses. These two arrays are known to correspond to two types of motion: a traveling wave that beats in a rear to front (peristaltic) fashion, and a synchronous state.
It has been suggested [20] that this network can be represented by two globally coupled arrays: one with inhibitory (repulsive) coupling, corresponding to oscillator interneurons, and another one with local diffusive (excitatory) coupling, representing premotor heart interneurons. The global interaction term corresponds to the mutual coupling of the two arrays via ganglion 5, and depends on the overall activity of the two arrays.
The overall topology of this network is shown in Fig. 3, with dynamics represented by the following set of equations: FIG. 3. Topology of the network described by Eqs. (14) and (15) for the N = 3 case. The x array corresponds to oscillator interneurons, with mutually inhibitory (repulsive) local coupling, represented by connections with filled-in ovals. The y array corresponds to the premotor heart neurons, with excitatory (diffusive) local connections, shown by solid black arrows. The two arrays are globally coupled to each other, with the strength of the coupling given by c and dependent on the overall activity of the two arrays.
where k l > 0, F(x) is defined in Eq. (2), and N is the number of oscillators in each array. The superscripts r and d denote repulsively and diffusively coupled arrays, respectively. Note that in the absence of global coupling, c = 0, Eqs. (14) and (15) are the same as Eq. (1). Therefore for small c, Eq. (14) generates a stable traveling wave (corresponding to peristaltic motion), while Eq. (15) has a stable synchronous state, oscillating at frequency ω.
It was experimentally demonstrated [19] that the bursting in a leech occurs within a narrow range of parameters, corresponding to a bifurcation. This narrow range is separated on either side by zones of tonic spiking or silence. As the global coupling strength c between the two arrays is increased, an onset of high-frequency, Nω, oscillations is observed in a diffusively coupled array, at a bifurcation value given by c = c bif [20,21], as shown in Fig. 4. Note that this onset corresponds to this experimentally observed transition from tonic spiking to bursting [19].
It was shown in Ref. [17] that the bifurcation value for globally coupled limit cycle arrays can be calculated by replacing the global coupling term c N k=1 |x k | with cN|x 0 |, where |x 0 | is the amplitude of the corresponding stable solution in the absence of global coupling. The bifurcation values for the two arrays are then given by where c bif andc bif correspond to a bifurcation for diffusively and repulsively coupled arrays, respectively. The bifurcation constants C b andC b can be found by calculating when the three real roots of the intersection of theẋ = 0 nullcline with a vertical line disappear [22], resulting in where γ = (α + k l )/3 andγ = {α + k l [1 + 2 cos(π/N)]}/3. Using Eq. (16) and substituting into c bif andc bif , as well as using the amplitudes of the steady-state oscillations for the 14) and (15), α = 1, k l = 0.2, ω = 1/2,N = 3. a. |c| < c bif . Both arrays oscillate at frequency ω. b. c bif < |c| <c bif . High-frequency, Nω, oscillations, which correspond to experimentally observed bursting behavior, occur in the diffusively coupled array. c. |c| >c bif . Oscillator death.
In contrast, there is no range of the global coupling constant for which high-frequency oscillations are observed if the traveling wave is created by rotational rather than repulsive coupling. This corresponds to replacing the nearest-neighbor coupling term in Eq. (14) by the rotational coupling term given by Eq. (13). In this case |x d 0 |, and therefore γ , remains the same. However, |x r 0 | is now obtained by solvingẋ = F(x), since the rotational coupling term drops out in a steady state forming a traveling wave, resulting in |x r 0 | = √ α andγ = α/3. Since in this case c bif >c bif , there is no range of c for which Eq. (18) is satisfied and the onset of high-frequency oscillations does not occur.
For phase coupling the amplitudes of both out-of-phase and synchronous oscillations are the same, irrespective of the coupling topology, since the dynamics is assumed to remain on the limit cycle due to the weakness of connections, resulting in γ =γ . Again, in this case Eq. (18) is not satisfied for any value of c and either both arrays oscillate at frequency ω or oscillator death occurs if c > c bif [23].
In conclusion, contraction theory was used to establish the stability of our network designs using either diffusive or repulsive coupling and yet with both phase and amplitude dependence. Recent work has shown that such dependence is needed to explain the different phases of the dynamics in a salamander and in a lamprey [1]. We show that, similarly, this more complicated form of coupling (as compared to only phase coupling) is also needed to account for the three different types of dynamics in a very different model of a leech heartbeat. The results suggest that phase oscillators, although very common in modeling neural networks, may nevertheless be insufficient to model many CPGs of interest.
We also considered two types of phase-and amplitudedependent coupling, namely, rotational and mutually inhibitory (repulsive) coupling. In nature, very few examples of rotational coupling, as compared to inhibitory coupling, exist. Unlike rotational coupling, inhibitory coupling does not require a precise phase difference and therefore is much more robust to perturbations. We showed using contraction theory that inhibitory coupling can also create a stable traveling wave, similar to the one created by rotational coupling. This suggests that inhibitory coupling is a better and a more biologically plausible alternative in designing traveling waves that commonly occur in CPGs.
Our results also suggest that arrays that use phase-and amplitude-dependent inhibitory and excitatory coupling can be combined to design biologically inspired models of CPGs. We demonstrate by analyzing a model of a leech heartbeat, showing that by combining these two types of arrays one can account analytically for certain essential features of experimentally observed dynamics.
A.S.L. acknowledges support from a Marie Curie International Incoming Grant within the FP7 Framework. | 2016-03-14T22:51:50.573Z | 2012-10-01T00:00:00.000 | {
"year": 2012,
"sha1": "da0ec71ab42b7396f836e7f8f35a4e9492c273b9",
"oa_license": "CCBYNC",
"oa_url": "https://dspace.mit.edu/bitstream/1721.1/76203/1/Landsman-2012-Control%20of%20traveling-wave%20oscillations%20and%20bifurcation%20behavior.pdf",
"oa_status": "GREEN",
"pdf_src": "Anansi",
"pdf_hash": "0ca576e26a4a86e7c25e74d14995aeb84b878e6c",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Medicine"
]
} |
25737546 | pes2o/s2orc | v3-fos-license | The Insect Hemolymph Protein HP19 Mediates the Nongenomic Effect of Ecdysteroids on Acid Phosphatase Activity*
The activity of acid phosphatase (ACP) in insect fat bodies is stimulated by the steroid hormone 20-hydoxyecdysone (20E) in vivo. However, in fat bodies kept in culture, a factor from the hemolymph is required to enhance the ACP activity. We identified the factor as a protein with a molecular mass of 19 kDa (HP19) from the hemolymph of a lepidopteran insect, the rice moth, Corcyra cephalonica. Western analysis of hemolymph proteins with denaturing and non-denaturing PAGE using antibodies raised against HP19 suggest that this protein exists as a monomer. It is synthesized by the hind gut-associated lobular fat body of the larvae and is released into the hemolymph. The stimulatory effect of HP19 on the ACP activity is developmentally regulated and exhibits its maximal effect shortly before the onset of metamorphosis. We cloned the HP19 cDNA by immunoscreening a hind gut-associated lobular fat body cDNA expression library. Analysis of the amino acid sequence shows that HP19 belongs to the family of glutathione S-transferase (GST) like proteins. However, affinity-purified GST from Corcyra failed to show any mediation effect on 20E-stimulated ACP activity, and HP19 lacks GST enzymatic activity. Notably, HP19 mediates the hormone-stimulated ACP activity in intact fat body tissue and homogenates even in the presence of inhibitors of transcription and translation, suggesting a nongenomic mode of action. In addition, we show that HP19 inhibits the 20E-induced phosphorylation of the hexamerin receptor protein.
Insect metamorphosis, the transition from the larval to the adult stage of insects, is controlled by ecdysteroid hormones (1)(2)(3). The ecdysteroids, like the steroids in vertebrates, regulate gene transcription by binding to the nuclear receptors, which are ligand-activated transcription factors that convert the hormonal stimulus into a transcription response (4 -7).
Metamorphosis involves the breakdown of larval structures and the formation of new tissues (1). As a part of cell remodeling during metamorphosis, acidic autophagic vacuoles accumulate in the cells of the fat body, and the activity of several lysosomal enzymes such as acid phosphatases increase and cause the lysis of larval tissues (8 -11). The fat body fills a large fraction of the insect, and its function has been considered equivalent to the role of the vertebrate liver in the intermediary metabolism (12). It has been demonstrated that the stimulation of the lysosomal activity is governed by ecdysteroids (11,(13)(14)(15)(16). There is an indication that in this case the hormone possibly acts on a nongenomic level (17). Although the molecular mechanism of the genomic mode of steroid action is well known, the mechanism of nongenomic steroid action remains unclear to date (18). Earlier studies show that 20E 1 stimulates ACP activity in fat bodies in vivo but not in vitro (19,20). This result suggests that 20E, the active form of ecdysone, requires an additional factor (or factors) to enhance ACP activity. Hence, we have focused on the process of acid phosphatase activation by 20E in the fat body cells of our model insect, the rice moth Corcyra cephalonica. We report the appearance of a stage and tissue-specificregulated protein, HP19, in the hemolymph of Corcyra responsible for the activation of the 20E-dependent stimulation of ACP activity. This hormone-triggered activation is independent of gene transcription.
EXPERIMENTAL PROCEDURES
Insects-Larvae of the rice moth, C. cephalonica (Stainton), were reared on coarsely crushed sorghum seeds at 26 Ϯ 1°C, 60 Ϯ 5% relative humidity, and a 14:10-h light:dark photo period. In the present study, the last (ϭVth) instar larvae, classified into early (ELI), mid (MLI), late-last instar (LLI) larvae and prepupae on the basis of body weight and head capsule size were used (21). The larvae were thoraxligated behind the first pair of prolegs by slipping a loop of silk thread around the head of the larvae as described earlier (20). The fat body and other insect tissue homogenates from ligated or unligated larvae were prepared as published earlier (22) and used after protein estimation in an aliquot of the homogenate (23).
ACP Assay-The enzyme assay was carried out according to the method of Henrickson and Clever (24). The reaction mixture contained 150 mM sodium acetate buffer (pH 5.0) and 20 g of tissue homogenate proteins. It was incubated at 37°C for 10 min to exclude glucose-6phosphatase activity (25). The reaction was initiated by the addition of 5 mol of substrate, p-nitrophenyl bisodium phosphate (Sigma) to the assay mixture followed by incubation for 1 h at 37°C. The reaction was terminated by the addition of 0.5 ml of 0.1 N NaOH, and the color was measured at 420 nm against a substrate blank. The p-nitrophenol was used for the preparation of a standard curve. The activity of the enzyme was expressed as nmol of p-nitrophenol released/h/g of fat body protein.
Studies on Fat Body Cultures-Ribbon-shaped visceral fat bodies from LLI larvae were dissected 24 h after ligation under sterile conditions in cold insect Ringer's solution and transferred to 100 l of TC-100 insect culture medium (JRH Biosciences Inc.) with traces of streptomycin sulfate. After rinsing, the tissue was transferred to 200 l of fresh culture medium, and 80 nM 20E was added while an equal volume of carrier solvent (ethanol) was added to the control cultures. The hormone 20E (Sigma) was dissolved in ethanol, the final concentration of which never exceeded 0.05% in any of the experiments. To study the hemolymph effect, the diluted (1:20) or fractionated hemolymph was added to the fat body culture in the presence or absence of 80 nM 20E. Studies with glutathione S-transferase (GST) were carried out by adding purified cytosolic GST from Corcyra to the fat body cultures in the presence of hormone. These cultures were then incubated for 4 h at 25°C with gentle shaking. At the end of incubation the tissue was removed, rinsed in ice-cold insect Ringer, homogenized, and used for ACP assay.
Fractionation and Purification of Hemolymph Proteins-Total hemolymph protein was loaded on a pre-equilibrated (10 mM Tris-HCl (pH 7.4)) Sephadex G-50 column and eluted with the equilibration buffer. The single fractions were checked for their ability to enhance the ACP activity in LLI fat bodies kept in culture. The apparent molecular mass of the active fraction was determined by gel electrophoresis. The purification of the active fraction was carried out by fractionating the hemolymph proteins followed by gel filtration chromatography. For fractionation we used molecular weight cut-off fractionators (Amicon Inc.). The hemolymph sample preparation (1 mg of protein/ml) was first transferred into 30-kDa cut-off fractionators (YM-30) and centrifuged at 4000 ϫ g for 20 min at 4°C to strip the protein fractions with mass above 100 kDa. The resultant filtrate as well as the retentate was collected separately. The filtrate with molecules Ͻ100 kDa was again subjected to fractionation using a 10-kDa cut-off fractionator (YM-10) for obtaining fraction of Ͻ30 kDa proteins. Once again, both filtrate and retentate were collected. Three fractions, having masses above 100 30 -100 and below 30 kDa thus obtained, were used to test their effect on 20E-dependent ACP activity as described above. For the purification of the active fraction, hemolymph from ϳ500 LLI larvae was fractionated using the 30-kDa cut-off fractionator. The filtrate was loaded on a Sephadex G-50 column pre-equilibrated with 10 mM Tris-HCl (pH 7.4). The protein fractions were eluted with the same buffer. An aliquot of each fraction was added to the fat body kept in culture in the presence of 20E (80 nM), and its enhancing effect on ACP activity was analyzed. Peak fractions that stimulated ACP activity were pooled, and the purity was checked by SDS-PAGE.
Production of Polyclonal Antibodies against HP19 -The HP19 protein band was electroeluted using a model 422 electroeluter (Bio-Rad) after slicing the protein band resolved by 12% SDS-PAGE and used as an antigen for raising polyclonal antibodies in 3-month-old male rabbits (New Zealand variety). The IgG fraction was purified by protein Aagarose chromatography (Bio-Rad).
Electrophoresis and Western Blotting-SDS-PAGE was carried out using 2.1% stacking and 12% resolving gel (26). Non-denaturing PAGE was carried out as described by Burmester et al. (27). The resolved proteins were visualized by silver staining (28). For Western blotting, the electrophoretically separated hemolymph proteins were transferred to a nitrocellulose membrane (29) and probed with polyclonal HP19 antibody to detect a polypeptide with the capability of mediating the ACP activity in the hemolymph as well as in different tissues. Visualization of the specific cross-reactivity was carried with anti-rabbit IgG coupled with alkaline phosphatase followed by nitro blue tetrazolium chloride/5-bromo-4-chloro-3-indolyl phosphate color reaction.
Preparation of a Hind Gut-associated Lobular Fat Body (HGLFB) cDNA Expression Library and Immunoscreening-The cDNA from the RNA of HGLFBs was generated and amplified using the SMART III TM kit (Clontech). Total RNA was isolated using the TriFast kit (peqLab). The library was cloned in a pTriplEx2 vector and screened with the IgG fraction of anti-HP19 antibody (30). By immunoscreening of 6 ϫ 10 9 recombinant phage plaques, 10 positives were obtained after 3 rounds, in vivo excised, converted into plasmids, and used for XL-1Blue cell transformation (Stratagene, La Jolla, USA). The restriction analysis revealed six of these clones to be of identical size.
Sequencing of HP19 Clone-The inserts were sequenced on a PerkinElmer 310 sequencer using forward and reverse primers provided with the kit. About 60% of the positives were found to be GST-like sequences (BLAST search). One such clone was analyzed in detail. A multiple sequence alignment of deduced amino acid sequence with best matching invertebrate GSTs was carried out using ClustalW and GeneDoc 2.6. Southern and Northern Blotting-The genomic DNA from total larval body was isolated as described in Birren et al. (31) and was digested with EcoRI and HinfI. The digested DNA was resolved on 0.8% agarose gel. The total RNA isolated from different insect tissue was resolved on 1.2% agarose/formaldehyde gel and transferred to a nylon membrane. The HP19 cDNA insert corresponding to whole open reading frame was excised from the pTriplEx2 vector by EcoRI and NotI digestion, gelpurified, and random prime-labeled (MBI Fermentas) using [␣-32 P]dATP (ϳ3000 Ci/mmol, Board of Radiation and Isotope Technology (BRIT), Hyderabad, India) to a specific activity of Ͼ10 9 dpm/g. The blots were worked up following standard procedures (30).
GST Assay-The GST activity was measured by the method of Habig et al. (32). A 1-ml reaction mixture contained 10 l of 100 mM 1-chloro-2,4dinitrobenzene, 10 l of 100 mM reduced glutathione (GSH), and 100 mM potassium phosphate buffer (pH 6.5). The reaction was initiated by the addition of the enzyme source and the product, i.e. the formation of thioether conjugate, was measured at 340 nm on a time scan of 0 -60 s.
Immunohistochemical Studies-The insect tissues were fixed in Carnoy's fixative (ethanol:chloroform:acetic acid, 6:3:1) for 4 h at room temperature and paraffin-embedded. Sections (5 m) were cut and mounted on glass slides. For immunohistochemical staining the sections were deparaffinized and treated with blocking solution (2% bovine serum albumin and 1% preimmune goat serum in 50 mM Tris-buffered saline (pH 7.4) containing 0.1% Triton X-100) for 1 h at 4°C. This was followed by anti-HP19 IgG treatment for 24 h at 4°C with gentle shaking. The slides were then treated with anti-rabbit IgG coupled with alkaline phosphatase for 1 h. Washing after each step was done with three changes of Tris-buffered saline. These slides were finally processed for staining using nitro blue tetrazolium chloride/5-bromo-4chloro-3-indolyl phosphate color reaction and mounted in glycerol gel (50% glycerol, 7.5% gelatin, and 0.1% azide in 0.1 M Tris-buffered saline) (33,34). The specificity of the antibodies was checked for all immunochemical experiments by treating parallel tissue sections using pre-immune rabbit serum.
[ 35 S]Methionine Incorporation Studies-In vitro incorporation of [ 35 S]methionine in the fat bodies kept in culture was carried out to compare the changes in the level of total protein synthesis and ACP activity stimulation in the presence of 20E and HP19. The fat bodies kept in culture were first preincubated for 2 h with 10 Ci of [ 35 S]methionine (ϳ1000 Ci/m mol, BRIT) at 25°C followed by incubations either with 20E (80 nM) alone, HP19 (40 ng) alone, or both along with actinomycin D (1 mM) or cycloheximide (1 mM) for 2 or 4 h. After incubation, the fat bodies were removed and homogenized as described above. Equal amounts of protein were used for radiolabel quantitation after trichloroacetic acid precipitation.
In Vitro Phosphorylation of Fat Body Proteins-Fat body homogenate from LLI larvae was incubated with 80 nM 20E and 40 ng of HP19 for 30 s, and phosphorylation was initiated by the addition of 4 Ci of [␥-32 P] ATP (ϳ3000 Ci/m mol, BRIT) as described earlier (22). The labeled proteins were separated by 10% SDS-PAGE. The gel was vacuum-dried and exposed to Kodak X-Omat AR film at Ϫ70°C for autoradiography.
Statistical Analysis-The mean and S.D. were calculated for the variables studied. The data were statistically analyzed by one way analysis of variance followed by comparisons of means by Tukey multiple comparison tests using Sigma Stat software (Jandel Corp.). The values were considered significantly different from each other when p Ͻ 0.05.
Identification of a Protein in the Hemolymph of Corcyra That
Mediates the 20E-stimulated ACP Activity in the Fat Bodies of Thorax-ligated, Hormone-deprived Larvae-When insect larvae are ligated behind the first pair of prolegs, i.e. behind the hormone producing glands, the posterior part of the animal is known to be relatively free of endogenous ecdysteroids (35)(36)(37). Fig. 1, a and b, shows the effect of thorax ligation and injection of exogenous 20E on the fat body ACP activity in LLI larvae. The ACP activity declined gradually from 6 to 72 h after ligation. Because the ACP activity in the fat body was significantly lower after 24 h of ligation, this time period was used for all hormone manipulation studies. Hormone injections of 80 nM 20E, i.e. the physiological concentration (22,36), to 24 h postligated LLI larvae caused a significant increase in the ACP activity in fat bodies after 24 h compared with the solventtreated larvae (Fig. 1b).
To study the effect of hormone on the ACP activity of fat bodies kept in culture, the tissue was dissected from 24-h post-ligated larvae and cultured for 4 h in the presence of 80 nM 20E. The results show that 20E did not elicit any stimulatory effect, and the activity was more or less the same as in the controls (Fig. 1c). However, the addition of hemolymph from LLI larvae together with 20E caused a significant increase in the ACP activity (Fig. 1c). This observation suggests that the hemolymph contains a factor (or factors) required by 20E to stimulate the ACP activity in fat body cultures. When the hemolymph was treated with alcohol, heat, acid, alkali, or protease, no stimulation of the ACP activity by 20E could be observed, suggesting the proteinaceous nature of the factor (Fig. 1d).
Purification and Characterization of the Hemolymph Factor as a Protein Mediating 20E-stimulated ACP Activity-After loading total hemolymph protein on a Sephadex G-50 column, we eluted several fractions (Fig. 2) and checked their ability to mediate the 20E-stimulated ACP activity. We found an active protein fraction with a molecular mass of ϳ22 kDa, calculated from the elution profile (Fig. 2, inset), or 19 kDa, calculated from the mobility on a SDS-PAGE (Fig 3b). On the basis of these results we purified the active hemolymph protein first by fractionating the total hemolymph protein followed by gel filtration chromatography. The fractionation was carried out using 30-and 10-kDa cut-off filters. The filtrate from the 10-kDa cut-off filter gave a protein fraction that mediated the 20Estimulated ACP activity of fat body cultures by an increase from 0.7 to 1.2 nmol of p-nitrophenol release/h/g of protein (Fig. 3a). However, the protein yield in the filtrate obtained from the 10-kDa cut-off filter was insufficient to proceed for further purification. Therefore, the filtrate from the 30-kDa cut-off filters, in which the HP19 was contained, was used for gel filtration. The protein fraction eluted from the Sephadex column that mediated the 20E-dependent ACP activity resulted in a contaminant-free pure polypeptide band of 19 kDa (Fig. 3b). Hence, the active hemolymph protein was named as HP19. Starting with 50 mg of total hemolymph protein, we obtained a 98.5-fold purification with 0.05% yield.
The Western blots presented in Fig. 3c show the specificity of the HP19 antibody both on denatured (lane 1) and non-denatured (lane 2) PAGE. A single protein band of 19 kDa suggested a monomeric structure of HP19. Southern analysis of genomic DNA (Fig. 3d) digested with EcoRI or HinfI probed with HP19-cDNA revealed HP19 as a single copy gene.
cDNA Cloning and Sequence Analysis of HP19 -To identify the cDNA encoding the HP19 protein, a cDNA expression library, prepared from the RNA of HGLFB of LLI larvae, was immunoscreened. We picked 10 positive cDNA clones for detailed examination. The restriction analysis revealed 6 of the 10 clones to be of identical size. Initial sequencing study dem- onstrated significant sequence similarity among these clones. Furthermore, they showed homology with invertebrate GSTs. One of our clones was sub-cloned and totally sequenced (Gen-Bank TM accession number AY369240). This HP19 cDNA was 634 nucleotides long, with an open reading frame of 585 bp, which encodes a protein of 195 amino acids. The calculated molecular mass of the translated unmodified protein was 22.95 kDa, which is close to the mass of HP19 detected in HGLFB, the tissue that synthesizes the protein (see Fig. 6e). The polypeptide comprises 12.3% basic (9 Arg, 1 His, and 14 Lys) and 13.3% acidic residues (10 Asp and 16 Glu) but no Cys residue. The estimated isoelectric point (pI) is 5.36. Comparison of the C. cephalonica HP19 (CcHP19) cDNA with the sequences in the GenBank TM showed 67% identity with Choristoneura fumiferana GST (CfGST) (38). Similarities of HP19 cDNA with other invertebrate GST were found to be less than 38%. The comparison of the amino acid sequences of CcHP19 with the four best matching invertebrate GSTs is shown in Fig. 4. Although the CcHP19 cDNA sequence revealed 67% identity with CfGST, affinity-purified GST from Corcyra had no enhancing effect on the 20E-dependent ACP activity when compared with purified HP19 or recombinant HP19 (Fig 5a). Furthermore, the hemolymph as well as the purified HP19 had negligible GST activity (Fig. 5b).
Tissue-specific Appearance of HP19 -Co-culturing of different larval tissues with fat body demonstrates that the HGLFB is the only HP19-synthesizing tissue. A stimulation of the ACP activity by 20E was only observed when it was co-cultured with HGLFB (Fig. 6a). The hemolymph used in all experiments was cell-free and, therefore, cannot be the site of HP19 synthesis. The tissue specificity of HP19 biosynthesis was further confirmed by immunohistochemical staining of different tissue sections using HP19 antibody. Again, HP19 was found to be localized only in HGLFB (Figs. 6, b and c). Western analysis of proteins from different tissues also revealed the presence of HP19 only in HGLFB (Fig. 6e, lane 2) and total larval body protein (Fig. 6e, lane 5). However, the apparent mass of HP19 in these was ϳ5 kDa higher than the HP19 present in hemolymph (Fig. 6e, lane 1). Northern hybridization with total RNA from different tissues and from total larval body displayed the tissue-specific expression of HP19 gene in HGLFB (Fig. 6f, lane 2). The faint band of HP19 in the Western blot (Fig. 6e, lane 5) and the apparent absence of HP19-mRNA in the total larval body, which comprises the HGLFB, is probably due to the limitation of the methods used.
Developmental Regulation of HP19 -The developmental profile studies of HP19 in Corcyra during the last (ϭVth) instar larval development suggest that only the hemolymph of LLI larvae could mediate the 20E-stimulated ACP activity (Fig. 7a). Western analysis of proteins from hemolymph (Fig. 7b) and HGLFB (Fig. 7c) of different developmental stages of Vth instar larvae show that HP19 is present at a maximal concentration in LLI (lane 3). HP19 is present in the hemolymph at all the developmental stages tested (Fig. 7b, lanes 1-4) but was not detectable in the HGLFBs of prepupae (Fig. 7c, lane 4). The intensity of the HP19 band in ELI and MLI was low compared with LLI (lanes 1-3). These results suggest that HP19 is synthesized throughout the complete last larval stage in HGLFB and is released into the hemolymph. The synthesis rate is low in ELI larvae, and the secretion into the hemolymph is rapid with exception of the LLI stage, when the HP19 synthesis is accelerated and is paralleled by the activation.
Nongenomic Regulation of 20E-stimulated ACP Activity by HP19 -When the fat bodies kept in culture were incubated with 20E and HP19 for 4 h, we observed that the stimulation of the ACP activity remained unaffected in the presence of transcriptional or translational inhibitors (Fig. 8a). The results in Fig. 8b further indicate the nongenomic regulation of 20Estimulated ACP activity by HP19 because the addition of protein directly to the fat body homogenate also mediated the steroid-stimulated action. This effect was rapid and could be observed within 30 s to 1 min (Fig. 8b). We further confirmed the nongenomic regulation of 20E-dependent ACP activity by incubating the fat bodies kept in culture first with [ 35 S]methionine for 2 h followed by incubation with hormone, HP19, and transcriptional or translational inhibitors. The results (Fig. 9) show that the total protein synthesis is induced under the influence of 20E, and this induction is inhibited by actinomycin D and cycloheximide (Fig. 9a). However, the inhibitors did not affect the ACP activity. The HP19 enhanced the 20E-stimulated ACP activity (Fig. 9b).
Phosphorylation of the Hexamerin Receptor by 20E-We have recently reported that 20E stimulates the phosphorylation of the Corcyra hexamerin receptor, which is partly mediated by a tyrosine kinase at a nongenomic level (22). Therefore, it was pertinent to check whether HP19 had any effect on hexamerin or tyrosine kinase phosphorylation. In vitro phosphorylation of fat body homogenate proteins from Corcyra resulted in the phosphorylation of mainly three proteins of masses 120, 60, and 48 kDa (Fig. 10). The addition of HP19 either in the absence (lane 3) or in the presence of 20E (lane 4) inhibited the basal as well as 20E-stimulated phosphorylation of the 120-kDa protein, which was identified earlier as the hexamerin receptor (22). DISCUSSION Insect development and reproduction are mainly controlled by a complex interaction between two groups of hormones, the ecdysteroids and the sesquiterpenoid juvenile hormones. The prominent functions of ecdysteroid hormones are their ability to trigger the temporal sequence of developmental process underlying molting and metamorphosis (1,2), one of the most dramatic event in the animal kingdom. The hemolymph, i.e. the insect "blood," is known as a source of several factors that do not only regulate ecdysteroid synthesis but also mediate ecdysteroid-dependent actions in larvae and pupae (39 -45). 20E has been shown to elicit effects on autophagic processes of the fat body by stimulating the lysosomal enzyme activity of ACP (10,11,(13)(14)(15)(16).
Thorax ligation that renders the posterior part in larvae relatively free of endogenous hormone showed a gradual decline in larval fat body ACP activity, pointing to the hormonal dependence of ACP activity (Fig. 1). The injection of a physiological dosage of exogenous 20E into ligated larvae caused a stimulation of ACP activity. However, when 20E alone was added to fat bodies kept in culture the hormone could not stimulate the enzyme activity. Only co-treatment of the cultured fat bodies with 20E together with hemolymph enhanced the ACP activity. This result confirmed our earlier findings of the presence of some hemolymph factor that mediated the 20E-dependent ACP activity in vivo (20). In this study we identified and characterized this hemolymph factor as a 19-kDa protein (HP19) and studied its synthesis as well as its developmental regulation. The novel function of the protein prompted us to devise a protocol for purification. Although a few nanograms of the protein were found to be sufficient for stimulation even in crude or partially purified fractions, the yield of the purified protein was very low. Other limitations were the requirement of a large quantity of hemolymph of a specific developmental stage (LLI) and removal of the major contaminating protein, hexamerin, which constitutes 75-80% of total hemolymph protein (46). Therefore, an antibody against HP19 was raised by electroeluting HP19, and its specificity against HP19 was 5. Effects of HP19 on the ACP activity in fat bodies compared with GST. a, ability of affinity-purified cytosolic Corcyra GST on 20E-dependent fat body ACP activity. Note that the presence of purified or cloned HP19 (CcHP19) mediated the 20E-stimulated enzyme activity, whereas the presence of GST did not show any effect. b, GST activity in different larval tissues and in purified HP19. Note that both hemolymph as well as purified HP19 have negligible GST activity. PNP, p-nitrophenol.
20E-stimulated ACP Activity Regulation by HP19
confirmed. When this antibody was added together with hemolymph to cultures of LLI fat body, the hemolymph failed to mediate the 20E-dependent action. Similarly, when the hemolymph was first immunoprecipitated using HP19 antibodies and the resulting complex and the supernatant were added to the fat body cultures, HP19 action was suppressed (data not shown).
Western analysis of denatured as well as non-denatured PAGE demonstrated that HP19 is a monomeric protein in Corcyra and is obviously the product of a single copy gene (Fig. 3d). Three independent methods, co-culturing, Western analysis, and immuno-histochemistry, revealed that HP19 is synthesized by the HGLFB, from where it is released into the hemolymph. This is further confirmed by the tissue-specific gene expression only in HGLFB. Western analysis provides evidence of a difference of ϳ5 kDa in the mass of HP19 in HGLFB (i.e. 24 kDa) and in hemolymph (19 kDa). The predicted mass of the unmodified translated protein from the HP19 cDNA was close to the HP19 synthesized in HGLFB. Furthermore, the 0.66kilobase HP19-mRNA (Fig. 6f) indicates a mass of about 24 kDa synthesized in HGLFB. The amino acid sequence deduced from HP19 cDNA did not show a typical signal peptide necessary for transmembrane transport (47), probably due to the lack of any hydrophobic sequence (38). Therefore, we conclude that HP19 might be cleaved before its release into the hemolymph. The presence of two putative N-glycosylation sites (Asn-51-Arg-52-Thr-53-Leu-54 and Asn-116 -Glu-117-Thr-118 -Ala-119) indicates that the protein can be secreted from the synthesizing cells.
The biosynthesis of HP19 takes place in the HGLFB during the total last larval instar. The protein is rapidly released into the hemolymph. The maximal HP19 concentration in the tissue as well as in the hemolymph could be observed in LLI. It is notable that only hemolymph from this developmental stage is capable of stimulating the 20E effect on ACP activity, i.e. the activity of HP19 is developmentally regulated. However, the molecular mechanism of activation of ACP by ecdysteroid hormone is unclear at this point.
To gain more insight into the nature and function of HP19, we produced and characterized CcHP19-cDNA clones and compared the sequence with those of related proteins, particularly with that of CfGST, which exhibits 67% identity. The CcHP19 does not contain specific amino acids that are responsible for glutathione binding or are involved in modulating the specific activity of mammalian , , and GSTs (38, 48 -50). The HP19 sequence also showed other putative post-translational modification sites like protein kinase C (189 -191), casein kinase II (81-90, 124 -127, 142-145, and 161-164), and tyrosine kinase (19 -25 and 33-41) phosphorylation sites as well as N-myristoylation (138 -143 and 150 -155) sites. However, in vitro phosphorylation studies with hemolymph did not reveal any phosphorylation of HP19 (data not presented).
GSTs with hormone-regulating actions are unknown, although few studies on vertebrate GSTs speculate about their steroid binding properties or their developmental and hormonal regulation (51)(52)(53)(54). In the present study we also checked the possibility of GST exhibiting HP19 function. For this purpose we tried to replace HP19 with affinity-purified GST and tested whether GST can exhibit the same effect as of HP19. The CfGST shows Ͼ60% sequence identity with HP19 (38), and the CfGST antibody revealed immunocross-reactivity with different tissue proteins in the mass range of HP19 (data not presented). The affinity-purified GST from Corcyra did not show any potentiation in 20E-stimulated ACP activity like that of HP19 purified from the hemolymph or recombinant HP19. In addition, the GST activity in the hemolymph was very low and negligible in the purified HP19. All these studies together show that HP19 has no GST activity. During the purification of cytosolic GST of Corycra by affinity chromatography, the flowthrough fractions that did not bind to the affinity matrix mediated the 20E effect on ACP activity, further indicating that HP19 is different from GSTs (data not presented).
A wealth of data on the molecular mechanism of ecdysteroid action shows that the transcriptional cascade leading to molting and metamorphosis is initiated when 20E binds to its nuclear receptor complex (2). For about four decades evidence has accumulated that some of the hormonally induced effects seemed to be too rapid for the classical model (18,55). This evidence casts doubt on the so-called genomic pathway as the sole mode of steroid action. Today, several modes for nongenomic steroid actions are examined. Most of them are thought to continuously modulate the long term program allowing cells or organs to adapt rapidly to environmental changes. Numerous experiments with many different species show that insect metamorphosis in general is under the genetic control of ecdysteroids. A few studies indicate that some events necessary for and accompanying metamorphosis are controlled by 20E at a nongenomic level. However, studies on these mechanisms are restricted to a small number of experimental systems, e.g. the activation of lysosomal enzymes and the hexamerin receptor (17,22,37).
To learn more about the molecular mechanisms of ecdysteroid in regulating the HP19-assisted ACP activity, the fat bodies in culture were incubated for several time periods with HP19 and 20E. A minimum incubation of 4 h was essential for the stimulation of the enzyme activity by 20E, time enough for a genomic hormone action. However, the measured stimulation was unaffected by transcriptional or translational inhibitors, indicating the independence of gene activation. Furthermore, the in vitro study with fat body homogenate showed a rapid stimulation (within seconds to 1 min) of the enzyme activity. That the homogenate preparation is essentially a cell or nucleus disintegrated fraction suggests cell/nuclear integrity is not an essential requirement for the effect of 20E on ACP activity. This possible nongenomic regulation was further strengthened by the results that in the presence of 20E fat body cultures showed a higher incorporation rate of [ 35 S]methionine, which was inhibited by transcriptional or translational inhibitors. This, however, had no effect on the ACP activity, and the presence of HP19 in the tissue culture rendered 20E capable of stimulating the enzyme activity even in the presence of inhibitors of transcription and translation.
The construction of adult tissues during metamorphosis requires large amounts of energy and building blocks, which are in all the insects investigated so far and are provided by "storage proteins." The major fraction consists of hexamerins, which are sequestered by the larval fat body by receptor-mediated endocytosis, stored in vacuoles, and utilized by the activity of lysosomal enzymes (46,56,57). The hexamerin receptor of Corcyra, a 120-kDa protein, is activated through phosphorylation by a tyrosine kinase in the fat body membrane (22). Therefore, it was tempting to extend our study and to check the effect of HP19 on the 20E-stimulated phosphorylation of the hexamerin receptor. As demonstrated, HP19 inhibits the 20E-induced phosphorylation of the hexamerin receptor. The physiological significance of this inhibition lies in the experimental evidence that uptake of hexamerin is a stage-specific event (46,56). As we have shown earlier, the receptor is present during the last larval instar in an inactive form and is not capable of sequestering hexamerins, i.e. binding its ligand (22,58). Furthermore, it is well documented that the ecdysteroid titer is very low during the last larval instar but increases rapidly (10-fold or more) before pupation (36). We could demonstrate that the hexamerin receptor becomes phosphorylated at this stage and is able to bind hexamerin. Moreover, we have shown that 20E is responsible for the phosphorylation (receptor activation) by enhancing the activity of a tyrosine kinase (22). From the present study it is evident that phosphorylation of the receptor is inhibited by HP19 in the larval stage, when hexamerin is not sequestered by the fat body. This suggests the existence of an inbuilt regulatory mechanism to prevent the ecdysteroid-dependent hexamerin uptake by the larval fat body cells, a stage when the fat body tissue actually synthesizes both the hexamerin and its receptor. Later, at the prepupal and pupal stage when hexamerin uptake occurs, HP19 gets inactivated, which is evident by its inability to mediate the 20E regulated ACP activity at this stage. However, like ACP, the dual requirement of both HP19 and 20E remains elusive, as HP19 alone could also inhibit the receptor phosphorylation. The receptor phosphorylation occurs in intact fat bodies in the membrane fraction from fat body as well as in the fat body homogenate, suggesting a nongenomic regulation of the phosphorylation of hexamerin receptor by 20E (22). The selective inhibition of the receptor phosphorylation in fat body homogenates suggests that HP19 most likely has a kinase inhibitory/phosphataseactivating effect on the phosphorylation of the hexamerin receptor and possibly acts by inhibiting the fat body tyrosine kinase activity. Further work needs to be done to broaden the understanding of nongenomic regulation by ecdysteroids and the regulatory role of HP19 on the 20E-stimulated acid phosphatase activity and herewith on the phosphorylation of hexamerin receptor in insects. | 2018-04-03T05:51:09.237Z | 2004-07-02T00:00:00.000 | {
"year": 2004,
"sha1": "97634cc7f0ec23108203b86032f2b373d3093fe3",
"oa_license": "CCBY",
"oa_url": "http://www.jbc.org/content/279/27/28000.full.pdf",
"oa_status": "HYBRID",
"pdf_src": "Highwire",
"pdf_hash": "30fb7db144c9cfca474a6709ffd79a5ad2e6da8e",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
235254271 | pes2o/s2orc | v3-fos-license | PINGAN Omini-Sinitic at SemEval-2021 Task 4:Reading Comprehension of Abstract Meaning
This paper describes the winning system for subtask 2 and the second-placed system for subtask 1 in SemEval 2021 Task 4: ReadingComprehension of Abstract Meaning. We propose to use pre-trianed Electra discriminator to choose the best abstract word from five candidates. An upper attention and auto denoising mechanism is introduced to process the long sequences. The experiment results demonstrate that this contribution greatly facilitatesthe contextual language modeling in reading comprehension task. The ablation study is also conducted to show the validity of our proposed methods.
Introduction
Reading Comprehension of Abstract Meaning (Re-CAM) (Zheng et al., 2021) is a cloze-style task, which takes a document and related human written abstract with one word replaced by a placeholder as input. The model is required to choose the best word from five candidates. The ReCAM consists of three subtasks. In subtask 1 and 2, participating systems are required to choose the best imperceptible concept and hyper-nyms concepts word respectively. Subtask 3 aims to evaluate performance of a system trained on one definition and test on the other.
Traditional cloze-style reading comprehension model (SA reader) (Kadlec et al., 2016) uses attention to directly pick the answer from the context, which makes model incapable to answer the question where the answer does not appear in passage. Furthermore, GA reader (Dhingra et al., 2017) adopts multi-hop attention mechanism to build query-specific representation of answer for ranking the candidates which is not part of passage.
Pre-trained language models (Radford et al., 2018;Devlin et al., 2019;Yang et al., 2019;Lan et al., 2020) have been widely adopted for context modeling in many Natural Language Processing tasks. These models are pre-trained on huge corpora with plain texts and can better model contextual dependencies of tokens, thus enhance the performance of downstream approaches. As described in (Lai et al., 2017;Zhang et al., 2020;Chen et al., 2019;Zhu et al., 2018), they improved the performance of single-choice reading comprehension tasks by introducing pre-trained model, but they takes excessive memory for concatenating each option with the question and the passage.
GPT2 (Radford et al., 2018) has outperformed the SOTA result on cloze-style task CBT (Hill et al., 2016). As stated in (Radford et al., 2018), GPT2 computes the probability of each choice and the rest of the sentence conditioned on this choice according to the pre-trained model, the answer is the choice with highest probability. GPT2 outperforms RNN-based models without fine-tuning on CBT task, we assume that pre-trained language model has better potential to address the cloze-style problem than fine-tuning the pre-trained model with an additional ranking network.
The most relevant work to our model is Pattern-Exploiting Training (PET) (Schick and Schütze, 2020a,b), which proposed to reformulate the sentence classification task to cloze-style task with defined golden answer word as supervising signal. The comparison between PET and our proposed method is reported in section 5. Different with PET, We propose a novel Auto Denoising Discriminator for Abstract Concept in reading comprehension (ADDAC) by fine-tuning the pre-trained discriminator of ELECTRA (Clark et al., 2020). Auto Denoising is introduced while processing long sequences. By fine-tuning the pretrained model on its own structure with the original pre-training loss , the tasks results is significantly improved even with small train dataset, we suppose the representations stored in the pre-trained model has been maximum reserved in this way.
Task Description
The task intends to answer a cloze-style question, the answer to which depends on the understanding of a context document provided with the question. The model is also provided with a set of possible answers from which the correct one is to be selected. This can be formalized as follows: The training data consists of tuples (q, d, a, A), where q is an abstract sentence of document d and one word in q is replaced with a placeholder. A is a set of possible options and a ∈ A is the golden answer. Both q and d are sequences of words and golden answer a does not appear in the article d.
ELECTRA vs ALBERT
Since the success of BERT (Devlin et al., 2019), pre-trained language models have adopted a large amount of parameters to achieve better modeling performance. ALBERT (Lan et al., 2020) uses factorized embedding parameterization and cross-layer parameter sharing to greatly reduce the amount of model parameters and achieved SOTA in multiple natural language understanding task. ALBERT outperforms other pre-trained language models which is trained by MLM(masked language model) in combination with PET method for this task.
ELECTRA (Clark et al., 2020) proposed the RTD (replaced token detection) task with adversarial learning as an alternative to the MLM task. A smaller generator is used to replace the special token [MASK] in training samples, and then a discriminator is trained to predict each word in the input is real or generated by generator. In section 3.1, we will elaborate the details about our proposed discriminator mechanisms based on ELECTRA. The performance comparison between ALBERT with PET and our proposed discriminator model on ReCAM task is reported in Section 5.
System overview 3.1 Pre-trained Discriminator
We approach the competition tasks as cloze-style task, which can be reformulated to masked language modeling (Devlin et al., 2019) problems. As shown in Figure 1, we replace placeholder with golden and negative options in question q denoted as q A , which is further concatenated with corresponding document d in pre-trained model input Figure 1: Discriminator overview, the placeholder in q is replaced by a which is one of candidate options, the label of golden option is 0 followed the ELECTRA pre-training setting.
style ([CLS]q A [SEP ]d[SEP ]
). We ignore the part of sequence which exceed the maximum length of input sequence . The input sequence is forward to ELECTRA discriminator and the hidden states are calculated by Equation 1 where F is the pre-trained 24-layer transformers and D is a linear layer which classify the hidden states of replaced word from ELECTRA Discriminator. H q A ∈ R N ×d are the hidden states of input sequence, in which N and d are the maximum length of input sequence and the dimension of hidden states. Then we only use the hidden states of placeholder H p ∈ R d selected from H q A as the input to D. BCE (Binary Cross Entropy) loss, which measures the Binary Cross Entropy between the golden label and the output, is used for binary classification as Equation 3.
where l p is the binary label of option word (replacing placeholder in input sequence). The label is set to 0 for golden option, 1 for negative options, which is the same as ELECTRA pre-training setting (Clark et al., 2020). While in inference, the option with lowest logit p is regard as the right answer to the question. The experiment results show that discriminator outperforms the ranking and PET implementation on this task (see section 5).
We also implement a ranking model based on ELECTRA for single-choice reading comprehension task to compare with the Discriminator approach. The question and document is combined with each option as the input of ELECTRA encoder, which is denoted as S ∈ R N ×m , where N is the length of sequences and m is the number of options. A linear layer and a softmax layer is added after ELECTRA encoder as the ranking network, which use the hidden states of [CLS] in S as input.
Processing Long Input Sequence
Most of the pre-trained models have a limitation in processing long sequences. The maximum sequence length of ELECTRA is 512, which is much shorter than the maximum length of input sequence (i.e. concatenation of question and document). We propose two methods for processing long input sequences: 1) document is segmented to shorter passages, which leads to the problem of mislabel samples. We introduce an auto denoising mechanism to address the problem. 2) We adopt an upper attention upon transformers.
Auto Denoising The whole document d is segmented into a set of fragments {s d 1 , s d 2 , · · · , s d k } with a fixed window size. Then, these fragments combine q a to form model input sequence. In the inference phase, the lowest predicted logit is selected from all results of fragments as Equation 4.
where q with placeholder replaced by a specific option from A denote as q A . However, this method causes the noisy-label problem. Supposed that the answer a just finds proof from fragment s d 1 , which results in other samples (0, [q a ; s d i =1 ]) to be mislabeled samples, which significantly impact training of model and decrease the prediction accuracy. Therefore, we take the advantage of a noise-tolerant loss (bi-tempered logistic loss, BT (Amid et al., 2019)) and a noise detection method (over-fitting to under-fitting, O2U (Huang et al., 2019)) in this work. The BT logistic loss lowers gradient on noisy samples which relieve the negative effects on model training via adjusting bi-temperature. The O2u makes full use of the property that model is easier to forget the mislabeled samples than the clean samples, to identify and filter the mislabeled samples.
Upper Attention
The long sequence of input is segmented into small segments with the length of 512 tokens and each segments are concatenated with same question to form the input sequences, which are encoded into hidden states H i ∈ R d×512 , where i is the index of segments of passage, the d is the dimension of hidden states and 512 is the sequence length. The hidden states of placeholder is denoted as H p i ∈ R d . We use a 1-layer multihead self-attention block to fuse the hidden states of placeholder from multiple segments output.
where k is the number of segments, A g is 1-layer multi-head-attention, without residual connection. H p f use is applied in Equation 2 and Equation 3 for training.
Optimizer
The ELECTRA-large which has large amount of parameters tends to over-fit on small training dataset. We utilize RecAdam to finetune the pre-trained model to address the overfitting problem. RecAdam optimizer is proposed to address the catastrophic forgetting problem of sequential transfer learning paradigm by introducing a recall and learn mechanism into Adam optimizer, which maintain the learned knowledge in pre-trained model while learning a new task.
Data augmentation
To further boost the performance of our proposed model, we conduct data augmentation.We random pick 3000 articles in CNN/DailyMail (Hermann et al., 2015), and craw 824 latest articles from BBC news website. The CNN news is much longer than the training samples, while the length of BBC news is approximately same. The extra training samples are generated in following steps: 1) The title of news article is used as abstract. 2) We pick one most meaningful word from abstract by TF-IDF scores and the origin word is used as golden option. 3) We use words with same category of POS (Part Of Speech) from other documents as negative options. We train models on the extra to dataset as warming up and further train the models with training dataset. Unfortunately, extra training data did not effectively improve the performance of our model on this task. We just use extra data in models ensemble phase.
Experimental setup 4.1 Data
We use the official released dataset of SemEval2021 Task4 for experiments. The dataset of subtask 1 contains 3227/837/2025 samples for train, dev and test data. The subtask 2 dataset contains 3318/851/2017 samples for train, dev and test data. The maximum/mean length of subtask 1 and subtask 2 training data are 2275/374.34 and 2274/578.31, respectively. The statistics of sample length shows that length of article in subtask 2 is much longer than the length in subtask 1. The rate of the sequences exceeding 512 tokens is 13.95% in subtask 1, 42.46% in subtask 2, which may cause that the methods for processing long sequences are more effective in subtask 2.
Since the provided dataset is small, we apply data augmentation in model ensemble to further improve generalization of the model (see Section 3.4).
Parameter settings
Our implementation is based on the Pytorch framework for transformer-based models (Wolf et al., 2020). We trained our model based on the pretrained ELECTRA-Large discriminator, and adopt the same model structure for subtask 1 and subtask 2. We use Adam optimizer with a learning rate of 1e-5, batch-size of 32 to train the baseline model, which is actually the ELECTRA discriminator. The max sequence length is 512 and the epoch of training is set to 5. To address the overfitting problem, we apply RecAdam with sigmoid annealing function, where the annealing rate is 0.01 and the annealing time-step is 500. The coefficient of the quadratic penalty is set to be 5,0000. For Auto Denoising, the two temperatures and label smoothing of BT are equal to 0.9, 1.5 and 0.1 respectively. The maximum learning rate, minimum learning rate and epochs in cyclical round about O2U are set to be 5e-5, 1e-6 and 5.0 respectively.
Since only 5 submissions are permitted in submitting phase, we trained multiple models under different settings for model ensemble. We also adopt 8-fold cross-validation training to improve the model generalization.
Ensemble
Two strategies are used for our final submissions on test data: 1) we ensemble all 8 models from 8-fold cross-validation training by averaging their outputs, which is trained on the train data of each subtask; 2) we trained multiple models on the train data and the augmented data with different model structures. 7 top different models are selected based on the dev accuracy for models ensemble, then average their outputs as the final output. Moreover, the model trained in cross-validation is Discriminator with Auto Denoising and RecAdam optimizer for subtask 2, and Discriminator with RecAdam for subtask 1. While in top ensemble, the techniques of Discriminator, Auto Denoising, RecAdam and Upper Attention are applied in different models.
Single Model Performance
We implement two baseline models for comparison with our proposed method. The ALBERT PET is the combination of PET method (Schick and Schütze, 2020a,b) and ALBERT-xxlarge model, which is trained by the MLM. The golden answer of training dataset is used as the target for MLM, but the negative ones are omitted in training. The ELECTRA Rank reformulates the cloze-style task to single-choice task, which is described in section 3.1.
Label accuracy is the official metric of the tasks and Table 1 shows the results on development dataset. Task3(1-2) is the subtask 3 which is trained on subtask 1 dataset and test on subtask 2, Task3(2-1) means train on subtask 2 and test on subtask 1. It is obvious that our proposed ELECTRA Discriminator significantly improve the performance of all tasks and outperform the best baseline by 4.7%/1.16%/3.76%/8.85% in subtask1/subtask2/subtask3(1-2)/subtask3(2-1) respectively. This confirms our hypothesis that pretrained language model has more potential for cloze-style task. The PET with ALBERT is not suitable for this task because it can not utilize the negative options. The ELECTRA Rank performance is unsatisfactory, suggesting that fine-tuning on ranking network damage the knowledge stored in the pre-trained model.
The results of ablation study are also reported in Table 1. Disc (Discriminator) with RecAdam achieves further improvement in all tasks by 0.25% / 0.53% / 0.25% / 0.26% in subtask1 / subtask2 / subtask3(1-2) / subtask3(2-1) respectively, which prove the RecAdam optimizer is more effective for pre-trained model and also promotes the model generalization. Disc+Upper, Disc+AutoDenoise+RecAdam achieve the best results in task2 and task3(2-1) respectively. This prove the validity of the two methods we proposed for long sequences. The Upper Attention achieve the best result on task2, while AutoDenoise achieve the best result on task3. We ensemble them to produce a stronger system. The Upper and Auto-Denoise do not effectively improve the baseline in subtask 1, since the mean length of subtask 1 is 374.34 which is much shorter than the max sequence length of original pre-trained model.
Ensemble Performance
The performances of ensemble models are shown on Table 2, which is obtained from the competition leader-board. Our system got the first place in subtask 2 and the second place in subtask 1. For the subtask 3, our ranking is first place in subtask3(2-1), and second place in subtask3(1-2), that indicates our system has strong transfer capability in abstractive reading comprehension tasks. Ensemble results on dev dataset are exhibited for comparison, and we do not experiment model ensemble on dev data of subtask 3.
Memory-Efficiency
We trained all compared models on 16GB Tesla-V100 GPU, except for ELECTRA Rank which takes 17GB with batch size set to 1. Our proposed Discriminator only take 9GB, since one option with article and question is considered as single training sample for binary classification task. In contrast, ELECTRA Rank is required to encode the input sequence with different options at once to learn the ranking function between positive and negative options. ELECTRA Discriminator is much faster than ELECTRA Rank to converge. The epoch of training ELECTRA Discriminator is less than 5 and ELECTRA Rank needs at least 10 train epochs to converge.
Conclusion
In this paper, we propose an effective framework of combining ELECTRA discriminator with denoising learning method to boost the performance of cloze-style reading comprehension task. Our proposed model outperforms all others participating system on subtask 2 and gets the second-placed on subtask 1. We have conducted an ablation study, demonstrating the validity of Discriminator, Upper attention and Auto denoising. Pre-trained models have made great performance gain compared to traditional neural network models in many natural language tasks and is able to build comprehensive hidden representation of input text. The above experiment results may suggest that the current pre-trained model mechanism still has room for improvement. | 2021-07-28T13:28:29.804Z | 2021-08-01T00:00:00.000 | {
"year": 2021,
"sha1": "7211e0bb527b6d86d94ea8b34f787e480417abe0",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.18653/v1/2021.semeval-1.109",
"oa_status": "HYBRID",
"pdf_src": "ACL",
"pdf_hash": "df2556b2ffef385c2d902b22a63a349c27490258",
"s2fieldsofstudy": [
"Computer Science",
"Linguistics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
18051911 | pes2o/s2orc | v3-fos-license | Speech recognition index of workers with tinnitus exposed to environmental or occupational noise: a comparative study
Introduction Tinnitus is considered the third worst symptom affecting humans. The aim of this article is to assess complaints by workers with tinnitus exposed to environmental and occupational noise. Methodology 495 workers went through an epidemiological survey at the Audiology Department of the Center for Studies on Workers’ Health and Human Ecology, from 2003 to 2007. The workers underwent tonal and vocal audiometry, preceded by a clinical and occupational history questionnaire. Two-factor ANOVA and Tukey were the statistical tests used. All the analysis set statistical significance at α=5%. Findings There was a higher prevalence of occupational tinnitus (73.7%), a predominance of female domestic workers (65.4%) in cases of environmental exposure, and predominance of male construction workers (71.5%) for occupational exposure. There was a significant difference in workers with hearing loss, who showed a mean speech recognition index (SRI) of 85%, as compared to healthy workers with a mean SRI greater than 93.5%. Signs and symptoms, speech perception, and interference in sound localization with the type of noise exposure (environmental versus occupational) comparisons found no significant differences. Conclusion Studied group’s high prevalence of tinnitus, major difficulties in speech recognition with hearing loss and the presence of individuals with normal hearing with both types of exposure justify the importance of measures in health promotion, prevention, and hearing surveillance. The findings highlight the importance of valuing the patients’ own perception as the first indication of tinnitus and hearing loss in order to help develop appropriate public policies within the Unified National Health System (SUS).
Introduction
Tinnitus is an endogenous, illusory sonorous sensation, result of an physiological alteration, that appears to be a sound, but is perceived in the absence of external sound stimuli. It is the first signal of a high sonorous stimulus. Tinnitus is commonly found along with hearing loss, either noise induced or not and can be classified according to its properties like its duration (seconds or minutes, intermittent or continuous), duration in time (days, months years) and graveness (annoyance level or life daily interference) [1]. Occasionally, everyone experience a transient sensation of tinnitus, which ceases spontaneously a few seconds later. However, when tinnitus becomes permanent, occurring frequently and having longer duration, may be a marker of quality of life worsening. It can be caused by numerous otological, metabolic, neurological, cardiovascular, pharmacological, dental, and psychological conditions, side effects of medications, and possibly the ingestion of drugs, caffeine, and alcohol [2], besides acting as a factor with major negative repercussions on the individual's quality of life, interfering with sleep, concentration on daily and professional activities, and social life [2]. It affects approximately 17% of the world population, including 33% of the elderly [3][4][5]. Based on these data, approximately 7.6 million Brazilians complain of tinnitus, considered the third worst symptom affecting humans, exceeded only by intense and intractable pain and dizziness [6].
Tinnitus is the single or principal symptom involved in various diseases that can jeopardize health and wellbeing. Thus, special attention is needed in the case of normal hearing, since tinnitus can be the first symptom of diseases that are generally diagnosed by the presence of hearing loss [7].
According to the World Health Organization (WHO), "Health is a state of complete physical, mental, and social well-being, and not merely the absence of disease or infirmity." Since the workplace is where workers spend major part of their time, maintaining health depends on a major extent of workplace healthiness. According to WHO, about 40% of European union (EU) population is exposed to environmental noise, 20% of these higher than 65 dBHL [8]. Unhealthy sound is defined under a specific national standard of the Brazilian Ministry of Labor, NR15, which characterizes occupational noise exposure, that is, worker is daily exposed to, at least, 85dBHL noise during his 8 hours labor day. Environmental noise exposure got its nomenclature after NBR 10152, which defines noise levels for acoustic comfort [9], more restrictive than occupational exposure ones. According to NBR10152, it's defined by acoustic comfort level a 30-65dBHL maximum noise exposure.
In relation to occupational exposure, noise is the most common cause of tinnitus, and is also the main harmful physical agent in the workplace [10], as corroborated by the World Health Organization [11], which considers it an important problem for workers worldwide, due to its high prevalence.
Some authors evaluated noise exposure psychosocial effects on workers [10,12]. They found a connection between tinnitus and hearing loss, and a dose-response relation. Similar results were also found on Singapore [13] and England [14] workers with occupational noise exposure history. Among workers with a history of occupational noise exposure in Singapore [13], 23.3% reported tinnitus, as compared to 29.7% in a study in England [14]. A random sample study at Salvador city, Brazil [15], analyzed 720 individuals 15-90 years range and showed a 25% tinnitus rate, which differs somewhat from the literature (showing a prevalence of 15% in the general population and 33% in the elderly). From the workers who reported tinnitus, 86% also had hearing impairment.
The auditory effects (cochlear hearing loss, irreversible) and extra-auditory effects of noise exposure result in a clear decrease in quality of life [12,16], especially in workers who may experience difficulties in perceiving sound signals and locating sound sources from distance, in addition to hindering social relations due to difficulty caused by the hearing loss [12].
Noise-induced hearing loss (NIHL) is considered one of the ten principal etiologies of hearing loss in population, and among all auditory injury causes, it is the one with the best possibilities for prevention [16,17]. In addition, 85 to 96% of individuals with tinnitus show some concurrent degree of hearing loss, and both can have important repercussions on daily life [18]. Another major aspect of human auditory function is the high percentage of workers complaints related to difficulties in oral communication, their capacity for speech recognition and intelligibility [19].
Considering the related above, and in search of a more precise evaluation of workers' health, the need to develop appropriate indicators for hearing health, and the institutional calling of a public research center in occupational audiology, a study was conceived, based on workers' self-reported tinnitus, to identify the complaints and audiometric profile as well as relate them to the sound exposure types (environmental or occupational). This epidemiological survey results might contribute significantly to the audiology study field. Nowadays in Brazil, occupational exposure-related studies characterize hearing loss profile, hearing compromising level and their connection with hearing loss risk groups, specifically. In this study the group aim to correlate not only the hearing loss presence, but analyze the speech recognition index (SRI) in occupational and environmental exposed workers as well.
Material and methods
This was an epidemiological cross-sectional survey, conducted from 2003 to 2007 at the Audiology Department of the Center for Studies on Workers' Health and Human Ecology (CESTEH/ENSP/FIOCRUZ), a federal public institution.
Sample
The potential study population consisted of 760 workers with hearing complaints who showed up spontaneously to evaluate their hearing condition. The final study sample included 495 workers (65.1% of the original total) that reported exposure to occupational or environmental noise plus tinnitus, all fit to undergo audiometric examination. The other 265 (34.9%) were excluded from the sample, since they did not report tinnitus complaints.
Classification of the origin of sound exposure, namely environmental [9] versus occupational [20], was based on the workers' report of their job/workplace concerning the presence of a specific noise source during the labor day. Were not included cases with simultaneous exposure (occupational plus environmental).
Occupational exposure selection criteria
According to NR 15 [20], workers undergoing unhealthy activities, which means, the ones daily exposed to high sonorous pressure (above 85 dBHL) for, at least, 8 hours per day in their workplaces, must be submitted to audiometric evaluation.
Environmental exposure selection criteria
For environmental exposure classification, including work areas that are known to be noisy, like construction, transportation, industry, or laboratories, it was observed that workers related it as lower noisy exposure.
Procedures
The workers underwent tonal and vocal audiometry, preceded by a clinical and occupational history (prepared by the department), inspection of the ear canal, and 14 hours of acoustic rest [21]. We sought to identify associations between the principal complaints, for which the following items from the case history were selected and analyzed statistically: socio-demographic data, hearing history, sonorous perception and perception difficulties, speech perception, and auditory signs and symptoms.
The tone thresholds were studied for air conduction at frequencies of 250 to 8000Hz and for bone conduction at frequencies of 500 to 4000Hz. The resulting hearing profile was classified as normal hearing or hearing loss: hearing thresholds were considered normal up to 25 dB HL [19,22], and logoaudiometric thresholds (SRI) were studied without the presence of competing noise to evaluate speech intelligibility [12]. This methodology aimed to select hearing status, distinguishing between normal hearing and hearing loss, so it did not involve the exclusive selection of profiles consistent with noiseinduced hearing loss. Independent of the sample age, the qualitative and quantitative SRI analysis takes into account the 500, 1000 and 2000Hz range frequency specter, therefore there are no interferences over this article's discussed data [19].
Instruments
Both instruments and audiometer were calibrated and checked according to the national [21] and international (ISO) [23] standards. The audiometric tests were performed using a Beltone 2000 audiometer (ANSI-69 standard) and TDH 39 earphones, all submitted to annual calibration according to the ISO R 389 standard (1998) for air conduction and ISO 7566 (1987) for bone conduction.
Questionnaire
The questionnaire was developed and adapted [24] by the speech therapists team at the Occupational Audiology Department, CESTEH/ENSP/FIOCRUZ, Rio de Janeiro, and has been applied since 2002 in over 2,000 patients. This questionnaire, which focuses on: sociodemographic data, personal habits, work market situation, social security benefits, environmental exposure, occupational exposure to noise and chemical substances, auditory identification (analysis of hearing acuity, onset of symptoms, difficulties on locating and noticing sound sources direction and speech perception), psychosocial evaluation, hearing history, hearing-related symptoms, non-auditory events, history of illness, and bother related by the worker when given the alternatives to define it according to the tinnitus duration, classifying it in mild, moderate or intense, in silence and after the workday. The three different tinnitus intensity levels identification was adapted based on Burden of disease from environmental noise, from WHO, which classifies tinnitus according to its attributes [1].
The following variables from the original questionnaire were analyzed: gender, work area (administrative, services, construction, domestic, industry, laboratory, health, transportation, and others), signs and symptoms like: difficulties on locating and noticing sound sources direction (bells and telephones), speech perception (closer, stronger, louder, in front, need for repetition, in a silent environment, background noise in front of radio or television, and when answering the telephone), tinnitus, and other non-auditory signs and symptoms like dizziness, vertigo, fainting, nausea, vomiting, and sweating, where the worker is allowed to report more than one symptom.
On the "workplace" variable the last activity related by the worker was considered, however all workers were exposed to occupational noise at any moment of their work lives.
Statistical analysis
The variables gender and work area were submitted to descriptive analysis according to occupational versus environmental exposure (frequency). Descriptive analysis (means and standard deviations) were used for the variable "speech recognition index" in the right and left ears, according to the hearing profile (normal hearing and hearing loss). Data were submitted to two-factor ANOVA and Tukey statistical tests. Tinnitus was the independent variable; "occupational and environmental exposure" and "hearing loss and normal hearing" were the dependent ones. All the analyses set statistical significance at α=5%.
The resulting data were keyed in, consolidated, and analyzed using EPI-AUDIO software, developed by the Occupational Audiology Department team, CESTEH/ ENSP/FIOCRUZ, Rio de Janeiro [25] based on Epi Info TM W , made available by the CDC (Centers for Disease Control and Prevention) [26].
The study was approved by the Institutional Review Board of FIOCRUZ, case number 107/5, and included the guidelines of Brazilian Ruling 196/96 on research involving human subjects.
Findings
Description of the study population It was thus observed a predominance of females (65.4%) in environmental exposure and males (71.5%) in occupational exposure. Concerning the workplace, approximately 30% of the workers with environmental exposure did domestic work, while for occupational exposure, 30% of the affected workers worked in construction jobs (Table 1). Both expositions had a, at least, 8 hours labor day.
Hearing profile analysis
According to Table 2, when we analyzed SRI for the groups of workers (hearing loss and normal hearing) and exposure (occupational and environmental noise), all the associations proved statistically significant. Workers with hearing loss showed a mean SRI of approximately 85%, as compared to mean SRI greater than 93.5% in workers with normal hearing, indicating that, irrespective of occupational exposure, there is speech perception impairment when the individual have hearing loss. When analyzing SRI according to the age groups a SRI reduction is found as the age progresses for the right (F = 12,36; p < 0,01) and left (F = 7,45;p < 0,01) ears.
As shown in Table 3, the associations' investigation shows that among normal hearing workers who reported occupational noise exposure and tinnitus, 83,1% reported difficulties in sound localization, speech perception, and the presence of one or more non-auditory symptoms. Non-auditory symptoms were considered: tachycardia, insomnia, anxiety, irritation and difficulties in concentration and attention. Among hearing loss workers with tinnitus and occupational noise exposure symptoms, 79,7% reported difficulties in sound localization, 51,5% in speech perception and nonauditory signs and symptoms. When non-auditory signs and symptoms, speech perception and interference in sound localization were compared to the type of noise exposure (environmental versus occupational), there were no statistically significant associations.
Discussion
The instrument used for data collection and analysis aimed to identify age range, gender and labor time, relate occupational/environmental sonorous exposure existence, in addition to characterizing difficulties in sound localization, speech perception, and the principal nonauditory signs and symptoms in workers with tinnitus as their principal complaint.
Other studies have been performed on similar populations [10,12]. The first article, on the psychosocial effects of noise, evaluated 32 male workers exposed to noise, with 20 years mean labor time. Of these, 62.5% had acquired neurosensory hearing loss, and the principal complaint was tinnitus (68.7%). The second author evaluated 284 workers also exposed to noise, without distinguishing industrial or non-industrial exposure, and found prevalence rates of approximately 48% for tinnitus and 63% occupational exposure. The author concluded that there is not only an association between hearing loss and tinnitus, but also a dose-response relationship. The study sample was 70.7% male and 29.2% female, with 15.6 years mean labor time and mean age 42.5 years.
Authors [14] have reported a tinnitus prevalence rate of some 20% in workers with a history of occupational noise exposure. In our sample, however, the prevalence rate was 65.1%, far higher than reported elsewhere in the literature. This high percentage is probably related to the fact that the study focused on workers with hearing complaints and the demand reported by workers during the case work-up, the symptoms' subjectivity, and the fact that no test was performed to assess the possibility of other influences (since it was not part of the objectives in the study design). This was also an experienced sample, with a mean age of 42.5 years and mean labor time of 15 years, despite the type of noise exposure.
In a Master's thesis on the influence of noise spectrum on the prevalence of noise-induced hearing loss and tinnitus [5], 192 hearing tests were performed, preceded by an occupational case history. The author found a 45.8% rate of reported tinnitus, and the highest prevalence of tinnitus (56.4%) occurred in workers (49.0% of the sample) that presented audiometric results consistent with noise-induced hearing loss.
The symptom is the factor that leads to the most difficulties and scarcity of data in studies on tinnitus, along with some other problems such as: the fact that tinnitus is a symptom rather than a disease; the lack of objective measurement methods; lack of adequate experimental models; and variations in the individual's emotional or physical status [10]. The evaluation that can be used to record tinnitus is acuphenometry [27], but there is still no specific test for diagnosing tinnitus [14]. However, speech therapy clinical practice relies on the patient's perception of tinnitus, which is subjective and displays wide individual variability in self-reporting. This highlights the importance of the case history and physical and audiometric examination.
In our study, the analysis of hearing quality in workers with tinnitus showed that in occupational noise exposure, the prevalence of hearing loss was higher (79.6%) than that of normal hearing (20.4%). The same was true for tinnitus and environmental noise exposure, that is, a higher prevalence of hearing loss (69.2%) as compared to normal hearing (30.8%). Similar results for hearing loss were detected in a retrospective study of 358 patients examined from January 1995 to June 1999 by the research group on tinnitus at the University of São Paulo (HCFMUSP) [28] in which clinical hearing loss was reported by 60.4% of patients.
A study on individuals with tinnitus and normal hearing [7] showed this is an important group to investigate, since the findings are not influenced by hearing loss. There are few studies in this field, and there are no longitudinal studies on the evolution of these patients. In a search for answers to these questions, 36 patients were selected from the same research group as in the previous study, from 1995 to 2003, who presented normal audiometry when they were included for follow-up. Although the study showed neither worsening of tinnitus over time nor significant changes in its characteristics, an important share of the sample evolved to hearing loss (7.4%).
In our study on workers with tinnitus, for both types of noise exposure (environmental and occupational), the majority of patients presented hearing loss; however, the proportion of patients with normal hearing was higher in cases of occupational noise exposure. This result is especially important when considering the findings that indicate evolution to hearing loss [7] and studies by other authors [5,16] highlighting tinnitus as the first symptom of hearing dysfunction.
A literature review shows numerous articles in the occupational field describing the prevalence of hearing loss, for example a study in Greater Metropolitan Salvador, Bahia State, Brazil [29] in 7,925 workers from 44 factories in nine different branches of industry, showing 45.9% prevalence of hearing loss.
In marble workshops in Brasília, Federal District of Brazil, the prevalence of hearing injury was 48.0% [30]. In the city of Goiania, a study in a metallurgical factory with 187 workers found a 22% prevalence of hearing loss, suggestive of noise exposure [31].
In São Paulo State, noise control programs in four metallurgical factories in the city of Piracicaba [32], with a total of 741 workers, were analyzed, showing 41% prevalence of hearing alterations, with workers' mean age 42.3 years and 16.7 years mean labor time.
Another highly relevant question is the measurement of hearing function in relation to speech intelligibility in the audiological battery normally used in audiometric tests. In most routine examinations the battery is considered incomplete, since it does not include measures of speech recognition [19]. Thus, in the attempt to understand these and other aspects, studies have been performed with the presence of competing noise and other factors.
One example [33] is a prospective clinical study in 60 adults, divided into 3 groups (20 with normal hearing, mean age 23.3 years; 20 with hearing loss, mean age 40.4 years; 20 elderly with hearing loss and mean age 66.8 years), aimed at investigating the effects of hearing loss and age on speech recognition in the presence of ipsilateral competing noise. The study showed a mean SRI proportion of 92% for all the groups studied, that is, no major difference was found between the three groups in terms of intelligibility performance, since the hearing thresholds at the speech frequencies were preserved. In our study, the mean SRI values in both the right and left ears were consistent with both the literature [19,34] and the observed hearing conditions, not concerning the type of exposure. The values varied from 84.2% to 86.7% for hearing loss and from 93.5% to 98.1% for normal hearing, thus representing a limited number of failures in speech intelligibility. Even with values close to 88%, some speech recognition impairment is expected [19].
The observed hearing loss values reflect a small percentage-wise alteration in speech intelligibility, consistent with the literature [19,34] and with the above-mentioned study [33], although the latter was not applied to individuals with tinnitus.
These minor alterations in speech recognition contrast with the high rates of difficulty in speech perception identified in normal hearing individuals complaining of tinnitus with occupational noise exposure (81.8%).
In 1986, Tyler & Baker [35] described the interrelationship between tinnitus and daily activities. In a public referral service for tinnitus, 49.3% of the 358 patients examined [28] reported impaired concentration in activities of daily living and 14.2% in social activities. In a previous study [36], the same service conducted a retrospective study of 150 patients in the Tinnitus Outpatient Clinic of the Department of Clinical Otorhinolaryngology at the FMUSP University Hospital and found that 76% reported some alteration in at least one activity of daily living, and that 47.3% reported difficulties in concentration. Our study found prevalence rates of 84.5% for sound localization difficulty (doorbell and telephone), 81.8% for speech perception (silent environment/presence of background noise), and 83.1% for otological signs and symptoms. These percentages in individuals with tinnitus with occupational noise exposure were similar to those found in environmental exposure. It is important to note that studies cited by Sanchez et al. [36] refer to a tinnitus specialist public ambulatory.
Study's constraints
Since it is a cross-sectional study, the chronological relation between the events may not be easily detectable. It is about occupational/environmental noise exposed workers, having specific complaints like tinnitus, welcomed at a specific workers' health Center. Age range shall not be considered as bias for hearing loss and tinnitus, since the age group includes hearing loss workers with ages below 30 years. Moreover, SRI evaluation methodology used corroborates for the findings.
As information bias, external factors as earplug use and vehicle, home and free-time noise may not be related.
Conclusion
In this study, when we analyzed the mean speech recognition index (SRI) for the groups of workers (with hearing loss versus normal hearing), age groups and types of exposure (occupational versus environmental noise), we found lower SRIs for workers with hearing loss, thus characterizing greater difficulty in speech recognition for cases of occupational hearing loss.
Another finding was the presence of individuals with normal hearing in both types of noise exposure, which highlights the importance of health promotion, prevention, and hearing surveillance measures, especially considering the fact that tinnitus can be the first symptom of diseases that are usually diagnosed on the basis of hearing loss. In relation to environmental exposure, this finding is especially important due to lack of recognition of environmental noise risks for workers. There is no legal control of workers' exposure to environmental noise, which emphasizes the need for awareness-raising campaigns in defense of environmental sound quality, regardless of the origin of the noise, as well as the need for surveillance measures to monitor hearing in the exposed population. The existing Brazilian legislation [20] only deals with the built-up environment.
The findings highlight the importance of valuing the workers' own reported complaints in developing public policies (or educational programs) for the early detection of tinnitus and hearing loss, given that self-perception can be the first indication of tinnitus and hearing loss. Workers' self-report is an inexpensive and preventive technique, which can minimize the costs for the Unified National Health System (SUS) in workers' treatment and rehabilitation. | 2016-01-15T18:20:01.362Z | 2012-12-22T00:00:00.000 | {
"year": 2012,
"sha1": "6735b5820a5536ad231b916bd01c8d350ca8e07d",
"oa_license": "CCBY",
"oa_url": "https://occup-med.biomedcentral.com/track/pdf/10.1186/1745-6673-7-26",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "241e39ff14d32becaee4d95b6284486405b489af",
"s2fieldsofstudy": [
"Environmental Science",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
44965253 | pes2o/s2orc | v3-fos-license | Serum NT-proBNP Levels Are Not Related to Vitamin D Status in Young Patients with Congenital Heart Defects
Context. Hypovitaminosis D frequently occurs in early life and increases with age. Vitamin D has been suggested to influence cardiac performance and N-terminal-pro-type B natriuretic peptide (NT-proBNP) release in adults with heart failure. Objectives. To assess the vitamin D status and the impact of hypovitaminosis D on circulating NT-proBNP levels in young patients with congenital heart defects (CHD). Design and Patients. This cross-sectional study included the assessment of serum 25-hydroxyvitamin D (25OHD), parathyroid function markers, and NT-proBNP levels in a series of 230 young in-patients (117 females, 113 males; 6.4 (4.0–9.1) years (median, interquartile range)) with CHD. Results. Serum 25OHD levels <20 ng/mL were detected in 55.3% of patients. Optimal 25OHD levels (>30 ng/mL) occurred in 25% of patients. Serum 25OHD levels inversely correlated with age (r = −0.169, P = 0.013) and height standard deviation score (r = −0.269, P = 0.001). After correction for age, 25OHD negatively correlated with serum PTH levels (β = −0.200, P = 0.002). PTH levels above the upper quartile (44 pg/mL) occurred in 32% of hypovitaminosis D patients. Serum NT-proBNP levels were not correlated with 25OHD and PTH levels. Conclusions. Half of the young CHD patients were diagnosed with 25OHD deficiency and a third of hypovitaminosis D patients experienced hyperparathyroidism. Nonetheless, serum NT-proBNP levels were not associated with hypovitaminosis D as well as hyperparathyroidism.
The N-terminal-pro-type B natriuretic peptide (NT-proBNP) is 76 amino acids derived from cleavage of a prohormone of 108 amino acids synthesized and released by cardiomyocytes. NT-proBNP can be used as an adjunctive marker in the integrated screening, diagnosis, management, and follow-up of children with heart failure caused by various acquired and congenital heart diseases [6]. The natriuretic peptides, NT-proBNP and BNP, correlate with various indexes of disease severity in children with congenital heart defects. The measurement of the circulating cardiac biomarkers BNP and NT-proBNP is now recommended by international guidelines [7,8].
Hypovitaminosis D is prevalent worldwide: around 37% of studies evaluating the vitamin D status in various 2 Disease Markers populations reported mean values below 20 ng/mL [9]. Pediatric patients with congenital heart defects might be a population at high risk of hypovitaminosis D due to their chronic disease and to the reduced exposure to sunlight. Vitamin D is key component of the calcium metabolism. Calcium and the calciotropic hormones, vitamin D and parathormone (PTH), exert direct and indirect effects on cardiomyocytes. Calcium dyshomeostasis associated with hypocalcemia, hypovitaminosis D, and secondary hyperparathyroidism can lead to the development of dilated cardiomyopathy with lifethreatening consequences in both infants [10,11] and adults [12]. Moreover, cardiomyocytes express both the vitamin D receptors and the PTH receptors (PTHR1), and studies in rodents have shown that vitamin D protects against cardiac hypertrophy and myocardial dysfunction [13][14][15][16]. In line with experimental data, vitamin D deficiency has been associated with a poor prognosis in adult patients with heart failure [17], and circulating PTH levels are related with an increased risk of cardiovascular events and mortality [18][19][20].
In the present study, we tested the hypothesis that vitamin D status might influence cardiac performance and therefore the circulating NT-proBNP levels in young patients with congenital heart defects. We investigated (1) the vitamin D status and (2) the correlations of calcium metabolism markers with serum NT-proBNP levels in a consistent series of young patients with various congenital heart defects.
Study Population.
Two hundred thirty young patients (117 females, 113 males; 6.4 (4.0-9.1) years (median age, interquartile range)) with CHD, referred to the Policlinico San Donato Pediatric Cardiosurgery between January 2007 and January 2009, were consecutively enrolled. All patients were studied at the admission to be evaluated for congenital heart defects and to undergo open or transcatheter surgical correction when haemodynamically significant CHD was diagnosed. In particular, 153 patients harboured defects determining increased volume overload (i.e., defects characterized by left-to-right shunt, such as ventricular septal defect, patent ductus arteriosus, truncus arteriosus, atrial septal defect, and atrioventricular septal defects); 62 patients had heart defects with pressure overload involving the left ventricle (i.e., aortic stenosis and aortic coarctation) or the right ventricle (i.e., tetralogy of Fallot and pulmonary stenosis); and 15 patients were affected with complex cyanotic CHD (i.e., univentricular heart and transposition of the great arteries).
Anamnestic and anthropometric parameters were recorded. Height was evaluated by height standard deviation score (HSDS) related to Italian growth curves for children aged 2-20 years [21]. HSDS for infants aged 0-2 years was calculated by using the Kabi-Pharmacia Growth Calculator, based on the British standards. Weight was evaluated by Weight-to-Height Index (WHP) expressed as the percentage of the median of the British standards using the Kabi-Pharmacia Growth Calculator. Clinical data are shown in Table 1. Exclusion criteria were considered overt cyanotic status, acute or chronic kidney failure, liver failure, infective and inflammatory diseases, acquired or metabolic cardiomyopathy, concurrent treatment with diuretics, amiodarone and steroids, ongoing calcium and/or vitamin D supplementation in the last 6 months, overt hypoparathyroidism, and overt or mild hypothyroidism.
Informed consent for personal data and blood collection was obtained by parents of all patients. The study was approved by the local ethical committee.
Measurements.
Venous blood sampling was collected the day before the diagnostic procedure after an overnight fasting, and biochemical and hormonal markers were measured in all patients. Serum calcium, phosphate, magnesium, and creatinine were measured by routine assays. Serum calcium levels were corrected for the serum albumin concentrations according to the formula: serum total calcium (mg/dL) − 0.8 × [serum albumin (g/dL) − 4.0]. Serum albumin concentrations were within the normal range in all patients indicating the absence of malnutrition. Ionized calcium concentrations were assayed by Liquichek Blood Gas (IL Synthesis Series, Bio-Rad Laboratories, Segrate, MI, Italy). Serum 25OHD concentrations were measured by a chemiluminescent assay (LIAISON test, DiaSorin Inc., Stillwater, MN, USA) with mean intra-and interassay coefficients of variation (CV) of 4.5% and 7.5%, respectively. Serum NT-proBNP and PTH were measured by Electrochemiluminescence on an Elecsys 2010 (Roche Diagnostics, Mannheim, Germany).
Statistical Analysis.
Continuous parameters following a nonnormal distribution were presented as median and interquartile range (IQ). Normally distributed parameters were presented as mean ± standard deviation; categorical data were given as percentages. Based on serum 25OHD concentrations, we formed four categories according to widely used cut-off values [22][23][24]: severe deficiency, less than 10.0 ng/mL; moderate deficiency, 10.0-19.9 ng/mL; insufficiency, 20.0-29.9 ng/mL; and 25OHD optimal range, 30.0-100.0 ng/mL. For nanomoles per liter, multiply by 2.496. Serum 25OHD, PTH, and NT-pro-BNP concentrations were logarithmically transformed before being used in parametric procedures. Comparisons among CHD groups were performed by analysis of variance (ANOVA) with for linear trend for continuous parameters or ANOVA on ranks for nonparametric data. A 2 test was performed for categorical variables. Simple correlation analyses (Pearson or Spearman correlations where appropriated) and multiple linear regression analyses including several independent variables were performed to examine whether 25OHD levels were associated with NTpro-BNP levels. A value <0.05 was considered statistically significant. Data were analysed using SPSS 22.0 statistical package (SPSS 22.0 Inc., Chicago, IL). serum 25OHD levels > 30 ng/mL, were detected in about oneforth of the CHD patients. Vitamin D status did not differ between males and females nor among CHD groups. Young CHD patients with 25OHD <10 ng/mL did not complain for symptoms of rickets (bowed legs, thickened wrists and ankle, and breastbone projection). Serum 25OHD levels negatively correlated with age ( = −0.169, = 0.013) (Figure 1(a)) and with height SDS ( = −0.269, = 0.001) (Figure 1(b)), also after adjustment for age ( = −0.200, = 0.002). Accordingly, CHD patients with severe vitamin D deficiency (<10 ng/mL) were older, heavier, and with higher median PTH levels than patients with an optimal vitamin D status (>30 ng/mL) ( Table 1), while serum 25OHD levels did not differ among the three CHD groups identified according the underlying congenital heart defect ( Table 2).
Parathyroid Function in Young CHD Patients.
Overt hypocalcemia and hypomagnesemia were considered exclusion criteria. Hypercalcemia, defined as elevated levels of serum albumin-corrected calcium and ionized calcium, was not detected in any patients. Hyperparathyroidism, defined as serum PTH levels in the higher quartile (>44 pg/mL), was diagnosed in 56 out of 230 CHD patients (24.0%). Serum PTH levels positively correlated with 25OHD levels ( = −0.216, = 0.001; Figure 1(c)) also after adjustment for age ( = −0.200, = 0.002). In the present series, renal function was conserved in all CHD patients as confirmed by serum creatinine levels in the normal range.
Effect of Vitamin D Status and PTH Secretion Alterations on Serum NT-proBNP Levels.
Serum NT-proBNP levels ranged from 5.1 to 5529.0 ng/mL in the present series of young CHD patients. Serum NT-proBNP levels did not show any significant correlation with age and did not differ among the different CHD groups (Table 2). Moreover, significant correlations between serum NT-pro-BNP levels and 25OHD (Figure 2(a)) as well as PTH and albumin-corrected calcium failed to be detected (Figure 2(b)).
Discussion
Circulating NT-proBNP is a cardiac biomarker for diagnosis, prognosis, and therapeutic monitoring. Data available so far in pediatric patients support the NT-proBNP measurement in specific cases [6]. Serum 25OHD and/or PTH levels have been shown to be independently associated with all cause and cardiovascular mortality in adult patients with heart failure [25][26][27][28]. Moreover, previous studies reported a negative correlation between serum 25OHD levels and NT-proBNP levels in adult patients with coronary artery diseases and heart failure [27,29,30]. Indeed, findings are controversial as an investigation in the setting of adult postacute myocardial infarction failed in detecting an association between nutritional vitamin D status and NT-proBNP levels [31]. Moreover, a recent interventional study reported any significant effect of oral vitamin D supplementation on circulating NT-proBNP levels in adult peritoneal dialysis patients [32].
In the present series of young CHD patients, we failed in detecting such a relationship between 25OHD and NT-proBNP levels as well as between PTH and NT-proBNP levels, though in the present young CHD cohort vitamin D deficiency and mild hyperparathyroidism frequently occurred.
Data about vitamin D status in young patients with CHD are scanty: Avitabile et al. [33] reported vitamin D deficiency, defined as serum 25OHD levels lower than 20 ng/mL, in 25% of a small series of children and young adults ( = 50) with Fontan physiology, while McNally et al. [34] reported hypovitaminosis D in about 40% of 58 young CHD patients. This is the first study investigating the preoperatively vitamin D status in a consistent series of young patients with various congenital heart defects. In the present series, hypovitaminosis D occurred in half of the patients. Children and adolescents have been reported to be potentially at high risk for vitamin D deficiency [35,36]. In Europe, where very few foods are fortified with vitamin D, children would appear to be at especially high risk [34]. Indeed, the prevalence of hypovitaminosis D in the present series of CHD young patients was similar to that found in a cohort of Italian healthy children and adolescents [37], suggesting that most of the CHD conditions do not represent a risk factor for the development of hypovitaminosis D in children and adolescents. This point is further supported by the lack of significant differences in median 25OHD levels among the three pathophysiological CHD groups. Moreover, serum 25OHD levels were inversely related to CHD patients' age and height, suggesting, in line with a previous report [38], that older children and adolescents as well as subjects with overweight experience more frequently hypovitaminosis D. Serum 1,25-dihydroxyvitamin D levels have been suggested to strongly and independently predict cardiovascular mortality in adult patients with chronic heart failure [39]. Therefore, 1,25-dihydroxyvitamin D levels might be more sensitive in detecting the correlation with NT-proBNP levels in young CHD patients; unfortunately, 1,25dihydroxyvitamin D determinations were not available in the present series of CHD patients.
The serum PTH threshold of 44 pg/mL, corresponding to the upper quartile of the PTH distribution in the young CHD patients, has been chosen to detect mild parathyroid function activation induced by vitamin D deficiency in young subjects with normal renal function [40,41]. Indeed, PTH levels in healthy children and adolescents have been found to cover a narrower range than the adult values [42]. Considering this PTH threshold, about one-third of hypovitaminosis D young CHD patients showed hyperparathyroidism. Sustained rises in plasma PTH lead to intracellular Ca 2+ overloading in diverse cells, including cardiomyocytes [43]. In turn, elevated energy ATP stores become depleted and ATPase-dependent Ca 2+ efflux reduced, together with an induction of oxidative stress that can threaten cardiomyocyte survival. Though the experimental and clinical data suggest an association between NT-proBNP and PTH levels and support their role as predictors of clinical outcomes in adult with cardiovascular diseases, any significant association could be detected between NT-proBNP and PTH levels in young CHD patients.
In conclusion, vitamin D deficiency occurred in half of the young CHD patients and mild hyperparathyroidism could be detected in quarter of the patients. The circulating biomarker NT-proBNP was not related with both vitamin D status and PTH increases, suggesting that calcium metabolism aberrations might not affect circulating NT-proBNP levels in young CHD patients. | 2018-04-03T05:57:49.753Z | 2016-02-03T00:00:00.000 | {
"year": 2016,
"sha1": "6354a9997295e01143fe8cfe38a3a6da4ac8609f",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/dm/2016/3970284.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "6354a9997295e01143fe8cfe38a3a6da4ac8609f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
216471323 | pes2o/s2orc | v3-fos-license | Methodology for detection of paroxysmal atrial fibrillation based on P-Wave, HRV and QR electrical alternans features
Received Oct 24, 2019 Revised Feb 19, 2020 Accepted Feb 25, 2020 The detection of paroxysmal atrial fibrillation (PAF) is a fairly complex process performed manually by cardiologists or electrophysiologists by reading an electrocardiogram (ECG). Currently, computational techniques for automatic detection based on fast fourier transform (FFT), Bayes optimal classifier (BOC), K-nearest neighbors (K-NNs), and artificial neural network (ANN) have been proposed. In this study, six features were obtained based on the morphology of the P-Wave, the QRS complex and the heart rate variability (HRV) of the ECG. The performance of this methodology was validated using clinical ECG signals from the Physionet arrhythmia database MIT-BIH. A feedforward neural network was used to detect the presence of PAF reaching a general accuracy of 97.4%. The results obtained show that the inclusion of the information of the P-Wave, HRV and QR Electrical alternans increases the accuracy to identify the PAF event compared to other works that use the information of only one or at most two of them.
INTRODUCTION
Atrial Fibrillation (AF) is the most clinically diagnosed cardiac arrhythmia, both in outpatients and hospitalized patients. Its prevalence and incidence increase with age reaching epidemic characteristics in senior citizens. The indicators of progress of paroxysmal atrial fibrillation (PAF) to a persistent or permanent one have not been fully identified, therefore, detecting an AF in its early form is important to avoid the risks of a stroke, heart failure and / or mortality [1].
The process of detecting an AF is performed manually by a cardiologist or electrophysiologist by interpreting the electrocardiogram (ECG) records. This process is highly demanding due to both the number of records to be analyzed and the fact that sometimes it is necessary to examine each beat individually to ensure the correct identification of the cardiac pathology. Thus, an automated method for classification and detection would improve the diagnostic and prevention of an AF [2][3][4][5].
To date, different authors have proposed methods that automate the detection of PAF. Some authors have reached a detection accuracy between 70% and 92% [6][7][8][9] using the characteristics of the P wave [10], others propose the use of heart rate variability obtaining an accuracy between 81.2% and 94.7% [7,[11][12][13][14][15][16][17][18], finally, [9] proposes the use of QR electrical alternation reaching an accuracy of 70%. According to this, the problem of appropriately detecting a FAP is not fully solved yet, due to results achieved by these methods are not definitive and can still be improved. So, in this paper a new methodology is proposed to address [19] on the characteristics of the P wave, heart rate variability and QR electrical alternation. The accuracy obtained using this new method was 97.4%.
A PAF is characterized by irregular movement of the left atrium that prevents the proper blood flow into the circulatory system and also by a reduction of the time that the ventricles valves have to receive and send blood to the lungs. In an ECG signal, these two characteristics have an impact in the morphology of the P-Wave and in the distance between the P-Wave and the R-Wave see Figure 1, therefore, it is important to locate the characteristic points P-Onset, P-Offset, P Width and P Height, as well as the PR segment, heart rate variability (HRV) and QR electrical alternation to fully describe a PAF. Different authors relied on one or two characteristics for the detection of PAF as illustrated in Table 1. In this paper, unlike other works reported in the literature, it is proposed to use the information of the three characteristics to cover all the symptoms of the PAF and extract six relevant features to improve the detection rates. Sensitivity, specificity and accuracy were used as performance metrics for the evaluation of the methodology proposed here. Table 1 lists some works that address the same theme and the characteristics used.
RESEARCH METHOD
In the proposed methodology, a previously digitized ECG signal is received as input. The signal is processed in four main stages (Preprocessing, characteristic points extraction, features extraction, detection) and it is determined whether or not a PAF exists, as illustrated in Figure
Preprocessing
The first part of the algorithm is prepared to receive as input the lead II of a standard 12-lead ECG signal. Due to the variable nature of the sampling frequency of the ECG signal, an 1170Hz resampling is performed to ensure a standard frequency for the subsequent application of a low-pass finite impulse response digital filter (FIR) and to allow each of the characteristic points of the signal to be established more precisely. This stage comprises three steps: Resampling, moving average and filtering.
Resampling
The proposed methodology uses six features for the recognition of atrial fibrillation that are based on the morphology of the signal. Therefore it is very important to preserve the frequency content as well as the shape of the signal during processing. For this reason, it is required to be represent each beat by a sufficient number of points that ensure a good detection and a good feature extraction.
The height and width of the P-Wave are features that need a good morphological representation, thus, in this paper we considered using a resample frequency to ensure that the P-Wave has at least 50 samples.Considering that there are documented cases of patients with PAF at the age of 22 [20], the maximum heart rate ( ) considered in this methodology was calculated using the proposed (1). The result obtained was 168 beats per minute (BPM).
Subsequently, it was found that the duration of the P-Wave is 43 ms using on (2).
Finally, a resampling frequency ( ) of approximately 1170 Hz was obtained through (3) by relating the 50 samples that represent the P-Wave with its duration.
Moving average
ECG signals normally have a baseline wander that must be corrected to reference the voltage levels of the signal to a zero DC level. The moving average given in (4) is commonly used to do this which requires specifying a window size ( ). This paper proposes to obtain M based on the most common heart rate value present in the signal. To find this value, we obtain the frequency with the highest energy value in the power spectral density of the signal bounded between 60 bpm and 200 bpm. Therefore, M is defined as the inverse of the frequency with the highest energy value rounded to the nearest even value. This is shown in (5). Said frequency was obtained applying the fast Fourier transform (FFT) to the signal autocorrelation given in (6).
Filtering
An ECG signal is represented by (7), where, ( ) is the signal generated by cardiac activity with a frequency range of 2.5 Hz and 45 Hz [21], r (n) is electrical noise and white noise with frequencies greater than 45 Hz and b (n) is baseline noise with frequencies less than 2.5 Hz [22].
Noise ( ) was already removed using moving average in the last step. In this step, a low-pass filter with a cutoff frequency of 45 Hz was designed to remove the noise ( ).
The preprocessing stage is summarized in Algorithm 1.
Characteristic point detection
In the second stage of the methodology the peaks P, Q, R, S, P-Onset, P-Offset and Q-Onset were found on each beat of the ECG signal. These points, shown in Figure 3, will later be used to extract the features of the beat.
R-Wave peak
In this step, a moving window four times the size M found in section 2.1.2 was used to find the R-Wave peak. The window moves throughout the ECG signal finding peaks that exceed 0.6 times the maximum amplitude in the window and have a separation between them of at least 353 ms, that is, the heart rate does not exceed the maximum value chosen in this methodology of 170 bpm.
P-Wave peak
The P-Wave peak was found based on the location of two consecutive R-Wave peaks. As seen in Figure 4, The P-Wave peak is the maximum value found within a defined search area between 70% and 90% of the distance between two consecutive R-Wave peaks.
Q-Wave peak
The Q-Wave peak is characterized by a negative peak located just before the appearance of the R-Wave, for this reason, a derivative was used as a search method for this peak. According to the proposed (8), the value of the derivative is calculated on each sample one at a time before the R-Wave. This process is done until a derivative with a negative value is found as seen in Figure 5. In this paper we propose a distance of eight samples to be used in order to avoid small variations that could have a negative derivative in the path.
S-Wave peak
The identification of the S-Wave peak was carried out following a procedure similar to that used with the P-Wave peak. This time, the minimum value was sought within a defined area between 0% and 10% of the distance between two consecutive R peaks as is shown in Figure 6.
P-Onset
The P-Onset point is defined as the sample where the P-Wave starts and ideally has a value of 0 mV. This point was found by evaluating each one of the samples prior to the P-Wave peak one by one until the condition set in the proposed (9) was met. This equation considers the fact that, in practice, the P-Onset has a positive value higher than the baseline, therefore, a value of 0.15 times the amplitude of the P peak was used to find it.
P-Offset
This characteristic point is defined as the sample where the P-Wave ends. To find this point, we proceeded in a similar way to the method used to find the P-Onset with the difference that the samples evaluated are located after the P-Wave peak.
Q-Onset
Q-Onset is the sample where the Q-Wave begins. To find this characteristic point, a similar method used in section 2.2.3 was considered. Each of the samples before the Q-Wave peak is calculated one by one on the derivative described by the proposed (10) until a positive value is found. In this case, a sensitivity greater than that required to find the Q-Wave peak is required, thus the distance was reduce from eight to four samples.
The characteristic point detection stage is summarized in Algorithm 2.
Features extraction
Once the characteristic points have been identified, the six features presented on the third stage of the methodology described in Figure 2 are extracted for each beat of the ECG signal. The first three features P-Wave height, P-Wave width and PR segment are the magnitudes of the P-Wave peak, the difference between P-Offset and P-Onset, and the difference between Q-Onset and P-Offset respectively. As for the fourth feature P-Wave Area, it is defined as the area under the curve between P-Onset and P-Offset. Considering that the ECG signal is discrete, a trapezoidal numerical integration is used as an approximation to the integral of the signal between these two points. The (11) describes this condition.
The fifth feature called Heart Rate Variability (HRV) is the number of beats per minute (bpm) that would be generated according to the distance between two consecutive R-Wave peaks. The (12) describes this process.
In (12), is the beat number.
( ) is the location of the R-Wave peak. is the resampling frequency, 1170 Hz in this case. The sixth and last feature called QR electrical alternans is defined as the difference between the amplitude of the R-Wave peak and the Q-Wave peak. The features extraction stage is summarized in Algorithm 3.
Detection
Detection is the final stage of the proposed methodology. To determine the presence of a PAF in the ECG, a feedforward neural network with two hidden layers each with 10 neurons was used as a classifier [23]. This neural network, whose training was carried out using 60% of the information in the database shown in Table 2, to identify the presence or not of a PAF in each beat of the ECG.
Performance metrics
Sensitivity (SN), specificity (SP) and accuracy (ACC), shown in (13)(14)(15) respectively, were calculated since these are the most widely used performance metrics to assess the probability of success of a classifier [24]. Table 3 shows the results of these metrics in different works reported in the literature.
RESULTS AND ANALYSIS
To evaluate the proposed methodology, the Atrial Fibrilation (afdb) and Normal Sinus Rhythm (nsrdb) databases from Physionet [25] were used. Each one has ECG signal samples from both sick and healthy patients. Each signal is processed using the methodology described before. Figure 7(a) shows an original ECG signal from the database, while Figure 7(b) shows the signal after preprocessing. Finally, Figure 7(c) shows the signal with its characteristic points obtained.
The extraction of characteristics was applied to each of the records in both databases. We obtained six features of a total of 99,002 beats as illustrated in Table 2. To ensure the linear independence of the features, the degree of correlation between each of them was determined through the correlation matrix. As it is shown in Table 4, the relation between the six features is low in all cases except between P-Wave area, P-Wave height and P-Wave width which is moderate. These results ensure that the features obtained through the proposed methodology are suitable for the training of a neural network.
The PAF was detected through a feedforward neural network whose training data corresponded to 60% of the information provided by the 99,002 beats obtained before. The network was trained on 10 different occasions and was obtained the SN, SP and ACC in each training. The calculations of the maximum, minimum, average and standard deviation of each performance metric are shown in Table 5.
A comparative analysis of the performance metrics between different classifiers used in similar works and the proposed methodology was done. The results are shown in Table 3. The proposed methodology obtained a minimum SN of 96.4% that is higher than the others. On the other hand, the SP reached a maximum value of 98.1% being surpassed only by [8], however, the ACC exceeds all the reported works even with its minimum value of 96.3%.
CONCLUSION
To date, different authors have proposed methods that automate the detection of PAF using the characteristics of the P wave, heart rate variability or QR electrical alternation. The accuracy reached by these methods vary between 70% and 94.7%. Thus, the problem of appropriately detecting a FAP is not fully solved yet, due to results achieved by these methods are not definitive and can still be improved.
This paper proposes a methodology to identify the presence of a PAF in patients by analyzing their ECG. The methodology includes both the identification of the characteristic points of the ECG signal and the methods to extract six features that allow a PAF to be detected through a classifier. The results obtained show that the inclusion of the information of the P-Wave, HRV and QR electrical alternans for the extraction of features increased the accuracy in the detection of a PAF to 97.4% on average. The SN obtained was higher than that obtained in other works, achieving at least a result of 96.4%. The SP was similar to the results obtained by the works consulted.The results obtained serve as the basis for the future implementation of a methodology that allows predicting the occurrence of a PAF in a given period of time. | 2020-04-27T16:22:59.509Z | 2020-08-01T00:00:00.000 | {
"year": 2020,
"sha1": "6e5589b56d4f3dab1cbc985f452e2b45f1505a71",
"oa_license": "CCBYSA",
"oa_url": "http://ijece.iaescore.com/index.php/IJECE/article/download/21303/14082",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "fce27baef65170665cd9f76088ddbc0539d5fb88",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
234338451 | pes2o/s2orc | v3-fos-license | FlingBot: The Unreasonable Effectiveness of Dynamic Manipulation for Cloth Unfolding
High-velocity dynamic actions (e.g., fling or throw) play a crucial role in our everyday interaction with deformable objects by improving our efficiency and effectively expanding our physical reach range. Yet, most prior works have tackled cloth manipulation using exclusively single-arm quasi-static actions, which requires a large number of interactions for challenging initial cloth configurations and strictly limits the maximum cloth size by the robot's reach range. In this work, we demonstrate the effectiveness of dynamic flinging actions for cloth unfolding with our proposed self-supervised learning framework, FlingBot. Our approach learns how to unfold a piece of fabric from arbitrary initial configurations using a pick, stretch, and fling primitive for a dual-arm setup from visual observations. The final system achieves over 80% coverage within 3 actions on novel cloths, can unfold cloths larger than the system's reach range, and generalizes to T-shirts despite being trained on only rectangular cloths. We also finetuned FlingBot on a real-world dual-arm robot platform, where it increased the cloth coverage over 4 times more than the quasi-static baseline did. The simplicity of FlingBot combined with its superior performance over quasi-static baselines demonstrates the effectiveness of dynamic actions for deformable object manipulation.
Introduction
: Cloth unfolding with dynamic interactions. Given a severely crumpled cloth, Fling-Bot uses a high-speed fling to unfurl the cloth with as little as one interaction. In this paper, we demonstrate that such dynamic actions can efficiently unfold cloths, generalizable to different cloth types, and improve the effective reach range of the system.
High-velocity dynamic actions play a crucial role in our everyday interaction with deformable objects. Making our beds in the morning is not effectively accomplished by picking up each corner of the blanket and placing them in the corresponding corners of the bed, one by one. Instead, we are more likely to grasp the blanket with two hands, stretch it, and then unfurl it with a fling over the bed. This fluid high-velocity flinging action is an example of dynamic manipulation [1], which is used to improve our physical reachability and action efficiency -allowing us to unfold large crumpled cloths with as little as one interaction.
From goal-conditioned folding [2] to fabric smoothing [3,4], prior works have achieved success using exclusively single-arm quasistatic interactions (e.g., pick & place) for cloth manipulation. However, these approaches require a large number of interactions for challenging initial configurations (e.g., highly crumpled cloths) or rely on strong assumptions about the cloth (e.g., predefined keypoints). Additionally, since the robot arm cannot manipulate the cloth at locations it can't reach, the maximum cloth size is greatly limited by the robot arm's reach range.
In this work, we focus on the task of cloth unfolding, a typical first step for many cloth manipulation tasks. The goal of this task is to maximize the cloth's coverage through interactions to expose the key visual features for downstream perception and manipulation. Therefore, an ideal cloth unfolding approach should be: • Efficient: the approach should reach a high coverage with a small number of actions from arbitrarily crumpled initial configurations.
• Generalizable: the algorithm should not rely on heuristics (e.g., grasp predefined key points, etc.). This is especially important when unfolding is only the initial stage of a cloth perception and manipulation pipeline, where key points are not visible or severely occluded, and when the system must handle cloth types unseen during training, which may not contain the predefined key points.
• Flexible Beyond the Workspace: the approach should work with cloths of different sizes, including large ones which lie outside the robot's physical reach range.
To achieve this goal, we present FlingBot, a self-supervised algorithm that learns how to unfold cloths from arbitrary initial configurations using a pick, stretch, and fling primitive for a dual-arm setup. At each time step, the policy predicts value maps from its visual observation and picks actions greedily with respect to its value maps. To provide the supervision signal, the system computes the difference in coverage of the cloth before and after each action -the delta-coverage -from the visual input captured by a top-down camera. FlingBot achieves over 80% coverage within 3 actions on novel cloths and increases the cloth's coverage by more than twice that of pick & place and pick & drag quasi-static baselines on rectangular cloths. Our approach is flexible to large cloths whose dimensions exceed the robot arm's reach ranges and generalizes to T-shirts despite being trained on rectangular cloths. We fine-tune our approach in the real world, where, on average, it increased the cloth coverage over 4 times more than the quasi-static pick & place baseline did. In summary: • Our main contribution is in demonstrating the effectiveness of dynamic manipulation for cloth unfolding through our self-supervised learning framework, FlingBot.
• We propose a parameterization for the dual-arm grasp of our fling primitive, which enables the application of the simple yet effective single-arm grasping technique [5,6,7] to dual-arm grasping while satisfying safety constraints of dual-arm systems. The simplicity of FlingBot combined with its superior performance over quasi-static baselines further emphasize the effectiveness of dynamic actions for deformable object manipulation.
• We present a custom simulator 1 built on top of PyFlex [8,9], a CUDA accelerated simulator, which supports the loading of arbitrarily shaped cloth meshes. We hope this opensource simulator expands cloth manipulation research to more complex cloth types.
Related Work
A convincing argument for the addition of dynamic actions to previously exclusively quasi-static cloth manipulation pipelines would need to demonstrate their superior performance on a core cloth manipulation skill. As a typical first step for many cloth manipulation tasks, cloth unfolding is a popular and important problem setting for studying deformable object manipulation. The goal of cloth unfolding is to maximize the coverage of the cloth on the workspace, which exposes key visual features of the cloth for downstream applications. However, achieving a fully unfolded cloth configuration from a severely crumpled initial configuration remains a challenging problem. Additionally, doing so efficiently for many different types of cloths, some larger than the system's reach range, is extremely challenging.
Quasi-static Cloth Manipulation with Expert Demonstrations. Prior works have explored using heuristics and identifying key points such as wrinkles [10], corners [11,12,13], edges [14], or combinations of them [15], but fail to generalize to severely self-occluded cloth configurations, to when required key points aren't visible, or to non-square cloths. Recent reinforcement learning approaches [3,16] relied on cloth unfolding expert demonstrations in quasi-static pick-and-drag action spaces, While they sidestep the exploration problem, expert demonstrations can be sub-optimal and brittle if from hard-coded heuristics [3] or expensive if from humans [16].
Self-supervised Quasi-static Cloth Manipulation. Bypassing the dependency on an expert, selfsupervised cloth manipulation has been demonstrated in unfolding with a factorized pick & place action space by Wu et al. [4] and goal conditioned folding using spatial action maps [7,6,5] by Lee et al. [2]. However, all these approaches operate entirely in quasi-static action spaces.
Dynamic Cloth Manipulation. In contrast to quasi-static manipulation, which are "operations that can be analyzed using kinematics, static, and quasi-static forces (such as frictional forces at sliding contacts)" [1], dynamic manipulation involves operations whose analysis additionally requires "forces of acceleration". Intuitively, dynamic manipulation (e.g., tossing [6]) involve high-velocity actions which build up objects' momentum, such that the manipulated objects continue to move after the robot's end-effector stops. Such dynamic manipulation results in an effective increase in reach range and a reduction in the number of actions to complete the task compared to exclusively quasi-static manipulation. However, prior works on dynamic cloth manipulation have either relied on cloth's vertex positions [17] (so is only practical in simulation), a motion capturing system with markers combined with human demonstrations [18], or custom hardware [19,20].
While prior cloth unfolding works report high average final coverages from relative high average initial coverages, we emphasize that we consider severely crumpled initial cloth configurations, which are much more challenging yet still realistic (see section 4.2). Additionally, no prior works have considered unfolding cloths larger than the robot's reach range (indeed, Ganapathi et al. [16]'s workspace is significantly larger than their maximum cloth size due to physical reachability limitations). Our algorithm achieves high performance in these challenging cases through self-supervised trial and error and learns directly from visual input without requiring expert demonstrations or ground truth state information.
Method
The goal of cloth unfolding is to manipulate a cloth from an arbitrarily crumpled initial state to a flattened state. Concretely, this amounts to maximizing the cloth's coverage on the workspace surface. Intuitively, dynamic actions have the potential to achieve high performance on cloth unfolding by appropriately making use of the cloth's mass in a high-velocity action to unfold the cloth (Sec. 3.1).
From a top-down RGB image of the workspace with the cloth, our policy decides the next fling action (Sec. 3.2) by picking the highest value action which satisfies the system's constraints (Sec. 3.3). It predicts the value of each action with a value network (Sec. 3.4) which is trained in a selfsupervised manner to take actions that maximally increases the cloth's coverage. The supervision signal is computed directly from the visual observation captured by the top-down camera. After training in simulation, we finetune and evaluate the model in the real world (Sec. 3.5).
Advantages of the Fling Action Primitive
Quasi-static actions, such as pick & place, rely on friction between the cloth and ground to stretch the cloth out. The complexity of friction forces between the workspace and the (nonvisible, groundfacing) surfaces of the cloth means the cloth's final configuration may be difficult to predict from only visual observations, especially for novel cloth types or significantly different friction forces (i.e.: simulation v.s. real). In addition, such systems can't manipulate points on the cloth to regions outside of their physical reach range, which limits their maximum cloth dimensions. Meanwhile, dynamic actions largely rely on cloths' mass combined with a high-velocity throw to do most of its work. Since a single dynamic motion primitive can effectively unfold many cloths, dynamic unfolding systems can learn a simpler policy which generalizes better to different cloth types. In addition, they can reach higher coverages in smaller numbers of interactions and throw corners of cloths larger than the system's reach range, effectively expanding their physical reach range.
We propose to use a dual-arm pick, stretch, and fling primitive. Here, a dual-arm system stretches the cloth between the arms, which unfolds it in one direction, and then flings the cloth, which makes use of the cloth's mass to unfold it in the other direction. The combination of stretching and flinging ensures that, given two appropriate grasp points, our motion primitive should be sufficient for single step unfolding when possible. Figure 2: Action Primitives. The dynamic Fling primitive starts with a two-arm grasp at the left L and right R grasp locations with center point C, followed by a fixed stretch, fling, and place routine. A good fling could unfold a highly crumpled cloth in as little as a single step. In contrast, the quasistatic Pick & Place primitive, which grasps at the start location S, lifts, moves, then places at the end position specified by the arrow tip, requires many steps for such challenging cloth configurations. The quasi-static Pick & Drag primitive is just a Pick & Place without the lift step.
Fling Action Primitive Definition
To achieve an efficient, generalizable, and flexible cloth unfolding system, we argue that the system requires two arms operating in a dynamic action space. To this end, we've designed a pick, stretch, and fling motion primitive for a dual-arm system, where each arm is placed on either side of the cloth workspace, as follows. First, the arms perform a top-down pinch-grasp on the cloth at locations L, R ∈ R 3 . Second, the arms lift the cloth to 0.30m and stretch the cloth taut in between them. Third, the arms fling the cloth forward 0.70m at 1.4m s −1 then pull backwards 0.50m at 1.4m s −1 , which swings the cloth forwards. Finally, the arms place the cloth down to the workspace and release their grips. The entire motion primitive is visualized in (Fig. 2a).
Instead of parameterizing and learning all steps of our primitive, we can fix the stretching step to always stretch the cloth as much as possible without tearing the cloth. We also fix the fling speed and trajectory from the observation that the real world system could robustly unfold the cloth using a wide range of fling parameters (i.e.: fling height, fling speed) if given a good grasp (Fig. 7). Thus, the problem of learning to pick, stretch, and fling reduces down to the problem of learning where to pick, which is parameterized with only two grasp locations, one for each arm.
Constraint Satisfying Fling Action Parameterization
A dual-arm grasp is parameterized by two points L, R ∈ R 3 , which denotes where the left and right arm should approach from a top-down grasp respectively, and requires 6 scalars. Without loss of generality for the purposes of grasping from a top-down RGB-D input, the third dimension could be specified by depth information. This reduces L and R to two points in R 2 , each representing pixels to grasp from the visual input and uses only 4 scalars in total.
However, to minimize collisions between two arms, we wish to impose a constraint that L is always left of R, and vice versa (Fig 9a). Additionally, the grasp width (i.e., the length of the line L − R) must not exceed the physical limit of the system and be smaller than the minimum safe distance limit between the two arms ( Fig. 9b). Directly using L and R will entangle these two contraints (see supplementary materials), making them difficult to satisfy. To make these constraints linear and independent, we propose an alternative 4-scalar parameterization, which consists of pixel position of the point C ∈ R 2 at the center of the line L − R, an angle θ ∈ R denoting the planar rotation of the line L − R, and a grasp width w ∈ R denoting the length of the line in pixel space. To constrain L to be on the left of R, we can constrain θ ∈ [−90 • , 90 • ], while w can be directly constrained to appropriate system limits.
Learning Delta-Coverage Maximizing Fling Actions
The naive approach for learning the action parameters C x , C y , θ, w by directly predicting these 4 scalars does not accommodate the best grasp's equivariances to the cloth's physical transformations. However, the learner should be equipped with the inductive bias that cloths in identical configurations have identical optimal grasp points for flinging relative to the cloth, regardless of the cloth's translation, rotation, and scale. ...
.4 m /1 .5 x
2.0x .. To this end, we propose to use spatial action maps [5,6,7]. By predicting grasp values with constant scale and rotation in pixel space from the transformed images, spatial action maps can recover grasp values with varying scales and rotations in world space by varying the transformation applied to the image. Concretely, given a visual observation from the top-down view (Fig. 3a), we generate a batch of rotated and scaled observations (Fig. 3b) then predict the corresponding batch of dense value maps (Fig. 3c). Each pixel in each value map contains the value of the action parameterized by that pixel's location, giving C, and its observation's rotation and scaling, giving θ and w respectively (Fig. 3f). The value network is supervised to regress each pixel in the value map to the ground truth delta-coverage -the difference in coverage before and after the action. Thus, by picking the grasping action with the highest value (Fig. 3d), the system picks grasp points which it expects would lead to the greatest increase in cloth coverage. By its architecture, the value network is invariant to translation. By rotating and scaling its inputs, it is also invariant to rotation and scale.
a) Workspace b) Rotated & Scaled Images c) Predicted Value Maps d) Highest Value
In our approach, we use 12 rotations, which discretizes the range [−90 • , 90 • ], and 8 scale factors in the range [1.00, 2.75] at 0.25 intervals. Our value network is a fully convolutional neural network with nine residual blocks [21] and two convolutional layers in the first and last layer, and takes as input 64 × 64 RGB images. To evaluate actions within a useful range of scales, we crop the image such that the cloth takes up at most two-thirds of the image width and height before shrinking it by the inverse of the 8 scale factors above. We also filter out and reject all grasp point pairs which are out of reach for either arm.
Self-Supervised Learning. The value network is trained end-to-end with self-supervised trials in simulation then finetuned in the real world. In each training episode, a random task is sampled. Each fling step is labeled with its normalized delta coverage, which is computed by counting cloth mask pixels from a top-down view, then dividing by the cloth mask pixel count of a flattened cloth state. While our experiments have plain black workspace surface colors and non-textured colored cloths for easy masking, our approach can be combined with image segmentation or background subtraction approach to obtain the corresponding cloth mask to work with arbitrary cloth and workspace textures. Note that this supervision signal (i.e., cloth mask) is only needed during training.
The policy interacts with the cloth until it reaches 10 timesteps or predicts grasps on the workspace. The latter stopping condition indicates that the policy does not expect any possible action to further improve the cloth coverage. All policies are trained in simulation until convergence, which takes around 150,000 simulation steps, or 6 days on a machine with a GTX 1080 Ti. The network is trained using the Adam optimizer with a learning rate of 1e-3 and a weight decay of 1e-6.
Experiment Setup
Simulation Setup. Our custom simulation environment is built on top of the PyFleX [8] bindings to Nvidia FleX provided by SoftGym [9]. On top of the functionality provided by PyFleX and Soft-Gym, our custom simulator 2 can load arbitrary cloth meshes, such as T-shirts, through the Python API. While our simulation environment does not support loading URDFs of robots, we found it sufficient to represent only the arms' end effectors, and apply appropriate physical constraints to their locations. The observations are rendered using Blender 3D where the cloth HSV color is sampled uniformly between [0.0, 1.0], [0.0, 1.0], and [0.5, 1.0] respectively, and the background is dark grey with a procedurally generated normal map to mimic the wrinkles on the real-world workspace surface. We further apply brightness, contrast, and hue jittering on observations to help with the transfer to real. We manually tuned our simulation fling speed to qualitatively match the real-world cloth dynamics (see Sec. 6.3 on simulation sensitivity and real world robustness to fling parameters).
Real World Setup. Our real-world experiment setup consists of two UR5s, where one is equipped with a Schunk WSG50 and the other with an OnRobot RG2, facing each other and positioned 1.35m apart. The top-down RGB-D image is captured with an Intel RealSense D415. To help with pinch grasp success in real, we used a 2-inch rubber mat as the workspace and added sponge tips to the gripper fingers. To mitigate the reliance on specialized and expensive force sensor, we chose to implement stretched cloth detection using only a second Intel RealSense D415 capturing a frontal view of the workspace. By segmenting the cloth from the front camera RGBD view of the workspace, then checking whether the top edge of the cloth mask is flat as a proxy for the cloth being stretched, we can pull the arms' end effectors apart until the top of the cloth is no longer bent.
Evaluation
We design a series of experiments to evaluate the advantage of dynamic actions over quasi-static actions in the task of cloth unfolding. The system's performance is evaluated on its efficiency (i.e., reaching high coverages with a small number of actions), generalization to unseen cloth types (i.e., flattening different types of T-shirts when only trained on rectangle cloths), and reach range (i.e., on cloths with dimensions larger than its reach range). Finally, we evaluate the algorithm's performance on the real-world setup. Please visit https://flingbot.cs.columbia.edu for experiment videos.
Metrics
The algorithm performance is measured by the final coverage, delta-coverage from initial coverage, and the number of interactions. Final coverage measures the coverage at the end of an episode, whereas delta-coverage is final coverage minus initial coverage. All our coverage statistics are normalized (i.e., divided by the maximum possible coverage of the cloth in a flattened configuration) and can be easily computed from a top-down camera, while the number of interactions is the episode length. To evaluate our policy, we load a task from the testing task datasets then run the policy for 10 steps or until the policy predicts grasps on the floor.
Task Dataset Generation
Each task is specified by a cloth mesh, mass, stiffness, and initial configuration. The cloth mesh is sampled from one of three types: 1. Normal Rect, which contains rectangular cloths with size within the reach range. Edge lengths are sampled from [0.40m, 0.65m]. 2. Large Rect, which contains rectangular cloths with at least one edge larger than the reach range (0.70m). Otherwise, edge sizes are sampled from [0.40m, 0.75m], which means the shorter edge could still be smaller than the reach range. 3. Shirt, which contains a subset of shirts sampled from CLOTH3D's [22] test split, all of which are resized to be within the reach range. These shirts include tank tops, crop tops, short and long sleeves. The cloth mass is sampled from [0.2kg, 2.0kg] and an internal stiffness from [0.85kg/s 2 , 0.95kg/s 2 ]. Finally, the cloth's initial configuration is varied by holding a randomly grasped the cloth at a random height between [0.5m, 1.5m] then dropping and allowing the cloth to settle (similar to Lee et al. and tier-3 in Seita et al.), resulting in a severely crumpled configuration.
To calculate the normalized coverage, we use the maximum possible coverage of the cloths in their flattened configurations. For rectangular cloths in simulation, the flattened configuration can be analytically calculated using the undeformed vertex positions of the grid mesh, which means the normalized coverage could still be higher if the cloth rests in a stretched position due to friction. For shirts in simulation, we opted to calculate its maximum possible coverage as its outer surface area divided by 2, since a qualitatively flattened T-shirt may not actually maximize coverage. While this choice also makes it possible for the normalized coverage to be greater than 1, it will still preserve performance rankings. For real world experiments, the flattened configurations are manually set.
We emphasize the difficulty of our unfolding tasks, where cloths in Normal Rect, Large Rect, and Shirt have average initial coverages of 28.8%, 27.1%, and 46.4% in simulation respectively, and Step 2 Step 3 Step 3 Red and green circles represent grasps by left and right arms, placed above and below field of view, respectively. FlingBot discovered through trial-and-error a two-arm corner and edge grasp when corners and edges are visible. While the pick & place baseline discovered a similar strategy, its performance is inherently limited by quasi-static actions, requiring significantly more steps to achieve a final coverage lower than FlingBot's.
26.1%, 33.6%, and 24.0% in the real world respectively. In contrast, simulated cloths from Seita et al. start at 77.2%, 57.6%, and 42.0% for easy, medium, and hard tasks, while real world cloths from Ganapathi et al. starts at 71.4% and 68.4% on two different real-world trials. Yet, the challenging cloth configurations found in our tasks (Fig. 4) are realistic and prevalent in typical households.
In simulation, the policy is trained on 2000 rectangular cloths sampled evenly between Normal Rect and Large Rect, and evaluated on 600 novel tasks split evenly between Normal Rect, Large Rect, and Shirt cloths. In real, the simulation policy is deployed to collect real world experience on 150 Normal Rect episodes (257 steps), optimized on both simulation and real world data, then evaluated on 10 novel test tasks in each cloth type. [FlingBot], but directly predicts the action parameters C x , C y , θ, w from visual inputs instead of exploiting the task's equivariances. Its policy is trained using DDPG on deltacoverage rewards to maximize single-step returns. However, from Tab. 1, [Fling-Reg] completely fails to perform the task, demonstrating the advantage of encoding inductive biases which leverage equivariances in the problem structure.
Results
Better Efficiency. In this experiment, we compare the unfolding efficiency between quasi-static and dynamic actions. From the Normal Rect column in Tab Increased Reach Range. In this experiment, we investigate these approaches' performance on cloths which have dimensions larger than the robot arm's reach range. These tasks are not only challenging for quasi-static baselines, because the arms can't manipulate the cloth at locations beyond its reach range, but also for our flinging policies, because the arms can't fully stretch or lift the cloth off the ground. Despite these challenges, [Fling-Bot] achieves 79.2% (Tab. 1, column Large Rect). Compared to the quasi-static baselines, [FlingBot] increases the coverage by +52.0%, which is roughly twice that of the quasi-static baselines ( +27.1%, +24.8%, +23.1%). These results show how high-velocity actions could effectively expand the physical reach range, allowing the system to be more flexible with extreme cloth sizes.
Generalize to Unseen Cloth Types. In this experiment, we investigate how well these approaches, trained on only rectangular cloths, can generalize to unseen cloth types (i.e.: T-shirts). Qualitative real world (Fig. 4) and simulation (Fig. 6) results suggest that our flinging policy has learned to grasp keypoints on cloth (i.e., corners, edges) when it sees them, or otherwise fling to reveal these features. FlingBot's generalization performance (93.3% in Tab. 1, column Shirt) to novel cloth geometries can be attributed to this strategy, since a cloth of any type can be unfolded by grabbing one of its edges, stretching, then flinging it in a direction perpendicular to the edge. Through self-supervised exploration, our approach discovered grasping strategies which were manually designed in heuristic based prior works, while being more generalizable to different cloth configurations and types. Meanwhile, all quasi-static baselines exhibited worse cloth unfolding efficiency. The Sim Shirts plot in Fig. 5 shows that our flinging policies take only 3 actions to reach their maximum coverages, while the quasi-static baselines take upwards of 8 steps to reach lower maximum coverages. Evaluating Real-World Unfolding. Finally, we finetune and evaluate our simulation models from Tab. 1 with real-world experience on a pair of UR5 arms. Task generation is automated using the robot arms by randomly grasping the cloth at height 0.50m then dropping it back on the workspace. We use a 0.35m × 0.45m cloth for Normal Rect, a 0.40m × 0.70m bath towel for Large Rect, and a 0.45m × 0.55m T-shirt for Shirt. The system collected 257 experience steps over 150 cloth tasks for finetuning in total. The performance is reported averaged over 10 test episodes, where real-world grasp errors are filtered out (see "Real World Failure Cases" below). From Tab. 2, we report that our policy achieves over 80% on all cloth types, which outperforms the quasi-static pick & place baseline by over 40%. Additionally, we report that the pre-finetune performance of FlingBot on Normal Cloths is 69.8%, justifying our decision to finetune to get a 12.1% improvement. Overall, our flinging primitive takes a median time of 16.7s. While the pick & place primitive only takes a median time of 8.8s, it is unable to reach the high coverages even with many more interaction steps (Fig. 5, bottom row). Both primitives incur additional overheads of 2.5s for preparing the transformed observation batch and 2.9s for a routine which checks whether the cloth is stuck to the gripper after the gripper is opened (details in Sec. 6.4).
Real World Failure Cases. Grasping failures, where the policy specified a grasp point on the cloth but the grippers failed to successful pinch grasp, constituted all of our real-world pipeline failure cases. Our real world system uses its frontal RGBD view to detect grasp failures after the dual arm lifts the cloths up and automatically discards these episodes. The average grasp success rate is 78.0%, 45.0%, and 75.8% for normal rectangular, large rectangular, and shirts respectively. We used a bath towel for our large rectangular cloths, which is significantly thicker and stiffer than the tea towel and T-shirt we used for normal rect and shirts. Therefore, pinch grasps with large cloths failed significantly more. Additionally, even on successful grasps, the gripper may hold the cloth tight in its crumpled state, rendering all flings ineffective. However, we do not discard of these episodes. We discuss more of real world grasp failures in Sec. 6.4 and show simulation failure cases in Fig. 8.
Conclusion and Future Work
We proposed a dynamic fling motion primitive and a self-supervised learning algorithm for learning the grasp parameters for the cloth unfolding task. The cloth unfolding policy is efficient, generalizable, and works with cloth sizes beyond the reach range of the system in both simulation and the real world. While we've demonstrated the effectiveness of dynamic actions for cloth unfolding, dynamic actions alone are insufficient for more complex deformable object manipulation tasks, such as goalconditioned folding. In this direction, future work could explore learning in combined action spaces (i.e., dynamic and quasi-static, two-arm and single-arm, etc.) and integrating it with vision-based cloth pose estimation [25] for goal-conditioned cloth manipulation. For normal rectangular cloths, the most common failure case is when a dual-arm corner grasped is slightly misaligned and becomes a single-arm grasp and fling instead, resulting in a low coverage configuration. For large rectangular cloths, cloths could fold in half almost perfectly, thus appearing completely unfolded, causing the policy to terminate the episode. For shirts, the self-discovered dual-arm corner and edge grasp for flinging which is effective for rectangular cloths fail on shirts in two main ways. First, if the sleeves of the shirt get stuck in the shirt's collar, FlingBot will be unable to pull the sleeves out. This failure case motivates future work on combining quasi-static and dynamic actions for cloth manipulation. Second, dual arm grasps where one grasp is on the outer surface and another grasp is on the inner surface of the shirt usually flings to low coverages. While this failure case is expected to the differences between rectangular cloths and shirts (presence of holes, inner/outer surfaces, etc.), FlingBot's performance on shirts still suggested generalizable cloth manipulation abilities.
Real world fling parameter robustness
In designing our motion primitive, we optimized fling parameters (waypoints, velocities, acceleration) to maximize coverage assuming a good grasp (e.g.: a dual arm grasp on a normal rectangular cloth in a stretched state). We observed that the real world flinging setup system could robustly unfold the cloth using a wide range of fling parameters (i.e.: fling heights and speeds) if manually given a good grasp (Fig 7), while the simulated system was highly sensitive to such parameters. This sim2real gap was bridged in our work by tuning the simulated fling parameters such that flings from good grasps in simulation would also lead to high coverages like in the real system, which results in a variable fling height and a different fling speed in simulation compared to a fixed fling height and speed in real. Crucially, this gap underscores the importance of real world results for cloth manipulation as well as motivates future work on fast and accurate cloth simulation engines.
Another large sim2real problem was poor collision handling in simulation. There were cloth configurations that the real world system experienced but was not observed in simulation, such as cloth twisting. While Nvidia Flex is decent at preventing self penetration, it does so at the cost of unrealistic collision handling. Qualitatively, this unrealistic collision handling means cloths untwist themselves when twisted (or are in any other state with high self-collision). Another case where this self-unfolding behavior was observed was for shirts due to the two layers of the shirts colliding with each other. We hypothesize that poor collision handling is the main reason why performance for the simulated shirt benchmark is higher across the board for all approaches, despite being unseen cloth types.
Real world Failures
All failure cases in our real world experiment were due to failed grasps (where the policy specified a grasp point on the cloth but the grippers failed to pinch grasp the cloth). Cloth grasping failure is a common problem when working with real world cloth manipulation. For instance, Ganapathi et al. also observed that "the most frequent failure mode is an unsuccessful grasp of the fabric which is compounded for tasks that require more actions". However, we believe this issue can be mitigated by using specialized gripper hardware (like in Ganapathi et al., Seita et al.) or incorporating grasp success estimation.
In addition to grasping failures discussed in Sec 4.4, we also observed occasional cloth stuck errors, where cloths get stuck to the gripper after the gripper opens in an attempt to release the cloth. To enable the system to automatically recover from this issue, after opening the gripper, the arms move to a predefined height above the workspace. Then, we use the frontal RGB-D view (used to implement the stretching primitive) to check whether the cloth is detected above a manually set height threshold above the workspace.
Here, we summarize techniques for implementing a real-world cloth manipulation pipeline: 1. High friction gripper finger: In our dual-arm system, we used an OnRobot RG2 and Schunk WGS50 gripper, where the former had a rubber tip, and the latter had a metal tip. After observing a significantly lower pinch grasping success with the WSG50, we added a rubber fingertip to the WSG50, which improved its grasp successes significantly. Alternatively, picking a gripper that can apply lots of pressure using its fingers, like the da Vinci Research Kit surgical robot in Seita et al., should also help.
Soft Workspace:
A successful pinch grasp should apply the right amount of pressure between the cloth and the workspace. Too little pressure and the cloth will not get grasped, while too much pressure could easily damage the cloth, gripper, and robot arm. Due to noise in-depth sensing, the arms may be asked to grasp points slightly below the surface of the workspace. Similar to prior works [16,3], we found that using a firm and thin rubber mattress was sufficient to address this problem.
3. Accurate depth sensing: Building on the previous point, hardware improvements on the sensing side could help significantly. In our real world experiments, the Azure Kinect v3 has significantly less noisy and more accurate depth images than the Intel Realsense D415. While the real world numbers we reported in this paper are only with Realsense cameras, the codebase for our Kinect/Realsense real world setup is publicly accessible at https://github.com/columbia-ai-robotics/flingbot.
4 DOF Grasp success prediction:
A grasp which is parallel to a small crease on the cloth is more likely to result in a successful pinch grasp. Therefore, we hypothesize that additionally considering gripper z-rotation, on top of the positional 3 DOFs we have in our system, and learning the optimal pinch grasp z-rotation using a grasp success predictor may result in higher overall grasp success. Future work could explore weighing task rewards with such grasp success predictions.
Designing dynamic motion primitives
To achieve the highest speed at the end effector while respecting the torque limit of each joint, the primitive must move upper joints (i.e.: wrist) more than the lower joints (i.e., base). We also found that adding a blending radius between each target joint configuration gave a much smoother flinging trajectory and cloth swinging, as opposed to a jerky cloth motion without blending radius. In designing our motion primitive, we optimized fling dynamics parameters (waypoints, velocities, acceleration) to maximize coverage assuming a dual arm grasp on a normal rectangular cloth in a stretched state. Automatically discovering dynamic motion primitives, such as flinging, and simultaneously learning their parameters is an important and interesting direction for future work.
Constraints to Dual-arm Grasps
Cross-over Constraint Grasp Width Constraint 6.6 Why is learning fling speeds unhelpful?
We hypothesize that light and thin cloths, whose air resistive forces to momentum during flinging ratio is significantly higher than the cloths tested, would require a higher fling speed. Therefore, fling speed learning may be helpful when cloths in the task dataset contain a wider variance in density and thickness. We also hypothesize that a successful fling speed prediction approach may require extra information about the cloths' physical parameters that could not be obtained through visual observation alone, which would be out of scope for this work. | 2021-05-11T01:16:18.341Z | 2021-05-08T00:00:00.000 | {
"year": 2021,
"sha1": "7597a55eac399cb76d23ea34273e6ff07424f0da",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "7597a55eac399cb76d23ea34273e6ff07424f0da",
"s2fieldsofstudy": [
"Computer Science",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
258917947 | pes2o/s2orc | v3-fos-license | How nursing staffs deal with burnout syndrome through job satisfaction and self-efficacy: the fight or flight mechanism
Background. During the COVID-19 pandemic, the psyche nursing staff might suffer from burnout syndrome. This subsequently leads to decreased working performance which might compromise the quality of care. Nurses experience depersonalization. Objective. The study aimed to determine the effect of burnout syndrome on job satisfaction among nurses and how self-efficacy can solve the problem. Materials and Methods. Mix method study was conducted among 79 nurses from October 2021 until February 2022. The quantitative data were collected using the Maslach Burnout Inventory questionnaire and Minnesota Satisfaction Questionnaire via proportional random sampling. The qualitative data were collected by thematic analysis to find out nurses’ self-efficacy during the pandemic. Results. The results of the Spearman rank test prove a significant value (p)=0.004 with coefficient correlation (r)=-0,315. It means that burnout syndrome can affect job satisfaction negatively. Nurses should be skillful at conducting nursing care according to their assigned duties (performance outcome). Improving communication ability, especially the ability to utilize the local language. Mastering international languages might also help to cope with the market’s demand in the future (verbal persuasion). Nurses should be careful in perceiving the faced situation (observational learning) and try to think positively (emotional arousal). Conclusion. The higher the level of job satisfaction, the lower the level of burnout syndrome. Therefore, nurses should improve their self-efficacy as they are the healthcare front liners during the pandemic. With better self-efficacy, the quality of care should improve as well.
Introduction
The COVID- 19 pandemic is yet to end. Currently, several countries worldwide are faced with a significant increase in case numbers, suggesting the emergence of the third wave. 1 In Indonesia, as it has been quite a while since the pandemic started in March 2020, burnout syndrome happens among nursing staff. 2 Burnout syndrome also leads to decreased working performance among nurses. In the workplace, two phenomena have been proven to affect the quality of patient care, namely, burnout syndrome and job satisfaction. 3,4 However, previous studies that aim to determine the relationship between burnout syndrome and job satisfaction were still inconsistent. Burnout syndrome is characterized by 3 dimensions of burnout according to the multidimensional theory by Maslach, such as emotional exhaustion, depersonalization, and lack of personal accomplishment. 5 The depersonalization arises as negative behaviors of workers toward co-workers and others as their cope mechanism for extreme tiredness. Nurses experience depersonalization. Therefore, improved self-efficacy (SE) is necessary as it might significantly benefit the nurses (being a protective factor) despite its possible risks. 6 If nurses could understand the importance of SE, this might prevent the occurrence of burnout syndrome.
Objective
The study aimed to determine the effect of burnout syndrome on job satisfaction among nurses and how self-efficacy can solve the problem.
Desain study
This study is a mixed method study with a cross-sectional design about the effect of burnout syndrome on nursing personnel's job satisfaction during the COVID-19 pandemic in Indonesia (quantitative measurement) and how self-efficacy can solve that problem (qualitative measurement).
Study respondents
The population of this study is nursing personnel who works at 3 major hospitals in Jember Regency, East Java, Indonesia (Dr Soebandi Hospital, Perkebunan Jember Klinik Hospital, Kaliwates Hospital). The study was conducted from October 2021 to February 2022. Slovin's formula calculated this study's sample size. A total of 79 respondents were recruited via a proportional random sampling technique. They should meet the researcher's criteria: i) nursing personnel with an associate and bachelor's degree in nursing; ii) providing health services to COVID-19 patients directly.
Respondents will be excluded from the study if they likely experience depression based on the PHQ-9 questionnaire scoring.
Quantitative measurement
Data was collected using 4 standard questionnaires that have been tested for validity and reliability (19)(20)(21). The questionnaire used in this study was obtained from previous studies and adapted to the Indonesian language (22)(23)(24)(25). It consists of a sociodemographic questionnaire, Maslach Burnout Inventory (MBI), Minnesota Satisfaction Questionnaire-Short form (MSQ-Short form), and Patient Health Questionnaire-9 (PHQ-9).
The sociodemographic questionnaire identifies respondents' sociodemographic data, including name, age, gender, marital status, years of experience, kind of job, working duration, and working unit.
The MBI questionnaire consists of 21 questions (6 items about emotional exhaustion, 6 items about depersonalization, and 7 items about lack of personal accomplishment) to evaluate burnout syndrome perception. Every item was interpreted by 4 points Likert scale, 1 for never, 2 for seldom, 3 for often, and 4 for always perception.
PHQ-9 is a questionnaire used to determine the symptoms of depression based on 9-item questions. Each item was interpreted by 3 points Likert scale, 0 for not at all, 1 for several days, 2 for more than half of the days, and 3 for nearly everyday perception. In this study, PHQ-9 was explicitly used as a tool to exclude the respondent from reducing the study's bias caused by overlapping burnout syndrome and depression. Respondents will be excluded from this study if he gets a total score > 4.
Qualitative measurement:
Qualitative data was collected by using thematic analysis. Selfefficacy domains consist of performance outcome, verbal persuasion, observational learning, and emotional arousal.
Statistical analysis
Sociodemographic characteristics, burnout syndrome perception, and job satisfaction perception were presented by frequency and percentage. At the same time, Spearman Rank Test was used to test the effect of burnout syndrome on job satisfaction with an ordinal data scale. The effect of burnout syndrome on job satisfaction is significant if the significant value P<0.05, while correlation strength is determined by correlation coefficient (r) with some classification, very weak correlation (r=0.00-0.199), weak correlation (r=0, 20-0.399), moderate correlation (r=0.40-0.599), strong correlation (r=0.60-0.799), and very strong correlation (r=0.80-1000). Data were analyzed by Microsoft Excel 2010 and Statistical Package for the Social Sciences version 24 software.
Ethical coincidence
This study has passed the ethical coincidence test by the Ethics Commission of Medical Faculty, Jember University, as certified under ethic No: 1544/H25.1.11/KE/2021. All the study respondents were informed about the study and voluntarily participated.
Distribution of sociodemographic characteristics
As shown in Table 1, most respondents (55.7%) are around age 31-40 years old. The gender distribution shows that 38 respondents (48.1%) are male, and 41 (51.9%) are female. Among the respondents, 75 were married (94.9%), and 4 others were single. In this study, 24 respondents (30.4 %) from the COVID-19 Treatment unit, 16 respondents (20.3%) from the non-COVID-19 treatment unit, 10 respondents (12.7 %) from the emergency unit, etc. During this research, most of the workers had worked for > 10 years (58.2%), and 55 respondents (69.6 %) had a work duration of ≤8/day. Table 2 shows that the perception of burnout syndrome is dominated by mild levels (83.7%), with no respondents who experience moderately high and high levels of burnout syndrome.
Burnout syndrome's perception
The mild level of burnout syndrome is likely related to the sociodemographic characteristics of the respondents shown in Table 1. The characteristics of the respondents, including age (31-40 years), marital status (married), length of work (>10 years), and duration of daily work (≤8 hours/day), were found to associate with a low tendency of burnout syndrome. Young health workers tend to develop burnout syndrome more than older health workers during stressful conditions. 7 Older health workers were assumed to have more work experience and thus have adequate abilities in managing work stress ozkul. 7,8 The following characteristic, marital status, was also found to be associated with the perception of burnout syndrome in the respondents. Single physician personnel experienced more depersonalization and emotional exhaustion than married physician personnel. 9 Social support from family helps nursing personnel deal with work stress. 10 The third characteristic is the length of work. It has been reported that health workers with work experience of fewer than ten years are inclined to emotional exhaustion and decreased selfachievement. 11 Another study among nurses in Poland stated that fresh graduate nurses experience burnout syndrome more than experienced nurses while working during the COVID-19 pandemic. 12 The fourth characteristic is the duration of daily work. A daily work duration of more than 8 hours was associated with high emotional exhaustion and depersonalization in health workers in Egypt during the COVID-19 pandemic. 13 The perception of mild burnout syndrome can also be related to organizational support. The incidence of burnout syndrome can be reduced by improving modifiable risk factors, such as increasing organizational support to meet health workers' physical and emotional needs. 14 The perception of a mild level of burnout syndrome can also be associated with a short research period in this study. The time for collecting data was conducted after the second wave of COVID-19 in Indonesia (July-August 2021), and there has been a lowering in COVID-19 cases in Indonesia. Table 3 shows that 43 respondents (54.4%) had a high level of job satisfaction, 36 respondents (45.6 %) had moderate job satisfaction, and no low level was found.
Job satisfaction perception
Job satisfaction is a perception resulting from workers' evaluation of their work. Well-being at work is one of the critical aspects that can be used to make sure that the workers are safe, healthy, and engaged at their work. 15 A high level dominated the respondent's perception of job satisfaction according to the respondent's response to the 20-item Minnesota Satisfaction Questionnaire-Short form statement, which was dominated by a neutral and satisfying response. Most respondents were satisfied with how the leadership handles employees, the supervisor's competence, how company policies are implemented, communication between co-workers, and the working conditions. Furthermore, the response to the perception of the amount of salary was balanced between neutral and satisfied. In line with a previous study conducted in Aisyiyah General Hospital Ponorogo, Indonesia, several factors such as salary, work environment, the opportunity to grow, and leadership correlate with nursing personnel's job satisfaction during the COVID-19 pandemic. 16 Organizational and social support received by nursing personnel affects job satisfaction during the COVID-19 pandemic. 17,18 Perceptions of well-being in the workplace arise from evaluating all aspects, including the work environment quality and the support of the workplace organization.
Responses to the MSQ-Short form also showed that most respondents felt satisfied with the opportunity to develop their work and the praise they received. Respondents have been given training on health protocols to minimize the transmission of COVID-19 and handling patients with COVID-19 during the pandemic. Following the Two Factor theory by Hertzberg, the satisfiers (motivational) group, which includes job advancement, personal growth, and job recognition, affects job satisfaction by influencing work motivation in individuals. Promotions and job acknowledgments from superiors and co-workers fulfill the respondent's self-actualization needs to provide work motivation and influence perceptions of job satisfaction. 15,19 The effect of burnout syndrome on job satisfaction Table 4 shows that the result of the Spearman rank test proves a weak negative effect of burnout syndrome on job satisfaction with a significance value of P=0.004 and coefficient correlation (r)=-0.321. This study's results align with the research results by Andarini and Mijakoski, which concluded that burnout syndrome negatively affects job satisfaction. 20,21 The effect of burnout syndrome on job satisfaction was based on stressful conditions at work. Stress at work occurs when job demands are not matched with individual abilities and resources to cope with their job demands. 6 Workers who are required to work beyond their capacity tend to experience emotional exhaustion, characterized by the emergence of negative affect such as feeling emotionally tired and losing energy in carrying out their duties. 22 Additionally, workers who continuously experience fatigue tend to experience depersonalization as their coping mechanism to workplace stress. 2,50 Depersonalization arises as negative behaviors toward work and others, such as cynicism, irritability, ignorance of their duties, and ignorance of the work environment. Thus, it may create a bad organizational atmosphere. The perception of the workforce in the organizational atmosphere affects the evaluation of their welfare in the workplace. 23 Despite it, the effect of burnout syndrome on job satisfaction in this study was weak, (r) -0.315. A survey of social workers concluded that social support at work positively correlates with job satisfaction and mediates the negative correlation between burnout and job satisfaction. 24 Besides that, compensation and a safe work environment reduced the adverse effects of emotional exhaustion on job satisfaction. 25 Workers who receive proper compensation for the work that they have done will feel satisfied with their work even though they are emotionally tired. 26 In this study, respondents felt good perceptions of organizational support and social support from co-workers. It was marked by respondents' responses which were dominated by satisfying responses on how the leader handled employees, communication methods between co-workers, job praise received, salary received, and organizational support in creating a safe work environment.
Article
The qualitative results of the interview with Oswen (pseudonym), a 33-year-old nursing staff working in the intensive care unit, revealed that burnout syndrome was an actual threat to the nursing staff's mental well-being. Oswen had worked as a permanent employee in a type A COVID-19 referral hospital for eight years. Ever since the pandemic struck, Oswen had been one of the front-liners caring for patients with COVID-19. He was initially assigned to care for patients in the COVID-19 isolation ward, then transferred to the COVID-19 intensive care unit four months ago. On the first month of working in the COVID-19 isolation ward, he intended to transfer as he perceived the situation significantly made him feel physically and mentally burdened. During the peak of the pandemic's second wave, his feeling of suffocation and exhaustion seemed to be even more unbearable. Fortunately, the situation did not last long. He witnessed the fellow nurses who seemed to feel the same. As he found people to share his burden with, he gradually learned to be at peace with the situation. Frequent sharing and cheering on each other with fellow nurses seemed to help ease the perceived workload.
Having proper self-efficacy leads nurses to improved control of burnout syndrome. 27 Therefore, nurses should be aware of SE as a protective instead of a risk factor during the pandemic. SE might adversely become a risk factor when it makes nurses too confident and leads to forgetting their regular tasks and functions, along with their limitations. Hence, the positive context of the SE should be evaluated according to the place and the actual condition. Excessive SE might inadvertently become a threat to nurses during the pandemic state. Therefore, nurses should be able to properly utilize their SE, and recognize when and how to utilize the domains inherent in SE. SE domains consist of the performance outcome, verbal persuasion, observational learning, and emotional arousal.
Performance outcome
Nurses should be skillful at conducting nursing care according to their assigned duties. Their competence would support their patient caring performance, both the non-emergency and emergency COVID-19 patients. Wearing proper personal protective equipment could also increase their confidence when dealing with infectious and dangerous patients with COVID-19. 28
Verbal persuasion (improve community skills)
Improving communication ability, especially the ability to utilize the local language. Mastering international languages might also help to cope with the market's demand in the future. Be able to master English, Chinese, or other international languages. This would lead to better relatedness between the nurses and the patients and patient's families, which might reduce misunderstandings that could possibly lead to malpractices. 29,30 Observational learning Nurses should be careful in perceiving the faced situation. Emergency nursing-related sciences should be applied when faced with patients with COVID-19. One should be mindful of oneself
Emotional arousal
Make efforts to think positively. Don't add the burden of irrational thoughts and prejudices. This pandemic state is inevitably physically and mentally draining. Caring for patients with a poor condition and even the death that frequently happens might lead to negative memories. One should attempt to put upfront empathy as a health care worker instead of practicing the type of empathy that might adversely affect one's wellbeing. 28 Figure 1 demonstrated the priority order of SE domain problem solving from ones that are not important nor require immediate action to the ones that need immediate action due to its impending adverse consequences if nothing is done.
Conclusions
This current study found that burnout syndrome negatively affects job satisfaction among nursing personnel at 3 major hospitals in Jember Regency, Indonesia. The respondents experience mild burnout syndrome and a high level of job satisfaction. Having proper self-efficacy leads nurses to improved control of burnout syndrome by minimalizing depersonalization. | 2023-05-27T15:03:11.757Z | 2023-05-25T00:00:00.000 | {
"year": 2023,
"sha1": "e7b277d425d0c8954b91a05f7b116ce77f6eeffc",
"oa_license": "CCBYNC",
"oa_url": "https://www.publichealthinafrica.org/jphia/article/download/2551/872",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "9190f33a1eac8978a49fd50e485950ac0e891d2d",
"s2fieldsofstudy": [
"Psychology",
"Medicine"
],
"extfieldsofstudy": []
} |
18357250 | pes2o/s2orc | v3-fos-license | Are There Rab GTPases in Archaea?
A complex endomembrane system is one of the hallmarks of Eukaryotes. Vesicle trafficking between compartments is controlled by a diverse protein repertoire, including Rab GTPases. These small GTP-binding proteins contribute identity and specificity to the system, and by working as molecular switches, trigger multiple events in vesicle budding, transport, and fusion. A diverse collection of Rab GTPases already existed in the ancestral Eukaryote, yet, it is unclear how such elaborate repertoire emerged. A novel archaeal phylum, the Lokiarchaeota, revealed that several eukaryotic-like protein systems, including small GTPases, are present in Archaea. Here, we test the hypothesis that the Rab family of small GTPases predates the origin of Eukaryotes. Our bioinformatic pipeline detected multiple putative Rab-like proteins in several archaeal species. Our analyses revealed the presence and strict conservation of sequence features that distinguish eukaryotic Rabs from other small GTPases (Rab family motifs), mapping to the same regions in the structure as in eukaryotic Rabs. These mediate Rab-specific interactions with regulators of the REP/GDI (Rab Escort Protein/GDP dissociation Inhibitor) family. Sensitive structure-based methods further revealed the existence of REP/GDI-like genes in Archaea, involved in isoprenyl metabolism. Our analysis supports a scenario where Rabs differentiated into an independent family in Archaea, interacting with proteins involved in membrane biogenesis. These results further support the archaeal nature of the eukaryotic ancestor and provide a new insight into the intermediate stages and the evolutionary path toward the complex membrane-associated signaling circuits that characterize the Ras superfamily of small GTPases, and specifically Rab proteins.
Introduction
A major question in evolutionary biology is the origin of the Eukaryotic cell plan, which is characterized by a multitude of intracellular organelles, including the energy producing endosymbiotic organelles, complex endomembrane trafficking system, and a nucleus containing a large genome that encodes thousands of genes. The protein repertoires associated with these organelles have been found in most Eukaryotes, suggesting that they were already present in the Last Eukaryotic Common Ancestor (LECA) (e.g., Field and Dacks 2009;Schlacht et al. 2014). Like in other areas of the evolutionary biology, the search for intermediate, transitional forms has attracted the attention of many, and eukaryotic-like cellular features or gene repertoires have been identified in different prokaryotes, for example, having been termed as the "dispersed eukaryome" in Archaea (Koonin and Yutin 2014).
Inferring ancient events such as the origin of Eukaryotes or the origin of their specific molecular traits is a very challenging task given the timescale, data scarcity, and insufficient methods. Despite this, mounting evidence suggests that the ancestral host cell that accommodated the endosymbiotic bacteria, which gave rise to mitochondria, was from the archaeal lineage (Lake et al. 1984;Cox et al. 2008, reviewed in L opez-Garc ıa andMoreira 2015). This host cell may have in fact evolved from within Archaea (the TACK superphylum), rather than result from a much earlier branching as a sister group to all Archaea (Guy and Ettema 2011;Kelly et al. 2011;Williams et al. 2012Williams et al. , 2013Williams and Embley 2014;Raymann et al. 2015). This scenario suggests that the search for transitional states should be carried out within the archaeal domain, and specifically the TACK superphylum.
A recent metagenomic survey of a deep ocean sediment sample from the Arctic Mid-Ocean Ridge revealed the existence of a new archaeal phylum within the TACK superphylum, the Lokiarchaeota (Spang et al. 2015). The authors reported that several building blocks characteristic of Eukaryotes are present in this taxon, suggesting that Lokiarchaeota and Eukaryotes share a common ancestor and that Lokiarchaeota is a modern descendant of that ancestor. Small GTPase gene families are highly expanded in Lokiarchaeota compared with other Archaea, including many small GTPases from the RAS superfamily; they form several distinct clusters, yet their relationship to the eukaryotic GTPases remains unclear.
The eukaryotic RAS superfamily contains five major families Arf, Ras, Rho, Ran, and Rab that are involved in the intracellular signaling and share the common G domain core (GTPase activity), responsible for the switching mechanism between the GTP-bound active and GDP-bound inactive state. The Arf family is involved in regulation of vesicular transport, Ras in response to diverse extracellular stimuli, Rho in actin dynamics, and Ran in nucleocytoplasmic transport (reviewed in Wennerberg et al. 2005). Here, we focused on Rab GTPases, critical regulators of vesicular trafficking systems (Fukuda 2008;Stenmark 2009;Kelly et al. 2012;Pfeffer 2013), included in the list of eukaryotic signature proteins, that is, "proteins that are found in eukaryotic cells but have no significant homology to proteins in Archaea and Bacteria" (Hartman and Fedorov 2002). This family has experienced extensive universal and taxon-specific duplications associated with the emergence of major organelles and organelle specializations of the endomembrane system; each Rab subfamily provides specificity to a particular component of the trafficking system and this function is generally conserved throughout evolution (Dacks and Field 2007;Dacks et al. 2009;Brighouse et al. 2010;Diekmann et al. 2011). They form the largest RAS family, with more than 60 Rab homologues in human , and several studies point to the existence of a rich Rab repertoire at the LECA (Diekmann et al. 2011;Elias et al. 2012;Klöpper et al. 2012); however, they have been so far restricted to the eukaryotic domain. Here, we test the hypothesis that Rab GTPases predate Eukaryogenesis, by investigating the small GTPase repertoire in Archaea, and in particular the expanded small GTPase family in the recently described Lokiarchaea.
Multiple Rab-like Sequences in Archaea
In the original metagenomic study by Spang et al. (2015) the assembly of a complete archaeal genome defined a novel archaeal phylum, the Lokiarchaeota. In this Lokiarchaeum genome, more than 90 members of the RAS superfamily were predicted, yet it is unclear whether these proteins belong to any specific, previously described RAS family or constitute a novel group. Here, we systematically searched all complete archaeal genomes, including the Lokiarchaeum, for members of the RAS superfamily of small GTPases and specifically annotated Rab-like proteins. We used the Rabifier (Diekmann et al. 2011), a bioinformatic pipeline that runs a series of consecutive classification steps as follows: 1) determining if a protein contains the small GTPase domain, 2) whether it belongs to the Rab family or another member of the RAS superfamily, and 3) what is the most likely Rab subfamily assignment of the protein. We detected a total of 3,152 proteins containing the small GTPase domain, of which 133 within the Lokiarachaeum genome (the remaining an average of 13.6 6 3.4 proteins per genome). Of this total, 42 were predicted as Rab-like GTPases without any specific subfamily annotation, that is, none of the Rab-like proteins is sufficiently similar to any of the established eukaryotic subfamilies. Among the 42 Rab-like proteins 37 belong to Lokiarchaeum, the remaining five (one copy per species) were identified in Thermofilum pendens, Thermofilum sp., Caldiarchaeum subterraneum, Thermoplasmatales archaeon, and Aciduliprofundum sp. These species are distributed across Archaea, they belong to one of two major superphyla, Euryarchaeota and TACK. This raises a question about the origin of these Rab-like proteins, as their phylogenetic profile
Inconclusive Phylogenetic Positioning of Archaeal Rab-like Sequences
Our bioinformatic analysis confirms the presence of many small GTPases in Archaea and identifies multiple Rab-like GTPases in diverse archaeal species, yet without any subfamily assignment. To determine the position of archaeal Rab-like proteins within the superfamily of small GTPases and their relationship to eukaryotic Rabs, we conducted a phylogenetic analysis of archaeal Rab-like proteins together with the eukaryotic Rabs which are likely present in the LECA (Diekmann et al. 2011;Elias et al. 2012), also including representative sequences of other RAS families. We used both Bayesian and Maximum Likelihood approaches for the phylogenetic inference (see Materials and Methods for details).
As previously observed (Dong et al. 2007;Rojas et al. 2012), trees of small GTPases have very weak statistical support for FIG. 1. Phylogenetic profile of the Rab family in representative species of eukaryotes (magenta), Archaea (red), and bacteria (blue). The remaining archaeal species that were used in the analysis, without Rablike protein predictions, are not shown in the figure. A full (hollow) square indicates the presence (absence) of at least one predicted eukaryotic Rab protein (black) or archaeal Rab-like protein (gray). The total number of Rab homologues is shown next to the square. TACK refers to the superphylum that comprises the Thaumarchaeota, Aigarchaeota, Crenarchaeota, and Korarchaeota phyla. Tree topology is consistent with Spang et al. (2015). To gain a more detailed view on the Rab-like family structure, we constructed a phylogenetic tree using only archaeal Rab-like sequences (supplementary fig. S3, Supplementary Material online). Although the deep branching pattern could not be reliably resolved, we observed that most of the sequences cluster within several highly supported groups. Short terminal branches suggest recent duplication of several Lokiarchaean proteins. Proteins from Thermoplasmatales, Aciduliprofundum, and Caldiarchaeum form long branches indicating very divergent sequences, which do not cluster together with Lokiarchaeum. In contrast, proteins from both Thermofilum species form a distinct cluster with two other Lokiarchaean sequences.
Overall, this analysis suggests that phylogenetic methods alone are insufficient to determine the relationship between archaeal Rab-like GTPases and the eukaryotic members of the RAS superfamily. This, however, raises the question of why these sequences were classified as Rab-like.
Rab-like Proteins Contain Typical Eukaryotic Rab Motifs
We next analyzed sequence properties of archaeal Rab-like GTPases at the family level to further assess their similarity to other members of the RAS superfamily. We constructed a sequence model for each family (Rab, archaeal Rab-like, Ran, Rho, Ras, Arf). We first built multiple sequence alignments using representative sequences for each family and a seed alignment of the small GTPase domain (Pfam:PF00071) to guide the alignment process and improve an overall quality of the alignment, the seed sequences were then removed from the final alignment. The alignments were subsequently used to construct profile hidden Markov models (pHMMs) and generate plurality-rule consensus sequences that describe each family.
We first calculated the overall, pairwise similarity between the families (supplementary table S2, Supplementary Material online) and observed a remarkable similarity of 78% (60% identity, local alignment) between eukaryotic Rab and archaeal Rab-like GTPases (71% and 55%, respectively, for global alignment, supplementary table S3, Supplementary Material online), much higher than between the archaeal Rab-like family and any other member of the RAS superfamily. We subsequently focused on a more specific comparison between Rab-like and Rab proteins; we compared amino acid variation along the sequence across Rab paralogues in Lokiarchaeum and representative species from different major eukaryotic groups (Homo sapiens, Trypanosoma brucei, and Guillardia theta). We observed similar patterns of variation for all analyzed species (supplementary fig. S4, Supplementary Material online): regions of both low and high sequence conservation belong to the corresponding positions in the Rab sequences from different species, suggesting that archaeal Rab-like sequences are evolutionarily constrained in the same regions as the eukaryotic Rabs.
We next tested the hypothesis that sequence conservation between archaeal and eukaryotic sequences is associated with the RabF motifs-sequence motifs unique to the Rab family that are important for the interaction with Rab effectors (Pereira-Leal and Seabra 2000). The results of this analysis are summarized in figure 3. All positions that correspond to the RabF1 and RabF2 motifs in eukaryotic Rabs are conserved in the archaeal Rab-like sequence. For comparison, in other families at most two amino acids are conserved at the corresponding positions. In the remaining three motifs most of the residues are identically conserved between Rab and Rab-like sequences, some are similar, for example, positively charged arginine and lysine in RabF4, aliphatic isoleucine and leucine in RabF5, and aromatic tyrosine and phenylalanine in RabF5 (tyrosine is also the second most common amino acid at this position in the archaeal sequences). From the sequence perspective, archaeal Rab-like proteins have all the hallmarks of Rabs, including the motifs involved in binding Rab regulators and effectors. The major difference between eukaryotic Rab and archaeal Rab-like sequences is the absence of C-terminal cysteine residues, the prenylation sites of the eukaryotic Rabs, in all of the analyzed archaeal sequences. Rab-like sequences tend to have a shorter C-terminal sequence, missing most of what is termed the (flexible) hypervariable region in eukaryotic Rabs, known to be involved in associations with the membrane.
Rab-like Proteins Are Structurally Similar to Eukaryotic Rabs
Given a high level of the primary sequence similarity between the archaeal Rab-like proteins and their eukaryotic counterparts, we modeled a putative 3D structure of a Rab-like GTPase and compared the location of Rab-specific features at the structural level. We chose a Lokiarchaeum sequence that contains all five RabF motifs (GenBank:KKK40223), as predicted by the Rabifier. To ensure a high quality of the model, we selected four templates from different Rab subfamilies that both have a high level of sequence identity to the archaeal homologue and a good crystallographic resolution of the 3D structure: Rab8 (H. sapiens, PDB:4LHW), Rab26 (H. sapiens, PDB:2G6B), Rab30 (H. sapiens, PDB:2EW1), and Ypt1 (Saccharomyces cerevisiae, PDB:1YZN). All template structures were in the active state, that is, bound to a GTP molecule. We used Modeller (Sali and Blundell 1993), a homology modeling platform to predict a putative structure of the archaeal protein (using all four templates simultaneously) and subsequently assessed its quality and stability. We obtained a similar structure using Phyre2 (Kelley et al. 2015), an automatic server for protein structure prediction and analysis (not shown). Figure 4 shows structures of both the model and the yeast template. Rab motifs are highlighted in blue (RabF motifs) and orange (guanine nucleotide-binding residues). Both structures are very similar (0.41 Å root-mean-square deviation of the Ca atomic coordinates), motifs are localized at the same structural elements and similarly exposed to the environment. We also compared the location of hydrophobic ( fig. 4b) and charged (fig. 4c) amino acids at the protein surface and observed a similar distribution of the residues in both structures.
We assessed the putative GTPase activity and the nucleotide-dependent conformational change of the archaeal Rablike protein by analyzing its thermodynamic stability at both the GDP and GTP-bound state and predicting interactions between the protein and the phosphate groups of the nucleotide. In addition to the model of the GTP-bound state, we modeled the structure of the GDP-bound form, again using several templates belonging to different Rab subfamilies: Rab1 . The interaction between the phosphate groups and the protein is stabilized by several residues present in the protein active site. The presence of Gln68 and its relative position to the GTP molecule enables the interaction between a water molecule and the phosphate, necessary for the GTP hydrolysis (Dumas et al. 1999). The analysis of structural models of the archaeal Rab-like GTPase indicates that it can exist in two stable conformations and it is able to cycle between an "on" and "off" state like other small GTPases and, in particular, eukaryotic Rabs.
A Rab Escort Protein/GDP Dissociation Inhibitor Ancestor in Archaea
Our analysis so far suggests that Rab-like sequences predate Eukaryogenesis. Surprisingly, we found motifs in archaeal Rablike sequences that are known to mediate interactions between eukaryotic Rabs and their regulators and effectors. Eukaryotic Rabs are prenylated on the C-terminus, a posttranslational modification catalyzed by the enzyme Rab geranylgeranyltransferase, which requires a chaperone termed REP (Rab Escort Protein) Leung et al. 2006); a paralogue of REP, termed GDI (GDP dissociation Inhibitor) recycles Rabs in and out of membranes ; fig. 5a). Binding of Rabs to REP and GDI is mediated by residues in the RabF motifs (Rak et al. 2003(Rak et al. , 2004Goody et al. 2005). The same regions are involved in binding other general Rab regulators-Rab activity is regulated by guaninenucleotide-exchange factors (GEF) that turn Rabs "on" by promoting the GDP to GTP exchange, and by GTPaseactivating proteins (GAP) that increase GTP hydrolysis rate and turn Rabs off. Both sets of proteins interact with Rabs with residues included in the RabF motifs (those within the switch regions). The identification of RabF motifs in Archaea raises the hypothesis that such proteins and interactions could also predate Eukaryogenesis. We used two approaches to test if homologues of these eukaryotic proteins can be detected in Archaea, indicating that some of the complex Rab regulatory cycles could predate Eukaryogenesis. First, we used sequences of several human regulators (GEFs, GAPs, FNT, PGGT1B, REP, RABGGT), performed BLAST (Altschul et al. 1990) similarity searches against archaeal genomes and found only hits with insignificant sequence similarity (not shown). As BLAST is known to lack sensitivity to detect remote homologies, we then used a more sensitive approach based on pHMM. We retrieved pHMMs (Pfam) of the domains that are found in Rab-binding proteins (Mss4, Sec2, VPS9, DENN, RabGAP-TBC, GDI/REP, prenyltransferase, PPTA), which we then used as queries for a similarity search using the HMMER package. In most cases, we found only scattered hits on the tree with marginal sequence similarity ( fig. 5b), suggesting that either canonical Rab regulatory proteins are absent from Archaea or their sequences diverged from the eukaryotic counterparts beyond the detection level of standard automated methods. In one case, however, that of REP/GDI, even though the statistics of the hits were poor, we observed repeated positive hits, which we then investigated further.
We manually inspected putative GDI/REP domains in Archaea. The primary sequence of GDI and REP domain containing proteins is generally weakly conserved in Eukaryotes, both within each family and between GDI-REP paralogues (e.g., 30% human and fruit fly REP, 21% human GDI1 and REP1, local alignment identity). Hence, given the evolutionary distance between Eukaryotes and Archaea we expect that any putative archaeal homologs would be within the "twilight zone" of sequence similarity, which precludes any automatic sequence-based analysis. We used a fold recognition method (Jones 1999) with the best scoring (HMMER) archeal GDI/REP protein to detect candidate proteins with determined 3D structures. The best predictions belong to eukaryotic GDIs and archaeal proteins without experimentally determined function (top three hits correspond to proteins from Bos taurus PDB:1D5T, Pyrococcus furiosus PDB:3NRN, and S. cerevisiae PDB:2BCG). These structures are also very similar to FAD-containing monooxygenases and oxidases (Schalk et al. 1996), including archaeal geranylgeranyl reductases. While the sequence identity between putative archaeal GDI/REP and eukaryotic GDI is very low, at the structural level both domains (3NRN and 1UKV, a yeast GDI in complex with YPT1) are similar, including the Rab-binding platform ( fig. 5c); our structural comparison revealed several residues that may form interactions with Rab switch regions (not shown). Our results strongly support the existence of a REP/GDI-like molecule in the TACK group, whose function implies an isoprenyl-binding ability.
We further used the same strategy to investigate whether the isoprenylation machinery, specifically the two subunits (a and b) of the eukaryotic isoprenyl transferases, is present in Archaea. Both approaches were inconclusive to determine the existence of the a subunit, as the tetratricopeptide repeat that characterizes this domain is widespread and functionally promiscuous, precluding any conclusion about function. However, we detected archaeal proteins whose predicted fold matches several isoprenoid metabolism enzymes including the geranylgeranyl transferase subunit b. We found multiple instances of genes containing these domains, observing some species where they co-occur ( fig. 5d).
Discussion
In this work, we investigated the hypothesis that the separation of the eukaryotic signature Rab sequences predates the emergence of Eukaryotes. This hypothesis follows from the recent discovery of a new archaeal group, the Lokiarchaeota, that was claimed to be a sister group of Eukaryotes. Our Rabifier pipeline identified 42 candidate Rab-like sequences that have multiple features related to eukaryotic Rabs, they exist in several Archaea of both the TACK group and Euryarchaeota but are particularly abundant in Lokiarchaeum. Although phylogenetic methods Surkont and Pereira-Leal . doi:10.1093/molbev/msw061 alone were insufficient to determine the position of Rab-like proteins within the RAS superfamily, our results indicate that these GTPases may be Rab precursors. Surprisingly, we also found evidence for a GDI/REP-like protein existing in Archaea, raising the possibility that this interaction predates Eukaryogenesis.
Small GTPases are well known to exist in prokaryotes, where they mediate diverse functions, for example, MglA regulates cell polarity and motility by accumulating at a cell pole in its active GTP-bound state (Zhang et al. 2010). The closest group to eukaryotic Rab/Rho/Ras/Ran are the Rup proteins (Ras superfamily GTPase of unknown function in prokaryotes; Wuichet and Søgaard-Andersen 2015). Phylogenetic analysis is not able to resolve the relationship between eukaryotic small GTPases and prokaryotic ones, so no claim can be made whether these sequences are Rup-like We concentrated on characterizing sequence and structural features that could shed light on the relationship between these sequences and eukaryotic Rabs. At the family level, they are more similar to the Rab family than to other eukaryotic small GTPases (Arf/Ras/Rho/Ran). We found extensive RabF motifs conservation, motifs that in Eukaryotes are diagnostic of this family, and that mediate important protein interactions characteristic of Rabs. On the structural models of archaeal Rab-like proteins, these motifs map to the same positions as their eukaryotic counterparts, suggesting that they could mediate similar interactions, which lends further support to their Rab-like classification. Our results thus point to Archaea having Rab-like sequences, which although not being full-fledged Rabs, as we will discuss below, are already differentiated intermediates to this small GTPase family.
The presence of Rab motifs that are known to mediate interactions with other Eukaryote-specific Rab regulators was puzzling and led us to test the hypothesis that one or more of these interactions could have predated eukaryogenesis. Using sensitive methods we found convincing REP/GDI-like proteins in multiple Archaea that are involved in the biosynthesis of membrane lipids (geranylgeranyl reductase, EC 1.3.1.101). An archaeal form of this enzyme had its crystal structure solved and aligns well with the crystal structure of GDI:Rab complex. It is thus very probable that the conservation of the RabF motifs in archaeal Rab-like sequences points to an established interaction with this enzyme. The functional meaning of this interaction is unclear, but the fact that this enzyme is involved in the synthesis of the isoprenoids that are used in the lipid modification of eukaryotic small GTPases is highly suggestive. Inspection of the structure of the archaeal enzyme suggests that although it has a binding pocket able to shield the lipid groups from the cytosol as REP and GDI do, it is in a different orientation, suggesting that it cannot chaperone lipid-modified eukaryotic Rabs that have longer C-termini than the archaeal Rab-like sequences.
In Eukaryotes REP/GDI are chaperones of the lipidmodified Rabs, that deliver them to the membranes, where REP is doing so in the context of the lipid modification reaction, as an accessory protein to the RabGGTase complex, and where GDI recycles Rabs in and out of membranes. The presence of a REP/GDI homologue in Archaea raises the hypothesis that membrane association of small GTPases via prenylation may have preceded the emergence of Eukaryotes. There is, at least, one report claiming isoprenylation of proteins in Archaea (Konrad and Eichler 2002). However, the absence of an extended C-terminal region beyond the GTPase globular domain together with the absence of the prenylateable C-terminal cysteine residues points against this. Furthermore, we found no evidence of a polybasic region that is known to mediate membrane association (Williams 2003), nor of any other membrane association signal. Our results thus suggest that these Rab-like sequences are unlikely to associate with membranes via lipidation. It is, however, interesting to note that archaeal homologues of both the alpha and beta subunits of eukaryotic prenyltransferases are common, although there is no evidence that they are able to form a heterodimer with the prenyltransferase activity. The beta subunit homologues are involved in the isoprenoid metabolism and their structure is predicted to be similar to eukaryotic prenyltransferases, which further supports the notion that some components of the prenylation complex are present in Archaea.
Small GTPases are molecular switches that can cycle between two membrane-associated states, as well as cycle in an out of membrane. Our results suggest that these Archaea represent a snapshot of the evolution of this circuit, that resolves part of the evolutionary path into membrane-associated protein trafficking regulators. The Rab protein family is already individualized, even though we lack any known internal membranes in the TACK Archaea. These proteins are apparently active GTPases able to cycle between two structural states, but it is unclear if they do it in the cytosol or if an "in" and "out" of membrane switch was already established. In this scenario, an interaction with the protein that will become the chaperone that catalyses this second part of the Rab cycle is already established, but in the absence of lipid modification. It is plausible that localization to membranes may exist via protein-protein interactions. Finally, the building blocks for a protein prenylation machinery are also found in multiple Archaea, suggesting that even the emergence of this component of the Rab cycle may also predate eukaryogenesis.
Our conclusions are possible because we were able to go beyond phylogenetic methods, which are clearly insufficiently sensitive to resolve events at this order of temporal divergence, using instead our motif/domain-based tool to identify Rabs, the Rabifier. It is important now to look into other small GTPase families, as our preliminary data suggest that other members of the Ras/Rho/Ran/Rab clade may have already been individualized in Archaea. It is also important to investigate whether the interaction we predict here between Rablike and REP/GDI-like sequences does in fact exist, and what is the subcellular localization of these small GTPases. Lokiarchaeota, are unlikely target organisms for these experiments, as they exist in a difficult to reach environment. However, organisms that are routinely cultured in the laboratory have these sequences (see fig. 5), which makes these experiments tractable. Furthermore, we found that other environmental (marine) samples (Kawai et al. 2014) also possess Lokiarchaeota-like small GTPases and specifically abundant Rab-like sequences (117 proteins in the analyzed sample), which makes the possibility of isolation and culture of these organisms more plausible. Our study gives further support to the notion that Eukarya emerged from within Archaea, and may be construed to support the notion that it was from within organisms close to the recently identified Lokiarchaeum. We are convinced that in the near future we will be able to resolve the origin of the in-out of membrane cycle of small GTPases, and their association with specific eukaryotic processes. It is possible that this cycle emerged in Archaea, even before the specific system they regulate in Eukaryotes has emerged, and that have later been co-opted.
Sequences
All complete archaeal proteomes (231) were downloaded from the UniProt database (The UniProt Consortium 2015), all Lokiarchaeum proteins (5,384) were downloaded from GenBank (Benson et al. 2014). The complete list of species is shown in the supplementary table S1, Supplementary Material online. Eukaryotic and bacterial genomes were downloaded from Ensembl (Cunningham et al. 2015).
Protein Sequence Alignments
Multiple sequence alignments were built with MAFFT 7.221 (Katoh and Standley 2013) using a high accuracy mode (-maxiterate 1,000 -localpair). TrimAl v1.2 (Capella-Gutiérrez et al. 2009) was used to remove gap-rich regions from alignments. Pairwise sequence alignments were constructed with water (the Smith-Waterman local alignment algorithm) and needle (the Needleman-Wunsch global alignment algorithm) from the EMBOSS package (Rice et al. 2000). Jalview 2.8.2 (Waterhouse et al. 2009) was used for alignment visualization.
Phylogeny Reconstruction
Phylogeny reconstruction using the Bayesian inference was conducted with MrBayes 3.2.5 (Ronquist et al. 2012) using the mixed amino acid model with gamma-distributed rate variation across sites. Two parallel runs with four chains each (Metropolis coupling) were run until the topologies converged (standard deviation of split frequencies is below 0.05), first 25% generations were discarded as the burn-in. RAxML 8.1.22 (Stamatakis 2014) was used for tree reconstruction using the maximum likelihood method, a discrete approximation to the gamma distribution with four categories was used to model across-site rate heterogeneity, the bestfitting substitution model (LG, Le and Gascuel 2008) was selected using ProtTest 3.4 (Darriba et al. 2011). ETE2 (Huerta-Cepas et al. 2010) and Dendroscope3 (Huson and Scornavacca 2012) were used for tree visualization.
Sequence Analysis pHMMs of protein families were build from sequence alignments using hmmbuild from the HMMER 3.1b2 software package (http://hmmer.org, last accessed April 6, 2016), plurality-rule consensus sequences were generated with hmmemit. Sequence logos were generated with WebLogo 3.4 (Crooks et al. 2004) from multiple sequence alignments.
Amino acid variation was calculated for each position in an alignment of paralogous proteins as the entropy of that position, HðXÞ ¼ À X n i¼1 pðx i Þlog 2 pðx i Þ, where pðx i Þ is the fraction of the residue x i at the X column in the alignment.
Protein Structure Prediction MODELLER v9.15 (Sali and Blundell 1993), a program which implements a homology-based method for structure modeling, was used to predict protein structures given templates with known structure that share a high level of sequence identity to the modeled protein. Model quality and stability were evaluated with the DOPE potential (Shen and Sali 2006), ProSA (Sippl 1993;Wiederstein and Sippl 2007), and Verify3D (Lüthy et al. 1992). PyMOL (The PyMOL Molecular Graphics System, Version 1.7.4 Schrödinger, LLC.) was used for structure visualization. | 2018-04-03T04:24:39.746Z | 2016-03-31T00:00:00.000 | {
"year": 2016,
"sha1": "826bd55583d5e1cc82b84cca25cd5fed2af07a59",
"oa_license": "CCBY",
"oa_url": "https://academic.oup.com/mbe/article-pdf/33/7/1833/17473593/msw061.pdf",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "a21e0ac46b1a68cf83324caec82443758132614a",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
118875438 | pes2o/s2orc | v3-fos-license | Symmetry transformations for magnetohydrodynamics and Chew-Goldberger-Low equilibria revisited
Being motivated by the paper [O. I. Bogoyavlenskij, Phys. Rev. 66, 056410 (2002)] we generalise the symmetry transformations for MHD equilibria with isotropic pressure and incompressible flow parallel to the magnetic field introduced therein in the case of respective CGL equilibria with anisotropic pressure. We find that the geometrical symmetry of the field-aligned equilibria can break by those transformations only when the magnetic field is purely poloidal. In this situation we derive three-dimensional CGL equilibria from given axisymmetric ones. Also, we examine the generic symmetry transformations for MHD and CGL equilibria with incompressible flow of arbitrary direction, introduced in a number of papers, and find that they cannot break the geometrical symmetries of the original equilibria, unless the velocity and magnetic field are collinear and purely poloidal.
I. INTRODUCTION
Two of the most important models widely applied to describe plasma equilibria are the isotropic ideal magnetohydrodynamics (MHD) and the anisotropic Chew-Goldberger-Low (CGL) [1] model. In [2][3][4][5][6][7][8][9] methods for constructing new continuous families of equilibria in the framework of the above mentioned models, once a given equilibrium is known, are introduced. More specifically, in [2][3][4][5] three sets of equilibrium transformations in the framework of MHD model were presented. The first set is applied to given equilibria with incompressible flow of arbitrary direction, while the second one to both static equilibria and stationary equilibria with field-aligned incompressible flow. The third set of transformations concerns plasma equilibria with compressible flow. In addition, in [6] symmetry transformations that produce an infinite class of anisotropic CGL equilibria, on the basis of prescribed CGL ones are introduced; also in [6][7][8][9] are presented symmetric transformations mapping static or stationary MHD equilibria into CGL ones. All these symmetry transformations depend on a number of scalar functions which have to be constant on the magnetic field lines. This implies that the new equilibria resulting from the transformations depend on the structure of the magnetic fields of the original ones, and thus, the topology of the original equilibria is essential for these transformations.
A magnetic field line before closing to itself may either cover a surface, if such a surface exists, or fill a volume. Any surface that is traced out by a number of magnetic field lines is called magnetic surface. In plasmas of fusion devices, however, the name is usually reserved for nested toroidal surfaces. The generic structure of the magnetic field can be either "open", in the sense that it closes to itself through infinity, as for example in magnetic mirror, screw pinch and earth's magnetosphere, or closed if it remains in a spatially finite region, as for example in the central region of tokamak and stellarator. It can be proved that if the magnetic field lines lie on some closed surfaces contained in a bounded region and do not have any singularities, then they must be toroids (topological tori) [10][11][12]. Hamiltonian theory guarantees the existence of magnetic surfaces in systems with three kinds of continuous geometrical symmetry: axisymmetry, as in an ideal tokamak, helical symmetry which can approximately describe a "straight stellarator" without toroidal curvature, and translational symmetry in which the system is unbounded along the symmetry direction. However, the latter category can represent a "straight tokamak" when the magnetic field is periodic along the direction of symmetry and therefore can be considered as toroidal field, since a single period of such a field is topologically equivalent to a torus. For such symmetric systems the magnetic surfaces are well-defined by the level sets of a function ψ(q 1 , q 2 ), with the third coordinate q 3 being ignorable. In non-symmetric devices, on the other hand, magnetic surfaces do not exist rigorously everywhere because the magnetic field may cover regions of finite volume. Open-ended systems, such as magnetic mirrors, do not possess magnetic surfaces that are traced out by one single line. This leads to a considerable degree of arbitrariness.
The lines of force lying on nested toroidal magnetic surfaces encircle the magnetic axis. This encirclement is characterized by the rotational transform which is defined as the ratio of the number of poloidal transits (the short way around the toroid) to the number of toroidal transits (the long way around it) of a field line. If the rotational transform is a rational number then the magnetic field lines close upon themselves on surfaces that are called rational surfaces, leaving finite parts of them with vanishing magnetic field. If not, the surfaces are ergodic (or irrational) and the field lines cover them densely everywhere.
In systems without geometrical symmetry there might exist stochastic regions in which the magnetic field lines do not lie on any surfaces but are chaotic, e.g. near a separatrix.
Such regions are undesirable for MHD equilibrium and stability. If the symmetry of the field is violated, for example by superimposing a perturbation having a different symmetry, then an analogous function to ψ may not be found, and magnetic surfaces will no longer be uniquely defined or defined at all [13]. In [14] it was proved that all smooth steady MHD equilibria with field-aligned incompressible flows possess (open) magnetic surfaces, with possible exception the force-free or Beltrami equilibria. Also, in [15] it was proved the existence of (open) magnetic surfaces of three-dimensional equilibria with field-aligned flows.
In [2][3][4][5][6][7][8][9] it is claimed that the symmetry transformations presented therein can break the geometrical symmetries of the original equilibria either static or with parallel incompressible flows. In this context, the respective symmetry transformations were applied to both the magnetic analog of Hill's vortex static axisymmetric equilibria with purely poloidal magnetic field [16,17], and to static helically symmetric equilibria with magnetic field along the symmetry direction [5,18] in order to model nonsymmetric astrophysical jets.
In the present work we make an extensive revision of the transformations presented previously in Refs. [2][3][4][5][6][7][8][9] concerning equilibria with incompressible flows. In Section II we introduce a symmetry transformation that can be applied to any known anisotropic CGL equilibria with field-aligned incompressible flows (or static equilibria) and anisotropy function constant on magnetic field lines, and produce an infinite family of anisotropic equilibria with collinear velocity and magnetic fields, but density and anisotropy functions that may remain arbitrary. These transformations consist a generalisation of the ones introduced in [2] for field-aligned MHD equilibria. We also prove that all transformations presented in [2][3][4][5][6][7][8][9] can break the geometrical symmetries of a known given equilibrium, static or with fieldaligned flow, if and only if its magnetic field is purely poloidal. In Section III we construct three-dimensional (3D) equilibria by applying the introduced transformations to known axisymmetric equilibria with field-aligned incompressible flow, pressure anisotropy, and purely poloidal magnetic field, related with the symmetry breaking. In Section IV we examine the In Section IV of Ref. [2] transformations between MHD equilibria with parallel flows are presented. Specifically, it is stated therein that if { B, v, p, ̺} is a solution of the ideal MHD equilirium system of equations with field-aligned incompressible flow: then { B 1 , v 1 , p 1 , ̺ 1 } defined by the following symmetry transformations, that depend on the arbitrary functions a( r), b( r), c( r), consist a new solution to the MHD equilibrium set of equations with field-aligned flows: The above special transformations are defined only when the velocity and magnetic field of the original equilibria are related through v = (λ/ √ µ 0 ̺) B, and are also valid in the static limit, v = 0. Their reductive form for constant a, b, c, and λ was first derived in [4] from given axisymmetric equilibria found in [19]. According to [2] the functions a( r), b( r), c( r), and pressure anisotropy function, constant on magnetic field lines. Then { B 1 , v 1 , ̺ 1 , p ⊥ 1 , p 1 } given by the following transformations: where a( r) = 0, b( r), c( r), and n( r) = 0, are arbitrary functions, define a solution to the CGL set of equilibrium equations with field-aligned flows, if and only if the functions are constant on the magnetic field lines of the original equilibria.
Proof. The original equilibria { B, v, ̺, p ⊥ , p } satisfy the CGL equilibrium equations with field-aligned flows (3): where the CGL pressure tensor is defined as with the function σ d measuring the pressure anisotropy. It is assumed that the flow is incompressible, ∇ · v = 0, which by the continuity equation implies that the mass density is constant on streamlines, v · ∇̺( r) = 0; it is also assumed that the anisotropy function is constant on the magnetic field lines, B · ∇σ d ( r) = 0. When the equilibria possess some geometrical symmetry, the latter hypothesis for the function σ d in conjunction with incompressibility, lead to the derivation of a single Grad-Shafranov (GS) equation that governs them [20][21][22]; also, according to [23] this assumption on σ d may be the only suitable for satisfying the boundary conditions on a fixed, perfectly conducting wall. It may be noted that for the given field-aligned equilibria, the vectors v and B are collinear (parallel) and therefore the magnetic field lines are the same as the velocity streamlines. It follows that the function λ( r) must be constant on magnetic field lines, B · ∇λ( r) = 0. Also, the force balance equation of the set (7) can be cast into the useful form In order for the new solution (5) to be valid it must satisfy the following set of CGL equilibrium equations: where Note that systems (7) and (10) are reductions of the generic CGL equilibrium equations since for field-aligned flows it holds v × B = v 1 × B 1 = 0, and therefore the electric field vanishes by Ohm's law.
Substituting (5) into (10) yields With the use of Eq. (9) and assuming that σ d = 1 (in which case v 1 = 0, C = 0 and the transformations (5) are not invertible), Eq. (14) takes the form Now with the aid of (6), Eqs. (12), (13) and (15) assume the forms Since the first term on the lhs of (18) vanishes, it is apparent that if Eqs. (16) and (17) are valid, then (18) is trivially satisfied. Thus, we conclude that in order for transformations (5) to be valid, Eqs. (16) and (17) are "open", i.e. they approach infinity in one direction, defined by a variable q 3 , then the magnetic field should be finite as q 3 → ∞. The function λ( r) should depend on two transversal variables when the third one goes to infinity, λ(q 1 , q 2 , q 3 → ∞) = λ(q 1 , q 2 ), and thus, must depend on these two variables in the whole plasma domain for magnetic surfaces λ(q 1 , q 2 ) = const. to exist. In both of the above kinds of magnetic field lines (bounded and "open"), the functions g( r), f ( r) have to be functions of two transversal variables, i.e. q 1 , q 2 .
One could suggest that this does not restrict the functions a( r), b( r), c( r), n( r) to have the such that f ( r) = A(q 1 , q 2 )K(q 1 , q 2 )). However, equation ∇ · (̺ 1 v 1 ) = 0 yields B · ∇a( r) = 0, which means that a = a(q 1 , q 2 ) and consequently c = c(q 1 , q 2 ). Then from the definition of the constant C it follows that b = b(q 1 , q 2 ), and as a result n = n(q 1 , q 2 ). Thus, if the magnetic field lines are finite closed loops or go to infinity in some direction, all functions of transformations must, in general, depend on two variables transversal to this direction.
If the magnetic field lines cover densely everywhere (ergodically) closed magnetic surfaces, λ( r) = const. (which are toroids), then the functions g( r), f ( r) must be constant on them, and so must be all four functions of the transformations. In this situation, if the field possesses some geometrical symmetry, with ignorable variable q 3 , the surfaces λ( r) = const.
are nested, with λ = λ(q 1 , q 2 ). Then all functions a( r), b( r), c( r), n( r) have the same symmetry (i.e. are functions only of q 1 , q 2 ). However, there exists an exception; the one when the original equilibrium has some known geometrical symmetry with purely poloidal magnetic field to be examined as follows.
Axial symmetry: Consider the case that the original equilibria are axially symmetric with field-aligned incompressible flows and anisotropy function constant on magnetic surfaces [21].
Employing cylindrical coordinates (ρ, z, φ) we have where the function I relates to the toroidal magnetic field and ψ(ρ, z) = const. labels the Set (20) is satisfied either if functions g, f are constant on the magnetic surfaces, or I = 0.
The latter case implies that transformations (5) can break the axial symmetry of field-aligned equilibria with purely poloidal magnetic field. The same statement holds for translationally symmetric equilibria with field-aligned flows [24], while the more generic case of helical symmetry will be studied separately below.
Helical symmetry: Consider now that the original equilibria are helically symmetric with field-aligned incompressible flows and anisotropy function constant on magnetic surfaces for which the following relations hold [22]: Here (r, u, ξ) are helical coordinates defined through the usual cylindrical ones (ρ, φ, z) as r = ρ, u = mφ − kz, ξ = z; I relates to the helicoidal magnetic field and ψ(r, u) = const. labels the magnetic surfaces; the vector h = (m/(k 2 r 2 + m 2 )) g ξ points along the helical direction, where (k, m) are integers, and the covariant helical basis vectors, g i , i = r, u, ξ, are defined through the respective cylindrical unit vectors as g r =ρ, g u = (r/m)φ, g ξ = (rk/m)φ +ẑ. For the adopted non-orthogonal helical coordinates, the u-and ξ-covariant and contravariant components of a given vector A differ from each other, A i = A i (i = u, ξ).
The magnetic field written in contravariant components is The definition of a usual flux function ψ(r, u) so as equation ∇ · B = 0 to be satisfied reduces (22) into [see also [25]]: where g ξ =ẑ is the contravariant basis vector. Observe that This in fact dictated us to define the helical vector h that points into the symmetry direction.
Then the magnetic field is written in the form (21) with If we define the poloidal magnetic field as then the field on the plane normal to h is expressed as Now let f = f (ψ, ξ), g = g(ψ, ξ). Then satisfaction of Eqs. (16)- (17) require Respectively to (20), Eq. (28) leads to Thus, transformations (5) can also break the helical symmetry of the original equilibrium with field-aligned incompressible flow and pressure anisotropy, if and only if the magnetic field is purely poloidal.
Finally, it may happen that λ = constant and consequently, ∇λ = 0 in the whole plasma domain. In this situation the force balance equation (9) is written in the form wherep := (p ⊥ + p )/2 is defined as an effective isotropic pressure. In this case a family of magnetic surfaces w( r) = const., where w ≡p+λ 2 B 2 2µ 0 can be defined, in which both magnetic field lines and velocity streamlines lie on, B · ∇w( r) = 0. Analogous considerations can be made on the structure of these surfaces. Now it may happen w = const. with ∇w = 0 if J = y( r) B, that is the current density is parallel to the magnetic field, or equivalently ∇ × v = t( r) v, that is the velocity is parallel to the vorticity. This is the case of force free or Beltrami equilibria. Then magnetic surfaces y( r) = const. can be yet defined, B · ∇y( r) = 0. But in the particular case y ≡ const.
(everywhere) and therefore ∇y = 0 (and then as well t ≡ const., with ∇t = 0), we finally escape from the topological constraint that magnetic field lines lie on surfaces. The lines of force may be chaotic (space-filling) in this case, and all functions a( r), b( r), c( r), n( r) have to be constant.
The above conclusions lead us to formulate the following corollary: All conclusions derived herein concerning the validity of the transformations, the structure of the arbitrary functions and the symmetry breaking, also hold for the respective transformations (2) with isotropic pressure (cf Remark 1). In this respect, the symmetry breaking of the static helically symmetric equilibria related with astrophysical jets, examined in Section VIII of Ref. [2], should be revised, since the magnetic field of the original equilibria [cf Eq. (8.1) therein] is not purely poloidal unless the constant α is equal to zero.
III. CONSTRUCTION OF 3D CGL EQUILIBRIA WITH FIELD-ALIGNED FLOWS
Consider axisymmetric equilibria [21] with field-aligned incompressible flows, pressure anisotropy and purely poloidal magnetic field. In this case the equilibrium quantities are expressed as and the steady sates obey the following generalized GS equation Here the elliptic operator is defined as ∆ * := ρ 2 ∇ · ∇/ρ 2 ;p s is the effective pressure in the absence of flow, and the functions ̺, σ d , M p are constant on the magnetic surfaces ψ = const. Applying the symmetry transformations (5), with λ = M p (ψ), we obtain the following 3D equilibria: where the functions a, b, c, and n may depend, in addition to ψ, on the toroidal angle φ.
However, if either of the functions n( r) or a( r) remain constant on magnetic surfaces, the breaking of the geometrical symmetry of the original equilibria remains unaffected. Note that the transformed current density J 1 has a component perpendicular to the magnetic surfaces which is undesirable for confinement but this component vanishes when the function g = b/n is φ-independent. This choice, however, yields special equilibria with purely poloidal magnetic field, B 1 = κ(ψ) B and permits only 3D variations of velocity and pressure.
To construct a specific equilibrium let us make the following choice for the arbitrary functions: Then from (33) we obtain the exact equilibria with purely poloidal magnetic field, incompressible flows, and anisotropy function varying on the magnetic surfaces: We note that the above equilibria do not obey a GS-like equation analogous to (32). In order for equilibria (35) to be physically plausible we requirep 1 > 0, which yields the following where β s :=p s /(B 2 /2µ 0 ) is the poloidal beta in the absence of flow. It is also recalled that If the plasma is confined for example in an axisymmetric device then the physical quantities should be periodic in the angle φ. This yields for δ: Though it is well known that toroidal plasma confinement is not possible with a purely poloidal magnetic field, it is interesting that in that case transformations (5) can break the geometrical symmetry and yield 3D equilibria. These equilibria may be of astrophysical interest.
A. Review of transformations between MHD-MHD and CGL-CGL equilibria
In [2,3,6] symmetry transformations that produce an infinite family of MHD (CGL) equilibria with arbitrary incompressible flow once a respective MHD (CGL) equilibrium with incompressible flow is given, were introduced as follows.
MHD into MHD:
In the case of isotropic pressure, suppose that { B, v, p, ̺} is a known solution of the MHD equilibrium system with flow of arbitrary direction where Φ is the electrostatic potential. The flow is assumed to be incompressible, ̺ = ̺(ψ), and the function ψ labels the common magnetic and velocity surfaces, if such surfaces exist.
Note that these two sets of surfaces should coincide for flows of arbitrary direction because of the Faraday's and Ohm's laws. Then according to [2,3], { B 1 , v 1 , p 1 , ̺ 1 } defined by the following symmetry transformations (depending on the arbitrary functions a( r), b( r), c( r)) consist a new family of solutions to the MHD equilibrium system.
CGL into CGL:
For anisotropic pressure let { B, v, ρ, p ⊥ , p } be a given solution of the CGL equilibrium system of equations with arbitrary incompressible flow implying ̺ = ̺(ψ), and anisotropy function constant on magnetic surfaces, σ d = σ d (ψ). Then, according to [6], { B 1 , v 1 , ̺ 1 , p ⊥ 1 , p 1 } defined by the following symmetry transformations is also a solution. Note that transformations (41) depend on the arbitrary functions a( r), b( r), c( r), n( r), and reduce to the respective ones for isotropic pressure given by the set (39) when σ d = 0 and n( r) = 1.
As stated in [6] the functions a( r), b( r), c( r), n( r) have to be constant on the magnetic surfaces. Below we examine the validity of these transformations and whether they can break the geometrical symmetry of the original equilibria.
Validation of equilibrium equations for the transformed fields
In order for the new solution (41) to be valid it must satisfy the following set of CGL equilibrium equations where P 1 and σ d 1 are given by (11).
Expressing in (42) the transformed fields in terms of the original ones by means of (41) leads to the following system of equations: It is apparent that if all four functions appearing in the symmetry transformations are constant on the magnetic surfaces, Eqs. (43)-(48) are trivially satisfied; otherwise the above system of six equations for the four functions a( r), b( r), c( r), n( r) is in general overdetermined. However, if the functions a( r), b( r), c( r), n( r) are chosen so that being satisfied when then (45) Since b = ±c for the transformation to be invertible, Eq. (51) is satisfied only for parallel Note that the field-aligned equilibria (52) and (3) For the special equilibria with field-aligned flows satisfying (52) and a( r), b( r), c( r), n( r) generally not constant on magnetic surfaces, transformations (41) reduce to For equilibria with isotropic pressure satisfying (39), being recovered from (41) for σ d = 0 and n = a = b + c = 1, the choice (50) leads to With the aid of (53) and (54) we observe that in the presence of pressure anisotropy the transformed velocity and mass density differ from the respective, original ones. Now suppose that the original equilibrium is helically symmetric [22]. Then the following relations hold where the function Θ relates to the helical velocity field and Φ = Φ(ψ). Equation (50) implies I = ( √ 1 − σ d / √ µ 0 ̺)(Θ/̺) and dΦ/dψ = 0. These relations lead to the following which implies either I = 0 or M 2 p + σ d = 1. It turns out again that symmetry breaking is possible only for purely poloidal parallel flows (I = 0). The relation M 2 p + σ d = 1 is connected to the Alfvén singularity. The same conclusion holds for axially and translationally symmetric original equilibria with or without pressure anisotropy.
Arbitrary functions constant on magnetic surfaces
In the above Subsection we found that the symmetry transformations (41) (and the and thus, the magnetic surfaces through the transformation are preserved: where Equality of the poloidal velocity components in (63) and (66) yields As already mentioned in Subsection II B every equilibrium static or stationary with incompressible flows, which has some geometrical symmetry with pressure anisotropy function constant on magnetic surfaces, is governed by a GS equation for the flux function ψ [20][21][22][24][25][26]. Such an equation contains a quadratic term as | ∇ψ| 2 . For this reason an integral transformation is applied as: Transformation (69) Therefore the transformed equilibria differ from the starting ones only by a constant factor C 1/2 , in agreement with the conclusions drawn in the previous Sections; the geometrical symmetry of the original equilibria can break only for purely poloidal magnetic field, otherwise the transformed equilibria retains the original symmetry.
B. Transformations between MHD-CGL equilibria
In Refs. [6][7][8] transformations that produce CGL anisotropic equilibria from given isotropic MHD ones, are introduced as follows: If { B, v, p, ̺} is a known solution of the MHD equilibrium system (38), then the following symmetry transformations where C 0 and C 1 are arbitrary constants, produce an infinite family of CGL equilibria satisfying (42). Transformations (71) are also valid in the static limit, v = 0. Let us examine their validity.
Substituting (71) into (42) we obtain Thus, in order for transformations (71) to be valid, the functions f 1 ( r) and g 1 ( r) must be constant on the magnetic field lines and velocity streamlines of the original equilibria, and respective considerations on their structure can be made as those in Section II. Therefore it turns out again that the only way that the geometrical symmetry of the original isotropic equilibria can break is if and only if the magnetic and velocity fields are collinear and purely poloidal. In this context, conclusions for the breaking of the helical symmetry of astrophysical jets with magnetic field lines going to infinity in connection with the coordinate z, examined in [6,8] in the static limit, should be reconsidered.
V. CONCLUSIONS
In the present work we made an extensive revision of the symmetry transformations previously introduced in a series of papers, [2][3][4][5][6][7][8][9], which once applied to known MHD and/or CGL equilibria produce an infinite new continuous families of respective equilibria. These transformations contain some arbitrary scalar functions, the structure of which depend on the topology of the given equilibria. We examined both transformations that map MHD into MHD, CGL into CGL, and MHD into CGL equilibria, either with field-aligned or arbitrary incompressible flows, particularly as concerns their validitity and applicability. In addition, we examined whether these transformations can break the geometrical symmetry of the original equilibria.
In Section II we presented a new set of symmetry transformations that can be applied to any known CGL equilibria with special field-aligned incompressible flow satisfying (3) and pressure anisotropy function, σ d , constant on the magnetic field lines, to produce an infinite class of equilibria with collinear v 1 and B 1 fields, and with ̺ and σ d functions that in general may be arbitrary. These transformations consist a generalisation of the ones introduced in [2] for the same kind of field-aligned incompressible flow and isotropic pressure, and can also be applied to static anisotropic equilibria.
In addition, we examined the structure of the arbitrary scalar functions included in the | 2018-10-16T08:18:30.000Z | 2018-10-16T00:00:00.000 | {
"year": 2019,
"sha1": "1c49fd26c15ccb9a38d267a182c97d7140ddae38",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1810.06868",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "1c49fd26c15ccb9a38d267a182c97d7140ddae38",
"s2fieldsofstudy": [
"Geology"
],
"extfieldsofstudy": [
"Chemistry",
"Physics"
]
} |
255515029 | pes2o/s2orc | v3-fos-license | Active Ageing Index in Russia - Identifying Determinants for Inequality
This paper is aimed at the development of a tool analysing the AAI results for the Russian older citizens from different population groups, as well as at identifying factors underlying the inequalities in active ageing outcomes by calculation the AAI on the national and individual levels. The adaptation of the methodology of the AAI to the individual-level data and the limitations of the approach are explicitly explained. The older generations of Russia show relatively high levels of education, financial security and engagement in family care, especially in the care to children. The most significant potential for development have employment, volunteering, political engagement, physical activity, lifelong learning and use of the Internet. The calculation of the AAI at the individual level has revealed significant inequalities in the degree of realisation of potential in different areas of active ageing. The results of the project provide scientific evidence for the implementation of policy measures in the target groups. The high correlation of the index values with human capital indicators (health and education) underlines the importance of the early interventions aimed at promoting and supporting human capital at the earlier stages of the life course till the old age. The substantial positive connection of employment with other forms of activity stresses the necessity of developing a package of activation policy measures aimed at the retention of older adults in the labour market. At the same time, the statistical analysis showed the absence of a “dilemma of choice” between certain types of activity of the older generation, for example, between caring for grandchildren and employment, or employment and volunteering - the potential in different areas may be increased simultaneously.
Introduction
The realisation of the potential of the older population is currently most often considered within the framework of the concept of active ageing , which is a dominant strategy in the development and implementation of social policies for the older generation. In accordance with it, many international (United Nations 2002; World Health Organization (WHO) 2002) and national documents have been adopted that legally establish the principles and policies of active ageing. Russia is no exception: in February 2016, the Action Strategy in the interests of older citizens until 2025 and the Action Plan on the implementation of the first stage of the Strategy for 2016-2020 were adopted (though, currently there is no official definition of the concept of "active ageing" or "active longevity"). In 2019, a draft Concept of Active Ageing Policy in the Russian Federation was developed. After the broad discussion withing ministries at the federal and regional level, its final version is going to be adopted by the government in 2020.
Indices of active ageing and well-being serve as a tool for assessing progress in realising the potential of the older population. One of the most actively used indices is the Active Ageing Index (from now on referred to as the AAI) -a multidimensional index developed in the framework of a joint project of the European Commission on Employment, Social Affairs and Integration and the Population Division of the United Nations Economic Commission for Europe (UNECE). The AAI estimates both the achieved level of the current realisation of the potential of older people to monitor the "overall progress with respect to active ageing" (World Health Organization (WHO) 2002) and the extent to which the existing environment provides the opportunity to maximise their potential (UNECE / European Commission 2015). The AAI helps to identify strengths and weaknesses of active ageing in the areas of employment, social interaction, political participation, physical activity, education, etc. Thus, the AAI provides information to political institutions on areas requiring additional measures, as well as on the effectiveness of the existing policies (UNECE / European Commission 2016; UNECE / European Commission 2017).
Since 2014, the Institute for Social Policy of the HSE has been calculating the AAI for Russia (Varlamova et al. 2017). In 2018, data were available for the AAI 2010AAI , 2012AAI , 2014AAI , 2016AAI and 2017. Firstly, the AAI was estimated based on the multiple databases, including Federal state statistic surveys and some international surveys conducted in Russia (ESS, GGS). 1 It allowed researchers to combine the most appropriate wording from different surveys, to compare the values of the indicators obtained from different databases and to find the indicators comparable on the international level (Zaidi and Stanton 2015). However, this approach often leads to problems with temporal comparability since waves of different surveys do not coincide by years. Although the analysis shows good comparability of the results obtained for Russia with the EU data, full-time comparability has not yet been achieved. The estimation of the robustness of the indicators to the data source and question-wording reveals that both can be a vital source of the variation of the results of the individual indicators but have no significant influence on the total value of the AAI (Varlamova et al. 2017).
In 2018, a fundamentally new approach was applied to calculating the AAI for Russia -the index was calculated based on a single database (RLMS-HSE), 2 which allows analysing the individual values of the final index by multidimensional statistical methods. To achieve maximum comparability with the original AAI methodology, we have included the AAI original wording of the questions to the RLMS-HSE surveys in 2016 and 2017. Into the 2016 wave, the questions on voluntary activity and care to children and older adults were integrated. In 2017, two more questions (political activity and material deprivation) were added into the questionnaire. In 2017 the analytical sample covered 3817 respondents aged 55 years old and over.
Compared with the index, based on the multiple datasets, the AAI at the individual level has several limitationsit is impossible to compare it with the EU results (calculations based on the SHARE are using explicit weights and question-wording that is different from the original (Barslund et al. 2017)). Currently, the AAI for Russia on the individual data cannot be calculated in dynamics (for several periods) as the necessary questions were introduced into RLMS-HSE only recently. Nonetheless, the individual-level AAI allows more accurate identification of the target groups that require social support in each of the domains of active ageing and provide a better understanding of inequalities in active ageing experiences, which is crucial for designing effective social policies. Understanding of individual differences in employment, social activity, material well-being, etc. in general and in separate domains allows better adaptation of social policy to the needs of older people and makes it more targeted. The index based on a single dataset also examines more deeply the features of potential realisation in the context of socio-demographic and socioeconomic characteristics. This paper is aimed at estimating the AAI individual results for the Russian older citizens from different population groups, examining the degree of the existing inequality in active ageing potential as well as at enabling factors behind it. Our research questions are the following: 1) To what extent the individual values of the AAI vary with the sociodemographic characteristics of older adults (age, sex, place of residence, subjective health status, marital status and education)? 2) What are the characteristics of different groups of older people with high and low values of the AAI? 3) What is the relationship between participating in employment and other indicators of the active ageing potential?
Methodical Approach
The calculation of the individual index implies the need for specific changes in the original methodology (UNECE / European Commission 2018; Zaidi et al. 2013). First, some AAI indicators have different age limits. In essence, the employment rates are calculated for four 5-years groups, the independent living indicator includes people aged 75 years old and over, financial indicators are dedicated to the 65+ age group, and the use of the Internet and the educational attainment indicators have an upper boundary of 74 years old. When calculating the individual values on the same methodology as the national level index, it would be necessary to recalculate the weights of the indicators, depending on the list of variables applied to the selected age group. In order to avoid calculating individual indices based on different weighting mechanisms, which would limit comparability between different age groups, and to maintain the concept of the value of active ageing principles regardless of the age of the individuals, we have decided to calculate all the indicators for all individuals over 55 years. Unlike previous research on the individual-level data (Barslund et al. 2017), employment level is not calculated by the age group, as a person can not belong to several age groups, which means, that on domain level his participation in the labour market will be accounted as 1 in his age group and as 0 in the three other age groups. Thus, the final contribution of employment to the total value of the index is four times reduced. In case of implicit weights (see below), the employment in the younger age groups will have more impact on the total active ageing value, as the implicit weights decrease with age, which does not necessarily reveal the individual value of employment.
The second change concerns weighing mechanisms. The AAI methodology assumes two types of weights -explicit and implicit (UNECE / European Commission 2018; Zaidi et al. 2013). Implicit weights represent the expert assessment of the significance of the spheres in the overall index. Explicit weights are the final values that are tailored to the magnitude of the values of the indicators and domains, providing their occurrence with weights closest to those proposed by the experts. In case of the individual AAI values, most indicators are binary variablesa person can be employed (1) or not (0), participate in an organised voluntary activity (1) or not (0) and so on. Thus, the average value of the EU indicator, taken into account by the explicit weights, is irrelevant here. Hence, unlike (Sidorenko and Zaidi 2013), we calculate the individual values of the AAI for Russia based on implicit weight (see 6 for figures), although it limits the direct comparability of the individual values with the results at the national level, but corresponds to the internal logic of the index methodology.
The mechanism for dealing with missing values when calculating the index at the national level in the multi-base approach assumes that if there are less than 5% of missing values and answer options such as "Difficult to answer", "Don't know", "No answer", these options could be dropped. In case of the single-base approach that would significantly limit the number of cases, thus the missing values are set to zero if this does not contradict the logic of the question, for example, if individuals do not know whether they care for grandchildren, they most probably do not. The question about social connectedness was not asked in the 2017 questionnaire, and for this indicator, 2016 data were used. The data are attached through the unique identifiers of the respondents. The number of missing is 268 respondents, for them, the weights for the 4th domain, which includes the question on social connectedness, are recalculated so that the relationship between the other five indicators did not change. This strategy is an example of indirect internal imputation of data in cases where the level of the excluded indicator is significantly different from the level of other indicators of the domain. However, a small number of respondents with data gaps (7%) do not have a significant effect on the distribution, even in the most critical cases. The number of domains (4), as well as the interpretation of the results, did not change.
The analyses of statistical relationships and group differentiation are made based on regression modelling, correlation analysis, the construction of contingency tables and SEM modelling. Further on, we analyse Russia's indicators for four the AAI domains, how they relate to each other and socio-demographic characteristics of the population. The analyses follow the division of the domains used in the index. First, the overall index scores are presented, following by the employment domain, the participation in the society domain, the independent and secure living domain, and the capacity and enabling environment for active ageing domain. The analysis of individual results is preceded by a summary of the relevant national-level multi-base results to supply the relevant context.
The Overall Index
The overall Russian AAI values gradually decline from 2010, except for 2016, although this uptick could be considered statistically non-significant (Table 1). In 2017 the AAI value was 29.8 points which equal to the 27th ranking place (30.6 for men and 29.2 for women). There is significant unrealised potential in the first and the second domains in Russia. However there is also much room for improvement in creating opportunities for independent and secure living and generating higher capacity and a stronger enabling environment for active ageing. Russia is well-below the EU averages in care to older adults, political participation, use of the Internet, voluntary activities, lifelong learning, physical exercise and employment of people aged 60-64 and 70-74 years old, and what is fundamentalin life expectancy, as no active ageing is possible without the opportunity to age.
The comparison between 2017 to 2010 is limited by the use of different databases for several questions. In general, the most significant changes have been caused by the employment domain (minus 15% since 2010) and social participation (minus 18%, but mainly due to a change in the methodology). The independent and secure living . That could have a double-sided effect, as the higher proportion of older ages in the structure of the population of 55+ could be associated with higher social disengagement, but at the same time life expectancy is included as one of the indicators and its rise contributes to the total index value. The AAI at the individual level shows a double-humped distribution, with two main groups of older people: the average realised potential for the first group is 34 points, for the second -66 points. The division is made by k-means clusters to provide a clear division into two groups. The minimum value is 13 points; the maximum is 87 points, the standard deviation is 15 points with an average of 42 points, the interquartile range is 19 points with a median of 37 points (Fig. 1).
The groups are selected mostly by the difference in the employment status of the individuals ( Fig. 2) with only 2.7% of the unemployed fell into the AAI group with high rates of realised potential, and only 0.4% of employed were in the group with low indicators of realised potential (Chi-squared test is statistically significant, Cramer's V is 0.941). People with employment have higher active ageing scores, and employment participation decreases with age. Therefore the calculation of the AAI without employment was conducted.
The calculation of the AAI at the individual level excluding employment status gives a distribution close to normal with a median and average -49 points, an interquartile range of 15 points and a standard deviation of 10 points (Fig. 3). On average, people over 55 realise their potential by 49% (± 10%), if paid employment is not considered. The AAI results decrease with age ( Fig. 4, Appendix Table 4) of respondents and depend on gender (indicators for women are higher than for men To evaluate the factors making for inequality in active ageing outcomes, we estimate the effects of the respondent's age, sex, employment status, educational level, place of living, marital status and health on the AAI without employment using regression modelling. 3 The type of settlement appeared to be not statistically significant. The level of education is included in the AAI as one of the components, that is why its inclusion into the list of explanatory variables can be questioned. However, since the scale used in the explanatory variable is different, we believe it is possible to include it in the model. At all accounts, it does not change the estimates for employment. The results presented in Table 2 show a weak positive effect of employment on the total AAI score (calculated without employment) while controlling for the age of the respondent, gender, marital status, health status and educational level. The same results are obtained when we additionally limit the regression estimation of the AAI without employment to the 55-69 age group, except for the age, that loses the level of its significance and the difference between higher education and secondary general education that becomes weaker.
To reveal further the differences between groups of older people with high and low levels of the AAI, we compare the characteristics of respondents falling in the first (lowest values of the AAI, less than 36%) and tenth (highest values, more than 63%) deciles of the AAI distribution without employment (Table 3). A group with higher active ageing potential is younger, better educated and more often employed. Hence, they also have higher incomes. The most crucial differences are in mental health status. However, seniors with higher active ageing potential also estimate their health state much better, and they have higher subjective well-being (measured by life satisfaction).
In most cases, indicators of the individual AAI have a weak positive correlation, up to 0.5 (Fig. 5), which suggests the possibility of combining various types of social and economic activities and statuses for the older generation, many indicators have a positive relationship with employment. A weak negative correlation is observed between indicators of engagement in family care and voluntary activity and independent living (independently living people report providing care and being involved in volunteering less frequently). In the next sections, the AAI results for Russia by domains and individual indicators, both on national and individual levels are considered.
The Employment Domain
The domain of employment of the national-level AAI in Russia amounts to 25.2 points for both men and women aged 55-74 (28.8 points for men and 22.9 for women), which corresponds to the 23rd place among EU countries and is below the EU average (31.1 points). This value is close to Malta and Spain (Appendix Table 5).
All the age groups show results below the EU average, except for the 70-74 years old. The primary reasons include low pension age (55 years old for women and 60 years old for men for the survey years), which is the lowest among 29 countries, numerous options for early retirement as well as poor health of the Russian older people. Since 2010 the negative dynamic of the values of the employment indicators is observed, which is even more intensified in 2017 due to restrictions on indexation of pension benefits for working pensioners introduced in 2016. For the period 2010-2017, the value of the domain estimated for both sexes decreased by 4.4 percentage points (2.3 . Compared to 2016, the decline in employment covered almost all age groups, except for women aged 65-69. Although federal statistical surveys give higher employment levels of the older generation and its dynamics, it is important to stress that observed negative trends are the consequences of both economic crises from 2014 and changes in the pension legislation regarding working pensioners (abolition of pension indexation from 2016). It is highly probable that the employment rates of older people will decline further in Russia.
At the individual level, 22.7% of respondents aged 55 years and above are employed, the employment rate strongly depends on the respondent's age and sex (Fig. 6). However, even in the younger age groups, there is a significant potential for improvement. For instance, only 60.4% of men and 46.6% of women aged 55-59 are employed. In the 60-64 age group, the employment rates decline to 32.8% for men and 29% for women. In all age groups, older women are less employed than men. Employment also varies by the level of health and education. Significant differences in the employment rates by the type of settlement can be observed only between cities and villages, in the latter the level of employment is lower for all age and sex groups. The marital status reveals no significant difference and is not shown on the graph.
The Participation in the Society Domain
The second domain of the AAI assesses the engagement of the older people in social activity, both within the family and outside it, and consists of 4 individual indicators: -Informal social activity: In 2017 there was a modification of the respective EQLS questions in the original methodology: the caring for children indicator changed the filter of respondents, also two new categories were added to the question of caring for adults: neighbours and friends. The wording of the questions used for the Russian AAI corresponds to the previous version of the methodology (used for the AAI-2012 and the AAI-2014), which allows tracing the changes in time for the multi-base AAI but limits the comparability with the latest EU results.
The results for the participation in the society domain for Russia equals 12.81 points for both sexes (11.27 points for men and 13.60 points for women), which places Russia to the 24th place of the total ranking among Austria, Poland and Greece. Thus, the potential to participate in public life is fulfilled by 48% compared to the "best EU practice" -Belgium. The significant decrease of the second domain value (Table 1) is foremost connected with the change of the database for the political activity indicator from ESS with the design weight to RLMS-HSE, which led to a three-times decrease of the indicator.
At the individual level, the average value of the second domain of the AAI is 18.5 points, which is mostly the results of the different weight system applied to the domain estimation. 4 However, the distribution is highly skewed to the left -59% of respondents do not have relevant social activities, 34.5% realise their social potential at half of the maximum possible or lower, and only 6.5% demonstrate high social involvement.
Overall, Russian older people are much more often engaged in family care activities than in formal social activities outside the family (Fig. 7). In caring activities, involvement in childcare is three times higher than in caring for elderly or disabled relatives, which most probably reflects the lower prevalence of very old relatives due to the earlier mortality of the Russian population. The number of respondents engaged in voluntary activity is too small to analyse it by socio-demographic characteristics. All indicators have a significant, but very weak positive correlation among themselves.
Women are more engaged in all the activities of the domain in Russia, which is quite unusual for the EU, where men tend to be more active outside of the family. Engagement in any social activity decreases with age and deteriorating subjective health status (Fig. 8, only significant factors are presented). The sample does not show significant differences in social participation by the place of residence, apart from slightly higher results for residents of the urban-type settlements (significance at the level of 0.1), which can be an artefact created by some sample peculiarities. Although some indicators of engagement in social activity vary with education and employment (e.g. involvement in political activity varies from 6.8% for people with higher education to 1.4% for people without secondary education and from 8.5% for working respondents to 2.5% for non-working), engagement in any (of four) social activity significantly does not depend on respondent's employment, income or education when controlling for the other variables. This result may indicate that respondents do not have the "work-help-grandchildren/children" choice dilemma.
The Independent and Secure Living Domain
The third domain of the AAI "Independent, healthy and secure living" is aimed at assessing the material security and independence of the older generation, their safety and comfort of living. The domain consists of eight indicators, three of which measure the financial independence of the respondent.
The value of the independent, healthy and secure life domain for Russia in 2017 is 60.5, which brings Russia to the 28th ranking place with only Latvia being below. Men show results considerably higher than women -66.0 and 58.0 correspondingly, which is a common situation in the EU, but the gender gap in Russia is one of the largest. From 2010 Russia's values of the domain almost did not change, despite a slight reduction in the relative median income and a small improvement in the indicators of access to medical and dental care, physical safety and independent living.
At the individual level, the best results are demonstrated by the indicators of the absence of the risk of poverty and material deprivation, as well as access to medical Fig. 8 Distribution of values of the second domain of the AAI depending on the respondents' sociodemographic characteristics (individual-level data) care (Fig. 9). Areas requiring significant development are physical activity (including walking) at the "every day or almost everyday" level and lifelong learning. Lifelong learning in Russia shows meagre rates, given only 1% of the population aged 55+ has attended any courses in the last 12 months (mostly employed women aged 55-64 living in regional centres or cities), which reflects the narrowness of the modern understanding of the concept of lifelong learning exclusively as professional retraining. The low level of older people participation in formal lifelong learning programs is a consequence of many factors. Among them, participation in lifelong learning can hardly help older adults to keep their job or to find a new one. Employers are not interested in retraining people near the pension age since they can easily replace them with younger workers or migrants. Also, the fact that many older adults have tertiary education can help them to fill the gap in the required knowledge or skills by self-education, which explains higher involvement in self-education. Identically to life long learning, a relatively higher indicator of physical activity is also demonstrated by women living in the regional centres. However, the age of maximum physical activity falls on the 64-74 age group, after which the rate decreases again. The indicator varies from 13.3% for respondents with good health to 7.9% for respondents with poor health, from 21.5% for persons with higher education to 6.7% for those without secondary education, from 17% for residents of regional the centres, up to 7.1% for residents of villages, which may indicate a lack of sport facilities for this age group, especially outside the cities. The independent living indicator in Russia for the individual level and 55+ age group is only 56.5% (the national-level indicator is lower than the EU minimum value, only 61.2% of the older people aged 75 years old and over live independently). This indicator aims to capture decisional autonomy regarding one's own life in older age, which means being selfsustaining, running the finances independently, being a household head. This indicator is strongly conditioned by dominating cultural patterns, and the low value for Russia could be explained by the widespread occurrence of multigenerational families. Such cohabitation can be a source of regular social contacts, inclusion into the social life and the feeling of being needed. However, very often it reveals difficulties of buying a separate housing and the forced living with the grown-up children. This is partly confirmed by the results of the correlation analysis -the lowest values for the independent living are observed in regional Independent, healthy and secure living Fig. 9 The value of indicators of the third domain of the AAI (individual-level data) centres and urban settlements. Women have a significantly lower probability of living alone or with a partner while having higher education, on the contrary increases the indicator. At the individual level, the values of the third domain are quite high, that is partly a consequence of the changed weights; the average value is 74.7 points, the median is 73 points, and the standard deviation is 17.4 points. The domain values are lower for older people with poor health status (based on subjective estimation), lower levels of education and lack of employment. The value of the domain is higher for urban residents than for those living in regional centres due to the indices of independent residence and the availability of medical services. Higher values of these indicators are also shown in the villages (possibly due to lower requirements for access to medicine), along with self-assessments of safety and the absence of risk of deprivation (Fig. 10). The effect of living in an urban-type settlement is not statistically significant on the value of the third domain when controlling for the other socio-demographic variables. Living with a partner have a positive influence, while widowhood is associated with a decrease in independent, healthy living when compared to those who were never married (not on the graph). The indicators of the third domain have a multi-directional correlation, but the strength of the connection is deficient, except for the "poverty riskrelative median income" pair, which corresponds to the logic of the AAI.
The Capacity and Enabling Environment for Active Ageing Domain
The indicators of the fourth domain, which measures substantive opportunities and empowerments to enhance active and healthy ageing, form the basis for realising the current potential of the older population, measured in the first three domains. The indicators can be divided into three main groups by their relation to active ageing: available assets (life expectancy); health capital (healthy life expectancy, mental health); human capital and opportunities favourable for active ageing (education, social connections, use of the Internet).
The value of the fourth domain at the national level is 51.4 points for Russia (23rd place in the ranking for both sexes), which is 72% of the best EU practice -Sweden. The closest neighbours in the ranking list are Slovakia and Poland. According to the capacity for active ageing, the potential of Russian men (49.6 points) is slightly lower than that of women (53.0 points), which is due to the lower male life expectancy, less frequent social contacts and lower use of the Internet, but men show a higher level of mental well-being. Since 2010 there is a gradual increase in the life expectancy at 55 years, use of the Internet, mental well-being and educational attainment (as later cohorts pass the index' lower boundary). The share of healthy life expectancy trend is unstable with men having the maximum value in 2012 with a gradual decline afterwards and women with an increase in 2012 and 2016 (the max achieved value), partly that could be explained by the subjectivity of the question used. 5 The peculiarity of Russia is a pronounced gender discrepancy of life expectancy and the excessive mortality of men before the age of 55, which leads to one of the highest values of the share of healthy life indicators, since many Russians, unfortunately, die before the onset of diseases that limit the capacity and significantly worsen the subjective health assessments. The current gap between the life expectancy of men and women equals to 10.6 years for the life expectancy at birth and 6.6 at 55 years. While the first indicator is slowly decreasing, life expectancy at 55 years remain stable, meaning that the reduction of mortality of man is happening mostly in the workingage (Federal State Statistics Service 2010; Federal State Statistics Service 2017).
Since the mortality tables cannot be estimated at the individual level, to calculate the fourth domain of the AAI at the individual level, life expectancy and the proportion of healthy life expectancy are taken as fixed values in order to preserve the original structure of the index. This methodological decision reduces the socio-demographic change of the individual values of the fourth domain, given that there are differences in life expectancy and health, related to the type of residence, education and income. In the previous work on the AAI on the individual level, the share of healthy life expectancy was substituted by the health self-esteem (Barslund et al. 2017), but this reduces the possibilities of using this indicator in socio-demographic variables. Thus the fixed macrolevel gender values are preserved for these two indicators, which lead to certain underestimation of the inequality of the active ageing capital.
Two indicators that have the most significant effect on the fourth domain value are the level of education (above the secondary general) and the proportion of healthy life expectancy (fixed to the national level), which is associated with the peculiarities of the Russian mortality pattern (Fig. 11). The area that requires the most attention is the use of the Internet since only 29% of respondents reported using the network at least once a week. The latter indicator varies considerably depending on the age of the respondent ( of people aged 55-59 years old and only 0.6% of the other), job availability (56.1% of working respondents and only 20.9% of non-working), education level (54.2% of people with higher education and 5.8% of those without secondary education), place of living (34.9% of older people in regional centres and 19.1% in villages). At the individual level, the average value of the fourth domain equals to 51.9 points. The standard deviation is 12.5 points; the median is equal to the mean. Because of the peculiarity of the methodological approach (two fixed indicators), the minimum value of the fourth domain does not fall below 30 points, while the maximum is 75 points. The value of the domain varies depending on the subjective health status, employment, level of education, and age (Fig. 12) The capacity and enabling environment for active aging Fig. 12 The distribution of the values of the fourth domain of the AAI, depending on the respondents' sociodemographic characteristics (individual-level data) positive correlation. The strongest correlations (at the level of 0.26) is observed between the education level and the Internet use.
Discussion and Conclusions
The research presented in this paper aimed to analyse the degree of the existing inequality in active ageing potential and to reveal defining factors behind it by estimating the Active Ageing Index (AAI) at the individual level supplied by the national-level context. The results show the significant variation of the AAI and its domains by the older peoples' age, sex and subjective health status. Health status has a statistically significant impact on the total score of active ageing while controlling such socio-demographic characteristics as age, employment status, gender, education level, marital status and place of living. In many cases, the education level correlates positively with engagement in different activities, and consequently, the index and its domains' values. Besides, we demonstrate the positive correlation between employment and engagement in other activities at the individual level. Surprisingly, the place of living correlates significantly with the values of only the first and third AAI domains. However, it can be a result of the contradictory relations between different indicators within domains with the place of living. The most important factor differentiating the AAI at the individual level into two groups is employment. That is why we decided to analyse the data without information on the employment variable. The group with higher values of the AAI without employment includes more women, people with better health and better education, living in cities. To the contrary, the group with lower values of the AAI without employment includes more people with poor health, secondary general education or lower. The latter group should be a focus of policy attention.
The results of the project provide evidence for the implementation of policy measures in the target groups. The high correlation of the index values with human capital indicators (health and education) underlines the importance of the early interventions aimed at promoting and supporting human capital at the earlier stages of the life course till the old age. The substantial positive correlation of employment with other forms of activity as well as positive correlations of the latter with human capital variables stresses the importance of developing a package of activation policy measures aimed at the 50-55 age group to increase the likelihood of their active participation in all areas of active ageing while they become older. Furthermore, the weak positive correlation of the indicators of the AAI indicates the absence of a "dilemma of choice" between certain types of activity, for example, between caring for grandchildren and employment, or employment and volunteering. It proves the right balance for the index as a tool for monitoring the complex activity of the older generation. High dependence of the indicator values on age demonstrates the need to diversify policy measures depending on the age characteristics of older adults. Older people (70+ or 75+) have a higher probability of having lower values of the AAI and hence should be a target group of active ageing policies.
Besides, calculation of the individual values of the AAI could also be used for promotion of the active ageing principles for the general public. Being by the essence a questionnaire with yes/no answer options, it could be used as a personal checklist assessing the contribution of an individual to the realisation of his or her potential for the benefit of the society.
The comparison of the AAI values estimated for Russia with the EU countries allows revealing the strengths and weaknesses of the current situation with active ageing in Russia. The strengths of Russia in the context of realising the potential of the older citizens include the level of education of older adults, financial indicators and the engagement in family care, especially in childcare. To the contrary, the maximum potential for the development of active ageing today shows indicators of employment, volunteering, political engagement, physical activity, lifelong learning and the use of the Internet.
Both the individual and the national levels analyses indicate the importance of health improvement and increasing the life expectancy of older citizens in order to move Russia further in active ageing. Reducing health-related restrictions will create a solid basis for increasing the AAI in the first three domains, and vice versa, maintaining the current level of mortality and morbidity will reduce the effects of social policy.
An important factor that is not directly taken into account by the AAI, but significantly influences the activity of the older people in the social sphere, is the accessible and age-friendly environment, and in particular, the accessibility of transport, which allows maintaining social contacts, participating in public life and using infrastructure facilities without help. More active involvement of the older people in work, retraining and providing alternative job positions based on the needs and potential of this age group will not only have a positive effect on financial security but will also increase the level of realised potential in other areas (the third and fourth domains show a statistically significant relationship with employment).
We should also underline that the results of the calculation of the AAI for Russia are immensely relevant for monitoring the effectiveness of new and existing social policy measures aimed at improving the quality of life and increasing the activity of the older generation. The AAI indicators coincide with the objectives and directions set in the Strategy of Action for Older Citizens until 2025 and a new National Project on Demography (and its componentfederal project Old Generation), and, hence, can be used to assess progress and identify areas requiring the attention of politicians and the society.
Given high interregional diversity of Russia, further research should be devoted to the regional studies. Although the analyses showed a weak or no significance of place of living, the hypothesis of the dependence of the index value on the region of residence should be checked in the following works. It would also highlight the most successful regions for the dissemination of their experience. Calculations of indicators for individual social groups may be of interest to both the academic community and politicians. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. | 2023-01-08T14:11:05.683Z | 2020-05-12T00:00:00.000 | {
"year": 2020,
"sha1": "b65c4c17ab72a7d1868d09765f68cc0b9b71331e",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s12062-020-09277-4.pdf",
"oa_status": "HYBRID",
"pdf_src": "SpringerNature",
"pdf_hash": "b65c4c17ab72a7d1868d09765f68cc0b9b71331e",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": []
} |
226052992 | pes2o/s2orc | v3-fos-license | Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting
<p>Skillful streamflow forecasts provide a key support to several water-related applications. Ensemble forecasting systems are gaining a widespread interest, since they allow accounting for different sources of uncertainty. Because of the critical impact of the initial conditions (ICs) on the forecast accuracy, it is essential to improve their estimates via data assimilation (DA). This study aims at assessing the sensitivity of the DA-based estimation of forecast ICs to several sources of uncertainty and to the update of different model states and parameters of a conceptual rainfall-runoff model. The performance of two sequential ensemble-based techniques are compared, namely Ensemble Kalman filter and Particle filter, in terms of both efficiency and temporal persistence of the updating effect through the assimilation of observed discharges at the forecast time. Several experiments specifically address the impact of the meteorological, model state and parameter uncertainties over 232 catchments in France. Results show that the benefit of the DA-based estimation of ICs for forecasting is the largest when focusing on the level of the model routing store, which is the internal state the most correlated to streamflow. While the EnKF-based forecasts outperform the PF-based ones when accounting for the meteorological uncertainty, the representation of the model state uncertainty allows greatly improving the accuracy of the PF-based predictions, with a longer-lasting updating effect (up to 10 days). Conversely, the forecasting skill is undermined when accounting for the parameter uncertainty, due to the change in the hydrological responsiveness through the update of both the production and routing store levels. A further effort is focused on assessing the impact of the spatial resolution of the hydrological model on the predictive accuracy of DA-based streamflow forecasts.</p>
Skillful streamflow forecasts provide key support to several water-related applications. Because of the critical impact of initial conditions (ICs) on forecast accuracy, data assimilation (DA) can be performed to improve their estimation.
• sensitivity to several sources of uncertainty Daily discharge measurements at watershed outlets ( ) are assimilated. The uncertainty in observations is assessed as a function of the streamflow rate (Weerts and El Serafy, 2006;Thirel et al., 2010).
EnKF SIR-PF • Production store level (S) • Routing store level (R) • Unit hydrograph (UH) • Capacity of production store (X 1 ) • Capacity of routing store (X 3 ) #shareEGU20 Piazzi, Thirel, Perrin, Delaigue Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting > Uncertainty in meteorological forcings
Methodology
Probabilistic meteorological forecasts are generated by stochastically perturbing the SAFRAN meteorological reanalysis with multiplicative stochastic noise (Clark et al., 2008).
Model state variables
• Potential evapotranspiration (E) • Precipitation (P) • Production store level (S) • Routing store level (R) • Unit hydrograph (UH) • Capacity of production store (X 1 ) • Capacity of routing store (X 3 ) #shareEGU20 Piazzi, Thirel, Perrin, Delaigue Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting >
Methodology
After the analysis procedure, model states are perturbed through normally distributed null-mean noise (Salamon and Feyen, 2009).
Model state variables
• Potential evapotranspiration (E) • Precipitation (P) • Production store level (S) • Routing store level (R) • Unit hydrograph (UH) • Capacity of production store (X 1 ) • Capacity of routing store (X 3 ) #shareEGU20 Piazzi, Thirel, Perrin, Delaigue Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting > Uncertainty in model parameters
Methodology
Model parameters are jointly updated with state variables, according to the augmented state vector approach, and perturbed (Moradkhani et al., 2005).
Model state variables
• Potential evapotranspiration (E) • Precipitation (P) • Production store level (S) • Routing store level (R) • Unit hydrograph (UH) • Capacity of production store (X 1 ) • Capacity of routing store (X 3 ) #shareEGU20 Piazzi, Thirel, Perrin, Delaigue Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting > Compared to PF, EnKF-based ICs guarantee a greater improvement in predictive accuracy (PF affected by ensemble shrinkage during no-rain periods).
Both the EnKF and the PF schemes reveal an effective usefulness to improve predictive accuracy by the assimilation of observed discharges. When dealing with a conceptual hydrological model, the main interest is on the routing dynamics to derive the most benefit from the DA-based ICs.
A comprehensive representation of both meteorological and state uncertainties allows for a more efficient improvement of predictive skill.
PF-based ICs are greatly enhanced thanks to a larger spread of the ensemble simulations. While the PF-based updating effect is longer lasting, the benefit of larger corrective terms for the EnKF rapidly decreases within a short lead time.
High sensitivity to the parameter estimation, as store capacities define the simulated hydrological responsiveness of the basin.
Parameter values estimated at the forecast time may not be the optimal ones to represent the model response over the forecast horizon. The equifinality issue can affect the parameter estimates, especially in PF. #shareEGU20 Piazzi, Thirel, Perrin, Delaigue Assessing sensitivity and persistence of updated initial conditions through Particle filter and EnKF for streamflow forecasting > Ongoing and future perspectives This study has been recently submitted to the Water Resources Research journal: Piazzi, G., Thirel, G., Perrin, C., Delaigue, O. Sequential data assimilation for streamflow forecasting: assessing the sensitivity to uncertainties and updated variables of a conceptual hydrological model.
An R package providing the DA schemes will be soon available.
The authors thank Météo-France and SCHAPI for providing climate and streamflow data. The first author received financial support from SCHAPI and the RenovRisk-Transfer project. This work contributes to the SPAWET project funded by the CNES-TOSCA program.
Introduction
Forecasting system
Results
Conclusions & perspectives 6 References | 2020-03-12T10:36:59.849Z | 2020-03-09T00:00:00.000 | {
"year": 2020,
"sha1": "a091d31ac229ff1d957b3d9965f206fb5553a03a",
"oa_license": "CCBY",
"oa_url": "https://hal.inrae.fr/hal-03266313/file/EGU2020-18694-print.pdf",
"oa_status": "GREEN",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "99fc6d66fdf23794b788ef8f0abaabf4f9c73691",
"s2fieldsofstudy": [
"Environmental Science",
"Engineering"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
182381627 | pes2o/s2orc | v3-fos-license | Intrahepatic Cholestasis in a Sickle Cell Patient Unresponsive to Exchange Blood Transfusion
With the advent of hydroxyurea, the sickle cell population has been enjoying a prolonged life span as compared to the pre-hydroxyurea era. Traditionally, acute complications of sickle cell disease includes acute chest syndrome, MI and stroke. In this report we present a case of an elderly man with sickle cell disease who presented with intrahepatic cholestasis (SCIC); a rather rare and fatal complication of sickle cell hemoglobinopathy. The patient presented with jaundice and elevated bilirubin up to 53, his hospital course was complicated by coagulopathy and encephalopathy, and expired on day 43 of presentation after failing multiple therapeutic interventions including exchange transfusion. In this report, we will provide literature review and discuss the underlying pathophysiologic mechanisms of intrahepatic cholestasis in the sickle cell population highlighting the need for immediate recognition and institution of therapy for this fatal complication of sickle cell disease, particularly in elderly populations with low metabolic reserve.
Introduction
SCD is characterized by ischemic infarcts of multiple organ systems, as a result of a sickling deformity of red blood cells. Hepatomegaly, mild to moderate indirect hyperbilirubinemia as well as pigmented biliary stones are relatively common liver related complications in sickle cell disease. [1] Cell hypoxia is believed to be the underlying mechanism for intrahepatic cholestasis. [3] With progressive damage to hepatocyte and sludging of red blood cells (RBCs) into the hepatic sinusoids, end organ damage begins to take place in the liver. Microscopy could reveal ballooned and necrotic hepatocytes, microinfarcts, and canaliculi plugged with bile.
Exchange transfusion (ET) is the only effective treatment for SCIC; early intervention with exchange transfusion helps to minimize sickling thus lowering the fraction Hemoglobin S (ideally below 30%). [2] ET is the only intervention which decreases patient mortality as compared to solely supportive measures. [2] Cholelithiasis and viral hepatitis must be ruled out before a diagnosis of SCIC can be made (often a diagnosis of exclusion). [4] The disease can present in its most severe form when the patient is already coagulopathic with renal failure and ultimately fulminant hepatic failure. [4] We discuss a case of an elderly male who initially presented with hyperbilirubinemia likely secondary to SCIC; his bilirubin continued to trend upward and he ultimately went into liver failure. The patient received exchange transfusion and had a decrease in Hgb S to less than 30%; however his clinical status did not improve and the patient succumbed to his disease.
Case Report
A 64 year old man with medical history of Hemoglobin S/β thalassemia, hypertension, Atrial fibrillation, CAD, and ESRD on hemodialysis presented with anuria, fever, vomiting, and icterus to the emergency room. The patient was alert and oriented, and his exam was negative for abdominal tenderness. Patient was not taking any hepatotoxic medications or supplements; he had been dialyzed at his center the day prior to presentation. The vital signs revealed a blood pressure of 148/74, heart rate of 93 beats per minute, respiratory rate of 18 breaths per minute, temperature of 98.9°F, and oxygen saturation of 97% on admission.
Initial labs were significant for cholestasis with total bilirubin of 20 mg/dL, and a direct bilirubin > 10 mg/dL, anemia with hemoglobin of 6.8 g/dL (patients baseline is 8.0 g/dL), and elevated liver enzymes with aspartate aminotransferase (AST) of 109 U/L, alanine aminotransferase (ALT) of 18 U/L and alkaline phosphatase (ALP) of 202 U/L. A viral hepatitis panel was unremarkable (for labs, see Table 1). Patients Hemoglobin S was 78.7%, his lactate dehydrogenase (LDH) was 1083 U/L and haptoglobin was <30 mg/dL.
On right upper quadrant sonogram ( Figure 1) the liver was noted to be 18.4 cm with normal echogenicity, no surface nodularity, and patent veins with normal flow. The intrahepatic ducts were normal, and the common bile duct diameter was 3mm at the porta hepatis; the gallbladder was not visualized. The CT Abdomen and Pelvis with IV contrast (Figure 2) was consistent with a surgical history of cholecystectomy and no evidence of biliary ductal dilation. The liver was of a normal contour, with patent portal veins.
The hematology service was consulted for further management of the patient; they initially believed that the hyperbilirubinemia was secondary to sickle cell hemolysis and underlying liver disease in the setting of elevated LDH, decreased haptoglobin and imaging negative for obvious obstructive disease. They also suggested exchange transfusion if the patient did not clinically improve. The nephrology service was also consulted to facilitate hemodialysis, which was maintained three times a week as scheduled during the patients' hospital course. The patients' bilirubin continued to rise and peaked at 52.8mg/dL on day 13 of his admission (Figure 3). The international normalized ratio (INR) also continued to uptrend to a peak of 3.2 (from a baseline of 1.2) on day 12 of his admission. Towards the end of his 12 th day of the hospital stay, the patient became altered and developed tender hepatomegaly.
He was promptly transferred to the intensive care unit for exchange transfusion (ET). ET was carried out with 7 units of PRBCs via his A-V fistula; the HbS concentration on HPLC was noted to be <30% after the ET. The patients' total bilirubin decreased to 40 mg/dL following ET, however he remained altered and coagulopathic. Since the patients HbS concentration decreased below 30% further ET was not warranted.
Patient continued to have altered mental status, however a CT Head was negative for acute intracranial pathology. Hyperbilirubinemia causing bilirubin neurotoxicity was proposed as the plausible explanation of the mental status alteration, particularly in the setting of normal ammonia levels. The patient was started on vitamin K for coagulopathy and ursodiol for hyperbilirubinemia. The patient continued to decompensate further and palliative care evaluation recommended supportive therapy. On day 43 patient expired.
Discussion
SCIC is a rare, but severe complication in patients who have either hemoglobin (Hb) SS or Hb S/β thalassemia disease states. Patients with SCIC commonly present with right upper quadrant pain, hepatomegaly, elevated conjugated bilirubin, and elevated transaminases, as was seen in our patient. While the incidence of SCIC is higher in children than in adults, adults tend to experience a more aggressive disease course with mortality rate as high as 50%. [5] Diagnosis of SCIC calls for primary biliary disease to be ruled out, which is most commonly done using ultrasound which has a positive predictive value of 95% in diagnosing acute cholecystitis. [7] Since our patient had a history of cholecystectomy, cholecystitis was not in the differential diagnosis, and there was no evidence of biliary obstruction on either abdominal ultrasound or CT scan. Infectious workup for viral hepatitis was unrevealing and there was no clinical suspicion for drug-induced liver injury.
According to Ahn et al., SCIC can be classified as either mild or severe disease; mild disease presents with a mean direct bilirubin level of 27.6 mg/dL without manifestations of significant hepatic damage. Severe disease presents with a mean direct bilirubin of 76.8 mg/dL, presence of a coagulation disorder or altered mental status. [8] Based on the above, our patient initially presented with mild disease, however progressed to severe disease. With this progression to severe disease, there is an increased mortality rate as high as 64%, compared to only 4% in mild disease. [11] Once the diagnosis of SCIC has been made, prompt efforts to treat must be undertaken as disease progression is rapid and fatal. According to Ahn et al., ET is an important treatment modality that resulted in success in 7 out of 9 patients reported in the study. On the other hand, in those who did not have ET, twelve out of thirteen patients expired. Their findings collectively support the empiric use of ET pending randomized control trials. In our patient with severe disease, despite receiving ET with 7 units of packed RBCs resulting in a decrease of Hb S to 30%, our patient did not survive, underscoring the high mortality rate among this vulnerable population. Other treatment options such as albumin dialysis have shown some benefit in limited case reports, however its efficacy has not been clearly documented or studied. [11] Liver transplantation has also been used to manage SCIC, however this treatment strategy has not been widely evaluated as these patients are generally poor candidates for transplant due to multiorgan failure, as was the case with our patient. Furthermore, in reported cases of liver transplantation for SCIC, outcomes are poor and transplantation is not successful due to sequela of ongoing SCD including vascular complications.
Conclusions
As the sickle cell population continues to age with better symptom control and availability of effective therapeutics options including hydroxyurea, rare complications such as SCIC come to the forefront of SCD management. As demonstrated in our report, SCIC, while a rare occurrence, is a highly fatal complication of SCD that calls for early diagnosis and initiation of ET therapy, if possible within 48 hours of diagnosis. To our knowledge, our report represents the oldest patient with documented SCIC that presented with a rather gradual onset, making the diagnosis hard to ascertain given the broad differential of hyperbilirubinemia. This atypical presentation has resulted in delayed initiation of ET; the only known viable therapeutic option available to date. In addition to late presentation, advanced age and end-stage renal disease in our patient likely contributed to his deteriorating status. CT Abdomen and Pelvis with IV Contrast with evidence of cholecystectomy and no evidence of biliary ductal dilation. The liver was of a normal contour, with patent portal veins | 2019-06-07T21:13:32.299Z | 2019-04-25T00:00:00.000 | {
"year": 2019,
"sha1": "9195cabad8732d675504d5d54b706a98114fb277",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.12691/ajmcr-7-4-4",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "b99c67c593f23680606bb60056fef372eced186c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
244793705 | pes2o/s2orc | v3-fos-license | Phospholipase A1 Member A Activates Fibroblast-like Synoviocytes through the Autotaxin-Lysophosphatidic Acid Receptor Axis
Lysophosphatidylserine (lysoPS) is known to regulate immune cell functions. Phospholipase A1 member A (PLA1A) can generate this bioactive lipid through hydrolysis of sn-1 fatty acids on phosphatidylserine (PS). PLA1A has been associated with cancer metastasis, asthma, as well as acute coronary syndrome. However, the functions of PLA1A in the development of systemic autoimmune rheumatic diseases remain elusive. To investigate the possible implication of PLA1A during rheumatic diseases, we monitored PLA1A in synovial fluids from patients with rheumatoid arthritis and plasma of early-diagnosed arthritis (EA) patients and clinically stable systemic lupus erythematosus (SLE) patients. We used human primary fibroblast-like synoviocytes (FLSs) to evaluate the PLA1A-induced biological responses. Our results highlighted that the plasma concentrations of PLA1A in EA and SLE patients were elevated compared to healthy donors. High concentrations of PLA1A were also detected in synovial fluids from rheumatoid arthritis patients compared to those from osteoarthritis (OA) and gout patients. The origin of PLA1A in FLSs and the arthritic joints remained unknown, as healthy human primary FLSs does not express the PLA1A transcript. Besides, the addition of recombinant PLA1A stimulated cultured human primary FLSs to secrete IL-8. Preincubation with heparin, autotaxin (ATX) inhibitor HA130 or lysophosphatidic acid (LPA) receptor antagonist Ki16425 reduced PLA1A-induced-secretion of IL-8. Our data suggested that FLS-associated PLA1A cleaves membrane-exposed PS into lysoPS, which is subsequently converted to LPA by ATX. Since primary FLSs do not express any lysoPS receptors, the data suggested PLA1A-mediated pro-inflammatory responses through the ATX-LPA receptor signaling axis.
Introduction
Phospholipase A1 member A (PLA1A, also known as PS-PLA1) belongs to the pancreatic lipase family and was purified initially from the thrombin-activated platelet-rich plasma of rats [1]. PLA1A shows specific substrate preference for sn-1 fatty acids on phosphatidylserine (PS) or lysophosphatidylserine (lysoPS). This feature distinguishes PLA1A from other classical pancreatic lipase family members [1]. PS localization in the
PLA1A Levels in Plasma, Synovial Fluids, and Human Primary FLSs
We first examined by ELISA the concentrations of PLA1A in platelet-free plasma from 12 healthy donors, 38 early-diagnosed arthritis (EA) patients, and 62 SLE patients undergoing treatment as well as in hyaluronidase-treated synovial fluids from five RA patients, three psoriatic arthritis (PsoA) patients, three osteoarthritis (OA) patients, and three gout patients. PLA1A concentrations in plasma from EA patients were 4.46 ± 2.03 ng/mL ( Figure 1A). Compared to healthy donors (0.09 ± 0.09 ng/mL) and clinically stable SLE patients (0.6 ± 0.2 ng/mL), EA patients showed a higher concentration of PLA1A in their plasma. There was no sex difference in the plasma concentration of PLA1A between male (3.44 ± 2.45 ng/mL) and female (5.49 ± 3.3 ng/mL) EA patients ( Figure 1B). The concentrations of PLA1A in female SLE patients (0.67 ± 0.23 ng/mL) and males (0.19 ± 0.18 ng/mL) were not significantly different ( Figure 1C).
hibitors or antagonists of the ATX-LPA receptor axis to study the mechanisms by which PLA1A activates FLSs. We demonstrate that the pro-inflammatory effects of PLA1A in FLSs were, at least in part, dependent on the ATX-LPA receptor axis.
PLA1A Levels in Plasma, Synovial Fluids, and Human Primary FLSs
We first examined by ELISA the concentrations of PLA1A in platelet-free plasma from 12 healthy donors, 38 early-diagnosed arthritis (EA) patients, and 62 SLE patients undergoing treatment as well as in hyaluronidase-treated synovial fluids from five RA patients, three psoriatic arthritis (PsoA) patients, three osteoarthritis (OA) patients, and three gout patients. PLA1A concentrations in plasma from EA patients were 4.46 ± 2.03 ng/mL ( Figure 1A). Compared to healthy donors (0.09 ± 0.09 ng/mL) and clinically stable SLE patients (0.6 ± 0.2 ng/mL), EA patients showed a higher concentration of PLA1A in their plasma. There was no sex difference in the plasma concentration of PLA1A between male (3.44 ± 2.45 ng/mL) and female (5.49 ± 3.3 ng/mL) EA patients ( Figure 1B). The concentrations of PLA1A in female SLE patients (0.67 ± 0.23 ng/mL) and males (0.19 ± 0.18 ng/mL) were not significantly different ( Figure 1C).
In synovial fluid samples, PLA1A was 35.08 ± 7.56 pg/mL in RA patients, 636.8 ± 521.4 pg/mL in PsoA patients, 10.04 ± 0.98 pg/mL in OA patients, and 18.74 ± 9.03 pg/mL in gout patients, respectively ( Figure 1D). Compared to gout and the non-rheumatoid disease OA, the inflammatory synovial fluids of RA patients showed a higher concentration of PLA1A. The concentration of PLA1A in PsoA was higher than in RA synovial fluids. 3 OA patients, and 3 gout patients and were treated with 0.5 mg/mL hyaluronidase for 5 min at room temperature before measurement. For statistical comparative analyses, PLA1A concentrations in EA patient plasma were compared to that in healthy donors and in SLE patients. PLA1A concentrations in synovial fluid of RA patients were compared to that in PsoA, OA, and gout donors.
In synovial fluid samples, PLA1A was 35.08 ± 7.56 pg/mL in RA patients, 636.8 ± 521.4 pg/mL in PsoA patients, 10.04 ± 0.98 pg/mL in OA patients, and 18.74 ± 9.03 pg/mL in gout patients, respectively ( Figure 1D). Compared to gout and the non-rheumatoid disease OA, the inflammatory synovial fluids of RA patients showed a higher concentration of PLA1A. The concentration of PLA1A in PsoA was higher than in RA synovial fluids.
Human tissues, such as skeletal muscle, kidney, liver, and testis, express the PLA1A transcript. However, the expression of PLA1A in human FLSs remains unknown. Higher PLA1A levels in synovial fluids of RA patients might come from FLSs. To assess this possibility, we performed a semi-quantitative RT-PCR to examine the expression of PLA1A in normal human primary FLSs using two different pairs of PLA1A oligo primers and the total RNA from human testis as a positive control. As shown in Figure 2A, PLA1A mRNA was not detectable in normal FLSs under our experimental conditions. However, using western blotting and flow cytometry, we detected PLA1A at the protein level in cultured FLSs ( Figure 2B,C). In the whole-cell lysate, the amount of PLA1A in serumstarved cells was slightly decreased compared to serum-cultured cells, thereby suggesting that fetal bovine serum (FBS) is an extracellular source of PLA1A in cell culture. Flow cytometry without permeabilization failed to detect the PLA1A signal on the FLSs surface (data not shown). Flow cytometry with permeabilization detected the PLA1A signals in cells cultured with or without serum, and the amount of PLA1A was slightly lower in serum-starved cells. Human tissues, such as skeletal muscle, kidney, liver, and testis, express the PLA1A transcript. However, the expression of PLA1A in human FLSs remains unknown. Higher PLA1A levels in synovial fluids of RA patients might come from FLSs. To assess this possibility, we performed a semi-quantitative RT-PCR to examine the expression of PLA1A in normal human primary FLSs using two different pairs of PLA1A oligo primers and the total RNA from human testis as a positive control. As shown in Figure 2A, PLA1A mRNA was not detectable in normal FLSs under our experimental conditions. However, using western blotting and flow cytometry, we detected PLA1A at the protein level in cultured FLSs ( Figure 2B,C). In the whole-cell lysate, the amount of PLA1A in serum-starved cells was slightly decreased compared to serum-cultured cells, thereby suggesting that fetal bovine serum (FBS) is an extracellular source of PLA1A in cell culture. Flow cytometry without permeabilization failed to detect the PLA1A signal on the FLSs surface (data not shown). Flow cytometry with permeabilization detected the PLA1A signals in cells cultured with or without serum, and the amount of PLA1A was slightly lower in serumstarved cells. FLSs from normal donors were tested for the expression of PLA1A by semi-quantitative RT-PCR. Two pairs PLA1A oligo primers were tested. GAPDH was used as a house-keeping gene. Total RNA Figure 2. Presence of PLA1A in cultured human primary FLSs. (A) Total RNA from human primary FLSs from normal donors were tested for the expression of PLA1A by semi-quantitative RT-PCR. Two pairs PLA1A oligo primers were tested. GAPDH was used as a house-keeping gene. Total RNA from human testis tissue was used as a positive control. (B) The presence of PLA1A in human primary FLSs cultured in the absence or presence of 10% FBS were examined using western blot. GAPDH was used as an internal loading control. Recombinant PLA1A protein was used as a positive control. The blots shown are representative of five independent experiments with similar results. Amount of PLA1A was quantified densitometrically and was normalized with respect to total GAPDH. The data were the means ± SE from five experiments. (C) The presence of PLA1A in human primary FLSs cultured in the absence or presence of 10% FBS were examined using flow cytometry. Cells were permeabilized with digitonin, and rabbit IgG isotype was used as a control. The figures shown were representative of three independent experiments with similar results. The data were the means ± SE from three experiments. MFI, mean fluorescence intensity.
Pro-Inflammatory Effects of PLA1A and the Detection of Its Substrate
To verify the potential pro-inflammatory effects of PLA1A, we incubated serum-starved cultured FLSs with or without recombinant PLA1A for 24 h (Figure 3). The basal concentration of IL-8 secretion in non-stimulated serum-starved FLSs was 4.03 ± 1.53 pg/mL. The addition of 0.2 µg/mL PLA1A increased the concentration of IL-8 to 17.76 ± 4.58 pg/mL. PLA1A induced IL-8 secretion in a concentration-dependent manner, and the highest concentration tested (0.5 µg/mL) had no cytotoxic effect as assessed using PI labeling.
Cells were permeabilized with digitonin, and rabbit IgG isotype was used as a control. Th shown were representative of three independent experiments with similar results. The d the means ± SE from three experiments. MFI, mean fluorescence intensity.
Pro-Inflammatory Effects of PLA1A and the Detection of Its Substrate
To verify the potential pro-inflammatory effects of PLA1A, we incubated starved cultured FLSs with or without recombinant PLA1A for 24 h (Figure 3). Th concentration of IL-8 secretion in non-stimulated serum-starved FLSs was 4.03 pg/mL. The addition of 0.2 µg/mL PLA1A increased the concentration of IL-8 to 4.58 pg/mL. PLA1A induced IL-8 secretion in a concentration-dependent manner, highest concentration tested (0.5 µg/mL) had no cytotoxic effect as assessed usin beling. The amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 prod FLSs cultured in serum-free DMEM was 4.03 ± 1.53 pg/mL, and for each experiment, b production was set at 100% for data normalization. The data shown are the means ± SE fro experiments. For statistical comparative analyses, cytokine amount in cells stimulated with were compared to that in non-stimulated cells. ** p < 0.01. PLA1A specifically hydrolyzes PS, which usually localizes in the inner leafle cell membrane. When cells are activated or undergoing apoptosis, they expose PS outer leaflet of the cell membrane. To assess the externalization of PS during cel tion, we treated FLSs with 100 ng/mL TNF for 1, 5, 15, 30, and 60 min and monit exposure on FLSs by flow cytometry using FITC Annexin V. Cell viability was not under those conditions, with dead cells (PS-and PI-positive) representing less tha the population. FITC Annexin V-positive and PI-negative cells were considered cells presenting PS on their outer leaflet. As shown in Figure 4A, the percentag positive FLSs increased from 13.98% to 23.6% (p-value > 0.05, n = 4), following stim with TNF for 30 min, and remained elevated up to 60 min, the last timepoint teste PLA1A has an affinity to surface heparin sulfate proteoglycan, and added can competitively bind to PLA1A, thereby preventing the cell-surface-exposed P The amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 produced by FLSs cultured in serum-free DMEM was 4.03 ± 1.53 pg/mL, and for each experiment, basal IL-8 production was set at 100% for data normalization. The data shown are the means ± SE from three experiments. For statistical comparative analyses, cytokine amount in cells stimulated with PLA1A were compared to that in non-stimulated cells. ** p < 0.01. PLA1A specifically hydrolyzes PS, which usually localizes in the inner leaflet of the cell membrane. When cells are activated or undergoing apoptosis, they expose PS on the outer leaflet of the cell membrane. To assess the externalization of PS during cell activation, we treated FLSs with 100 ng/mL TNF for 1, 5, 15, 30, and 60 min and monitored PS exposure on FLSs by flow cytometry using FITC Annexin V. Cell viability was not affected under those conditions, with dead cells (PS-and PI-positive) representing less than 2% of the population. FITC Annexin V-positive and PI-negative cells were considered viable cells presenting PS on their outer leaflet. As shown in Figure 4A, the percentage of PS-positive FLSs increased from 13.98% to 23.6% (p-value > 0.05, n = 4), following stimulation with TNF for 30 min, and remained elevated up to 60 min, the last timepoint tested.
PLA1A has an affinity to surface heparin sulfate proteoglycan, and added heparin can competitively bind to PLA1A, thereby preventing the cell-surface-exposed PS from hydrolysis [35]. Next, we added increasing amounts of heparin to FLSs 30 min before the treatment with or without 0.2 µg/mL PLA1A. As shown in Figure 4B, heparin concentrations less than 200 µg/mL enhanced PLA1A-mediated IL-8 production. At higher concentrations (400, 800, and 1600 µg/mL), heparin showed an inhibitory effect on IL-8 production. Since cells treated with heparin at a concentration higher than 800 µg/mL showed cell apoptosis (as observed through a microscope), we used heparin at 800 µg/mL in the subsequent experiments. treatment with or without 0.2 µg/mL PLA1A. As shown in Figure 4B, heparin concentrations less than 200 µg/mL enhanced PLA1A-mediated IL-8 production. At higher concentrations (400, 800, and 1600 µg/mL), heparin showed an inhibitory effect on IL-8 production. Since cells treated with heparin at a concentration higher than 800 µg/mL showed cell apoptosis (as observed through a microscope), we used heparin at 800 µg/mL in the subsequent experiments. Heparin was added to cell culture 30 min prior to PLA1A. The amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 produced by FLSs cultured in serum-free DMEM was 5.37 ± 1.43 pg/mL, and for each experiment, basal IL-8 production was set at 100% for data normalization. The data were the means ± SE from three experiments.
Pro-Inflammatory Effects of LysoPS and PLA1A in FLSs
PLA1A specifically hydrolyzed surface-exposed PS to produce lysoPS, which can bind and activate the lysoPS receptors in an autocrine or a paracrine manner. To determine whether the pro-inflammatory effects induced by PLA1A in cultured FLSs were dependent on the production of lysoPS and the activation of its receptors, we performed semi-quantitative RT-PCR to examine the expression of lysoPS receptors (GRP34, GPR174, and P2RY10) in human primary FLSs. As shown in Figure 5A, none of the lysoPS receptor transcripts was detected under our experimental conditions. In contrast, we detected the expression of GRP34, GPR174, and P2RY10 mRNAs in the human CD4 + T memory cells used as a positive control.
Pro-Inflammatory Effects of LysoPS and PLA1A in FLSs
PLA1A specifically hydrolyzed surface-exposed PS to produce lysoPS, which can bind and activate the lysoPS receptors in an autocrine or a paracrine manner. To determine whether the pro-inflammatory effects induced by PLA1A in cultured FLSs were dependent on the production of lysoPS and the activation of its receptors, we performed semi-quantitative RT-PCR to examine the expression of lysoPS receptors (GRP34, GPR174, and P2RY10) in human primary FLSs. As shown in Figure 5A, none of the lysoPS receptor transcripts was detected under our experimental conditions. In contrast, we detected the expression of GRP34, GPR174, and P2RY10 mRNAs in the human CD4 + T memory cells used as a positive control.
Effects of PLA1A and ATX in the Presence of Albumin
According to our data, the pro-inflammatory effects of PLA1A relied on the activity of ATX provided by FLSs. To define whether exogenous ATX promotes further cytokine production, we incubated the cultured FLSs with PLA1A (0.2 µg/mL) and recombinant ATX (5 nM) in combination. The basal level of IL-8 released by unstimulated FLSs cultured in media supplemented with or without 1% fatty acid-free BSA was 7.08 ± 2.45 and 8.05 ± 2.04 pg/mL, respectively. As illustrated in Figure 6A, in albumin-free condition, the addition of recombinant ATX alone had no effect on the production of IL-8, and PLA1A , and heparin were added to cell culture 30 min prior to LPA, lysoPS, and PLA1A. The amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 produced by FLSs stimulated with LPA, lysoPS, and PLA1A was 18.17 ± 7.19, 55.65 ± 22.6, and 23.11 ± 7.44 pg/mL, respectively. For each experiment, stimulated IL-8 production without inhibitors was set at 100% for data normalization. The data are the means ± SE from three experiments. For statistical comparative analyses, cytokine amounts in cells stimulated with LPA or lysoPS or PLA1A were compared to that in cells stimulated in the presence of Ki16425 or HA130 or heparin. ** p < 0.01; *** p < 0.001.
Effects of PLA1A and ATX in the Presence of Albumin
According to our data, the pro-inflammatory effects of PLA1A relied on the activity of ATX provided by FLSs. To define whether exogenous ATX promotes further cytokine production, we incubated the cultured FLSs with PLA1A (0.2 µg/mL) and recombinant ATX (5 nM) in combination. The basal level of IL-8 released by unstimulated FLSs cultured in media supplemented with or without 1% fatty acid-free BSA was 7.08 ± 2.45 and 8.05 ± 2.04 pg/mL, respectively. As illustrated in Figure 6A, in albumin-free condition, the addition of recombinant ATX alone had no effect on the production of IL-8, and PLA1A alone stimulated the production of IL-8 by approximately 2.5-fold. The combination of PLA1A and ATX increased the IL-8 production by 3.8-fold compared to non-stimulated cells, by 4.1-fold compared to ATX-treated cells, and by 1.5-fold compared to cells activated by PLA1A alone. However, differences were not statistically significant (p-value > 0.05). The addition to the culture media of 1% fatty acid-free BSA, which protects LPA from degradation and increases its half-life, enhanced the production of IL-8 in cultured FLSs. Compared to FLSs incubated in albumin-free media ( Figure 6A), the presence of albumin ( Figure 6B) increased ATX-induced IL-8 production by 2.3-fold, increased PLA1A-induced IL-8 secretion by 1.5-fold, and increased combination of ATX and PLA1A-induced IL-8 release by 1.8-fold. alone stimulated the production of IL-8 by approximately 2.5-fold. The combination of PLA1A and ATX increased the IL-8 production by 3.8-fold compared to non-stimulated cells, by 4.1-fold compared to ATX-treated cells, and by 1.5-fold compared to cells activated by PLA1A alone. However, differences were not statistically significant (p-value > 0.05). The addition to the culture media of 1% fatty acid-free BSA, which protects LPA from degradation and increases its half-life, enhanced the production of IL-8 in cultured FLSs. Compared to FLSs incubated in albumin-free media ( Figure 6A), the presence of albumin ( Figure 6B) increased ATX-induced IL-8 production by 2.3-fold, increased PLA1A-induced IL-8 secretion by 1.5-fold, and increased combination of ATX and PLA1A-induced IL-8 release by 1.8-fold. Amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 produced by FLSs cultured in serum-free DMEM supplemented with or without 1% fatty acid-free BSA was 7.08 ± 2.45 and 8.05 ± 2.04 pg/mL, respectively. For each experiment, basal IL-8 production was set at 100% for data normalization. The data are the means ± SE from five (A) and three (B) experiments, respectively. For statistical comparative analyses, cytokine amounts in cells stimulated with PLA1A and/or ATX were compared to that in non-stimulated cells. ** p < 0.01.
Discussion
High PLA1A serum level and tissue expression are associated with cancer metastasis, acute coronary syndrome, and autoimmune diseases, such as SLE. The current study is the first to explore the pro-inflammatory effects of PLA1A in RA disease processes, using primary human FLSs. In detail, we show that (1) PLA1A was elevated in the plasma of EA patients and synovial fluids from RA patients; (2) even if PLA1A mRNA was not detected in FLSs, a high level of cell-associated PLA1A was detected in FLSs cultured in medium supplemented with FBS; (3) PLA1A hydrolyzed PS exposed on the outer leaflet of FLSs and enhanced secretion of IL-8 through a pathway independent of lysoPS receptors; (4) the pro-inflammatory effects of PLA1A and lysoPS on FLSs were induced through the ATX-LPA receptor (LPAR) axis; and (5) the combined effects of PLA1A and ATX on cytokine secretion were more than additive, and the addition of fatty acid-free BSA during cell stimulation further enhanced IL-8 production by FLSs.
In our study, we first observed elevated concentrations of PLA1A in plasma of EA and synovial fluids from RA patients. PLA1A was increased in the circulation (49.5-fold increase in EA patients compared to healthy donors) and in the joint compartment (3.5fold increase in RA patients compared to OA patients). There was no gender difference of plasma PLA1A concentration among EA patients. In clinically stable SLE patients, the plasma concentration of PLA1A was slightly higher in females. Differences in PLA1A concentration between active autoimmune disease and steady or non-autoimmune disease were also remarkable, as highlighted in a previous clinical study with SLE patients [8]. Amounts of IL-8 released in the supernatants were monitored using ELISA. IL-8 produced by FLSs cultured in serum-free DMEM supplemented with or without 1% fatty acid-free BSA was 7.08 ± 2.45 and 8.05 ± 2.04 pg/mL, respectively. For each experiment, basal IL-8 production was set at 100% for data normalization. The data are the means ± SE from five (A) and three (B) experiments, respectively. For statistical comparative analyses, cytokine amounts in cells stimulated with PLA1A and/or ATX were compared to that in non-stimulated cells. ** p < 0.01.
Discussion
High PLA1A serum level and tissue expression are associated with cancer metastasis, acute coronary syndrome, and autoimmune diseases, such as SLE. The current study is the first to explore the pro-inflammatory effects of PLA1A in RA disease processes, using primary human FLSs. In detail, we show that (1) PLA1A was elevated in the plasma of EA patients and synovial fluids from RA patients; (2) even if PLA1A mRNA was not detected in FLSs, a high level of cell-associated PLA1A was detected in FLSs cultured in medium supplemented with FBS; (3) PLA1A hydrolyzed PS exposed on the outer leaflet of FLSs and enhanced secretion of IL-8 through a pathway independent of lysoPS receptors; (4) the pro-inflammatory effects of PLA1A and lysoPS on FLSs were induced through the ATX-LPA receptor (LPAR) axis; and (5) the combined effects of PLA1A and ATX on cytokine secretion were more than additive, and the addition of fatty acid-free BSA during cell stimulation further enhanced IL-8 production by FLSs.
In our study, we first observed elevated concentrations of PLA1A in plasma of EA and synovial fluids from RA patients. PLA1A was increased in the circulation (49.5-fold increase in EA patients compared to healthy donors) and in the joint compartment (3.5-fold increase in RA patients compared to OA patients). There was no gender difference of plasma PLA1A concentration among EA patients. In clinically stable SLE patients, the plasma concentration of PLA1A was slightly higher in females. Differences in PLA1A concentration between active autoimmune disease and steady or non-autoimmune disease were also remarkable, as highlighted in a previous clinical study with SLE patients [8]. Our data showed that compared to clinically stable SLE patients, plasma PLA1A in EA patients was increased by 7.4-fold. These data suggested that the elevated concentrations of PLA1A might contribute to the inflammatory conditions in the EA patients. In this regard, PLA1A could be a promising biomarker to monitor inflammatory disease activity [8]. In the study by Sawada et al., there was no significant increase in PLA1A level in the serum of RA patients [8]. However, our study monitored PLA1A in EA patients (symptoms ≤ 12 months) who had not received any disease-modifying antirheumatic drugs. Furthermore, joint manifestations in patients with early disease are difficult to distinguish from other forms of inflammatory polyarthritis. It is not excluded that the early stage of the disease and uncontrolled inflammation could contribute to the high plasma PLA1A level in EA patients.
In healthy donors, we detected 0.09 ± 0.09 ng/mL of PLA1A in the platelet-free plasma, whereas 33.8 ± 16.6 µg/L of PLA1A in serum were measured using a newly developed enzyme immunoassay in a previous study [37]. The cellular source of plasma PLA1A is yet to be established. We cannot exclude a role for activated platelets as a source of plasma PLA1A in humans. Although PLA1A mRNA was not detectable in human platelets, the transcripts were present in skeletal muscle, kidney, small intestine, spleen, testis, and in several human cell lines [38]. Extracellular PLA1A could be associated with platelets or stored intracellularly, as reported for ATX [39]. In rat platelets, PLA1A stored in α-granules is secreted into the extracellular environment upon activation with thrombin [1,40]. In addition, the liver tissue might be a source of human serum PLA1A. The serum PLA1A level in patients with liver injury (hepatitis, cirrhosis) was increased in comparison to that of healthy subjects [41]. Furthermore, to show that liver tissue could be the source of serum PLA1A, Uranbileg et al. studied patients with hepatocellular carcinoma, a condition associated with higher levels of serum PLA1A in comparison to healthy subjects [41]. In those patients, serum levels of PLA1A were correlated with the expression of PLA1A mRNA in the liver tissue without any carcinoma (or background tissue) but were not correlated with the PLA1A mRNA expression in carcinoma tissues [41]. According to our RT-PCR results, primary human FLSs do not express the PLA1A mRNA. However, cultured FLSs may internalize the PLA1A present in culture medium supplemented with FBS. The PLA1A in synovial fluids may derive from plasma leakage associated with inflammation or from immune cells that infiltrate synovial tissues.
PS has been widely involved in physiological and pathological conditions, such as blood coagulation, phagocytosis by macrophage, and activation of intracellular enzymes, such as protein kinase C [40]. Our results showed that TNF activation enhanced the exposure of PS in primary human FLS. This finding was consistent with previous reports suggesting that during platelet activation, cell apoptosis, and cytokine stimulation, PS externalized on the outer membrane leaflet is a potential substrate for secreted PLA1A [42]. Thus, PLA1A could hydrolyze externalized PS and contribute to IL-8 secretion through the ATX-LPA receptor axis (Figure 7). IL-8 is known to be a neutrophil chemotactic factor [43], and inhibition of IL-8 can reduce the neutrophil infiltration in RA joints [44]. The inhibition of IL-8 synthesis could also protect the cartilage and bone destruction and lead to disease suppression in collagen-induced arthritis model [45]. The N-terminal domain of PLA1A drives recognition and binding of the serine residue of PS and is responsible for its catalytic activity [38,40]. PLA1A belongs to the lipase family and has an affinity for heparin [46]. We should highlight that a low concentration of heparin enhanced PLA1A-mediated IL-8 secretion. Although the addition of heparin likely prevents PLA1A from binding to heparan sulfate lipid expressed by FLSs [47], we cannot exclude the possibility that PLA1A binding to heparan sulfate would modulate its catalytic activity. Our data highlighted that PLA1A had pro-inflammatory effects through hydrolyzing PS exposed on the surface of FLSs.
LysoPS receptors are expressed in immune cells. For instance, GPR34 is highly expressed in mast cells, whereas P2RY10 and GPR174 are expressed in lymphoid organs [15]. Our results showed that human FLSs expressed none of these three lysoPS receptors, thereby suggesting that lysoPS stimulated IL-8 secretion through an alternative pathway in FLSs and not through lysoPS receptor-induced signaling. ATX and LPA1/3 are widely expressed in human FLSs [36,48]. Indeed, our study showed significant inhibition of PLA1A-and lysoPS-mediated IL-8 production in the presence of an LPA1/3 antagonist (Ki16425) or an ATX inhibitor (HA130). Thereby, the data suggested that PLA1A and lysoPS mediated their pro-inflammatory effects through the ATX-LPAR axis. Thus, ATX metabolized extracellular PLA1A-derived lysoPS into LPA, which initiated a signaling cascade through activation of LPA1 and LPA3 expressed in FLSs [36].
ci. 2021, 22, x FOR PEER REVIEW 10 of 16 thereby suggesting that lysoPS stimulated IL-8 secretion through an alternative pathway in FLSs and not through lysoPS receptor-induced signaling. ATX and LPA1/3 are widely expressed in human FLSs [36,48]. Indeed, our study showed significant inhibition of PLA1A-and lysoPS-mediated IL-8 production in the presence of an LPA1/3 antagonist (Ki16425) or an ATX inhibitor (HA130). Thereby, the data suggested that PLA1A and lysoPS mediated their pro-inflammatory effects through the ATX-LPAR axis. Thus, ATX metabolized extracellular PLA1A-derived lysoPS into LPA, which initiated a signaling cascade through activation of LPA1 and LPA3 expressed in FLSs [36]. Figure 7. Schematic illustration of IL-8 production in FLSs stimulated with PLA1A. Activated FLSs exposed PS on outer leaflet of cell membrane and presented this specific lipid substrate to extracellular PLA1A to produce lysoPS, which was subsequently converted by additional or FLSs-secreted ATX to produce LPA. LPA stimulated IL-8 production in an autocrine manner through activating LPA1/3 associated signaling pathway. IL-8 is a neutrophil chemotactic factor that plays a role in inflammation. Added heparin competitively bound to PLA1A to prevent surface-exposed PS from hydrolyzation. ATX inhibitor HA130 and LPA1/3 antagonist Ki16425 decreased secretion of IL-8 in FLSs through inhibiting ATX-LPAR axis.
We also found that ATX and PLA1A in combination enhanced IL-8 secretion in FLSs compared to cells stimulated with ATX or PLA1A alone. In albumin-free condition, the combination of ATX and PLA1A can increase the production of IL-8 to a certain extent but cannot largely increase the IL-8 production as in the albumin-containing condition. It is a clinically significant observation since both PLA1A and ATX are present in RA synovial fluids, and synovial fluid is an albumin-containing biofluid. Surprisingly, in an incubation media devoid of albumin, the addition of ATX had no significant effect on IL-8 secretion by FLSs. Previous studies reported that ATX-derived LPA could activate cells in an autocrine/paracrine manner [49]. On the one hand, cells release a significant proportion of newly synthesized ATX in a vesicle-bound form rather than a soluble form [50]. ATX binds to exosomes through its NUC domain and is catalytically active [50]. On the other Figure 7. Schematic illustration of IL-8 production in FLSs stimulated with PLA1A. Activated FLSs exposed PS on outer leaflet of cell membrane and presented this specific lipid substrate to extracellular PLA1A to produce lysoPS, which was subsequently converted by additional or FLSs-secreted ATX to produce LPA. LPA stimulated IL-8 production in an autocrine manner through activating LPA1/3 associated signaling pathway. IL-8 is a neutrophil chemotactic factor that plays a role in inflammation. Added heparin competitively bound to PLA1A to prevent surface-exposed PS from hydrolyzation. ATX inhibitor HA130 and LPA1/3 antagonist Ki16425 decreased secretion of IL-8 in FLSs through inhibiting ATX-LPAR axis.
We also found that ATX and PLA1A in combination enhanced IL-8 secretion in FLSs compared to cells stimulated with ATX or PLA1A alone. In albumin-free condition, the combination of ATX and PLA1A can increase the production of IL-8 to a certain extent but cannot largely increase the IL-8 production as in the albumin-containing condition. It is a clinically significant observation since both PLA1A and ATX are present in RA synovial fluids, and synovial fluid is an albumin-containing biofluid. Surprisingly, in an incubation media devoid of albumin, the addition of ATX had no significant effect on IL-8 secretion by FLSs. Previous studies reported that ATX-derived LPA could activate cells in an autocrine/paracrine manner [49]. On the one hand, cells release a significant proportion of newly synthesized ATX in a vesicle-bound form rather than a soluble form [50]. ATX binds to exosomes through its NUC domain and is catalytically active [50]. On the other hand, secreted-ATX can interact with cell integrins through its SMB domain [51], which localizes ATX on cell surfaces [50,52] and contributes to the synthesis of LPA in the vicinity to its cognate receptors [53]. Besides, LPA can repress the expression and activity of ATX, and LPA binding to delipidated albumin could not affect this inhibitory effect [54,55]. However, a high concentration of ATX substrate and inflammatory cytokines could attenuate the inhibitory feedback loop of LPA [55]. Our data suggest that ATX-derived LPA should be either transported to activate its receptors and/or should be protected from degradation by cell-surface lipid phosphate phosphatases [51]. Our results also illustrate the importance of albumin in the maintenance of LPA functional responses during cell activation.
Unsaturated LPA species are produced by PLA1 isozymes and have been reported to be more potent than saturated LPA [33,46] in inducing proliferation [56], de-differentiation [42], and phenotypic modulation [57] of smooth muscle cells both in vivo and in vitro. ATX prefers lysophospholipids with unsaturated fatty acids according to the structure of the hydrophobic lipid-binding pocket [51,58]. Previous studies on the structure-activity relationship of cloned LPA receptors have reported differential activation of LPA receptors by different LPA species [33]. These differences include fatty acid length, degree of unsaturation, and linkage to the glycerol backbone. Cells and tissues express distinct patterns of LPA receptors. For example, LPA3 shows a higher affinity for C18:1, 18:2, and 18:3 LPA species compared to 20:4 LPA, whereas LPA1 shows a broad ligand specificity for either saturated or unsaturated LPA species [33]. These findings underline the potential importance of PLA1A as a source of unsaturated lysophospholipid species in various pathological conditions. Nevertheless, the origin of secreted PLA1A and the functions of PLA1A in healthy conditions and chronic rheumatic autoimmune diseases need further investigation.
Human Plasma and Synovial Fluid Samples
The CHU de Québec-Université Laval SARD Biobank and Data Repository provided the platelet-free plasma samples (12 healthy donors, 38 EA patients, and 62 undergoing treatment SLE patients). Patients with disease duration ≤ 12 months were considered EA patients. Human synovial fluid samples were obtained from five RA patients, three PsoA patients, three OA patients, and three gout patients. The CHU de Québec's Ethics Committee approved the study. All samples were collected after informed consent and were stored at −80 • C until the measurements.
Cell Treatment and Viability
FLSs seeded in 24-well plates were starved with serum-free medium for 24 h. Cell treatments were in fresh serum-free medium containing various concentrations of the tested compounds. Where indicated, Ki16425, HA130, and heparin were added to the culture medium 30 min before cell activation with LPA, lysoPS, and PLA1A. Cell supernatants were collected after 24 h and stored at −80 • C until IL-8 measurements by ELISA. Cell viability was evaluated by PI using flow cytometry. Cells were detached using trypsin/EDTA and incubated with PI (5 µg/mL). PI negative FLSs were considered viable.
ELISA
Synovial fluid samples were treated with 0.5 mg/mL hyaluronidase for 5 min at room temperature before measurement. PLA1A levels in plasma and synovial fluid samples and IL-8 levels in cell culture supernatant samples were monitored in duplicate, according to the manufacturer' protocol together with a standard curve (0.25-8 ng/mL for plasma samples: 78.1-5000 pg/mL for synovial fluid samples and 12.5-800 pg/mL for cell culture supernatant samples). Optical densities were determined using a SoftMaxPro40 plate reader at 450 nm.
Flow Cytometry
For detection of PLA1A, FLSs were cultured in a medium indicated with or without serum for 24 h and cells were detached using accutase cell detachment solution. Suspended cells were blocked with 1% fatty-acid free albumin and fixed with 0.5% PFA for 20 min at room temperature in the dark. Permeabilized cells (10 µg/mL digitonin) were stained with anti-PLA1A (1 µg for 10 6 cells) or rabbit IgG isotype control (1:5000 dilution) for 40 min on ice, followed by PE-conjugated anti-rabbit IgG secondary antibody (1:50 dilution) for 30 min on ice. For measurements of externalized PS, FLSs were stimulated with TNF for indicated times and were detached using accutase cell detachment solution. Suspended cells were stained with PI (5 µL for 10 5 cells) and FITC Annexin V (5 µL for 10 5 cells) for 15 min at room temperature. FITC Annexin V-positive and PI-negative identified the PS-positive viable cells.
Semi-Quantitative RT-PCR
Total RNA from human primary FLSs, CD4 + T cells, and testis tissue (provided by Dr. Belleannée of CHU de Québec) were isolated using Ribozol and reverse-transcribed into cDNA with EcoDry premix. Semi-quantitative RT-PCR was conducted using human oligo primer sequences (Table 1) and SYBR Green SuperMix. Amplification conditions were as follows: 40 cycles at 95 • C (denaturation, 15 s), 56.7 • C (annealing, 30 s), and 60 • C (extension, 30 s). Table 1. List of primers used in this work.
Western Blot
Where indicated, FLSs were cultured with or without serum for 24 h and detached using accutase cell detachment solution. Detached cells were lysed in boiling 2× Laemmli sample buffer (125 mM Tris/HCL (pH 6.8), 17.5% (v/v) glycerol, 10% (v/v) betamercaptoethanol, 8% (v/v) SDS, 0.025% bromophenol blue, 5 mM sodium orthovanadate, 0.025 mg/mL aprotinin-leupeptin protease inhibitor) for 7-10 min. Equal amounts of protein samples were separated using 12% SDS-polyacrylamide gel electrophoresis and transferred to methanol activated Immobilon PVDF membranes. Membranes were blocked with 5% skim milk at room temperature for 1 h and incubated with PLA1A antibody (overnight at 4 • C) or GAPDH antibody (one hour at room temperature), followed by appropriate HRP-conjugated secondary antibodies (30 min at room temperature). The Western Lightning chemiluminescence reagent, used according to the manufacturer's instructions, visualized the antibody-antigen complexes.
Statistical Analysis
Unless otherwise stated, data were from three independent experiments and the results were expressed as mean values ± SE. All statistical analyses used GraphPad Prism 8. Student t-test (two-tailed p-value) assessed the statistical significance between two experimental conditions (treated vs control). Multiple comparisons used one-way ANOVA test. A p-value less than 0.05 was considered statistically significant.
Conclusions
In summary, our study demonstrated for the first time that the inflammatory responses induced by PLA1A were dependent on the ATX-LPAR axis in cultured human primary FLSs. As presented in the image (Figure 7), PS exposed on the outer leaflet of FLSs can be hydrolyzed by extracellular PLA1A to produce lysoPS, a substrate of ATX. Then ATX-derived LPA stimulated the production of IL-8 in an autocrine manner through LPA1/3-associated signaling pathways. IL-8 is a neutrophil chemotactic factor that plays a role in inflammation. The high level of plasma PLA1A in EA patients suggested a possible implication of PLA1A in RA pathogenesis and qualified PLA1A as a potential biomarker and therapeutic target in chronic rheumatic autoimmune diseases. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: All data are included in the manuscript. | 2021-12-02T16:23:22.411Z | 2021-11-24T00:00:00.000 | {
"year": 2021,
"sha1": "51d3757a4d657e4a3d3616fcd28801aa3794a7af",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/22/23/12685/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "64b790d7215ff007f5e1775be2a9a19419ba915c",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
7776474 | pes2o/s2orc | v3-fos-license | The prevalence of women’s emotional and physical health problems following a postpartum haemorrhage: a systematic review
Background Postpartum Haemorrhage (PPH) is a leading cause of maternal mortality with approximately 225 women dying as a result of it each day especially in low income countries. However, much less is known about morbidity after a PPH. This systematic review aimed to determine the overall prevalence of emotional and physical health problems experienced by women following a postpartum haemorrhage. Methods Eight databases were searched for published non-randomised, observational, including cohort, primary research studies that reported on the prevalence of emotional and/or physical health problems following a PPH. Intervention studies were included and data, if available, were abstracted on the control group. All authors independently screened the papers for inclusion. Of the 2210 papers retrieved, six met the inclusion criteria. Data were extracted independently by two authors. The methodological quality of the included studies was assessed using a modified Newcastle Ottawa Scale (NOS). The primary outcome measure reported was emotional and physical health problems up to 12 months postpartum following a postpartum haemorrhage. Results Two thousand two hundred ten citations were identified and screened with 2089 excluded by title and abstract. Following full-text review of 121 papers, 115 were excluded. The remaining 6 studies were included. All included studies were judged as having strong or moderate methodological quality. Five studies had the sequelae of PPH as their primary focus, and one study focused on morbidity postnatally, from which we could extract data on PPH. Persistent morbidities following PPH (at ≥ 3 and < 6 months postpartum) included postnatal depression (13 %), post-traumatic stress disorder (3 %), and health status ‘much worse than one year ago’ (6 %). Due to the different types of health outcomes reported in the individual studies, it was possible to pool results from only four studies, and only then by accepting the slightly differing definitions of PPH. Those that could be pooled reported rates of acute renal failure (0.33 %), coagulopathy (1.74 %) and re-admission to hospital following a PPH between 1 and 3 months postpartum (3.6 %), an appreciable indication of underlying physical problems. Conclusion This systematic review demonstrates that the existence and type of physical and emotional health problems post PPH, regardless of the volume of blood lost, are largely unknown. Further large cohort or case control studies are necessary to obtain better knowledge of the sequelae of this debilitating morbidity. Electronic supplementary material The online version of this article (doi:10.1186/s12884-016-1054-1) contains supplementary material, which is available to authorized users.
Background
Postpartum Haemorrhage (PPH) is recognised as one of the leading causes of maternal mortality worldwide, accounting for 1 in 4 maternal deaths globally, but much less is known about morbidity post PPH. At least 225 women die from PPH every day, the majority of these deaths occurring in low income countries [1]. In the most recent Confidential Enquiries into Maternal Deaths in the UK, PPH was identified as being responsible for 10 % of direct maternal deaths [2]. Historically, maternal mortality has been used as an indicator of the quality of maternity services; however, as women's general health improves, and preventative care and treatment for PPH become more accessible, more women are surviving PPH, making the evaluation of morbidity a useful complementary measure.
Although PPH is itself a maternal morbidity, the outcomes for women who experience a PPH have not been studied widely. A surveillance study conducted in 27 obstetric units in Brazil found the maternal near miss ratio to be 2.2 per 1000 women with PPH [3]. Wagner et el. [4] describe a review of 11 studies, showing that PPH resulted in anaemia with a summary ratio of 2.35 (95 % CI 1.44-3.84), albeit with considerable heterogeneity. Emotional outcomes occur also, with the majority of women who experience PPH having negative memories of the birth including, for some, a persistent fear of dying [5]. A systematic review of women's experiences of all types of severe maternal morbidity showed that women found these incidents physically and emotionally distressing [6], and therefore worthy of study.
The prevalence of PPH varies, with the worldwide prevalence thought to be between 6 and 11 % [7]. In England, the incidence increased from 7 % in 2004/5 to 13 % in 2011/12 [8]. In an eleven-year population-based cohort study in Ireland, the rate of PPH increased from 1.5 % in 1999 to 4.1 % in 2009 [9]. In Australia (NSW), the PPH rate increased from 6.1 % in 2003 to 8.3 % in 2011 [10].
More recently, the rate in one large Dublin maternity hospital was 14.6 % in 2014, a 4-fold increase from the 3.3 % incidence reported in 2008 [11]. Whilst some of these increases may be a result of more accurate estimation of blood loss, or because PPH was redefined with a lower threshold of blood loss, it never-the-less means that more women are experiencing PPH.
Given the prevalence of PPH and approximately 135 million births occurring worldwide each year, approximately 8-11 million women will experience a PPH and many will suffer some health sequelae. The focus of this systematic review, therefore, was on determining the overall prevalence of emotional and physical health problems experienced by women following a postpartum haemorrhage.
Defining postpartum haemorrhage PPH is defined in four different ways by WHO [1], RCOG [12], ICD-10 codes [13] and NICE [14]. Primary (sometimes referred to as immediate or early) occurs within 24 h of birth, whilst secondary (sometimes referred to as delayed or late) occurs after 24 h and up to 12 weeks postpartum.
In these four definitions, the common component is the inclusion of blood loss greater than 500 mls for a primary PPH. For secondary PPH, only the ICD-10 codes [13] categorise blood loss as being greater than 500 mls. The terms used in two other definitions state 'abnormal or excessive bleeding from the birth canal' [12] / or 'of the amount that adversely affects the maternal physiology' [13]. WHO [1] or NICE [14] do not define secondary PPH.
Other terms used to define PPH include major obstetric haemorrhage, defined not only according to blood loss (>2500 mls), but also by number of transfused units of blood (n = 5) or requiring coagulopathy treatment (e.g., fresh frozen plasma, fibrinogen concentrate substitution therapy, platelets) [15,16].
However, despite the definition used, being able to estimate blood loss accurately is recognised as being almost clinically impossible [17], and studies have shown that blood loss is overestimated at low volumes and underestimated when the volume is large thus underestimating the severity of the PPH [18][19][20].
Defining maternal morbidity -women's physical and emotional health problems The term 'maternal morbidity' encompasses a wide range of acute and chronic conditions, ranging from 'near death' or 'near miss' to non-life threatening events. A WHO review, undertaken to create a consensus around a definition of maternal morbidity, found significant discrepancies in the literature and amongst experts [21]. More recently, population based studies have reported prevalence of severe maternal morbidity and, according to these reports, 1 in 140 pregnant women will suffer from a severe morbidity, the most common being PPH [16]. In Ireland, major obstetric haemorrhage was also the most common cause of severe maternal morbidity, with a rate of 2.55 per 1000 maternities [15]. These rates are for severe haemorrhage only, meaning that the prevalence of any degree of PPH in the wider maternity population is much greater.
Aim and objectives
This systematic review aimed to determine the overall prevalence of emotional and physical health problems experienced by women following a postpartum haemorrhage. The objectives were: To determine the prevalence of women's emotional health problems following a postpartum haemorrhage up to 3 months, >3 months and ≤6 months, and >6 months and ≤12 months postpartum; and To determine the prevalence of women's physical health problems following a postpartum haemorrhage up to 3 months, >3 months and ≤6 months, and >6 months and ≤12 months postpartum.
Outcomes
Primary outcome measures were the reported emotional and physical health problems up to 12 months postpartum following a postpartum haemorrhage. The type of haemorrhage experienced included: 1. Primary / immediate PPH; 2. Secondary / delayed PPH; and 3. PPH according to definition (as defined by WHO [1], RCOG [12], ICD-10 codes [13] and NICE [14]). When studies did not cite any of the above references, their definition was abstracted and assessed.
Secondary outcome measures were women's physical and/or emotional health problems following PPH according to parity, age, BMI and mode of birth and amount of blood lost/severity of PPH.
Inclusion and exclusion criteria
Inclusion and exclusion criteria were determined a priori. Eligible participants were postpartum women, irrespective of parity, who had given birth to a baby of at least 24 weeks gestation and who experienced a postpartum haemorrhage. An individual study had to report primary data on prevalence of emotional and/or physical health problems at any time point up to 12 months postpartum, to be included.
Published non-randomised, observational including cohort primary research studies, reporting on prevalence of emotional and/or physical health problems following a PPH were included. Intervention studies were included and data, if available, were abstracted on the control group.
Case reports, reviews and qualitative studies were excluded.
Search and selection strategy
A computerised search of the following electronic databases was performed from their foundation to April 2015: CINAHL , PubMED (MEDLINE not searched as included in PubMED), Maternity and Infant Care, Psy-cINFO, EMBASE (also includes deduplicated MEDLINE citations from 1966 to the present), Web of Science, Social Science Index and The Cochrane Library. Key words, search terms and string were developed, tested and adapted for each database. The search was not restricted by year or English language publications. Search terms were developed for the three key terms: 'postpartum haemorrhage' , 'postnatal period' and 'women's emotional and physical health' (Table 1). Search strings for individual databases were created (Additional file 1: Search Strings for Databases).
Following an initial scoping search, the search was limited to 'title' OR 'abstract'. The appropriate search terms were combined with the Boolean operands ' AND' and 'OR' as appropriate.
Reference lists of papers retrieved for full text review were also reviewed for potentially eligible papers that may not have been captured by the electronic search strategy [22].
Data management processes
The review followed the processes recommended by Higgins and Green [23]. Following agreement on the final search string, the selection of eligible studies and data extraction were performed by three people independently (MC, CB and DD), and the final results compared, thus minimising the likelihood that errors would go undetected. MC, CB and DD have expertise in midwifery, CB has systematic review and statistical expertise, DD has systematic review expertise.
Quality assessment of included studies
The Newcastle Ottawa Scale (NOS) [24] was modified and a scale of 0-6 stars was used to assess methodological quality of included studies ( Table 2). Six stars was deemed as strong, four to five stars was deemed as moderate and three or fewer stars was deemed as weak methodological quality. Studies judged to have weak methodological quality were excluded from the data extraction process. All items were regarded as equally important, and no weighting was applied to any item or category.
Data extraction management
The following data were abstracted and inserted in a data collection form, designed a priori.
Primary outcome measures
i. Country of study, study design, date of data collection, and methods used ii. Definition of PPH, if any, used.
The number of women reported as having a PPH, and who also reported a particular physical and/or emotional health problem, and the postpartum time it was reported.
Data analysis
Where possible, data reported on similar outcomes at the same time points were pooled.
Results of search and selection strategy
Following removal of duplicates, a total of 2210 papers were retrieved (Fig. 1), and 978 were excluded on title screening. A further 1111 were excluded on abstract screening, and the remaining 121 papers were subjected to full text review. Citations not in English were screened at this stage using Google translate. A total of 12 papers appeared to meet the inclusion criteria for this review [25][26][27][28][29][30][31][32][33][34][35][36]; however, at data extraction stage it was realised that we were unable to abstract data from six papers [26,29,[32][33][34][35] because the outcomes reported in these papers all related to management or treatments of PPH, and no maternal physical or emotional health outcomes were reported. Postnatal period TI postnatal OR AB postnatal OR TI post natal OR AB post natal OR TI post-natal OR AB post-natal OR TI postpartum OR AB postpartum OR TI post partum OR AB post partum OR TI post-partum OR AB post-partum OR TI "postnatal period" OR AB "postnatal period" OR TI "post natal period" OR AB "post natal period" OR TI "post-natal period" OR AB "post-natal period" OR TI "postpartum period" OR AB "postpartum period" OR TI "post partum period" OR AB "post partum period" OR TI "post-partum period" OR AB "post-partum period" OR TI puerperium OR AB Puerperium OR TI peripartum OR AB peripartum OR TI peri partum OR AB peri partum OR TI peri-partum OR AB peri-partum OR TI "peripartum period" OR AB "peripartum period" OR TI "peri partum period" OR AB "peri partum period" OR TI "peri-partum period" OR AB "peri-partum period" OR TI peripartum period OR AB peripartum period
Characteristics of included studies
The characteristics of the included studies are illustrated in Table 3. Five studies had the sequelae of PPH as their primary focus. The remaining study [27] focused on morbidity postnatally and we were able to extract data on the population of women who experienced a PPH. The majority of studies took place between 1995 and 2007, with one study conducted in 1989 (Table 3). Four studies collected data over 1 to 3 years, and two studies over a 5-9 year time-frame ( Table 3).
Results of methodological quality assessment
All included studies were judged as having strong or moderate methodological quality ( Table 3). The two studies judged as having only moderate methodological quality were deemed so because PPH and/or the postnatal period were not defined in the publication.
Results of data extraction
The six included papers reported data from six individual studies conducted in seven countries. Five studies were conducted in OECD countries and one in a non-OECD country. Detailed characteristics of included papers are presented in Table 3. Five studies reported the outcomes of primary PPH [25,27,28,30,31] and one reported outcomes after secondary PPH [36].
Definitions of PPH
Studies varied in how they defined PPH, and these were compared with the definitions provided by WHO [1], RCOG [12], ICD-10 codes [13] and NICE [14]. One study, which was reporting outcomes after severe or major haemorrhage, stating it as ≥1,500mls, defined PPH in the traditional manner, by amount of blood lost [27]. Two studies [28,30] used the ICD-9 codes to define PPH which also includes "the amount that adversely affects the maternal physiology, such as blood pressure and haematocrit", although these parameters are not defined. Three studies defined PPH in a somewhat similar fashion, including treatment required, e.g., "bleeding … persisting after manual exploration of the uterine cavity and requiring I.V. prostaglandin administration" [25], "findings of pelvic examination, condition of the uterus" [31], "a fall in haemoglobin, for example, to 7 g/dl and/or by 4 g/dl or more" [25,27].
Outcomes of PPH
One study reported on outcomes post secondary PPH, providing a definition similar to the RCOG's (2014) [36]. In total, 44 outcomes were reported in the six included papers (see Tables 4, 5 and 6 for a list and timeline of the health outcomes). The number of reported health problems varied from one (the rate of re-admission at ≥1 but <3 months postpartum [30]), to 31 [27]. Four studies reported health outcomes in the 1st postpartum month [25,28,31,36], two at ≥1 < 3 [27,30], and one at ≥3 < 6 [27]. Five reported outcomes following primary PPH [25,27,28,30,31] and one following both primary and secondary PPH [36]. The health outcomes reported in the early postpartum period tended to be acute, i.e., renal failure, respiratory failure, prolonged ventilation, coagulopathy/Disseminated Intravascular Coagulation (DIC), sepsis, anaemia, Superficial Venous Thrombosis and perforation of the uterus. Health problems reported in later months tended to be less severe, and some had become chronic.
Thompson et al. [27] was the only study that reported the prevalence of both emotional and physical health problems following a significant (>1,500mls) primary PPH between one and six months postpartum (Tables 5 and 6). The emotional health problems reported were anxiety, postnatal depression, fatigue, post-traumatic stress disorder and general health status, measured using Speilberger State-Trait Anxiety Inventory, Edinburgh Postnatal Depression Scale, Milligan's 10-item Scale for fatigue, 17-item PTSD Checklist and 360 item short form General Health Survey. The physical problems reported were infection (perineal, caesarean section (CS) wound, uterine, breast), pain (perineal, site of CS scar), urinary tract infection, incontinence (stress urinary, faeces, flatus), constipation, backache, headaches (frequent) and physical exhaustion. Women rated these as 'not a problem' , 'a minor problem' , or 'a major problem'.
Only two studies compared women who had primary PPH with those who did not. Chauleur et al. [25] found that women who experienced PPH were at increased risk of superficial venous thrombosis (adjusted relative risk: 5.3, 95 % CI 1.6 to 17). Hoveyda & MacKenzie [36] analysed data from 657 women with primary PPH and found 33 (5.02 %) had a secondary PPH compared with only 99 out of 18,479 women who did not have primary PPH (0.54 %), giving an odds ratio of 9.3 (95 % CI 6-2 to 14). One other study [27] provided data on those women re-admitted to hospital following a PPH at ≥ 1 month and < 3 months postpartum (4108 out of 113,861, 3.61 %). We compared these results with their data on the cohort of women who did not have PPH and were re-admitted (30,229 out of 2,538,865, 1.19 %). The difference was statistically significant with an odds ratio of 3.1 (95 % CI 3.01 to 3.21), p < 0.0001.
Data on sub-groups of women, grouped by parity, age and BMI, following postpartum haemorrhage were not reported in any paper. Due to the different types of health outcomes reported in the individual studies, it was possible to pool results from only four studies, and only then by accepting the slightly differing definitions of PPH. Combining the results of Bateman et al. [28] [31] was conducted on 50 women who had primary PPH, of whom 10 (20 %) had a subtotal hysterectomy performed, a population rate of peripartum hysterectomy (1.42 %) far in excess of the prevalence found in other countries. It is likely that the high incidence of acute renal failure and other complications seen in this study are due to the surgical procedure, rather than PPH per se. The benefit of combining these results is that a more realistic overall prevalence is reached. The combined results of Liu et al. [30] and Thompson et al. [27] in relation to those re-admitted to hospital following a PPH at ≥ 1 month and < 3 months postpartum, gave a total of 4125 re-admissions out of 114,032 (3.6 %).
Discussion
There are few papers that report on health outcomes following a postpartum haemorrhage, despite the reported increase in the incidence of PPH and more severe PPH, across the globe [8][9][10]. Many papers focus only on the risks for, and prevention of, maternal mortality due to PPH, with scant regard for surviving women's physical From the 2210 papers identified in this review, only six reported on outcomes following PPH. Of these, we were unable to pool and meta-analyse most data because of differing definitions of PPH, differing timing of measurement or differing outcomes measured. Those that could be pooled demonstrated a 0.33 % rate of acute renal failure and a 1.74 % rate of coagulopathy. The readmission rate to hospital following a PPH between 1 and 3 months postpartum was 3.6 %, an appreciable indication of underlying physical problems. Single studies included in this review showed other types of morbidity to be common; for example, depression (13 %) and a health status at 6 months postpartum, as measured by SF36, as 'much worse than 1 year ago' (6 %) [27].
Few studies of emotional sequelae of PPH within the six month postpartum period were found. Some work has shown that 68 % of 68 women surviving severe PPH had negative memories of the event (such as fear of dying (35 %)) when interviewed 4 to 17 years later) [5]. Sixty percent of the 15 women who went on to have another pregnancy suffered severe anxiety throughout, indicating the tenacity of such feelings. If emotional morbidity from PPH could be detected in the early postpartum months, and treated, it is possible that such long-term psychological disability could be averted. The first step, however, is to measure the prevalence and severity of such morbidity and an observational study is in progress in France that may provide more information in this area [37].
Secondary PPH is poorly defined, yet it seems illogical to define it as a blood loss of 500 mls, the amount used to define primary PPH, because this is considered to be in excess of what is considered a physiologically normal blood loss during the third stage of labour. We suggest that the definition should more correctly be: "any blood loss from the gentital tract in excess of normal lochia at any time period after 24 h post birth to 6 weeks postpartum." It is clear that more research is needed, to ascertain the amount and degree of both physical and emotional postpartum morbidity following PPH. Before that, though, it is important to develop a core outcomes set [38] for measuring maternal health outcomes postpartum, perhaps as a subset of the maternity core outcome set developed by Devane et al. [39].
Strengths and limitations of this review
The findings from this review are based on a comprehensive systematic literature search of seven electronic databases, and rigorous critical appraisal of included studies. Two studies with large sample sizes [28,30] influenced the overall results. Unfortunately there was high heterogeneity between all studies, which is a major limitation. We tried to be as inclusive as possible, but we were unable to pool and meta-analyse most of the findings, thus limiting the review's potential to add new knowledge to clinical practice. Despite this limitation, the findings demonstrate that the prevalence of physical and emotional health problems experienced by women post PPH remain largely unknown, indicating the need for further research in this area.
Conclusion
This review demonstrates that there are few studies examining the prevalence of maternal physical and health problems post PPH, and those that were found tended to be either small and/or focussed on a limited number of morbidities. PPH, estimated to affect as many as 8-11 million women annually, is likely to impact negatively on women's health, and it is conceivable that existing health problems experienced may be exacerbated. As these health problems remain virtually unknown, however, many women may be suffering sequelae from PPH during the postpartum period in the community, without proper diagnosis or treatment. Further research is required, but because of the variations in definitions used, and outcomes reported, studies will require a clear definition of PPH, and valid and reliable measuring instruments, based on a core outcome set, to ascertain the timing, prevalence and severity of health problems.
Additional file
Additional file 1: Search strings for databases. (DOCX 24.7 kb) | 2018-04-03T03:42:25.205Z | 2016-09-05T00:00:00.000 | {
"year": 2016,
"sha1": "d69ddd560d9ae5b9939410c0c95838d378818f98",
"oa_license": "CCBY",
"oa_url": "https://bmcpregnancychildbirth.biomedcentral.com/track/pdf/10.1186/s12884-016-1054-1",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d69ddd560d9ae5b9939410c0c95838d378818f98",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
238222972 | pes2o/s2orc | v3-fos-license | Virtual reality for relatives of ICU patients to improve psychological sequelae: study protocol for a multicentre, randomised controlled trial
Introduction Intensive care unit (ICU) admission of a relative might lead to psychological distress and complicated grief (post-intensive care syndrome–family; PICS-F). Evidence suggests that increased distress during ICU stay increases risk of PICS-F, resulting in difficulty returning to their normal lives after the ICU experience. Effective interventions to improve PICS-F are currently lacking. In the present trial, we hypothesised that information provision using ICU-specific Virtual Reality for Family members/relatives (ICU-VR-F) may improve understanding of the ICU and subsequently improve psychological well-being and quality of life in relatives of patients admitted to the ICU. Methods and analysis This multicentre, clustered randomised controlled trial will be conducted from January to December 2021 in the mixed medical-surgical ICUs of four hospitals in Rotterdam, the Netherlands. We aim to include adult relatives of 160 ICU patients with an expected ICU length of stay over 72 hours. Participants will be randomised clustered per patient in a 1:1 ratio to either the intervention or control group. Participants allocated to the intervention group will receive ICU-VR-F, an information video that can be watched in VR, while the control group will receive usual care. Initiation of ICU-VR-F will be during their hospital visit unless participants cannot visit the hospital due to COVID-19 regulations, then VR can be watched digitally at home. The primary objective is to study the effect of ICU-VR-F on psychological well-being and quality of life up to 6 months after the patients’ ICU discharge. The secondary outcome is the degree of understanding of ICU treatment and ICU modalities. Ethics and dissemination The Medical Ethics Committee of the Erasmus Medical Centre, Rotterdam, the Netherlands, approved the study and local approval was obtained from each participating centre (NL73670.078.20). Our findings will be disseminated by presentation of the results at (inter)national conferences and publication in scientific, peer-reviewed journals. Trial registration number Netherlands Trial Register (TrialRegister.nl, NL9220).
INTRODUCTION
An intensive care unit (ICU) admission is known to be a stressful experience for both patients and their relatives. As a result, relatives of ICU patients are at risk of developing several psychological sequelae, such as symptoms of post-traumatic stress disorder (PTSD), anxiety, depression and complicated grief in the unfortunate event of a patient dying during ICU treatment. Clinically relevant symptoms of PTSD occur in 21% of relatives of ICU patients, especially in relatives of adult patients, clinically relevant symptoms of anxiety occur in 40% and clinically relevant symptoms of depression occur in 23%. [1][2][3][4][5][6][7][8][9][10][11] These impairments are collectively referred to as the post-intensive care syndrome-family (PICS-F). 6 12 13 PICS-F frequently results in loss of employment, financial burden, lifestyle interference and a profound impact on quality of life. 14 These consequences often last a long time and already start during ICU stay of their kin. 3 Important risk factors for the development Strengths and limitations of this study ► A randomised controlled trial examining the effect of Intensive Care Unit-specific Virtual Reality for Family members/relatives (ICU-VR-F) on psychological well-being and quality of life using an innovative and uniform modality. ► ICU-VR-F represents an easy applicable, safe and immersive modality to improve communication through better information provision regarding treatment-related and environment-related information about the ICU, enabling relatives to receive uniform and complete information. ► ICU-VR-F is an innovative method that is generalisable and makes information easily accessible and immersive. ► Blinding of patients or investigators is not possible due to the nature of the intervention.
Open access of PICS-F are the unexpectedness of critical illness, the dramatic nature of the relatives' experience leading to emotional stress, the level of communication of the ICU staff and the use of medical jargon, that frequently makes it hard for the relative to understand the treatment explanation. 1 8 10 11 15 16 As such, relatives may witness invasive treatments with unfamiliar medical procedures and devices in an environment they do not understand. Therefore, communication between ICU staff and families is essential in the care process, and good communication and information provision improve the relatives' understanding of ICU treatment and satisfaction, limit lawsuits and are associated with lower prevalence of PTSD during ICU stay. 5 17 18 As such, good information provision to relatives of ICU patients is essential in improving the relatives' comprehension of ICU procedures and ICU environment during the ICU stay. During the COVID-19 pandemic, many hospitals worldwide disallowed visitors for all adult inpatients including all ICU patients with COVID-19 and without COVID-19. Relatives of ICU patients during the COVID-19 pandemic are therefore confronted with the impracticality of visiting their relative in the ICU or receiving good communication from the ICU staff, which may result in a higher psychological burden. 19 20 In the face of mounting PICS-F-related sequelae, several interventions, such as information brochures, family conferences and educational programmes for relatives, have been tested, but did not result in a clinically meaningful improvement in psychological well-being or quality of life. 21 22 The COVID-19 pandemic has resulted in the disruption of an integral aspect of care in most ICUs across the world and the importance of generalisable and on-demand information has been addressed. To date, a clinically meaningful, simple and generalisable intervention remains unavailable.
Virtual reality (VR) is a relatively new technique that allows the user to fully immerse within a virtual environment. As such, it allows relatives to experience what the patient is experiencing during ICU treatment, possibly leading to a better comprehension of an ICU stay. Information provision using VR has shown to decrease preoperative anxiety in both adult and paediatric patients, to help women and their partner to feel better prepared for caesarean delivery, to successfully deliver healthcarerelated information to adults with intellectual disabilities and to be an appropriate tool to deliver additional treatment-related information to increase patients' satisfaction. [23][24][25][26] Additionally, exposure through VR appears to be an effective treatment modality for several mental health disorders, including PTSD, depression and anxiety, in a non-ICU setting. [27][28][29][30] It provides an innovative modality that is generalisable and could improve the relatives' understanding of what is happening to long-stay ICU patients, without increasing staff workload. We hypothesised that offering treatment-related and environment-related information about the ICU via VR increases relatives' understanding of ICU treatment and environment and improves psychological well-being and quality of life.
METHODS AND ANALYSIS Study design and setting
This study will be a multicentre, clustered randomised trial conducted in the mixed medical-surgical ICUs of four hospitals in Rotterdam, the Netherlands. Cooperating hospitals are: the Erasmus Medical Centre (MC) (university hospital), Franciscus Gasthuis & Vlietland Hospital, Ikazia Hospital and Maasstad Hospital (all teaching hospitals). The Medical Ethics Committee (MEC) of the Erasmus MC approved this study (NL73670.078.20, approved 14 December 2020), and local approval was obtained from the institutional ethics review boards of each participating hospital, that is, the Franciscus Gasthuis & Vlietland Hospital, the Ikazia Hospital and the Maasstad Hospital. The study will be conducted from January to December 2021. Participants will be followed for 6 months after patients' ICU discharge. Any modifications to the study protocol, which may impact the conduct of the study or participant safety, including changes of the study objectives, study design, study population, sample size, study procedures or significant administrative aspects, will be sent for approval to the MEC of the Erasmus MC, and local approval will be obtained from the institutional ethics review boards of each participating hospital prior to implementation. Accordingly, the health authorities will be informed in accordance with local regulations.
Study participants
We aim to include relatives, or close friends in absence of relatives, of 160 ICU patients. Relatives ≥18 years of age, who are a first/second-degree relative of the ICU patient, are responsible for decision-making or sharing the same household are eligible for inclusion. Additionally, relatives should be able to understand the Dutch language to understand ICU-specific Virtual Reality for Family members/relatives (ICU-VR-F) and should in possession of smartphone, tablet or computer to watch ICU-VR-F at home. Multiple relatives per patient can participate; the primary contact person of the ICU patient will be approached first and will be invited to share the study information with other relatives who could be interested in participation. There is no maximum number of relatives per patient who can participate. In the case of multiple relatives of the same patient participating, relatives of the same patient will be clustered to the same randomisation allocation. Relatives with no formal address or relatives of patients with an expected ICU length of stay less than 72 hours will be excluded. Close friends are eligible for inclusion in the case that no relative is available. Close friends are considered close friends if they address themselves as close friends and are responsible for decision-making. Relatives of patients who die during ICU treatment will retrospectively be excluded from the main analysis.
Open access
Intervention Patients will be randomised to receive standard care with additional ICU-VR-F (intervention group) or standard care alone (control group).
The ICU-VR-F intervention was based on the previously described ICU-VR intervention for ICU patients and was designed by an interdisciplinary team of three intensivists, a psychologist, a former ICU patient and a VR/film director. Based on focus group meetings with this team and previous studies, the following information was included in the intervention: (1) an introduction by an intensivist and an ICU nurse to welcome the relative to the ICU and VR environment explaining daily movements at an ICU, (2) explanation of monitors and noises in an ICU room, (3) information regarding mechanical ventilation, intubation and tracheal tube suction, (4) information and necessity of central/peripheral lines and intravenous/ drips, (5) information and necessity of the treatment team and ICU workflow. 31 32 The ICU-specific VR intervention was designed with the aim of showing relevant and truthful treatment-related and ICU environmentrelated information, and was hospital specific. The point of view for the camera was the field of vision of the mock patient lying in a hospital bed. The hospital-specific ICU-VR-F from the Erasmus MC can be found https:// www. youtube. com/ watch? v= OakhhQ32jLs, from the Franciscus Gasthuis & Vlietland can be found https:// www. youtube. com/ watch? v= cvMffRVrOE4, and from the Ikazia Hospital can be found https://www. youtube. com/ watch? v= cvMffRVrOE4. The uniform video script can be found in the online supplemental data.
Standard care comprises either (1) a family meeting with the treating ICU physician during the first week of ICU admission, and (2) biweekly meetings with the treating ICU physician when patients have a stay of more than 14 days according to a hospital's local protocol. Additionally, family members will always be offered a digital/ hardcopy ICU diary according to national guidelines.
Study procedures
Outcome variables will be collected at each time point (see figure 1). The primary contact person of the ICU patient will be approached by an investigator of the research team within 2 days after ICU admission and will be asked to share the study information with other relatives. In case that other relatives are interested in participation, their contact details will be shared by the primary contact person with the investigator so informed consent can be obtained. A translation of the information for participants and the informed consent form can be found in the online supplemental data. After inclusion, participants will receive a first set of questionnaires (T0) consisting of a self-composed questionnaire regarding demographics, and validated questionnaires to assess psychological well-being and quality of life. Participants are asked to fill in the first set of questionnaires retrospectively, in order to obtain a measure of participants' anxiety and depression levels and quality of life prior to the current episode of the patient's illness leading to ICU admission. Hereafter, randomisation will be conducted.
During ICU treatment, all relatives will receive standard care, which comprises either: (1) a family meeting with the treating ICU physician during the first week of ICU admission, and (2) biweekly meetings with the treating ICU physician when patients have a stay of more than 14 days. Additionally, family members will always be offered a digital/hardcopy of an ICU diary.
After randomisation, participants in the intervention group will additionally receive ICU-VR using headmounted display VR (Oculus Go, Irvine, California, USA, CE: R-CMM-OC8-MH-A). Thereafter, they receive cardboard VR glasses and an access link to watch ICU-VR-F at home, which can also be used without the cardboard VR glasses. Participants who are not allowed to visit the hospital due to COVID-19 regulations, that is, mandatory self-quarantine, inability to visit the ICU or a limited number of visitors, will only receive ICU-VR-F using cardboard VR glasses via the access link. The number of times a participant watches ICU-VR-F will be logged. Participants will have access to the intervention during Open access the entire study period, including follow-up. Participants will receive a second set of questionnaires during ICU discharge of their relative to assess their understanding of ICU procedures and environment, and will receive follow-up questionnaires at 1 month, 3 months and 6 months after ICU discharge (table 1).
The study procedures of participants in the intervention group who are allowed to visit the hospital are presented in figure 2 and for those who are not allowed to visit the hospital in figure 3.
Randomisation and masking
Randomisation will be on a 1:1 ratio, clustered based on the ICU patient (that is, if multiple relatives of one ICU patient participate, they will all be assigned to the same group), stratified for study site and the ability to visit the hospital with regard to COVID-19 regulations. Randomisation will be performed using a centralised internet-based randomisation procedure (Castor EDC, Amsterdam, the Netherlands). Due to the nature of the intervention, blinding is not possible.
Outcomes and measurements
The primary endpoint is the effect of ICU-VR-F on psychological well-being and quality of life in participants up to 6 months after ICU discharge. Psychological well-being will be expressed as the presence and severity of PTSD-related, anxiety-related and depression-related symptoms, and will be assessed using the Impact of Event Scale-Revised (IES-R) and Hospital Anxiety and Depression Scale (HADS). 33 34 Quality of life will be assessed using the RAND-36. 35 36 The secondary endpoint is the participants' understanding of the ICU environment and procedures, that is, devices, monitors, sounds (alarm noises) and daily work practice. Understanding of ICU procedures will be assessed using a subset of the Consumer Quality Index-Relatives in the ICU (CQI-Relatives in the ICU). 37 Additional outcomes are the perceived stress factors during ICU treatment and the perspectives of participants about ICU-VR-F, assessed using the Caregivers Strain Index (CSI), a self-composed 'perceived stress factors' questionnaire, and a self-composed 'perspectives on the ICU-VR intervention' questionnaire.
The IES-R comprises 22 items, assesses subjective distress caused by a traumatic event and has been previously validated in ICU survivors. 38 39 The IES-R yields a total score (ranging from 0 to 88, with higher scores indicating more severe symptoms), and subscale scores can be calculated for symptoms of intrusion, avoidance and hyperarousal. An IES-R sum score ≥24 will be considered as PTSD. 40 41 The HADS comprises 14 items and is commonly used to determine the levels of anxiety and depression that a person is experiencing. A sum score >8 on either the depression (seven questions) or anxiety (seven questions) subscale will be classified as depression and anxiety, respectively. 33 42 43
Open access
The Research and Development 36-item Questionnaire (RAND-36) consists of 8 scaled scores, which are the weighted sums of the questions in their section. Each scale is directly transformed to a scale ranging from 0 to 100 on the assumption that each question carries an equal weight. The eight sections are vitality, physical functioning, bodily pain, general health perception, physical role functioning, emotional role functioning, social role functioning and mental health. In addition, a mental and physical component scale can be calculated, giving a perception of a person's physical and mental health. 44 The CQI-Relatives in the ICU was designed by the Healthcare Institute of the Netherlands in collaboration with several hospitals to measure the perceived quality of care by relatives of ICU patients. 37 The subset used in the present study was carefully tailored to the needs of the current study (online supplemental data). Therefore, unnecessary items for this study were removed and additional VR-specific questions were added. The subset consists of 38 items, distributed across 4 sections; (1) general questions, (2) questions regarding information provision and understanding of the ICU environment, (3) questions regarding care offered to relatives and (4) questions regarding the communication with the ICU staff.
The self-composed perceived stress factors questionnaire was based on existing literature regarding risk factors for the development of PICS-F, including time spent for visitation, worries about the physical, cognitive and psychological state of the patient, worries about family and familiarity with an ICU. The final questionnaire comprises 18 questions which can be answered on a Likert scale ranging from 0 (not at all) to 4 (a lot). The self-composed perspectives on the ICU-VR-F intervention questionnaire comprises 13 questions. Outcomes of these self-composed questionnaires will be used to determine different aspects of information that relatives were missing or were in need of in the current ICU-VR-F intervention. These data will be used to further improve the VR intervention and its content so it will better meet the needs of relatives. Translations of the self-composed questionnaires can be found in the online supplemental data.
Data management Data will be uploaded, stored and maintained on the electronic data capture (EDC) system of Castor (Castor EDC ( www. castoredc. com), Amsterdam, the Netherlands). The study team will be responsible for all data entry and quality control activities. The data will be checked by at least two persons from the study team and will be stored for at least 15 years on either the Castor EDC server or as a hardcopy in the ICUs of the participating hospitals. Questionnaires will be sent digitally using Castor EDC or hardcopy via postal mail whenever requested.
To maintain anonymity, data will be coded with a number and this number will be the only reference to identification. The principal investigator will be the only one in possession of the translation key, making it impossible to link data to the participant.
Sample size calculation
To the best of our knowledge, this study will be the first of its kind for which no previous conducted studies can be used to define the expected effect estimate. Due to expected non-normality of PTSD, depression and anxiety scores at 6 months after ICU discharge, this calculation could represent an overestimation of the effect estimate. Based on our clinical experience, and experience with a pilot study studying the effects of ICU-VR on ventilated ICU patients for which we found Cohen's d effect size of 0.77, we expect that a clinically meaningful Cohen's d effect size of 0.55 could be expected in relatives. 32 When taking this into account, using a two-sided alpha of 0.05, and a power of 0.80, assuming an expected loss to follow-up of 20%, we aim to include relatives of 160 ICU patients. We expect a needed time of 6 months based on the admission rate history of the participating hospitals.
Statistical analysis
Baseline demographics and treatment-related characteristics will be quantified using descriptive statistics. Continuous variables will be presented as mean (SD) or as median (95% range), based on the distribution of the variable. Categorical variables will be presented as absolute number and relative frequency.
A sensitivity analysis will be performed in which missing data (completely) at random will be dealt with using both multiple imputation according to the Markov-chain Monte Carlo and the Last Observation Carried Forward Method. 45 46 We will correct for multiple testing using the false discovery rate with a maximum of 5% false negatives. 47 For the primary outcome, the effect of ICU-VR on PTSD, anxiety, depression and quality of life, we will analyse differences in the IES-R sum score (PTSD), the HADS anxiety and depression score, and the RAND-36 subscales (quality of life) between participants in the intervention and the control group at each follow-up time point (that is, 1 month, 3 months and 6 months after ICU discharge) and throughout follow-up using a mixed-effects linear regression model with a random intercept for each study site and/or participants based on model comparisons using the Akaiki information criteria. In case of multiple relatives of one ICU patient participating, these participants will be considered as clustered, and a random intercept for each cluster will be used. Between-group differences in variables of interest throughout follow-up were studies by introducing the product of time×treatment group to the model.
Differences in the proportion of participants in the intervention group and participants in the control group with clinically relevant symptoms of PTSD (IES-R sum score ≥22), depression (HADS depression score >8) or anxiety (HADS anxiety score >8) will be analysed using a mixed-effects logistic regression model. Also, changes Open access from baseline will be computed dividing the parameter value at specific time points into the baseline value expressed as percentile changes (% of baseline). The magnitude of change among PTSD, depression, and anxiety at specific time points and differences will be tested using a mixed-effects linear regression model.
For the secondary outcome, understanding of the ICU and quality of care in the ICU, we will analyse differences between study groups per question using a mixed-effects logistic regression model. By combining the numerical values of the answers given, a sum score and subscales for the different sections can be calculated for each participant. The association between the intervention and these sum scores will be examined using mixed-effects linear regression models.
The explorative outcomes, the perceived stress factors and the perspectives of relatives on the ICU-VR-F intervention, will be described using descriptive statistics. Differences in continuous outcomes of the self-composed questionnaire regarding perceived stress factors and the sum score of the CSI will be analysed using mixed-effects linear regression models. Differences in categorical outcomes of the self-composed questionnaire regarding perceived stress factors will be analysed using mixedeffects logistic regression models.
The main analyses will be conducted per protocol. In these, all patients who have received ICU-VR-F, either both in the hospital and at home or only at home, will be compared with those who did not, and relatives of patients who have died during ICU treatment will be excluded. To determine whether there is a difference in effect between having watched ICU-VR-F the first time in the hospital and having watched the ICU-VR-F only at home, we will use dummy variables (ICU-VR-F in the hospital and at home/ICU-VR-F only at home/no ICU-VR-F) instead of the randomisation variables in the mixed-effects regression models, and determine whether that dummy variable has a significant contribution to the model. We will additionally perform an analysis in which (1) patients who did not watch ICU-VR-F in the hospital will be excluded and (2) patients who watched ICU-VR in the hospital will be excluded to determine whether there is a difference in effect.
All data will be gathered using Castor EDC. Analyses will be performed using SPSS (V.27.0) and R for Statistics (R Foundation for Statistical Computing, Vienna, Austria, 2015). A p value of ≤0.05 will be considered statistically significant.
Ethics and dissemination
This study will be conducted in accordance with the principles of the Declaration of Helsinki (version October 2013; www. wma. net) and in accordance with the Medical Research Involving Human Subjects Act and other guidelines, regulations and acts. We received approval from the MEC of the Erasmus MC, and local approval has been obtained from the institutional ethics review boards of each participating hospital, that is, the Franciscus Gasthuis & Vlietland Hospital, the Ikazia Hospital and the Maasstad Hospital. If deviation from the protocol is necessary, then it will not be implemented without the prior review and approval of the MEC of the Erasmus MC and each participating hospital's institutional ethics review board. Signed informed consent will be obtained from all participants. Previous research demonstrated that (ICU-)VR is safe. 23 31 32 48 Informed consent forms will be kept in a locked cabinet in a limited-access room at the Erasmus MC. Data will be archived for 15 years. The handling of personal data complies with the Dutch law. On completion of the study, its findings will be published in peer-reviewed journals and presented at national and international scientific conferences to publicise the research to healthcare professionals, health service authorities and the public. A summary of the results will be made available to the study patients if requested.
Patient and public involvement
Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Intensive Care, Maasstad Hospital, Rotterdam, The Netherlands Contributors JHV, JvB, E-JW, DG and MEVG conceived the study and initiated the study design. MEVG is the coordinating investigator and grant holder. DG is the principal investigator. TK provided statistical expertise in the clinical trial design, and JHV and TK wrote the statistical analysis plan. JvB, E-JW, JAML and AFCS are the local principal investigators at each study site. All the authors contributed to the refinement of the study protocol and approved the final manuscript. JHV and MPVB wrote the first manuscript draft. JHV and MEH composed the questionnaires used in the study. JHV and MPVB will collect the data and conduct the study.
Funding This study was supported by DSW (for the HORIZON-IC Project; no grant number available), Stichting Theia (grant number: 2020286), Stichting SGS (grant number: 2020355) and BeterKeten (for the HORIZON-IC Project; no grant number available).
Disclaimer
The funding sources had no role in the design of the study and collection, analysis, and interpretation of data nor in writing the manuscript.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is | 2021-09-30T13:08:19.328Z | 2021-09-01T00:00:00.000 | {
"year": 2021,
"sha1": "c826cbcb30f91aed991dceb9c135800c4c7ab6d1",
"oa_license": "CCBYNC",
"oa_url": "https://bmjopen.bmj.com/content/bmjopen/11/9/e049704.full.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c826cbcb30f91aed991dceb9c135800c4c7ab6d1",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
55789537 | pes2o/s2orc | v3-fos-license | Study of the existing problems and public consciousness of the in-service military officers in Royal Thai Armed Forces Headquarters
The public consciousness is an important characteristic which requires serious enhancement in Thailand. The objective of this study is to examine the existing problems and the public consciousness of the inservice military officers. As a result shows, there is currently no precise training system for public consciousness training at the Royal Thai Armed Forces Headquarters. The tasks are carried out by the chain of command. A study of the existing problems and the public consciousness could lead to the development of a training model to enhance public consciousness. If the in-service military officers, who would return to their hometown all over the country after being discharged, were to have public consciousness characteristics as a foundation for everyday life, the nation would head up the force to help society.
Introduction
Materialism is one of the critical problems in Thai society.People prioritize owning things over having a spiritual mind [1].We can observe the behavior of some people in society who take advantage of others rather than thinking of public utilization of resources to the public.People compete to earn more and more to the point that virtue and morality of the people have declined.. This may indicate that Thai people now lack public consciousness [10].Community is weakening.It can be seen in everyday life.One example is the way that some individuals claim "ownership" of the public footpaths for themselves and their own monetary gain.Another example is throwing tissue in the public toilets and leaving toilet seats dirty.These acts cause problems for others [7].Public consciousness is an important characteristic which needs serious improvement because it a Corresponding author: aimmy_cd29@hotmail.comwould propel social responsibility.Public consciousness is essential, in both community level and national level.Since it is knowledge and understanding of reasonable thinking, a sense of ownership in a public share, an awareness of the rights and duties without exploiting others, including to maintain a common public benefits, as well as taking care of disadvantaged person as appropriate without contrary to the common laws.
The military has a great responsibility in maintaining national security and help relieve the misery of the people [9].One of the most important groups in the military is the inservice military officers.In former times, this group was the first to fight the "enemies" in the battlefield.Now it could be considered just as important as the first group who will be first in the field to help the community when there is a critical situation.For example, after the tsunami in 2004 and Thailand's great flood in 2011, the in-service military officers were the first group to go and assist people in the critical areas.In-service military officer is a unit of military.Enhancing public consciousness of In-service military officer is also one of the 12 core values.The Thai government has announced the 12 core values of as they want to foster characteristic which considered to be important and necessary for Thai people.Instead of relieving the misery of people, there were some in-service military officers caused the misery by themselves such as in-service military officer of Army Air Defense Command whom being accused by Min Buri Provincial Court, for the offense of premeditated murder [4], etc.Thus it can be assumed that in-service military officers were lacked of public consciousness.He could not discern what was right or wrong, what he should do or should not do.This showed that in-service military officers were not aware of the social problems.According to the study, there is none of the research, focusing on improving public awareness of in-service military officer.Therefore our research would like to emphasize such problems, to understand the conditions and the actual level of public consciousness of in-service military officer, in order to develop them to be the productive citizens who maintain good life and concern for public and national benefits, not the trouble making and social burden.
Methodology
This is a qualitative study conducted by analyzing documents and using in-depth interviews with the commander of a trained unit and a military trainer to examine their training model.However it is also a quantitative study with the use of a questionnaire given to the in-service military officers to identify the public consciousness level of in-service military officers.
Result and Discussion
Thai men at the age of 18 years old must report to military bases.Once selected, they pass through the processes of joining the armed forces.The officers will be trained for certain skills or characteristics.The important skills and characteristics include military characteristics, postures and manners.traditions, love for the nation, religion, and the king, and knowledge of democratic ideology, discipline and citizenship.What seems to be lacking is self responsibility, social responsibility, and public consciousness.
Thai men 18 years of age must be registered to be an in-service military officer.After examination and the selection process, military officers learn to be a soldiers.The new inservice military officers begin toning their bodies and minds from living as civilians to being soldiers.The main objectives are: 1) To understand the features and gestures of the soldiers to improve our familiarity with military life, as well as to learn how to behave in order to live and work together within the military.
2) To understand the traditions of the military.It is required to maintain military characteristic t all times to serve as a military officer while being a military reservist.
3) To cultivate their minds to have good morals and ideals of patriotism and religion and a democratic system with the king as head of state.
4) To foster good manners and discipline, and to be good citizens of the country.They also aim to be respected by the public and later serve as leaders after being discharged from the military.
5) To enhance a healthy body to endure the arduous and patient work and also perform in various functions within the whole military.
The duration of the in-service military training can be divided as follows: During the first to the third months the officers aim to learn the general military subjects to obtain knowledge of communication, use maps and a compass, learn how to provide first aid and good hygiene, keep informed with basic military news, learn how to disarm a bomb and protect nuclear weapons, and use weapons.They also learn about combat operations as individuals and as a group, employing camouflage and concealment, knowing the characteristics of the soldiers, military courtesy, military morals and discipline, and to be good citizens to love and defend the country.
During the fourth to the sixth months, they learn the specific duties of the agencies to be able to operate according to the requirements of the unit.
During the seventh to the eighteenth months they go to work in the unit itself.When they later have six months to two years left before their discharge, the in-service military officers will be discharged to return to their homeland.They can choose to study general education and/or vocational education depending on their interests and individual knowledge.The general education and vocational education have been provided courtesy of the Ministry of Education and Ministry of Labor to supply teachers and trainers to the training.The in-service military officers finish their compulsory education and also gain vocational training in various fields such as agriculture, industry and service to the profession.Moreover, They will gain knowledge in areas including democracy, civics and leadership, hygiene, health, nutrition, family planning, disaster prevention, drug abuse resistance, and conservation of national culture and art and antiques [3].
In addition, from the interviews with the commander of the trained units, there was no precise training system of public consciousness for the Royal Thai Armed Forces.The training depended on the lecturers alone who were responsible for this course.In-service military officers still lacked of awareness in real situations.They tended to follow the line of command, thus lacked learning collaboratively.They did not gain awareness of public consciousness.These results are only from the qualitative study.A report of the quantitative results is still in progress.
As a result, the in-service military officers have no awareness in real situations.The work is carried out by the chain of command.There tends to be a lack of collaboarative learning exercises.Therefore, there is no real practical application of public consciousness.However, persons with public consciousness are able to solve problems by themselves even when it impacts themselves or the broader public.They have to be confident that can solve problems by learning together [6].The public consciousness can force people to be united in action and spirit [5].Military officers are adults, thus their training falls under adult learning which focuses on activities and experiences rather than learning through theory.Once adults do things by themselves, it will lead to the goals of individual learning [8].When the researcher studied the concepts of service learning of Belisle and Sullivan [2], they found that service learning is a model of experiencing learning by using activities that bond the human learning needs and community needs together in a reflective form to achieve the desired results.The model of public consciousness should start from learning through experience.This will make adults modify learn valuable knowledge which can be put to practical use.If a person becomes more aware in the field then it will act for the benefit of people and an environment around them [11].
Conclusion
As a mentioned before, the in-service military officer should have public consciousness which is an important characteristic for helping people in social and country.The researcher thinks that the study of the existing problems and public consciousness would lead to the development of the training model to enhance public consciousness for the inservice military officers.The concept will be service learning, adult education, training, and public consciousness.The researcher believes that if the in-service military officers have public consciousness characteristic as a foundation for everyday life, the nation will spread up the force to help the society everywhere.This will prevent the community through critical situations, which may occurs in the future.
Remark
This article is a part of "Development of A Training Model Based on Adult Learning and Service Learning Concepts to Enhance the Public Consciousness of In-Service MilitaryOfficers of Royal Thai Armed Forces Headquarters" research. | 2018-12-12T08:18:26.884Z | 2016-01-01T00:00:00.000 | {
"year": 2016,
"sha1": "58371873ea6c55f7c46f05f738b565feb343ccf4",
"oa_license": "CCBY",
"oa_url": "https://www.shs-conferences.org/articles/shsconf/pdf/2016/04/shsconf_erpa2016_01069.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "58371873ea6c55f7c46f05f738b565feb343ccf4",
"s2fieldsofstudy": [
"Political Science"
],
"extfieldsofstudy": [
"Engineering"
]
} |
269870174 | pes2o/s2orc | v3-fos-license | Efficacy of Cerebellar Transcranial Magnetic Stimulation in Treating Essential Tremor: A Randomized, Sham-Controlled Trial
Background and Purpose Repetitive transcranial magnetic stimulation (rTMS) of the cerebellar hemisphere represents a new option in treating essential tremor (ET) patients. We aimed to determine the efficacy of cerebellar rTMS in treating ET using different protocols regarding the number of sessions, exposure duration, and follow-up duration. Methods A randomized sham-controlled trial was conducted, in which 45 recruit patients were randomly allocated to 2 groups. The first (active group) comprised 23 patients who were exposed to 12 sessions of active rTMS with 900 pulses of 1-Hz rTMS at 90% of the resting motor threshold daily on each side of the cerebellar hemispheres over 4 weeks. The second group (sham group) comprised 22 patients who were exposed to 12 sessions of sham rTMS. Both groups were reassessed at baseline and after 1 day, 1 month, 2 months, and 3 months using the Fahn-Tolosa-Marin tremor-rating scale (FTM). Results Demographic characteristics did no differ between the two groups. There were significant reductions both in FTM subscores A and B and in the FTM total score in the active-rTMS group during the period of assessment and after 3 months (p=0.031 and 0.011, respectively). However, subscore C did not change significantly from baseline when assessed at 2 and 3 months (p=0.073 and 0.236, respectively). Furthermore, the global assessment score was significantly higher in the active-rTMS group (p>0.001). Conclusions Low-frequency rTMS over the cerebellar cortex for 1 month showed relative safety and long-lasting efficacy in patients with ET. Further large-sample clinical trials are needed that include different sites of stimulation and longer follow-ups.
JCN
dures. 4,5Therefore, it is essential to search for an effective noninvasive safe therapeutic solution for ET before resorting to invasive procedures.
Repetitive transcranial magnetic stimulation (rTMS) can be considered an ideal tool for modulating the output of the cerebellocortical pathway. 6It produced beneficial results for only 5 minutes when used once at a low frequency. 7However, Popa et al. 8 found that ET reduced for much longer (3 weeks) after five consecutive daily sessions of low-frequency (1 Hz) cerebellar rTMS.This suggested that the performance of the cerebellothalamocortical (CTC) pathway improves after only five consecutive sessions of rTMS, although that study was not sham-controlled.A recent sham-controlled study found no significant effect of cerebellar rTMS on ET, but it involved a small sample and applied rTMS for only five consecutive days. 9Furthermore, the application of several sessions of cerebellar rTMS to patients with Parkinson's disease resulted in a marked improvement of motor symptoms for several weeks. 10ive the diverse findings of these previous studies, we aimed to measure the efficacy of cerebellar rTMS in treating ET using different protocols as regards the number of sessions, exposure duration, and follow-up duration.
Study design and ethical considerations
This randomized sham-controlled trial recruited eligible patients who visited the specialized movement disorders outpatient clinics at both Al-Hussein University Hospital and Bab-Alshaarya University Hospital.The study was approved by the local institutional review board (IRB number 0000039) and its protocol was registered (number: NCT05157321).Each patient signed a written informed-consent form before being enrolled in the study.
Study participants
This study included adult patients older than 18 years and diagnosed with ET according to a consensus document for movement disorders. 11The included patients need to have residual tremors despite receiving the appropriate medical treatment.We excluded epileptic patients and those who were receiving tremor-inducing medications.Furthermore, we excluded patients with major neurological conditions other than tremors, those with implanted metallic or electronic devices, and those with contraindications to exposure to strong magnetic fields.
Sample-size calculation
Olfati et al. 9 compared the effects of active versus sham rTMS on ET and found that the Fahn-Tolosa-Marin tremor-rating scale (FTM) total scores in the groups receiving active and sham rTMS for ET after 1 month were 22.8±10.4and 30.2± 19.4 (mean±standard deviation), respectively.Based on this mean difference of 7.4, a type-1 error of 0.05, a type-2 error of 0.8, and 1 as the ratio between the two groups, we estimated that the minimum sample size using the estimation method proposed by Chow et al. 12 was 18 patients in each group.
Randomization and masking
This study was a single-blind randomized controlled trial in which patients were blindly allocated either to the active group or to the sham group at a ratio of 1:1 using simple randomization methods with the Clinical Trial Randomization tool of the National Institutes of Health (https://ctrandomization.cancer.gov/home).Patients were not aware of which group they were allocated to.The study investigators were not blinded since they were performing the rTMS procedure.
Allocation and assessment of patients
Forty-five of the 64 screened patients fulfilled the eligibility criteria and were randomly allocated to the 2 study groups.The active group included 23 patients who were exposed to 3 sessions per week for 4 weeks (a total of 12 sessions) of real active rTMS.The sham group included 22 patients who were exposed to 3 sessions per week for 4 weeks (a total of 12 sessions) of sham rTMS.All patients in both groups were maintained on their medication without any change to the types or doses.In addition to the demographic characteristics being assessed at baseline, the patients in both groups were subjected to clinical evaluations, radiological investigations with brain magnetic resonance imaging (MRI), and tremor scaling.The patients were subsequently reassessed after 1 day, 1 month, 2 months, and 3 months using the FTM 13 (Fig. 1).All patients were advised to continue on their medications without changes.
rTMS procedures
The lowest stimulus needed to induce a motor evoked potential was identified as the resting motor threshold (RMT).We aimed to identify the lowest stimulus that induced the most-acceptable long-term effect.We used the most-accepted protocol in the literature associated with better results, 8,9 but with some modifications to the site of stimulation and the frequency and number of sessions in order to achieve better long-term results.Therefore, patients in the active group received 900 pulses of 1-Hz rTMS at 90% of RMT daily on each side of the cerebellar hemispheres.The intensity of the stimulation was adjusted based on the distance between the scalp and the cerebellar cortex.This distance was obtained JCN from the MRI that was conducted at baseline.The rTMS was conducted using the therapeutic version of the Neuro-MS/D device (Neurosoft, Ivanovo, Russia) with a figure-of-eight coil with a winding diameter of 100 mm.This coil generates a maximum magnetic field of 1.6 T. The site of the stimulus was at one-third of the distance from the inion to the mastoid process bilaterally as described previously. 14The current flow was set from caudal to rostral with the coil angled at toward the midline.
The sham stimulation was conducted using a similar technique but with an inactive coil electrically stimulated at more than 2 mA to produce the same sound as the active rTMS.It was reported that this protocol of sham stimulation had no neuromodulatory effect. 15
Statistical analyses
Statistical analyses were conducted using SPSS software (version 26 for Windows, IBM Corp., Armonk, NY, USA).Cat-egorical data were presented as frequency and percentage values while numerical data were represented as mean and standard-deviation values or median and range values depending on whether the data conformed to a normal distribution.Comparisons between the groups were conducted using either Student's two-tailed t-tests or chi-square/Fisher's exact tests if the analyzed variable was parametric.The Mann-Whitney test was used for nonparametric variables.Furthermore, analysis of variance (ANOVA) was used to compare two groups regarding FTM scores at the different assessment times.Statistical results were considered significant when the p value was less than 0.05.
RESULTS
The included patients in the active-rTMS group were 35.73± 4.76 years old, while those in the sham group were 41.8±4.8 years old, with no significant difference between the two
JCN
groups.Similarly, there was no difference between the two groups regarding sex, the duration of tremors, the MRI findings, or the family history of ET (Table 1).All of the included patients were free of thyroid or parathyroid abnormalities and were receiving medical treatment for ET.There were no instances of patient dropout during the study.
Regarding the changes from baseline, there was a significant reduction in FTM subscore A throughout the assessment period starting after 1 day of stimulation up to 3 months in the rTMS group (p<0.001).Similarly, subscore B and the total score reduced significantly in the rTMS group during the assessment period (p<0.001).Despite subscore C being reduced significantly after 1 day and 1 month (p<0.001 and p=0.004, respectively), its values at 2 and 3 months did not differ significantly from that at baseline.There was no change from baseline in the sham group either in the subscores or in the total score (Fig. 2).
Comparisons between the active rTMS and sham groups revealed no difference at baseline in subscores A, B, or C or in the total score.However, subscore A differed significantly during the assessment period starting from after 1 day of stimulation (p=0.001) up to 3 months (p=0.031).Similarly, subscore B was significantly lower in the active-rTMS group than the sham group during the assessment period starting from after 1 day of stimulation (p<0.001) up to 3 months (p=0.011).Similar results were found for the total score and the initial assessment of subscore C, but the assessment of subscore C at 3 months revealed no difference between the two groups (p=0.236).Furthermore, the global assessment score at the end of the study was significantly higher in the active-rTMS group (p<0.001)(Table 2).The longitudinal analysis of the study period using repeated-measures ANOVA revealed significant differences from baseline during the assessment period in the three subscores and the total score (p<0.001).Fur- The "a" means the reference, "b, c" mean significant changes from the reference.When "a" is present in both blue and red lines; that is mean non-significant values.rTMS, repetitive transcranial magnetic stimulation.
JCN
thermore, the active group was superior to the sham group throughout the assessment period (p<0.001)(Table 2).
DISCUSSION
This sham-controlled trial investigated the efficacy of a 4-week course of 12 successive sessions of low-frequency rTMS over the cerebellar hemispheres in patients with ET and assessed the outcomes up to 3 months after the start of the stimulation protocol.Overall, there were significant reductions in both the subscores and the FTM total score in the active-rTMS group throughout the assessment period.All of the FTM subscores were significantly reduced after 1 month in the active-rTMS group; however, subscore C did not differ significantly between the two groups when assessed at 2 and 3 months.
It is hypothesized that the cerebellum plays a crucial role in the pathogenesis of ET. 16 This assumption was based on the coexistence of cerebellar features such as intentional kinetic tremors, dysarthria, cerebellar gait, and nystagmus. 17,18urthermore, ET improved after cerebellar stroke.Functional imaging studies have revealed cerebellar overactivation in patients with ET.However, the literature indicates that the cerebellum is not the exclusive center for ET, with the entire motor network in the CTC pathway being involved in the pathogenesis of ET, including the inferior olivary nucleus. 19oreover, functional MRI studies have revealed abnormal activities in the primary motor cortex (M1), sensory cortex, basal ganglia, thalamus, dentate nucleus, and cerebellar hemispheres. 20However, recent postmortem pathological studies found that only cerebellar anatomical changes were associated with ET. 21This is not contradictory since the cerebellum is a part of the CTC pathway.Therefore, this cerebellocortical output needs to be modulated by a valid noninvasive tool in patients with ET.
The results of the present study support the hypothesis that rTMS improves the CTC connectivity to result in ET reduction.Our results were similar to those in the trial conducted by Popa et al., 8 who used 900 pulses of 1-Hz rTMS, but only for five sessions, and revealed that rTMS showed efficacy for only 3 weeks in 11 patients with ET, with an FTM total score after 1 month of 31.8±12.However, the protocol applied in our study demonstrated that it was effective for more than 3 months.Moreover, Gironell et al. 7 found transitory clinical improvement in tremors after only a single session of cerebellar stimulation.That open-label trial used functional MRI to demonstrate that rTMS improved CTC connectivity.However, a previous double-blind sham-controlled trial used a similar protocol of rTMS, but both cerebellar hemispheres were stimulated for only five consecutive days, which revealed no significant efficacy at all of rTMS on ET in terms of the FTM total score: 22.8±10.4and 30.2±19.4 in the active and sham groups, respectively, after 1 month of stimulation. 9Another study conducted by Shin et al. 22 found that rTMS was not effective in improving the FTM score; they used a similar protocol of applying 1-Hz rTMS over both cerebellar hemispheres, but the stimulation intensity was lower than in our study.These discrepancies could be due to multiple confounders in these two trials, including fatigue, anxiety, 23 and caffeine usage. 24In addition, it is difficult to quantify tremor severity, especially using the FTM subscores, which could introduce more variability between individuals. 25Furthermore, the authors of those trial reports did not use neuronavigational tech-
JCN
niques to localize the true site of rTMS application. 9he prolonged favorable outcomes (for 3 months) in our study could be explained by the number of sessions used and their frequency.We used three sessions per week (one session every other day) for 4 weeks, for a total number of 12 sessions.The largest number of sessions reported in the literature before 2023 was five (one session daily for five successive days). 8,9Another recent trial found that applying 10 consecutive sessions of 600 pulses of 1-Hz rTMS showed short-term effectiveness relative to propranolol. 26That study used a simpler rTMS protocol than ours regarding the number of sessions and the pulses, but applied sessions at a higher frequency than in our study. 26Despite the effectiveness, that previous study did not examine the effectiveness of rTMS on patients who were resistant to propranolol, unlike in our study.Moreover, that study did not examine the long-term outcome of the protocol.This means that our protocol is superior regarding the long-term efficacy for drug-resistant ET.Furthermore, the total number of pulses used in our study was 1,800 pulses (900 over each hemisphere).Only Olfati et al. 9 used a similar number of pulses but at a different site (onethird of the distance from the inion to the mastoid process).Together these findings suggest that not only the cerebellum but also M1, basal ganglia, sensory cortex, and prefrontal cortex are involved in ET. 20 Therefore, these sites could be targets for inhibitory brain stimulation in future trials.
Unlike subscores A and B of the FTM, subscore C showed no difference between the two groups when the severity of the tremors was assessed after 2 and 3 months of stimulation.Subscore C of the FTM assesses the impact of tremors on the patient's quality of life, and so is subject to bias due to the presence of many confounders, including the recent deterioration in the socioeconomic status in most Egyptian patients.Similarly, Popa et al. 8 reported a nonsignificant difference between the two groups in subscore C even though the total score was significant.
Despite the superiority of our trial over the previous trials regarding the sample size and the used protocol, our study had some limitations.We did not use a cross-over design to ensure that all participants would receive the benefit and to avoid the effects of confounders.Furthermore, we did not use neuronavigational techniques to determine the exact site of stimulation, instead using previous literature to identify the posterior part of the cerebellar cortex as a potential site of stimulation. 14Moreover, the severity of tremors can change in different situations.We relied mainly on clinical assessments, which can cause bias in assessments.However, we tried to focus on all circumstantial environments of all assessments in order to decrease that bias as much as possible.We could not use a robust tool for assessing the tremors such as an ac-celerometer.The heterogeneity in our results reflects ET in our study encompassing many undetermined diseases. 11In addition, we did not assess the stimulation at other sites such as M1 or extend the follow-up beyond 3 months.
In conclusion, low-frequency rTMS over the cerebellar cortex for 1 month showed relative safety and long-lasting efficacy in patients with ET.This efficacy was greater for tremor severity than for the quality of life.Further large-sample clinical trials are needed to compare cerebellar stimulation and M1 stimulation with sham stimulation using the same protocol and a longer follow-up.
Fig. 2 .
Fig.2.Changes from baseline in the active-rTMS and sham groups for Fahn-Tolosa-Marin tremor-rating scale (FTM) subscore A (A), subscore B (B), subscore C (C), and total score (D).The "a" means the reference, "b, c" mean significant changes from the reference.When "a" is present in both blue and red lines; that is mean non-significant values.rTMS, repetitive transcranial magnetic stimulation.
Table 1 .
Demographic and characteristics of participants in the active and sham groups Data are n (%) or mean±standard-deviation values.*Mann-Whitney test and t-test between the two groups.ET, essential tremor; MRI, magnetic resonance imaging.
Table 2 .
Fahn-Tolosa-Marin tremor-rating scale score for ET of patients in the two groups at different times throughout the study | 2024-05-19T15:11:51.494Z | 2024-05-14T00:00:00.000 | {
"year": 2024,
"sha1": "e1603529affe3bde7ec042f38d5d29ab56d3a409",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.3988/jcn.2023.0348",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "9dac95c85051d908df97bb8c4222b3395b025edd",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
1270387 | pes2o/s2orc | v3-fos-license | A possible role for stochastic radiation events in the systematic disparity between molecular and fossil dates
Major discrepancies have been noted for some time between fossil ages and molecular divergence dates for a variety of taxa. Recently, systematic trends within avian clades have been uncovered. The trends show that the disparity is much larger for mitochondrial DNA than for nuclear DNA; also that it is larger for crown fossil dates than stem fossil dates. It was argued that this pattern is largely inconsistent with incompleteness of the fossil record as the principal driver of the disparity. A case is presented that given the expected mutations from a fluctuating background of astrophysical radiation from such sources as supernovae, the rate of molecular clocks is variable and should increase into the past. This is a possible explanation for the disparity. One test of this hypothesis is to look for an acceleration of molecular clocks 2 to 2.5 Ma due to one or more moderately nearby supernovae known to have happened at that time. Another is to look for reduced disparity in benthic organisms of the deep ocean..
Introduction
There has been a long acknowledged disparity between ages determined from the fossil record and those derived from molecular divergence dating. The sign of this disparity is nearly always in the direction of much older molecular ages. A variety of effects may cause biases of either sign in the molecular ages, but, due to incompleteness of the fossil record, fossil ages are (aside from outright errors) lower bounds to the age of the taxon, so that the ages of originations are always underestimates. Since this could in principle account for the entire effect, much discussion has centered on this option.
A systematic study of trends within the disparities in avian clades (Ksepka et al. 2014) asserts that variety of patterns, not to be enumerated here, are inconsistent with attributing all of the age disparity to gaps in the fossil record. We are urged to consider any possible systematic biases in molecular dates as well as calibration strategy. At this point it is relevant to mention that the whole basis of molecular dating has been accused of systematic underreporting of errors (e.g. Graur and Martin 2004). However, we shall for the purpose of this paper assume that they have a basis in reality.
One possible bias in molecular dates concerns a variable rate of molecular clocks due to changes in the mutation rate. An unknown but substantial fraction of mutations come from radiation of various kinds (Alpen 1997). The radiation background in the Earth's environment fluctuates strongly, so that it is expected to find events of increasing strength when looking further back in geologic time (Erlykin and Wolfendale 2009;Melott and Thomas 2011). There is a normal operational assumption that molecular clocks move at a constant rate, but fluctuating radiation backgrounds would vary this rate. In the absence of selection pressures, isolated communities, etc. this variable mutation rate need not correspond to a correlated variable rate of evolution.
I wish to stress the following: this is a physics based hypothesis; there are many other biological explanations that may explain all or part of the phenomenon. I do not intend to claim that it is the best possible explanation, but only to introduce it for discussion.
High energy events can produce air showers which have strong effects on the ground. Since such very strong events are not occurring now (see e.g Overholt et al. 2015), molecular clock rates determined from very recent data would not include this acceleration. The purpose of this note is to suggest further examination of this possibility, with attention to one specific test. Ksepka et al. (2014) reported that the disparity between fossil and molecular clocks is greater with mitochondrial DNA than with nuclear DNA. This observation is consistent with our hypothesis of a radiation link for the disparity, because mitochondrial DNA is more subject to damage from radiation and to oxidative stress, one of the primary mechanisms of radiation damage to DNA (Yakes and Van Houten 1997;Kam and Bonati 2013).
Radiation Events and the Earth
There are a variety of possible types of astrophysical radiation and likely sources for events at the Earth (Melott and Thomas 2011). Dominant among these are the Sun (see Wdowczyk and Wolfendale 1977) and other stars in our galaxy. It has been known for some time that supernovae and gamma-ray bursts from other stars in our galaxy are likely, based on their intensity and frequency of occurrence, causal agents in mass extinction every few 100 Myr. The Sun has X-ray flares, but the dominant form of radiation for biological consideration is in Solar Proton Events. There was in 775 AD an event indicated by 14C in tree rings which exceeds anything in the modern era (Miyake et al. 2012;Jull et al. 2014) and which is probably attributable to the Sun (Melott and Thomas 2012;Usoskin et al. 2013).
All of these are potentially dangerous sources, although interpreting the new data on the Sun and Sunlike stars is an emerging area. Although the atmosphere provides considerable shielding, effects on the ground can still include radiation in the form of muons (Atri and Melott 2011;Marinho et al. 2014) and neutrons (Overholt et al. 2013;Overholt et al., 2015). Most muons are stopped by a kilometer of water, so any potential muon damage would not include benthic organisms in the deep ocean. In addition, the ionizing radiation can deplete the stratospheric ozone layer (Thomas et al. 2013, and references therein), admitting increased damaging and mutagenic ultraviolet-B from the Sun. Nearby supernovae will bombard the Earth with much higher energy cosmic rays (protons) than are likely to come from the Sun.
Testing the Idea
Most of the astrophysical radiation including the indirectly increased UVB can be stopped by 10 m of water. However, the DNA of such organisms may be affected in a pelagic larval stage, especially if they are photosynthetic. Organisms which are shielded from the radiation should not show effects of strong fluctuations. In particular, deepwater benthic organisms should display more congruence between the fossil and molecular dating methods. Kspeka et al. (2014) noted that for very old dates (many times 10 Myr), there is better congruence than for younger dates. This would be consistent with a high rate of mutation in the not too distant past, and a return to a geologic mean rate over longer periods.
There is a large amount of new data in the form of 60 Fe in sediments which suggest that one or more supernovae went off within one or a few hundred light years of the Earth around the beginning of the Pleistocene (Fry et al. 2015;Wallner et al. 2016;Breitschwerdt et al. 2016;Melott 2016;Fimani et al. 2016;Binns et al 2016). This would indicate an enhanced radiation environment persisting for at least several thousands of years for each event. Therefore, if this idea has merit, there should be an acceleration of the molecular clocks relative to the fossil record around 2.5 Ma. The transport modeling work (Breitschwerdt et al. 2016) was published simultaneously and done without knowledge of the new data from the other studies. Improved modeling is in progress which should better constrain the dates. We have recently done computations of radiation transport in the galaxy and the atmosphere (Thomas et al. 2016) which suggest at least a factor of 10 increase in ionizing radiation at the surface of the Earth, persisting for thousands of years. This also should be further refined in near future. The other test of this idea, as noted, is the expected major reduction in the disparity for deep-sea organisms.
In closing, it should be emphasized that we do not claim that this is better than other existing explanations for this disparity, but that it is a possibility that should be considered, and can be tested by looking for the predicted acceleration near the beginning of the Pleistocene and better agreement between fossil and molecular ages for deep sea organisms. | 2016-04-22T08:34:03.594Z | 2015-05-29T00:00:00.000 | {
"year": 2017,
"sha1": "28872af623a8f4d2264b58cc549b66aff3ffd401",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1505.08125",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "9b390ccfcac6ca91a677a8a06243c3efde02e03d",
"s2fieldsofstudy": [
"Biology",
"Environmental Science",
"Physics"
],
"extfieldsofstudy": [
"Medicine",
"Biology",
"Physics"
]
} |
214294578 | pes2o/s2orc | v3-fos-license | Determinants of Value Added Tax Collection Performance in West Shoa Zone, Oromia Regional State, Ethiopia
The study was conducted to investigate the determinant factors of Value Added Tax collection performance in West Shoa Zone by giving emphasis on three selected districts i.e. Ambo city Revenues Authority, Dandi and Bako woredas Revenues Authority’s. The researchers used cross-sectional data covering a period in 2017. Since the data was survey in nature the researcher used both qualitative and quantitative (Mixed) research approaches to achieve objective of the study. The researchers used purposive and stratified sampling method to take a sample of Tax payers, Consumers and Revenues Authority Employees. Descriptive statistics and multiple regressions were used to analyze primary and secondary data to reach the result of the study using Statistical Package for Social Science (SPSS) version 20. The result of the study showed that Tax payers awareness on Value added Tax, Tax payers’ maintenance of account and Value added Tax Rate have positively influenced Value added Tax collection performance whereas Tax Evasion, Tax avoidance and Tax non-compliance variables have negatively affected Value added Tax collection performance for tax payers. The study results also showed that Value added Tax Assessment, Value added Tax Audit, Competence training, Adequate manpower, Tax payers identification and Penalization variables have positively influenced Value added Tax collection performance whereas External Legal Environment variable negatively affected Value added Tax collection performance for institutional variables. Thus, it’s recommended that Revenues Authority should fulfill necessary manpower and create continuous follow up in order to improve Value added Tax collection performance. Keywords: Value added Tax, Tax avoidance, Tax Evasion, performance, Tax payers DOI : 10.7176/RJFA/10-21-06 Publication date: November 30 th 2019
time frame (Mohd et al, .2011). Tax non-compliance can takes place in the form of either tax Evasion and/or tax avoidance. Tax evasion is an attempt to escape tax liability (wholly or partially) by breaking the tax law and it is a criminal act since it is achieved principally by making false declarations such as under-reporting income or over-reporting relieves and allowances. Soyode and Kajola (2006), defined tax avoidance as the arrangement of tax payers affairs using tax shelters in the tax laws, and avoiding tax traps in the tax laws, so as to pay less tax than ought to be paid, hence the person pays less by taking the advantage of the loopholes in the tax laws.
Empirical Review
Bisrat (2010) The VAT is simply a multistage sales tax that exempts the purchase of intermediate goods and services from the tax base. This study, focusing on the Ethiopian case was analyzing the causes of noncompliance by VAT taxpayers. It was analyzed the impact of low audit probability, the perceived fairness of VAT, dissatisfaction with the tax authority, and how business people think about the VAT money on tax compliance. To gather adequate information for analysis purpose, a combination of quantitative and qualitative methods is applied. The final major factor studied was how business people think about the VAT money. These are taxpayers" beliefs on the relationship between VAT money and their business fund, the equivalence of VAT payment with the value added by the business, and taxpayers" contribution in the VAT collection. The findings of the study suggest these are not really a major problems concerning tax compliance.
Hailemariam (2011) studied on identifying and assessing the problems rose in association with the implementations of VAT by the Ethiopian Revenue and Customs Authority. The researcher used both qualitative and quantitative descriptive research designs and a sample of taxpayers and employees of the authority were selected using stratified random sampling method. The researcher used questionnaires, Interviews and relevant documents to collect primary and secondary data from the data sources. Pie charts, graphs, table, Percentages were used in analyzing the collected data. The study identified misunderstanding of the public in general and business community in particular regarding the VAT laws, resistance against registrations for VAT by some traders, administration inefficiency from ERCA, provisions of understated financial statements, and nonissuance of invoices or issuance of illegal invoices exercised by registered business enterprises was the main problems. The study recommends that the authority should train the taxpayers about the rules and regulations of VAT continuously, recruiting new employees and give nonstop training for the existing once, delegate the tax authority to regional and city administration and it should also increase its follow-up and investigation to control noncompliance enterprises, as well as effectively and efficiently performing the tasks of identification of VAT taxpayers, processing of returns, controlling collections, making refunds on time, auditing taxpayers, recognizing genuine taxpayers and levying penalties to tackle the problems it has encountered.
From the review of available literature, it is inferred that the topics undertaken in these limited research studies mainly focused on the drawbacks of sales tax and make justification for the introduction of VAT. Similarly, problems which could arise in introducing VAT have also been explored in some studies. There are only few studies which explore the effects of VAT after its introduction.
A study made by Wollela (2008) examines VAT administration in Ethiopia and identifies key problems including lack of sufficient number of skilled personnel and gaps in the administration in such areas as refunding, invoicing and filing requirements. The paper suggests that in Ethiopia attempting to implement what is legislated in the main areas (such as refunds) deserves the government's due attention. The study also emphasizes the need to strengthen the administration capacity in general and the tax audit program in particular. Besides the above researcher's knowledge, one can see that empirically studied research under taken by the above authors emphasis on how VAT system implemented and administered. However the researchers were motivated to investigate the status of VAT collection performance rather than its implementation and administration, particularly in West Shoa Zone. As far as the knowledge of researchers in West Shoa Zone, no researches have been investigated on determinants of Tax collection performance emphasis on Value Added Tax. Therefore, the current study believed to add value to this subject. Vol.10, No.21, 2019 60 As shown in figure 4.2 in terms of area of occupation the share of West Shoa Zone VAT registered Tax payers for Hotel was 72 (32.1%) accounted the highest percentage and Manufacturing 26(11.6%) was the lowest percentage share of sector. It shows that majority of tax is come from the service nature of business. As indicated in figure 4.3, the highest level of average annual turnover of West Shoa zone VAT registered taxpayers were less than 500,000 birr which accounted 110(49.1 and the lowest one was more than 10, 000,000 birr that accounted only 12(5.4%). Table 4.1.summarizes the source of information on the concept of VAT presented for Tax payers. Out of total respondents 80(35.7%) and 57 (25.4%) VAT registered Tax payers got information from TV and radio respectively. 51(22.8%) and 36 (16.1%) Tax payers got the concept of VAT from Pamphlets and training respectively. This indicated that most of VAT registered Tax payers were got concept VAT through television and radio programs respectively. In terms of maintenance of accounts West Shoa Zone tax payers asked maintaining all Books of Accounts as per required procedures and guidelines of VAT Proclamation is cumbersome replay that (40)17.9% strongly disagree, (10)4.5% disagree, (85)37.8% agree,(70)31.3% strongly agree and (19) 8.5 % neutral. Also the respondents asked whether they maintain all Books of Accounts of VAT as per proclamation. (34)15.2% strongly disagree, (97)43.3 % disagree, (22)9.8% neutral and (71)31.7% agreed. This indicated that many of tax payers are not maintaining Book of account as per proclamation. Its observed that; most of the traders do not keep proper records of their transactions and those who keep proper records mostly default by keeping two different records to evade tax. Mostly they undervalue their turnover so that it does not qualify for VAT. Most small traders do not wish to register for VAT; because they will be forced to sale their products at a relatively higher price and so fail to compete effectively in the market. Therefore, the respondents say that, the West Shoa Zone Revenues Authority should employ effective methods in effective identification of the taxpayers as many of them do not wish to register voluntary.
Source: Survey Result, 2018
As it is indicated in table 4.5, the culture of tax compliance were observed, the result shows that few VAT Registered Tax payers would attempt tax evasion. Out of respondents (69)30.8% strongly agree, (25)11.2% strongly disagree and the other tax payers (35)15.6% disagree, (51)22.8% agree and (44)19.6% neutral. This shows that, most of respondents strongly agreed in society with a culture of high compliance with VAT law but some VAT Registered Tax payers attempt tax evasion. For the Tax payers should evade VAT if the VAT system (60)26.8% agreed, (72) 32.1% strongly agreed and for Splitting the business will affect the VAT the respondents respond that (32)14.3% disagree, (51)22.8% neutral, (56) 25% agreed and (85)37.9% strongly agreed. This shows that most of Tax payers agreed the Splitting the business will affect the VAT. The non-compliance of other taxpayers has a negative impact on compliance tax payers'. 0.0 0.0 3.6 29.9 66.5 Source: Survey Result, 2018 The respondents were asked there is Non-issuance of invoice to all buyers, the respondents respond that (120)53.6%agreed, and (104)46.4% strongly agreed. That implies that majority of tax payers agreed on Nonissuance of invoice. In addition the Tax payers asked the non-compliance of other taxpayers has a negative impact on compliance tax payers' (8)3.6% neutral, (67) 29.9% agreed and (149)66.5% strongly agreed. This means that most of Tax payers agreed on the non-compliance of other taxpayers has a negative impact on compliance tax payers. Respondents were asked questions related to Awareness on Mechanisms for raising objection against penalty is low. They respond that 45 (20.1%) disagreed, 57(25.4%) agreed and 122(54.5%) strongly agreed. It implies that the majority of respondents have insufficient awareness on Mechanisms/procedures for raising objection against penalty. Findings further revealed that respondents who have been penalized by West Shoa Revenues Authority because of late payment were 39 (17.4%) and 10(4.5%) of the respondents have been penalized due to Nonissuance of VAT invoice whereas a majority of respondents 175 (78.1%) have never been penalized.
Descriptive statistics of VAT Revenues Authority employees Variables
Data was also collected from employees of Revenue office and result obtained from them was summarized as follows: Internal VAT administration Table4 The respondents were asked in conducting Revenue Collection, the Revenues Authority has sufficient Staff Members. (47)63.5 % of the respondents disagreed, (17)23% of the respondents strongly disagreed, and (10)13.5% of the respondent's agreed. It revealed that the Revenues Authority did not have sufficient Staff Members. On the other hand there is shortage of resources and infrastructural facilities to implement and enhance the enforcement of VAT Law, (28)37.8% strongly agreed (32)43.2% agree and (5)6.8% disagreed and (9)12.2% strongly disagree. This shows that most of the respondents agreed on there are shortage of resources and infrastructural facilities to implement and enhance the enforcement of VAT Law. Respondents were asked on whether there is intense follow up and supervision over the identified VAT payers. The following were the responses; (41)55.4 % of the respondents disagreed, (10)13.5% agreed while (8)10.8% strongly agreed, (11)14.9% strongly disagreed and (4)5.4% remained neutral. It indicated that less close follow up and supervision by West Shoa Zone Revenues Authority on identified taxpayers. On the Tax payers' VAT report assessed on time, (20) 27% strongly disagreed, (29)39.2% disagreed, (5)6.8% neutral, and (20)27% strongly agreed. It implies majority of respondents disagreed on Tax payers' VAT report assessed on time. For the question VAT registered Tax payers have trust in assessment and collection procedures. (44)59.4% disagreed, (5)6.8% strongly disagreed, (9)12.2% neutral and (16)21.6% agreed. It shows that most of VAT registered Tax payers have not trust in assessment, and collection procedure. Respondents were asked there are tax payers yet to be registered for VAT but Revenues Authority so far not included. The following were the responses; (42)56.8% of the respondents strongly agreed, (21)28.4 % agreed while (7)9.5% disagreed, and (4)5.3% of the respondents were strongly disagreed on the tax payer's identification. It implies that there were tax payers yet to be registered for VAT but Revenue Authority so far not included. Also the respondents were asked the methods used by West Shoa Zone Revenues Authority to identify tax payers are adequate. They respond that 13.5% strongly disagreed, 64.9% disagreed and 21.6% agreed. This indicated that the methods used by West Showa Zone Revenues Authority to identify tax payers are not adequate. Respondents were asked that Tax payers who need to register for VAT when their annual turnover was more than Birr1 million. The Required turnover of Birr 1 Million is very high standard. (48) 64.8% of the respondents strongly agreed, (17)23% agreed while (9)12.2% disagreed on the registration threshold being high. In addition the respondents were asked it should be worthwhile to amend the possibility of reducing the requirement of 1Million Turnover. (2)2% strongly disagreed, (9)13% disagreed (38)51.4% agreed and (25)33.6% strongly agreed. Some of the respondents argued that this amount is higher and suggest that it should be lowered. They say that, small businesses wishing to secured businesses with registered businesses are unable because they are not registered and would wish to but the constrain remains to be the registration threshold. Some of the respondents say that, the nature of the businesses they are in requires them to register for VAT and be able to even get tenders but they are not in the position because they are not registered. Most respondents suggest that, the registration threshold should be reviewed to a position of being able to accommodate many businesses despite the fact that it will demand more administration for it to work effectively. shows that the respondents are asked there is lack of commitment in VAT experts to impose penalty on tax payers as per proclamation. The respondents respond that (5)6.8% strongly disagreed, (11)14.9 % disagreed, (2)2.6% neutral, (31)41.9% agreed and (25)33.8% strongly agreed. For the question there is lack of transparency and consistency in imposing penalties, (2)2.7% strongly disagreed, (15)20.3% disagreed, (40)54% agreed and (17)23% strongly agreed. Whereas for if there is no penalty or legal enforcement, there is a possibility that VAT registered Tax payers may not pay VAT they respond that (6)8.1% disagreed, (2)2.7% neutral, (12)16.2% agreed and (54)73% strongly agreed. It indicates that majority of respondents argued that there is lack of commitment in VAT experts to impose penalty on tax payers as per proclamation, lack of transparency and consistency in imposing penalties and if there is no penalty or legal enforcement, there is a possibility that VAT registered Tax payers may not pay VAT. The above table shows that out of 578 VAT registered Tax payers in Ambo city administration Revenues Authorities 13(2.55%) were penalized. That means 7(1.21%) by Non-issuance of invoice, 4(0.695) by Non-post advertisement and 2(0.35) by Obstruction to Revenues. In case of Dandi Woreda Revenues Authority out of 77 VAT registered Tax payers only 3(3.9%) were penalized by Non-issuance of invoice but in Bako Woreda Revenues Authority, out of 80 VAT registered Tax payers none of them were not penalized in 2017 year. In general it reviled that in West Shoa zone Revenues Authority there is poor practice on imposing penalty on VAT registered Tax payers. Source: Survey Result, 2018 VAT registered Tax payers Respondents were asked to give their views on the percentage change of total revenue collections from VAT is increasing from year to year, but not the meet the targets. 33.9% of the respondents strongly agreed, 43.3% agreed, 6.7% remained neutral while 12.1% disagreed and 4% strongly disagreed. In addition to that the Respondents were asked Revenue Authority witnessed unachieved VAT Targets,33.1% of respondents strongly agreed, 45.5% agreed, 13.8%neutral and7.6% disagreed on unachieved VAT Targets. Also the respondents asked VAT income is affected due to lack of intensive follow up, 36.6% strongly agreed, 52.2% agreed,8.1%neutral and 3.1% disagreed. This indicates that the percentage change of total VAT collections from year to year is not meets the target.
External Ethiopian VAT administration
Whereas the question for Revenues Authority employees asked the percentage change of total revenue collections from VAT is increasing from year to year, but not the meet the targets, 74.3% agreed and 25.7% strongly agreed. For the question Revenue Authority witnessed unachieved VAT Targets, 39.2% strongly agreed, 54% agreed and 6.8% neutral and finally the Revenues Authority employees were asked VAT income is affected due to lack of intensive follow up, 18.9% strongly agreed, and 56.8 % agreed, 10.8% neutral and 13.5% disagreed. this indicates that even if the revenue collections from VAT is increasing from year to year, but not the meet the targets due to lack of intensive follow up. Moreover data was collected from Consumers and result was summarized as follows: The above table shows that the respondents were asked to State level of awareness on VAT. They replied that 28(56%) have little aware, 15(30%) unaware, 5(10%) moderately aware and 2(4%) completely aware. Therefore majority of consumers respondents of the study have little aware of VAT. The respondents were asked as per VAT proclamation, all VAT registered tax payers should give invoice to consumers, but they are not practicing so. From the fifty respondents 21 (42%) agreed, 16(32%) strongly agree, 5(10%) disagreed and 8(16%) neutral. It indicated that most consumers respondents were agreed to as per VAT proclamation, all VAT registered tax payers give invoice to consumers, but they are not practicing. As it's shown from the above Table 4.27.consumers respondents were asked to express their view about Consumers are not willing to buy goods and services with VAT. Consequently, 23(46%) of consumers were replied that agreed, 13(26 %) of consumers were responded strongly agreed, 8(16%) of consumers were answered disagreed and 6(12 %) of consumers were replied that neutral. Therefore, from this finding, the majority of the respondents of Consumers were not willing to buy goods and services with VAT. It can conclude that the main reason for not willing to buy goods and services with VAT, by most consumers were lack of awareness about VAT.
Status of VAT income as a percentage of total tax revenue
For analysis purpose, VAT collection with its respective total tax revenue collection 2011to 2017 year for West Shoa Zone Revenues Authority showed in the following:
Regression Analysis
Regression analysis was conducted to empirically determine whether independent variables were significant with dependent variables. The regression analysis was done with VAT collection performance as dependent variable and determinants of VAT collection performance as independent variables (Tax payer's awareness of VAT, Tax payer's maintenance of account, Tax Evasion, Tax Avoidance, Tax Non-compliance, and VAT Rate) Tax payers point of view and VAT assessment, VAT audit, Competence training, Adequate manpower, Tax payers identification, External legal Environment and penalization on institutional point of view.
Model summery of Tax payers related variables
The regression results in the table4.30.indicated the goodness of fit for the regression between independent variables and dependent variables was satisfactory in the multiple regressions. An R squared of 0.63 indicates that 63% of the relationship explained by the identified six factors namely Tax payer's awareness of VAT, Tax payer's maintenance of account, Tax Evasion, Tax Avoidance, Tax Non-compliance, and VAT Rate. The rest 37% is explained by other factors in the VAT collection performance not studied in this research. Further, the adjusted R-square=0.62 shows that, the factors accounted for 62% of the variance in VAT collection performance.
Regression results in table 4.30.below indicated the coefficient for each of the variables as well as their significance levels in the model. As shown, Tax payer's awareness of VAT, Tax payer's maintenance of account, Tax Evasion, Tax Avoidance, Tax Non-compliance and VAT rate were determinants influence significantly VAT collection performance. The regression equation for Tax payer's variables was: VCP i = α0 + β1TPAWOV i + β2 TPMOA i + β3TE i + β4 TA i + β5 TNCOM i + β6 VR i + ɛi VCP i=2.6165+ 0.5893TPAWOV + 0.2676TPMOA -0.2949TE-0.1172TA-0.1379TNCOM + 0.1028VR + ɛi Moreover, the model summery also shows the significance of the model by the P value statistics (P=0.000) and F=62.78 makes known the sound explanatory power of the model. This statistics indicates that the overall model is significant in explaining the dependent variable since the associated probability is lower than 0.05. It also suggests that the relationship between the dependent variable and independent variables is linear. The beta (β) sign also shows the positive or negative effect of the independent variables coefficient over the dependent variable.
The findings shows that when Tax payer's awareness of VAT increase, VAT collection performance increase by 58.93% by controlling other factors constant. When Tax payer's maintenance of account increase, VAT collection performance increase by 26.76% by controlling other factors constant. When Tax Evasion increase, VAT collection performance decrease by 29.49% by controlling other factors constant. When Tax avoidance increase, VAT collection performance decrease by 11.72% by controlling other factors constant. When Tax Non-Compliance increase, VAT collection performance decrease by 13.79 % by controlling other factors constant. When VAT rate increase, VAT collection performance increase by 10.28% by controlling other factors constant. This implies that Tax payers awareness of VAT, Tax payers maintenance of account and VAT rate have a positive relationship with VAT collection performance and statistically significant (P-value =0.000, 0.000 and 0.034 respectively, refers to less than 0.05). But Tax Evasion, Tax Avoidance, Tax Non-Compliance have a negative relationship with VAT collection performance and statistically significant (P-value =0.000, 0.006 and 0.000 respectively, refers to less than 0.05). All variables related to Tax payers ; (Tax payer's awareness of VAT, Tax payer's maintenance of account, Tax Evasion, Tax avoidance, Tax non-compliance and VAR rate) are found significantly influencing VAT collection performance in the study area.
Model summery related to Institutional variables
The regression results indicated the goodness of fit for the regression between independent variables and dependent variables was satisfactory in the multiple regressions. An R squared of 0.78 indicates that 78% of the relationship is explained by the identified seven factors namely VAT assessment, Adequate Manpower, VAT audit, competency training, Tax payer's identification, External legal environment and penalization. The rest 22% is explained by other factors in the VAT collection performance not studied in this research. Further, the adjusted R-square=0.76 shows that, the factors accounted for 76% of the variance in VAT collection performance.
Research Journal of Finance and Accounting www.iiste.org ISSN 2222-1697(Paper) ISSN 2222-2847(Online) Vol.10, No.21, 2019 Moreover, the model summery also shows the significance of the model by the P value statistics (P=0.000) and F=34.22 makes known the sound explanatory power of the model. This statistics indicates that the overall model is significant in explaining the dependent variable since the associated probability is lower than 0.05.It also suggests that the relationship between the dependent variable and independent variables is linear. The findings show that when frequency of VAT Assessment increase, VAT collection performance increase by 15.67% by controlling other factors constant. When frequency of VAT audit increase, VAT collection performance increase by 13.16% by controlling other factors constant. When frequency of training increases, VAT collection performance increase by 22.94% by controlling other factors constant. When adequate manpower increase, VAT collection performance increase by 22.13% by controlling other factors constant. When Tax payers Identification increase, VAT collection performance increase by 35.12% by controlling other factors constant. When External legal Environment increases, VAT collection performance decreases by 15.09% by controlling other factors constant. When Penalization increases by, VAT collection performance increases by 37.92% by controlling other factors constant. This implies that VAT Assessment, VAT Auditing, Competence training, Adequate manpower, Tax payers identification and Penalization have a positive relationship with VAT collection performance and statistically significant in explaining VAT collection performance as p value=0.046, 0.037,0,002, 0.001, 0.000 and 0.006 respectively. But External legal environment has a negative relationship with VAT collection performance and statistically significant in explaining VAT collection performance as p value 0.014. Among the independent variables, Tax payer's identification has been found to be the most determinant and statistically significant.
In-depth interview results
The interviews were conducted by using semi-structured interview method to reach five VAT Administrations starting from Oromia Revenues Authority Assessment and follow up department head to woreda level. These were; Oromia Revenues Authority Assessment and follow-up department head (1), West Shoa Zone Revenue Authority Assessment and follow-up department head (1), Assessment and follow-up department heads of selected three woredas (3) were interviewed at different times.
There are tax payers yet to be registered for VAT but Revenue Authority so far not included, what are the reasons. They responded that many reasons, some of them are: lack of man power and resources, Tax payers not tell their exact annual income, lack of commitment of some employees at different level to identify tax payers, political pressure, lack of awareness of VAT, biased of employees and the instability of external part.
In-depth interview were asked about what are mechanisms for raising objections against penalties by tax payers in handling their grievances. Their responses were similar to the procedures written on proclamation. Appeal is made according to a constitutional right as far as the law. The VAT system is not exception to these rules. A tax payer under VAT, Who is aggrieved by the decision of tax authorities, may lodge his complaint to a body competent to hear and decide on such application. The aggrieved parts have two options: tax appeal committee or regular courts. The tax payer objecting the assessment by revise committee while taking appeal to Tax Appeal Committee is required to deposit 50% of the additional tax assessed. Then the committee gives the decision based on VAT proclamation. If Tax payers satisfied with decision given by committee the appeal can finalized at this stage unless go for regular courts.
For the interview question how the administrative officials are handling the challenges of VAT Levy and Collection by using appropriate VAT Management Techniques, The interviewee suggested that first we identify the areas of the problems. Then we discussed with committee and give the direction to solution. This means that make awareness creation to Tax payers, Employees and society as all. According to Oromia assessment and Follow up department leader said that there is person to do on VAT assessment and collection. In addition to we call other employees from different Zone to make operation side used as best techniques.
On what are the criteria used by the authority for sample selection of tax payers for audit and audit procedures. They responded that we follow audit criteria which are comprehensive and sample audit that have 13 points procedure. By using SIGTAS if the summation of thirteen point procedures greater than fifteen we used comprehensive audit and if the summation of thirteen point procedures less than fifteen we used sample audit. But according to west Shoa Zone assessment and Follow up department leader suggested that not follow this procedures rather used based on the year of VAT registered Tax payers. The first VAT registered Tax payers audited first.
Another question what are the major problems in relation to the assessment and collection of VAT were interviewed. The major problems they raised were lack of manpower and interruption of SIGTAS. Especially from the three selected woredas assessment and Follow up department leaders point of view there is no Auditor at woreda level to audit VAT registered Tax payer's document. Finally for what is the reason consumers are not willing to buy goods and services with VAT. They suggested that due to lack of consumers awareness of VAT, due to all Tax payers not registered for VAT and Consumers are interested to buy goods or services from nonregistered enterprise to get lower Price.
Conclusions
The study covered the interaction between Value Added Tax collection performance and exploratory variables of Tax payers (Tax payer's awareness of VAT, Tax payer's maintenance of account, Tax Evasion, Tax Avoidance, Tax Non-compliance, and VAT Rate) and Institutional variables (VAT assessment, Manpower, VAT audit, competency training, Tax payer's identification, External legal environment and penalization).The main objective of the study was to evaluate the determinants of VAT collection performance in West Shoa Zone. The researchers used quantitative and qualitative research approach and primary and secondary data type. Survey design was used with questionnaire and semi-structured interview as tool of data collection. In this study both purposive and stratification sampling techniques were used. The study analysis was used the descriptive statistics, Pearson correlation matrix and ordinary least square (OLS) estimation method. The outcome of regression result shows that classical linear regression model assumptions were not violated.
According to Pearson correlation result of Tax payers variables shows that Tax payer's awareness of VAT, Tax payer's maintenance of account and VAT Rate were positive correlated with VAT collection performance. Moreover, the coefficient estimates of correlation was 0.622, 0.5205 and 0.1837 from highest to lowest ρ<0.05 respectively. While Tax Evasion, Tax Avoidance and Tax Non-compliance were negative correlated to VAT collection performance with coefficient estimates of correlation was -0.5471, -0.3984 and -0.2854, ρ<0.05 respectively. On the other hand Pearson correlation result of Institutional variables shows that variables (Tax payers identification, Competence training, VAT assessment, VAT audit and Adequate manpower ) were positive and significantly correlated with VAT collection performance (r=0.7197, 0.7480, 0.6294, 0.5145 ,0.3463, ρ<0.05 respectively). But, two variables (External legal environment and penalization) were negative and significantly correlated with VAT collection performance (r=-0.3203,-0.2455, ρ<0.05 respectively). In general it's concluded that Value added tax have a great contribution in increasing government revenues, Hence the government authority/tax office has to focused on factors which negatively affect VAT and needs to take a corrective actions. | 2019-12-05T09:31:22.251Z | 2019-11-01T00:00:00.000 | {
"year": 2019,
"sha1": "2f8071d455719fb0ebb9c5d16736d7420b3d7265",
"oa_license": "CCBY",
"oa_url": "https://www.iiste.org/Journals/index.php/RJFA/article/download/50432/52089",
"oa_status": "HYBRID",
"pdf_src": "Adhoc",
"pdf_hash": "8da723bb372d4ff05ee85b9b7f06df56d27cebc5",
"s2fieldsofstudy": [
"Economics",
"Business"
],
"extfieldsofstudy": [
"Business"
]
} |
244189273 | pes2o/s2orc | v3-fos-license | An Obstructive Prostatic Urethral Calculus in a Patient with Urethral Strictures: A Case Report
Urinary calculi are the third most common affliction of the urinary tract only exceeded by urinary tract infections and pathologies of the prostate gland. Urinary tract calculi contribute to a major concern encountered in the practice of urology, it affects about 10-12% of the population with a variable incidence with respect to sex, age, occupation, geographical area, climate, dietary fluid intake, social class and race. Urethral calculus is always found on the site of prostatic urethra, bulbar and fossa navicularis. Primary urethral calculi are usually associated with urethral strictures, posterior urethral valve and a diverticulum. Urethral calculi represent 1-2% of all calculi in the urinary tract. This is a case of a 32-year-old farmer and fisherman who had a retrograde urethrocystography (RUCG) that showed an obstructive prostatic calculus, bladder wall calcification and thickening with contrast refluxing into the seminal vesicles bilaterally.
Introduction
Urinary tract calculi contribute to a major concern encountered in the practice of urology, it affects about 10-12% of the population with a variable incidence with respect to sex, age, occupation, geographical area, climate, dietary fluid intake, social class and race [1][2][3][4].
Primary prostatic urethral calculi are extremely rare especially the huge/giant forms and should not be confused with true prostatic calculi which develop in the acini or tissues of the prostate gland [7].
Urethral calculi are most commonly located in the prostatic urethra proximal to the narrow membranous portion [4].
Urinary calculi are the third most common affliction of the urinary tract only exceeded by urinary tract infections and pathologies of the prostate gland [3,4,6].
Urethral calculi usually originate from the bladder and less often from the upper urinary tracts, it is either primary or autochthonous urethral calculi arising proximal to urethral strictures or associated with a urethral diverticulum and urethrocele or secondary (migratory) urethral calculi from bladder or renal calculi [6,8].
Primary or autochthonous urethral calculi are less often seen in females and if seen are almost often accompanied with an underlying genitourinary pathology like stricture or urethral diverticulum [6,9,10].
Posterior urethral calculi were further classified into three categories by Swift Joly [11] as either: 1. Vesico-Urethral Stones: these calculi are located partly in the urinary bladder and partly in the posterior urethra. 2. Urethral Stones: these calculi are located in the urethra alone. 3. Urethro-Prostatic calculi: these calculi lie partly in the preformed cavity in the prostate gland. Prostatic urethral calculi occur more in younger male while true prostatic calculi are usually associated with older men aged beyond 50 years [12]. Urethral calculi are very rare in the female sex due to low incidence of bladder calculi and short urethra [7,13]. Predisposing factors for insitu formation of urethral calculi are the presence of strictures, diverticulum, hypospadias and urethral meatal stenosis [7,14,15]. Symptomatic prostatic urethral calculus often require treatment, this basically entail surgical removal following transurethral electro-resection loop or holmium laser [16].
Case Report
This is a case of a 32-year-old farmer and fisherman who presented with dysuria, frequency, terminal hematuria, reduced urinary volume, urethral discharge, feeling of a hard mass in the perineum and poor ejaculation for almost 2 years duration.
The patient is oriented and conscious, not pale, anicteric, not dehydrated, not in obvious respiratory difficulty and no pedal edema.
The case had a retrograde urethrocystography (RUCG) that showed bladder wall calcification and thickening with an oval opacity measuring about 25mm x 20mm x 20mm in craniocaudal, mediolateral and anteroposterior dimensions in the region of the prostatic urethra on the preliminary film of the RUCG (Figure 1).
Figure 4: Contrast image of an RUCG showing the annular strictures at the peno-bulbar urethral junction and bulbo-membranous urethral junction with an oval filling defect at the posterior urethral valve region; the calculus (left blue arrow). And contrast reflux in to the seminal vesicles bilaterally more marked on the left side (right and left red arrows).
All efforts to demonstrate the bladder became abortive due to patients' discomfort and excruciating pain during the examination.
Abdominopelvic ultrasound scan showed a thick-walled urinary bladder with calcific crust in the wall with bilateral hydroureteronephrosis (dilatation of the ureters and collecting systems of the kidneys).
The patient also came with result of microscopy, culture and sensitivity of the urine and urethral discharge that confirmed the presence of Schistosoma haematobium.
The patient was also been prepared for surgical removal of the prostatic calculus, but as at the time of the report he has not consented to the treatment option made available to him.
We report the radiological findings of this case due to its peculiarity in the practice of radiology and urology.
Discussion
Posterior urethral strictures are rare and are seen more in the male gender; this case is a male. Urethral calculi are also more common in the prostatic region following strictures or diverticulum; the index case is that of a prostatic calculus with associated strictures at the penobulbar and bulbomembranous urethra.
Posterior urethral calculi may follow infective conditions/processes, the index case had radiological, clinical and laboratory features of Schistosomiasis.
Urethral calculi may be associated with the occupation of the subjects as documented in most literatures; this patient happens to be a farmer and fisherman; these are occupational hazards towards infestation by Schistosoma. The index patient was a confirmed case of Schistosomiasis which is believed to have caused the urethral strictures predisposing to the prostatic calculus.
Urolithiasis especially in the urethra often present with varying symptoms like dysuria, hematuria, frequency, incomplete feeling of voiding, perineal hard mass, urethral discharge etc; our index case also presented with most features documented in the literature.
The radiological findings documented in the literature that is often associated with patients suffering from prostatic calculi may be demonstration of a calculus most often in the prostatic urethra, accompanying narrowing of the proximal urethra; strictures, bladder wall thickening, affectation of the upper tracts to mention but a few; the index case also had such findings with associated reflux of contrast in to the seminal vesicles bilaterally most likely from severe obstructive effect of the prostatic calculus that measured about 25mm x 20mm x 20mm in diameter.
Urethral calculi have been classified as either primary or secondary; the index case is most likely primary because no other associated calculi or features to suggest migration were demonstrated.
Long standing urethral calculus may be complicated by fistulous or sinus tracts, diverticulum and features of obstructive uropathy and subsequently obstructive nephropathy. The index case also presented with some of these features except for obstructive nephropathy.
In the literature, symptomatic prostatic urethral calculus often require treatment, this basically entail surgical removal following transurethral electro-resection loop or holmium laser. The index case was also advised on surgical removal of the prostatic calculus, thereby conforming to that reported in the literature.
Conclusion
Prostatic calculi may be associated with occupation of the patients, the commonest occupation in this environment is predominantly farming. Prophylactic intervention against Schistosoma may reduce the incidence of urinary schistosomiasis thereby reducing the probability of urethral stricture and subsequent calculi formation. | 2021-09-19T12:31:41.304Z | 2021-06-24T00:00:00.000 | {
"year": 2021,
"sha1": "6826bf948efe8c3170e34e508cb9a3b447d7222f",
"oa_license": "CCBY",
"oa_url": "https://www.auctoresonline.org/uploads/articles/1625245599JCRR-21-CR-178_Galley_Proof.pdf",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "6826bf948efe8c3170e34e508cb9a3b447d7222f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
247063750 | pes2o/s2orc | v3-fos-license | House screening for malaria control: views and experiences of participants in the RooPfs trial
Background The housing stock of rural sub-Saharan Africa is changing rapidly. With millions of new homes required over the coming decades, there is an opportunity to protect residents by screening homes from malaria mosquitoes. This study, undertaken in the Upper River Region of The Gambia, explores local perceptions of what a good house should provide for its inhabitants and responses to living in a house that has been modified as part of a randomized control trial designed to assess whether improved housing provided additional protection against clinical malaria in children (the RooPfs trial). Methods This descriptive, exploratory study was undertaken over 22 months using mixed-methods (informal conversations, observations, focus group discussions, photovoice, and a questionnaire survey) in a parallel convergent design. Analysis was conducted across the data sets using a framework approach. Following coding, the textual data were charted by a priori and emerging themes. These themes were compared with the quantitative survey results. The nature and range of views about housing and the RooPfs study modifications and the relationships among them were identified and described. Results The data were derived from a total of 35 sets of observations and informal conversations in 10 villages, 12 discussions with the photovoice photographers, 26 focus group discussions (across 13 villages) and 391 completed questionnaires. The study participants described a ‘good house’ as one with a corrugate-metal roof, cement walls (preferably cement block, but mud block covered with cement plaster was also an acceptable and cheaper substitute) and well-fitting doors. These features align with local perceptions of a modern house that provides social status and protection from physical harms. The RooPfs modifications were largely appreciated, although poor workmanship caused concerns that houses had become insecure. However, the long-term trusting relationship with the implementing institution and the actions taken to rectify problems provided reassurance and enhanced acceptability. Conclusion In developing housing to address population needs in Africa, attention should be paid to local perceptions of what is required to make a house secure for its inhabitants, as well as providing a healthy environment.
built of block or mud and cement plaster walls and metal roofs. In most instances, these changes are being undertaken by individual householders rather than by governments or private sector construction companies.
Improvements in housing have, in many areas, coincided with significant reductions of the malaria burden [2] and in areas of endemic malaria, the potential for further reducing transmission through housing improvements such as screening doors and windows has gained increasing attention [3]. Designing appropriate methods and metrics for evaluating the impact of such housing improvements on health and on malaria specifically is currently a matter for debate. Two recently conducted systematic reviews concluded that housing improvements may reduce malaria infection [4][5][6][7], resulting in the World Health Organization recommending the use of untreated screening of homes [8].
Alongside evaluating the effectiveness of housing modifications in reducing the malaria risk, several intervention trials have investigated individual and household views on the acceptability of these modifications. For example, a study on the social acceptability of two types of house screening interventions to reduce exposure to malaria vectors was conducted in The Gambia alongside a clinical trial evaluating the interventions' protective efficacy against malaria [9]. The study found that, besides reporting fewer mosquitoes, participants said that they liked the ceilings and screening as they prevented other insects and small animals from entering their houses and they felt 'more secure' in the screened environment. A recently published systematic review and metanalysis of randomized control trials of housing interventions to prevent malaria and Aedes-transmitted diseases reported that 11 of their selected studies incorporated a community acceptability component. A key finding was that house screening was perceived to enhance privacy as well as preventing the entry of mosquitoes [10]. These studies provided insights into the acceptability of the housing modifications within a trial context, but little attention has been paid to local perceptions of what features of house construction make a house desirable to live in, aside from the modifications undertaken as part of the trials, or how these 'disease prevention' modifications align with how and why house construction choices are being made in any given context.
It is widely recognised that to move from efficacy to effectiveness any disease control intervention needs to be accessible, affordable and acceptable [11,12]. Should a particular housing modification (or group of modifications) prove to have a demonstrable impact on reducing the burden of malaria, then it is important to understand the extent to which the modifications are not only acceptable in the context of a trial but are aligned with local views on what makes a house desirable to live in and have the potential to be incorporated into local construction practices.
This paper reports on a housing perceptions and experiences study undertaken in a rural area of The Gambia to understand local views about what makes a house desirable to live in; and to explore the experiences of living in a house that had been modified as part of a randomized control trial designed to assess whether improved housing provided additional protection against clinical malaria in children (the RooPfs trial) [13]. The trial was undertaken in an area of moderate malaria transmission with high coverage of insecticide-treated nets and indoor residual spraying. The RooPfs trial was a two-armed household-clustered randomized control study with 400 households enrolled in each arm across 91 villages with at least two control and two intervention houses per study village [13,14].
The housing perceptions and experiences study associated with the trial had three key objectives: (1) to explore local perceptions of what a 'good house' should provide for its inhabitants and what makes a house 'bad'; (2) to describe the most valued characteristics of a 'good house' and what is required (in terms of construction) to create those characteristics; and (3) to understand if the modifications introduced as part of the RooPfs trial were considered an improvement and which, if any, had contributed to creating a 'good house' .
Study setting
The housing perceptions and experiences study was undertaken over 22 months (April 2016 to January 2018) in the Upper River Region (URR) in The Gambia in the context of the RooPfs clinical trial. The URR is in the far east of the country, where malaria prevalence is the highest [15] and is one of the poorest regions in the country. Malaria transmission is seasonal, from July to December, and peaking in October-November. The rural communities are overwhelmingly farmers, with the predominant ethnic groups in the study villages being Fula (64%) and Mandinka (33%), with 3% 'other' [14]. In this area, polygamous marriages are a normative marital institution with a household often composed of multiple houses. Each house within the household is occupied by: (1) a man, (2) wife, usually with their young children, or (3) unmarried male youths. While men are usually the head of the household, the head of a house is often a woman.
For most of the twentieth century, the predominant rural house construction materials in the URR were mud and wattle or sundried mud bricks for the walls and grass for a thatched roof [16]. Traditional houses were round and constructed with open eaves (the gap between the top of the wall and the over-hanging eaves), two doors (front and back) and sometimes one or two small windows ( Fig. 1). Over recent years, in line with the changes that are happening across rural sub-Saharan Africa, thatched roofs in The Gambia are being replaced by corrugated-metal and there is an increasing shift to the construction of square houses using cement blocks for walls [16,17].
Only 'traditional' houses with a thatch roof and open eaves were eligible for recruitment into the RooPfs trial. There were relatively few of these houses since, in addition to the country-wide shift towards corrugate-metal roofs, the National Malaria Control Programme had, a few years earlier, run a campaign for home owners to close the eaves of houses to reduce malaria mosquito entry. The trial participants had not adhered to this campaign and represented a minority among the general population, often the poorest members communities in a poor region.
Prior to the modifications implemented by the trial, all enrolled houses had a single room, thatched roof, open eaves, mud walls in good condition and a front and back door. The RooPfs housing intervention consisted of removing the thatched roof and installing a corrugatemetal saddle-shaped roof and closed eaves with largescreened windows at the gable ends to help ventilate and cool the house (Figs. 2 and 3).
Original doors were also replaced with two screened doors (Figs. 4 and 5).
The intervention used locally available and purchased materials and the housing modifications were undertaken by locally employed masons and carpenters. House modifications were implemented during two dry seasons, from March to June 2015 and between December 2015 to May 2016. In July 2016, the occupants of all enrolled
Study design
The perceptions and experiences study reported here was a descriptive, exploratory mixed-method study employing a parallel convergent design [18], using qualitative (informal conversations, observations and focus group discussions-FGDs), participatory (photovoice-PV) and quantitative (questionnaire survey) methods.
Preliminary studies
Qualitative data collection started with a series of informal observations and conversations undertaken in April 2016 towards the end of the RooPfs house modification process in the remaining villages where the housing modifications were being completed. These informal conversations were held with the head of the enrolled intervention house, most of them women. While women were frequently the head of the house being modified, they were not necessarily the head of the broader household. This role was usually filled by their husband who, if present, also joined in the conversation. The purpose of this data collection exercise was to explore the roles of house dwellers and household heads in housing decisions, to gauge initial reactions to the modifications being made to the intervention houses and explore local perceptions of what constitutes a 'good house' . The data from these conversations were used to develop topic guides for subsequent FGDs.
Focus group discussions
FGDs were held with community members at least three months after the intervention had been completed in their village (Aug/Sept 2016). The FGDs were conducted to: (1) explore perceptions of what a 'good' house provides; (2) describe perceptions of the relationship between housing & health; and (3) explore initial perceptions and experiences of living in the modified houses. Participants in the FGDs were purposively selected from among households enrolled in the RooPfs study. Purposive maximum variation sampling, drawing on data from the clinical trial baseline survey, was used to identify villages and households likely to represent the range of participants involved in the RooPfs trial (using criteria such as predominant ethnic group in village, geographical location of village and number of households in village enrolled in the trial). Potential participants were approached at their homes by the Medical Research Council the Gambia (MRCG) research unit social scientist (AM) and social science fieldworker and asked if they would be willing/able to take part in a group discussion about their perceptions and experiences of living in a modified house. FGDs were held with one group of men and one group of women per village, between five and eight participants, and in a location and at a time convenient for all participants. Each FGD lasted 60 to 90 min with the discussion conducted in the local language, moderated by AM accompanied by a social science fieldworker. With the permission of the participants, the discussions were audio recorded and additional notes were taken by the social science fieldworker.
Photovoice
Photovoice (PV) data collection was conducted 12 to 24 months after the house modification (between May and July 2017) among additional villages, selected using the same purposive maximum variation strategy as employed for the FGDs. Photovoice is a participatory visual methodology that allows participants to identify and document objects, process and phenomena of relevance to them in relation to the topic of interest [19]. It has been used widely in public health research over recent years and is increasingly being employed in malaria research [20,21]. The method involves a series of steps to enhance its ethical use [19]. In April 2017, the potential study villages were visited by the study team and the photovoice activity discussed at a village meeting led by the Alkali (village head). All villages approached in this way agreed to participate in the study and the Alkali, together with members of the village, agreed on the selection of two participants from among the RooPfs study households (one male and one female) to act as photographers.
Following village enrolment, AM and a social science fieldworker returned to each village to carry out a sensitization and training visit lasting two days. During this visit, the two selected photographers received training on how to use the digital cameras and their ethical use. Following the training, the cameras were left in the village with the photographers (usually stored at the house of the Alkali). The photographers were asked to use the following two weeks to take pictures that to them represented a 'good house' and a 'bad house' . They were informed that, when the social scientists returned, they would be asked to upload their photos onto a computer and discuss the pictures taken. During these discussions, the photographers were asked to choose six pictures they agreed best represented a 'good' house, a 'bad' house and their perceptions of the RooPfs modified houses. During this visit, following the selection of the six photos, two FGDs were held with participants in the RooPfs project: one with men and one with women (mixed control & intervention houses). Participants were shown the six photographs and asked to discuss what they thought they represented and whether or not they represent a 'good' or a 'bad' house and why. The discussions also covered perceptions of how the RooPfs modified houses compare to 'good' houses and the challenges and benefits of living in a RooPfs modified houses. These FGDs were conducted using the methods described above in the post-intervention FGDs.
Qualitative data management and analysis
Audio recordings from the FGDs and PV FGDs were transcribed verbatim, translated into English, typed into a Word document and imported into Nvivo 10 for coding and analysis. Transcriptions and translations were undertaken by trained social science research assistants and quality checked by AM. Fieldnotes from the observations and informal conversations were also typed into a Word document and uploaded into Nvivo 10. All identifiers were removed during the transcription process. The selected photographs used in the FGDs were also uploaded into Nvivo 10. Analysis was conducted across the data sets using a framework approach [22]. The coding framework was developed a priori from the study aims and objectives as well as through codes and themes that emerged inductively from the data. The textual data were charted by the emerging themes with the charts subsequently being used, together with the photographs, to map the nature and range of views about housing and the RooPfs study modifications and identify relationships among them [23].
Questionnaire survey
Quantitative data collection was undertaken at the completion of the clinical trial in January 2018. The data were collected by fieldworkers from the trial study team as part of a survey conducted among all 800 study houses to assess the condition of the modified houses and to confirm with members of the control houses which modification they would like to receive. Two open questions of relevance to this paper were included at the end of the survey tool for those participants who had been living in the intervention (modified) houses. These were: (1) What do you like about the house modifications? And (2) What do you dislike about the house modifications?
The responses were extracted from the main data set and imported into an Excel spread sheet for descriptive quantitative analysis. The qualitative data were collected before the quantitative data and the two sets of data were analysed independently with the results subsequently compared for triangulation and interpretation [18].
Results
Qualitative data were collected from 23 study villages with quantitative data collected from all 91 villages recruited for the clinical trial (Table 1). During the data analysis it became apparent that there was little difference in nature and range of issues raised by the two predominant ethnic groups (Mandinka and Fula) and consequently the data in the results are presented across all study villages.
The questionnaire survey was administered to 391 households. However, for 8 of these households no data were recorded for the two open questions of relevance to this paper. The quantitative results are therefore drawn from 383 responses.
Perceptions of what makes a house desirable
In the initial informal conversations, FGDs and PV activity, the concepts of a 'good' and 'bad' house were discussed in terms of the structural qualities of the building and the environment that would exist inside a house built with specific materials and to particular standards. Both the type of construction materials and the quality of the build itself were key in shaping opinions about the likely nature of the internal environment and the desirability of living in any specific type of house.
Doors
Doors were a key concern for most participants. In the PV activity, pictures of doors appeared in all sets of photographs taken by the community photographers and were discussed at length in the FGDs and PV FGDs. The door illustrated in Fig. 6 is typical of the pictures taken by the photographers to illustrate a 'bad' door.
The key reasons for this categorization were the large gaps between the door and the door frame and between the door frame and the wall. Such gaps were of concern because they would allow the entry of insects and animals: When the door is not good you cannot call that house a good one because rats, all types of flies and even a snake can enter the house. PVFGD9-P4.
The gaps might also allow the entry of people who have not been invited into your house.
When you have such doors at your house when you lock your door and leave someone can tamper with your belongings in the house during your absence. Someone can come and open the door easily and steal so that makes the house unsafe. PVFGD3-P1. Furthermore, many participants across both types of FGD expressed concerns about privacy. If the gaps around the door were too large, then anyone would be able to see inside the house: The door had gaps and that shouldn't be. In a good house you should not be in either side and see the other side through gaps. Your house is your confidential place, but if it is not secured you cannot be protected. You cannot be healthy in such house where by what prevails outside prevails inside. Any house in such a manner is not a good house. PVFGD2-P9.
As one of the participants said, what is the point of a house without a door?
when a house has no door is better you sleep in the open. The reason to build a house, roof it and fix doors is for you to be protected but when there are no doors then you are not protected at all. PVFGD11-P11.
Furthermore, in one of the FGDs a participant mentioned that open doors not only allowed for the entry of human threats but also 'unnatural' threats such as demons and spirits; Having a good house with a door and a window is safety and prevention. As an old saying amongst Fula elders, if you are inside a house and close the front door and open the back door then "Satan" always enters through the back
Walls and roofs
The type and state of repair of walls and roofs were also frequently discussed in the FGDs in relation to what makes a good or a bad house. In the PV activity, pictures of roofs and walls (with open or closed eaves; Fig. 7) appeared in all except one of the sets of pictures chosen by the photographers to share and discuss with their community members. It was generally agreed that cement rendered or plastered walls were a key component of a good house. The primary reasons for this were that such walls would be secure, strong and durable, prevent the entry of insects and small animals, and not allow them to hide in any cracks inside the house.
When a house is not plastered, not only mosquitoes disturb you but also cockroaches and spiders can all be around to harm you directly or indirectly. PVFGD2-P10.
The non-plastered house can accommodate many insects and also the house lack quality. The guarantee of a house depends on cement because if not plastered you cannot be comfortable inside that much. PVFGD2-P8.
In addition, cement render, or mud plaster made to resemble cement render, was socially desirable and a sign of quality and status: When you properly observed the wall [in the picture] you'll think that it is plastered with cement instead of mud. When you don't have enough money it is more economic if you can plaster it like this…… When you are distant away from the house you'll think it is plastered with cement whilst it is plastered with mud and sprayed with white sand. PVFGD3_P1.
Open eaves were universally referred to as a sign of a 'bad' house. Key concerns about open eaves were the entry of mosquitoes and other small insects, rats, snakes and dust. An example of a picture taken by the photographers to illustrate their concern with open eaves is shown in Fig. 7, and is discussed in the following quote: This house [I] In the photovoice activity very few pictures were taken of a thatched roof. Where a picture of a thatched roof was presented, it was shown and discussed as a 'bad' house. In all the informal conversations that were held while the house modifications were taking place, as well as in the FGDs and in the PV FGDs, corrugate-metal roofs were seen to be a sign of a good house. Thatched roofs were associated with 'dirt' , 'rats' , 'insects' and low status.
It [a corrugate roof ] can protect us from disease. In a grass house many things can be hiding inside the grass that may harm you, if it's corrugate nothing can be hidden up there. FGD6_P42.
In addition, a grass roof required more upkeep and, as one participant explained, it was becoming more and more difficult to find the right kind of grass for the roof: A good house is always roofed with corrugate because people prefer corrugate than grass. By having a corrugate roof your husband is free from cutting grass to thatch your house. FGD4_P27.
Many female participants explained that it was difficult to keep the house clean if you have a thatched roof and they were concerned that a thatched roof conveyed 'low status' .
…if my relatives visited me in my thatched house I feel shy due to the bad conditions of the house …. FGD11_P68.
By contrast having a corrugate-metal roof provided a sense of pride: I am very happy to have this house because it is the only corrugate house in the compound…(Questionnaire respondent).
However, while corrugate-metal roofs were clearly more desirable, a few participants mentioned that thatched roofs tended to be cooler and were heavier so they were less likely to blow away in the high winds during the rainy season. In addition, where grass for the thatch was available it was possible for households to harvest it themselves. Drawbacks associated with a corrugate-metal roof were that it was noisy in the rains, hot in the dry season and expensive. Table 2 provides a summary of the key benefits and drawback of the different types of roof as described by the participants.
Clean, tidy, and aesthetically pleasing
The nature of the structural components of the house were discussed, not only in terms of durability and physical security but also in terms of being integral to the ability of its occupants to be able to create a clean, tidy, and aesthetically pleasing environment within the house. Even if the house was structurally sound, an untidy or unclean environment inside the house was discussed as being harmful to health and well-being. Untidy clothes or personal possessions could provide a hiding place for insects (including mosquitoes), snakes and small animals and facilitate the accumulation of dirt and dust, which were seen as harmful to health.
But when the house is not clean and things scattered all over you won't notice when scorpion, snake or even mosquitoes enter and hide inside. PVFGD3_P1.
When a house is not clean people don't feel healthy inside. You become inactive when you have visitors and yourself will not be comfortable. A house should be kept clean always. PVFGD4_P7.
In addition, in all three qualitative data sets, the importance of the house being aesthetically pleasing both on the outside and the inside was mentioned as a key feature of a 'good' house. Not only was this important for the direct comfort and security of the inhabitants but it was also mentioned as being an important signal of relative wealth. The response to the picture shared in PVFGD2 (Fig. 8) adds support to the idea that the appearance and content of a house are signals of wealth and social status.
This house is so good that I only aim it for myself. When you have such a house even if you put on ragged cloths people always feel that you have money. PVFGD2_P7.
In summary, from across the qualitative data sets the key emerging theme was that a good house will provide protection. Protection from natural (wind, rain, dust, light from lightning, insects and animals) and potential supernatural (spirits and devils) hazards as well humanmade threats (thieves and criminals); providing a secure environment that is structurally sound, durable, easy to keep clean and aesthetically pleasing. In such an environment people are safe from physical harm and disease and their well-being is enhanced. These attributes were unanimously described as being provided by a well-built structurally sound cement block house with a corrugatemetal roof-a 'modern' house in the URR. Such houses are perceived not only to offer more protection from natural and human-made threats, but are also more expensive to construct and as such are an aspirational goal for many of the households involved in the RooPfs trial. Living in such a house in the RooPfs trial communities confers social status on the inhabitants, providing a sense of social as well as physical protection.
The 'MRC' house modifications
The data on perceptions of the modifications to houses carried out in the RooPfs trial were drawn from the FGDs, the PV activity and PV FGDs and quantitative data from the end of trial survey.
Qualitative data from the informal conversations undertaken while the housing modifications were in progress suggested that while overall the participants were delighted with what they saw being done to their houses, there were a few concerns. The key positive improvements mentioned by the participants were that the houses looked beautiful (aesthetically pleasing) and strong, they had corrugate-metal roofs with closed eaves that would prevent the entry of dirt and insects and the expectation was that they would have more light inside and be easy to keep clean. There were concerns, however, related to the corrugate-metal roof not extending as low down the side of the house as the previous thatch, and the walls being exposed to the elements with potential consequences for the integrity of the mud walls, particularly during heavy rains. Citing the latter, several participants requested cement to render and protect their walls. There were also concerns expressed about the style of the back door (Fig. 4). The key concern was that it was not a solid door but rather a metal door frame with a central bar and covered in netting. It was feared that this door would let in too much light and cold air, and the netting could become torn.
In the subsequent FGDs, PV activities and the questionnaire survey, these positive and negative features of the housing modifications were key recurring themes. In all FGDs held at 3 months after the houses had been modified, participants raised some concerns about aspects of the modifications, while at the same time, expressing their gratitude to the MRCG for investing in improving their houses. While concerns were raised and discussed during the qualitative data collection, the end of study quantitative data suggested that the housing modifications were widely appreciated. Of the 383 participants who had lived in a modified house and provided a response in the questionnaire survey, 56% (214/383) had only positive things to say about the house modifications. An additional 160 participants (42%) had things that they liked and things that they disliked about the housing modifications. For example, many participants said they liked the doors because they reduced the mosquitoes and also the house was cooler at night, but disliked the walls not rendered with cement and so were getting washed away. Just 9 participants mentioned only things they disliked.
The positive responses in the end of trial survey are likely to have been influenced by the maintenance work carried-out during the trial by the local masons and carpenters at the request of the trial team. The trial implementation included a 'report and repair' system in which the condition of the intervention houses was assessed during and immediately after completing the modifications (May 2016) and again in August, 2016, and July, 2017 with repairs undertaken in December 2016, June 2017 and December 2017. Residents of houses in the intervention group were also encouraged to report any damage or malfunctioning of the interventions to the nurse field assistants who visited twice per week, from June to December in 2016 and 2017 (malaria transmission season), as part of the clinical trial.
A summary of the likes and dislikes of the modifications as expressed by the participants who had lived in the modified houses and took part in the end of project quantitative survey are provided in Table 3.
Corrugate-metal roof
In all FGDs the corrugate roof was seen as a positive asset because it meant the householder no longer had to find grass to repair the thatch, the house was lighter and cleaner inside and because of the status that a corrugatemetal roof bestows: I am happy with the MRC's support on these housing issues, before in our swampy areas there used to be grass called ''Nyantan'' which we used to roof our houses with and it takes almost ten years before changing it but now due to the low rainfall that type of grass does not grow. Our corrugate roofing does not get moldy and it can serve us for many years before changing it. FGD2_P13.
The happiness I have since I occupied the house is so great because before even if my relatives visited me in my thatched house I feel shy due to the bad conditions of the house….Now when my relatives come I don't feel shy because my house is in good conditions. FGD11_P68.
These sentiments were repeated in the PV activities where in all but one of the set of photos chosen by the photographers there were pictures of a corrugate-metal roof. A corrugate-metal roof was universally appreciated and preferred over a thatched roof (Fig. 9).
In this picture [ Fig. 9] it is a MRC house and is very good, good roof. Young and old, whoever sleeps in this house can be protected……. PVFGD9-P2. The information from the qualitative research was echoed in the responses in the quantitative survey. The most common response to the question: 'what did you like about the modified house' was the corrugate-metal roof (209/383; 55%). Reasons mentioned for liking the corrugate-metal roof were the same as those described in the informal conversations, FGDs and PV activity; that it was more durable and cleaner (less likely to become infested with insects and rats). A few of the questionnaire respondents (n = 15) said that they like the corrugatemetal roof because it was more secure. No one, in either the qualitative or quantitative data collection activities, expressed a preference for thatch over corrugate-metal roofing.
Despite the strong preference for a corrugate-metal roof, many participants in the FGDs and PV activities complained that the corrugate-metal roofs leaked during the rain: It is true that in any development there are constrains. My house leaks to the extent that all the beds get wet when raining and we have to move to another house ….
FGD1_P62.
A leaking roof was commonly mentioned as a problem with the corrugate-metal roof during the qualitative data collection activities (which took place in April, August and September 2016 and May and June 2017) but by the end of study survey (January 2018) only 23/383 participants mentioned the leaking as a 'dislike' . This is likely a consequence of the 'report and repair' system that operated during the trial, the last round of which was implemented in December 2017. Participants in the qualitative activities which took place earlier on in the trial are likely to have used the opportunity of having contact with MRCG 'staff ' to describe aspects of their house that needed attention: This is the right time to explain the problems we face with our new houses. We have no right over those carpenters because they are hired by MRC to do the work but we can lodge our complaints to you the MRC staff. FGD2_P20.
By the end of the trial most leaking roofs had been fixed. Interestingly, of the twenty-three participants who mentioned a leaking roof in the end of trial survey, 15 also expressed a liking for the corrugate-metal roofs and specifically mentioned the corrugate-metal roof in their response to 'what do you like about the modifications' . For example, a participant said they didn't like that their roof leaked but also said that "I like a house of corrugate and thank the MRC for giving me one".
While the corrugate-metal roof was clearly appreciated, there were also major concerns about the style of the new roof. A frequent complaint in the FGDs undertaken three months after the completion of the modifications was that the corrugate-metal roofs not only leaked but also did not extend far enough out from the walls of the house to give them protection from the rain.
MRC did not aim to cause any fault to our houses but hence you asked us to tell you what our constraints are towards the houses then that's normal. The houses were plastered with mud and the rain has washed it down and presently all the lower parts of the houses are soaked. I do spend the night with my wife and children in the newly roofed house but when raining we move to another house due to the leaking of the roof. Outside is wet, inside is wet and leaking we cannot be comfort in that situation. Presently two of us wanted to abandon our houses and move to another house due to these problems never did MRC aim it that way. We are therefore appealing for your urgent help. FGD5_P37.
In almost every FGD there were concerns that the mud walls of the houses would be 'washed away' by the rain causing their collapse.
[But] The house should not be odd like the house you modified for us…… The roofing should be lowered and cover some part of the wall to avoid rain water from soaking the wall which may lead to the falling of the wall. If the wall has a problem and fall down that can harm you and your children … FGD14_P83.
The only problem with the house is the roof. It's high on top, rain washes the wall down and as long as that is happening the wall may one day fall on us and that's why we abandoned our houses. FGD2_p9.
In two of the FGDs, participants reported that their walls had indeed collapsed: The problem with my roof is that it's high and without veranda. They brought only [one] bag of cement and give it my husband and he told them that the cement won't be Fig. 9 PV RooPfs corrugate-metal roof enough for plastering. I told them that the reason I am in a thatched house is because of poverty and you see the roof is lowered to protect the wall. They said to me that we will change the house to meet my demands. But when the house is ready the roof is so high without veranda. A bag of cement was given to me to plaster one side of it and as the roof is high rain washed the wall and the house fell down. FGD14_P83.
To prevent this type of accident happening more frequently, there was a widely held view that the MRC should provide sufficient cement render to apply to all of the mud walls of the houses since it was the modifications that had caused the problems.
We can roof the houses like the way you did but the reason why we didn't do it is simply because we can't afford the cement to plaster the wall. You came and changed the roofs and you did not replace them as they were, conceivably you can have a corrugate house without cement blocks but can't have one without plastering it with cement.
Your carpenters failed to lower the roofs and that caused fear amongst us. Some houses about four to six people sleep in there and that is why in mine when raining I don't sleep until the rain stops because I always think of the condition of the wall where my family is. FGD12_P81.
In all of the FGDs, participants requested cement so that they could render their walls and protect them from the rain: Only you have power to solve these problems the only thing we can do is to report the matter to you. The reason why you brought corrugate to roof the houses can make you buy cement for the houses too. FGD1_P5.
In response, the trial team provided one bag of cement to each house that requested assistance with this problem. By the time of the survey at the end of the study, only 26/383 participants mentioned that they still had a problem with the lack of adequate cement render.
Doors
During housing modification, several participants were concerned about the doors that were being installed and these concerns were echoed during the FGDs and the PV activity. During these qualitative data collection activities, the front doors were rarely mentioned and few of the photographs taken as part of the PV activity contained images of a front door. Any concerns about the front door related primarily to the way in which it had been installed, with a few participants reporting that there were gaps between the door frames and the walls, or the wooden frames being eaten by termites. By contrast, the back doors were a frequent subject of the photos and a topic of conversation in each of the FGDs.
Concerning the doors, the front doors are better than back doors and we would like you to change them for us.
If you cannot change them, you look for another type of net for back doors as the chicken wire is easily torn even by touching it. FGD7.
In addition to concerns about the thinness of the netting covering the backdoors that was easily ripped, by animals and children, the main concern about the backdoors was that they were insubstantial and open to dust, light, rain, wind and people. Figures 10 and 11 show pictures of a backdoor taken by one of the PV photographers. In each case participants in the PV FGDs described their concerns about the 'transparency' of the backdoors.
This door [ Fig. 10] is good and properly fixed but when there is heavy wind dust can enter the house through the door. A good house needs a good protective door which can prevent you from all sorts of harmful things. This door can prevent you from mosquitoes and heat but cannot prevent you from bad people and dust. PVFGD7_ P5. The houses that MRC modified here can prevent you from mosquitoes but when windy you cannot even stand in the house rather to sleep there. The doors can prevent mosquitoes from entering the house but when it is windy and dusty you vacate the house due to the dust. PVFGD11_ P7.
The house is good but the door is faulty because it's a screened door and even when it is raining water enters through. PVFGD7_P2.
..but the back door is transparent. The best thing to me when you are inside the house it should be dark and no one outside should see you inside. The way the door is made someone can glance at you inside and see everything clearly. PVFGD7_P7.
One of the participants in the FGDs said that she had been so worried about the transparent backdoor that her husband had covered it in corrugate: Yes, my husband changed it to a corrugate door.
FGD1_P6.
While the style of the backdoor was a concern, the way in which the doors (both front and back) had been installed was also causing some worries. Several of the photographs taken in the PV activity contained doors with large gaps around the frame (Fig. 11), gaps that would allow the entry of mosquitoes and other insects.
This house in the fifth picture [ Fig. 11] is somehow good but the only problem is the door which has a gap and cannot close well. …..that gap between the door and the wall if not sealed, not only mosquitoes but also many other things can enter through there to the house. PVFGD3_P11.
These concerns about the backdoor were, to some extent, echoed in the quantitative data at the end of the study. The most frequently mentioned dislike in the modified houses was the doors (72/383), but at the same time, substantially more participants (114/383) reported that this was a feature of the modified houses that they liked. Among those who reported disliking the doors, there were few concerns about the front door with only 12 participants mentioning concerns that primarily related to security due to ill-fitting doors and problems with termites that were eating the wooden door frames. The back door was more contentious, with 58 of the participants specifically mentioning that they didn't like the design of the backdoor. This was primarily because they felt that the screen design made the house too cold, it let in too much dust and too much light. There were also concerns about security with a few of the participants mentioning that thieves had entered the house through the backdoor over the previous months.
Reduction in mosquitoes and illness and the ITNs
In all FGDs, at least one participant mentioned either that there were fewer mosquitoes in their modified house, or that malaria among their children had reduced or both. This was perceived to be one of the main advantages of the house modifications: I am very happy with MRC because they provided us with good houses, free from leakage and the prevalence of mosquitoes reduced. We stay inside and sleep well.
FGD11_P71.
We are happy with these houses as they have saved us from constant visits to health centres due to malaria treatment. FGD5_P34.
As part of the trial, all participants were given a mosquito net and it is possible that this might have influenced the perceptions of fewer mosquitoes and less illness. In addition to reporting that they were paying fewer visits to the health facility participants also mentioned the presence of the ITNs: We are happy with the houses because all the houses have mosquito nets. FGD 5_P34 The presence of an ITN in the house was also the most common picture in the PV activity (Fig. 12). The presence of the net was universally perceived to be a sign of a 'good' house: In this fourth picture the bed is well organised and the net tugged round it. When the bed has net there will be no mosquito interference when sleeping. PVFGD3_P5.
Similar sentiments were found in the quantitative data. In the survey data the second most frequently mentioned reason for liking the house modification was the perceived reduction in mosquitoes inside the house mentioned by almost a third of the participants (118/383; 31%).
Interestingly, the data from the entomology conducted alongside the clinical trial suggested that the modified houses did not contain fewer mosquitoes than the unmodified, traditional thatch houses. However, enrollment in the trial did provide some participants with access to regular screening for malaria as well as being given an ITN. It maybe that there was less mosquito nuisance while sleeping due to the presence of relatively new ITNs: We are happy with the houses because all the houses have mosquito nets……Whenever our children are screened by the nurse they have malaria negative. FGD10_P64.
In line with the requirements of the ethical conduct of a clinical trial, all participants were taken through the informed consent process, providing them with information on the purpose of the modifications and the trial. This consenting process and additional trial benefits are likely to have had an influence on perceptions of mosquito nuisance and malaria incidence.
A lot of benefit is gained by sleeping in the houses such as reduction of sickness which is the purpose of the houses. FGD12_P73.
Responsibility for problems
While the overall response to the modifications was very positive, there were some considerable concerns about the type and quality of work that had been undertaken by the masons and carpenters hired by the MRCG to carry out the modifications. That is, problems with the houses were rarely blamed on the MRCG per se, rather they were blamed on the quality of the work undertaken by the contractors.
MRC knows that though am not saying that the contractors are not qualified but the beneficiaries are not very happy because many doors are not good, termites are grinding the roofing sticks to powder and the roof is leaking. FGD5_P38.
I am the one occupying the house but detected some problems there at the roofing level. When they roof the house I told them that water will enter from the corners of the house and they said no. I told them let me go inside and will throw water at the corners and see. We did that and the water entered then I asked them what happened? They say it had entered. When it is raining we know how we slept in our old houses and how we feel the same in the new house too. I told them that some of these houses will fall before rain stops and that's what happened. I told them that you will finish your work and go leaving us here to suffer with our houses. FGD12_P78.
Many of the houses required a considerable amount of maintenance, during the trial; fixing leaking roofs, mending the netting on the back doors and repairing poorly installed modifications (doors, gable windows and roofs). However, the report and repair system seemed to help identify and fix most of the problems so that by the end of trial survey most participants were happy with the house modifications and all participants in the control houses requested all modification components for their houses. It was clear from the FGD and PV data that most participants perceived that the responsibility for repairs or amendments lay with the MRCG.
Discussion
Th RooPfs housing perceptions and experiences study was undertaken alongside the RooPfs clinical trial designed to evaluate whether improved housing provided additional protection against clinical malaria among children living in a poor rural region of The Gambia [13,14]. This study focused on understanding local perceptions of what a 'good house' should provide for its inhabitants and what makes a house 'bad' , and on describing the extent to which the modifications were aligned with local perceptions and which, if any, had contributed to creating a 'good house' .
In The Gambia, as in many countries in sub-Saharan Africa, the past 70 years has seen changes in the design and materials involved in the construction of rural houses. For example, corrugate-metal was first used as a roofing material in the 1950s and has been gradually replacing grass as the material has become more widely available and householders are able to afford the cost. Mud bricks started to replace mud and wattle (krinting) walls in the 1960s and doors have also evolved with woven materials (bamboo or wattle/krinting) being replaced by one sheet of corrugate metal. The installation of ceilings, facilitated by the changes in roof and wall construction, became more common in rural Gambia in the early 1990s. The closing of eaves was encouraged in a campaign by the National Malaria Control Programme during 2013 following evidence of the impact on malaria [24] Today in the URR, the 'modern house' desired by the participants in this study, is one with a corrugatemetal roof, cement walls (preferably cement block, but mud block covered with cement plaster was viewed as an acceptable and cheaper substitute) and well-fitting doors. These features are aspirational for the study participants who were among the poorest households in the one of the poorest regions of The Gambia. Such a house provides visible signs of enhanced social status as well as increased physical security and comfort for the inhabitants. These data suggest that the changes that have been seen over the past 70 years in the Gambia, and elsewhere in Africa, have been influenced not only by social desirability, a modern house being a symbol of economic and social status as the materials need to be purchased rather than collected and constructed by the householder, but also by the enhanced security that houses with corrugate metal roofs and doors and solid block walls offer to their inhabitants. This sense of security is created through protection from the dangers posed to health and well-being by small animals, reptiles, insects and human as well as supernatural threats.
The findings that a 'well sealed' house is a desirable house as it offers protection from dust, rats, snakes, and mosquitoes, and helps keep the house 'clean' creating a healthy and socially desirable environment, is not new but rather echoes the findings of several other studies undertaken in The Gambia and elsewhere in Africa [9,10]. In their studies undertaken in The Gambia during the 1980s on the economic and cultural aspects of the use on ITNs, McCormack and colleagues found a clear preference for ITNs made of opaque materials that prevented the entry of rats, snakes and insects, protected against droppings falling from the roof and also provided greater privacy [25,26]. In these studies MacCormack and colleagues also found that the opaque nets were preferred because they helped protect against owl witches and spirits of the night [25]. A more recent ethnographic study undertaken in the URR and Central River Regions of The Gambia in 2013 and 2014, describes how in Gambian cosmology illness can either be caused by an organism inside the body detectable by biomedicine, or by supernatural forces outside the body that are invisible to biomedicine [27]. These supernatural forces include 'foul winds' , spirits and witchcraft. A well-sealed house that keeps out the wind is able not only to prevent illnesses caused by the 'cold' [27] but also prevent the entry of foul winds and spirits liable to cause illness and other harms. Security and privacy were also found to be important in a recent study undertaken in Tanzania of different types of housing designed to improve health in rural Africa [28]. In this study the timber clad house was preferred over the houses with bamboo or shade net cladding as they were perceived to be more secure, durable and to provide more privacy. In addition, double-storey houses with the sleeping space on the top floor, were preferred as they were perceived to offer greater protection from insects and crawling animals (snakes) as well as being cooler and providing more privacy.
The house modifications undertaken as part of the RooPfs trial were, in general, perceived to contribute towards creating a good, well-sealed house, with the corrugate-tin roof being most appreciated. However, there were concerns about both some of the modifications and the standards of workmanship. Perhaps unsurprisingly, the key concerns were around the modifications and standards of workmanship that were perceived to make a house less secure. For example, the small overhang of the corrugate-metal roof raised some concerns that the walls were more liable to rain damage. Poorly installed doors created concerns as they allowed the entry of insects, rats and snakes and could not be securely fastened against human or supernatural intrusions. However, by the end of the trial, many of these concerns had been addressed and most householders were happy with their modified house. The participants were all grateful that the MRCG had spent time and money on helping to improve houses, even if the houses weren't all perfect. It was widely agreed that the MRCG modifications had relieved a financial burden among these poorest with the equivalent re-roofing being unaffordable to trial participants due to price of corrugated metal.
The responses from the participants were overwhelmingly positive about the effects of the modifications on the presence of mosquitoes and malaria even though there was no evidence of this in the clinical trial data [14]. Participant awareness of trial involvement and the nature of involvement can affect participant responses and their perceptions of outcomes [29][30][31]. The process of trial enrollment creates awareness of the purpose of the trial which is likely to have a significant impact on participant perceptions and behaviours [30]. In this study the participants were aware that the purpose of the trial was to reduce the burden of malaria and so it is perhaps not surprising that several of the respondents in the FGDs reported that the burden of malaria had reduced. In addition, all sets of the PV pictures included a photograph on the inside of a house with an ITN with participants agreeing that a 'good house' required an ITN to cover the bed. This is perhaps not surprising when considering all participants were given a mosquito net. Several studies have shown that trust in the implementing organization can have unintended consequences in terms of expectations of the benefits of trial participation [32][33][34][35]. In the RooPfs trial it was clear that the long history of the MRCG in conducting health research in general, and malaria research in particular, had led to levels of trust and expectation among the participants that may have influenced their perceptions of the benefits and effects of the housing modifications. A complex range of factors influence the decision to participate in a trial and these 'trial effects' make it challenging to assess the extent to which the reports of trial participants are likely to reflect experience and perceptions under more routine conditions [30]. Nonetheless, at the end of this RooPfs study, all participants in the control arm requested the housing modifications, suggesting that overall these were acceptable to the study communities.
On current estimates, the population in sub-Saharan Africa will grow by 1.3 billion by 2035 [36]. This growing population will need adequate housing. To help ensure that such housing creates a healthy secure environment and fulfils local criteria for desirability, data is needed on the performance of different housing types as well as information on locally acceptable attributes. The data from this study suggests that in the URR of The Gambia, attributes that create a desirable house for local populations are those that provide a secure environment, but not necessarily a healthy internal one. However, modifications that create a more healthy internal environment but are perceived to decrease the integrity of the structure, or create a less secure environment are likely to be unacceptable to the local population.
Limitations
A key limitation of the study was that it was conducted in the context of the RooPfs trial implementation. As discussed, acceptability studies conducted in the context of a trial are likely to be affected both by social desirability bias and by trial effects. These effects were apparent in the data collected for this study (e.g., in the photographs taken in the PV activity which focused on those aspects of a house that were affected by the trial) and were interrogated as part of the analysis and interpretation of the data. While social desirability bias and trial effects were clearly present in the study, the presence of the 'report and repair' system encouraged participants to discuss their concerns about their modified houses during the qualitative data collection activities, in the expectation that these concerns would be taken on board and responded to. In addition, triangulation of data from multiple sources (informal conversations, FGDs, PV and questionnaires) and the longitudinal nature of data collection (revisiting communities across a 22-month period) allowed for the identification and interrogation of key recurring themes.
Conclusions
As the need for new housing in sub-Saharan Africa expands, interventions designed to create healthy houses need to consider local design preferences. In URR, householders were primarily concerned with the security and durability of their homes. Interventions such as replacing a short-lived thatch roof with a more durable corrugatemetal were perceived as providing enhanced security and were universally appreciated. Poorly-fitting doors and screened doors that could easily be torn caused concern. Where house modifications, or new house designs align with what is locally considered to be the features of a desirable house; providing security, built with durable materials and to a high standard, they are likely to be widely welcomed and accepted. However, interventions that make a house feel less secure for the inhabitants are unlikely to be acceptable and householders may make alterations, which potentially decrease the health of the indoor environment but enhance perceptions of security. | 2022-02-24T16:23:50.032Z | 2022-02-22T00:00:00.000 | {
"year": 2022,
"sha1": "50c9fd38bfe0d33ad28199ac9f6ccd9d6f065427",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Springer",
"pdf_hash": "44ee98ec50a4d23ab97faf98f3183ef58185208c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
59123177 | pes2o/s2orc | v3-fos-license | Electromagnetic and acoustic radiations before Kamchatka earthquake
The most probable cause of simultaneous anomalous radiations in electromagnetic and acoustic fields, different in nature, is intensification of deformation processes during strong earthquake preparation. To verify this fact, simultaneous observations of electromagnetic signals and acoustic emission in the frequency range from 0.1 Hz to 11 kHz were carried out in Kamchatka. In the result of the experiment, radiation bursts appeared about 24 hours before a seismic event with MLH=5.4. Such bursts were not observed on other days within this time of a day in August-September, 2013.
Introduction
Investigation of acoustic and electromagnetic fields during increased seismic activity were carried out in different regions [1,2], however, there are just a few simultaneous observations in natural conditions.The authors know only some cases of registration of anomalous radiations of these fields before seismic events with MLH≥5 [3].In the laboratory research [4] it was registered that during rock deformation stress, acoustic and electromagnetic radiations are generated.To detect natural electromagnetic and acoustic emissions assuming common deformation nature of their occurrence, simultaneous observations of these fields were carried out in the seismically active region of Kamchatka peninsular in August-September, 2013.
Observations of electromagnetic field variations were carried out at «Karymshina» site (52.82N,158.13E).Noise radiations were registered by a multichannel VLF-detector applying magnetic and electric antennas.The signals were recorded in a digital form.The sensitivity of VLF-detector magnetic component was not less than 2.10-7 nT/Hz½, and that of electric component was not less than 6.10-8 V/(mHz½).
We should note that the main noise sources of natural electromagnetic radiations in the North-East of Russia [5] are world lightning centers located in the near-equatorial region of the Earth.
Signals from the regional lightning were registered by VLF-detector at «Paratunka» site (52.97N,158.25E), located at the distance of 18 km from «Karymshina» site when the threshold level of 1 V/m was exceeded [6].
To register the acoustic emission, a complex based on a piezoceramic receiver (hydrophone) was used.It is installed at «Mikizha» site (52.99N,158.23E) at the distance of 20 km from «Karymshina» site by the bottom of a lake [7].The sensitivity of the hydrophone together with the pre-amplifier is the first hundreds of mV/Pa.The results of investigation of the acoustic emission in Kamchatka showed that acoustic radiation increase in the high-frequency range from hundreds of Hz to the first tens of kHz is determined by the intensification of rock deformations in the observation area including those associated with earthquake preparation [8].Thus, high-frequency acoustic emission can be applied as a sensitive indicator of intensification of the deformation processes preceding seismic events [8,9].
Detection of anomalies during electromagnetic and acoustic observation analysis
In the course of data analysis it was discovered that radiation bursts almost simultaneously occurred in electromagnetic and acoustic fields in the frequency range from 200 Hz to 11 kHz (Fig. 1, at the bottom) on September 1, 2013 between 04:00 and 09:00 UT at «Karymshina» and «Mikizha» sites.Such bursts were not observed on other days within the same period of a day.Azimuthal distribution of lightning strokes showed (Fig. 1, at the top), that their number increased significantly from the south-western direction (~ 2000 -2400).There were also radiations from the eastern direction (~ 450 -1350).It was assumed that the increased radiations were, probably, associated with deformation processes in the Earth crust.But there were doubts that the detected bursts are associated only with the intensification of lightning activity.To determine the real cause of anomaly occurrence, noise radiations registered at «Karymshina» site were compared with the direction-finding data obtained at «Paratunka» site and with the data of the World Wide Lightning Location Network (WWLLN) [http://wwlln.com].An on-line Earthquake Catalogue of Kamchatka Branch of Geophysical Service RAS [http://www.emsd.ru]was also considered.Fig. 2 shows the direction-finding observations at «Paratunka» site.Vertical lines mark the «anomalous» period which coincides with that in Fig. 1.It is clear from the graphs that the increased level of radiations is observed from the direction of 1800 -2700.Lightning locations from August 31 to September 2, 2013 were determined by WWLLN (Fig. 3).In order to check the reliability of the determined directions to the radiation sources, the real azimuths of «Alfa» radio station working at the frequencies of 11.90, 12.65 and 14.88 kHz were compared with those obtained during signal reception at «Karymshina» and «Paratunka» sites.The error did not exceed 3 degrees.Moreover, we also used the lightning sources registered by WWLLN, including the lightning source near Kluchevskaya sopka and Siveluch volcanoes which was observed on August 31, 2013 (Fig. 3).It is clear from Fig. 3 that lightning activity was quite high on August 31, on September 1 (during the anomaly) the lightning was not almost observed, and on September 2 it appeared again.The inconsistency of the number of received signals from lightning by WWLLN network and «Karymshina» and «Paratunka» sites is explained by the difference in instrumentation and radiation registration methods.It was established by the researches which had been carried out before [6] that in spite of the indicated differences, quite good agreement of the data was observed when lightning azimuths were determined.Thus, during the period of observation of anomalous bursts on September 1, there was no strong lightning.It is quite possible that some local radiation sources appeared, the power of which is sufficient to be detected by closely located electromagnetic sensors located at «Karymshina» and «Paratunka» sites but insufficient to be registered by remote sensors of WWLLN network due to signal strong attenuation.
Discussion of the obtained results
Based on the data of the on-line earthquake catalogue, analysis of the seismic state revealed that the earthquake with the energy class of К=12.3 (MLH=5.1)occurred on September 2, 2013 at 8:19 UT, that is about one day after the anomalous radiations, at the epicentral distance of 140 km from «Karymshina» site.For the earthquake energy classification, we used classes К according to S.A. Fedotov's scale [10].The relation of K to magnitude МLH is determined by formula МLH=(К-4.6)/1.5).We should note that this was the only earthquake with MLH≥5, which occurred in the first half of September in the south-eastern part of Kamchatka.Taking into account all the abovesaid, the following scenarios of anomalous acoustic and electromagnetic radiations are the most probabilistic: 1.A powerful remote source (not associated with the seismic event) became active.The radiation from it came from the south-western direction, in the result, VLF signal level increased during the time period indicated by the vertical lines shown in Fig. 1 and Fig. 2. But in this case, the cause of acoustic emission anomalies is unclear (Fig. 1).2. Remote source activity changed insignificantly but in the result of deformation processes, the propagation conditions changed (for example, the Earth's crust conductivity near the observation site changed) that caused intensification of the signals received from lightning.This fact may take place since considerable increase of electromagnetic field pulse flux was observed near the Earth crust fractures [11].3. A closely located source of electromangetic radiation, associated with deformation processes in the earth crust, appeared.This scenario is quite possible as long as radiations were observed at the same time both in acoustic and electromagnetic fields.Simultaneous anomalies of acoustic and electromagnetic radiations are often registered 1-2 days before quite strong earthquakes in Kamchatka.Based on the on-line earthquake catalogue, 11 seismic events with MLH≥5 occurred at the hypocentral distances up to 200 km from «Karymshina» site in 2013.From them, except for the case described above, radiation bursts were recorded before 6 seismic events (Table ) that results in 63.6% from all the earthquakes with MLH≥5.Examples of anomalies are shown in Fig. 4.
Conclusions
Thus, in the result of observations of acoustic and electromagnetic fields in 2013, it was discovered that more than a half of seismic events with MLH≥5 at the hypocentral distance up to 200 km are proceeded within 1-2 day interval by simultaneous bursts of acoustic and electromagnetic radiations in the frequency range from 200 Hz to 11 kHz.To the author's opinion, the most probable reason of occurrence of anomalous radiations in the fields different in nature is the intensification of deformation processes during earthquake preparation.
Fig. 1 .
Fig. 1.Simultaneous observations of acoustic and electromagnetic radiations on September 1, 2013 at «Karymshina» (VLF) and «Mikizha» (acoustics) sites.In the upper part of the figure is the azimuthal distribution of lightning stroke number AZ (dots) within a day.The azimuth for «Karymshina» site is counted clockwise from the northern direction.In the bottom part of the figure are the amplitude variations of the envelope of VLF-radiation US acoustic emission PS, in four frequency sub-ranges.Vertical lines indicate the time interval from 4:00 till 9:00 UT, within which anomalous bursts of radiation were recorded.
Fig. 2 .
Fig. 2. Direction-finding observations of electromagnetic radiations on September 1, 2013 at «Paratunka» site.Azimuthal distribution of the sources registered at «Paratunka» site is shown in the upper part of the figure.Hourly numbers of lightning strokes the radiations from which were received from the directions of 0 -360 degrees and 180 -270 degrees are in the bottom.Vertical lines indicate the radiation area from 4:00 to 9:00 UT.
Fig. 3 .
Fig. 3.A map with lightning strokes (circles) which occurred near Kamchatka from August 31 to September 2 and the earthquake (+) which occurred on September 2, 2013.Site locations are indicated by squares.
Fig. 4 .
Fig. 4. Simultaneous observations of electromagnetic (US) and acoustic (PS) radiations before earthquakes on May 3 (at the top) and on July 14 (at the bottom).Anomalous bursts of radiations are marked by ellipses.Arrows indicate the earthquake times.
Table .
Simultaneous anomalies of acoustic and electromagnetic radiations before earthquakes with MLH≥5. | 2018-12-15T03:43:20.572Z | 2017-01-01T00:00:00.000 | {
"year": 2017,
"sha1": "7169b979467c17cc01dd2e40aa3933eff407f03e",
"oa_license": "CCBY",
"oa_url": "https://www.e3s-conferences.org/articles/e3sconf/pdf/2017/08/e3sconf_strpep2017_03002.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "7169b979467c17cc01dd2e40aa3933eff407f03e",
"s2fieldsofstudy": [
"Geology",
"Physics",
"Environmental Science"
],
"extfieldsofstudy": [
"Geography"
]
} |
32209778 | pes2o/s2orc | v3-fos-license | Lost Trust: A Yellow Fever Patient Response
In the 19th century, yellow fever thrived in the tropical, urban trade centers along the American Gulf Coast. Industrializing and populated, New Orleans and Memphis made excellent habitats for the yellow fever-carrying Aedes aegypti mosquitoes and the virulence they imparted on their victims. Known for its jaundice and black, blood-filled vomit, the malady terrorized the region for decades, sometimes claiming tens of thousands of lives during the near annual summertime outbreaks. In response to the failing medical community, a small, pronounced population of sick and healthy laypeople openly criticized the efforts to rid the Gulf region of yellow jack. Utilizing newspapers and cartoons to vocalize their opinions, these critics doubted and mocked the medical community, contributing to the regional and seasonal dilemma yellow fever posed for the American South. These sentient expressions prove to be an early example of patient distrust toward caregivers, a current problem in clinical heath care.
IntroductIon
In the 19th century, yellow fever thrived in the tropical, urban trade centers along the American Gulf Coast. Industrializing and populated, New Orleans and Memphis made excellent habitats for the yellow fever-carrying Aedes aegypti mosquitoes and the virulence they imparted on their victims. Known for its jaundice and black, blood-filled vomit, the malady ter-rorized the region for decades, sometimes claiming tens of thousands of lives during the near annual summertime outbreaks. For these trade cities and the Gulf region, the harshest epidemics occurred in 1853 and 1878, when upwards of 10,000 and 20,000 people, respectively, died.
The yellow fever outbreaks suffered by the American South tested the region's scientists and policymakers, damaged the con-fidence of caregivers administering aid to patients and, as I will describe, disrupted an already fragile relationship between medical professionals and laypeople. Vocalizing their skepticism and frustration toward fever prevention and treatment methods, the critics of science communicated distrust and even publicly mocked caregivers in their efforts to resolve the epidemics, thereby heightening the marginalization of scientists in the late 1800s and deepening our understanding of yellow fever's impact on the South.
Medicine in the 1800s was far different than today. The 18th century discovery of cells and microorganisms using the microscope led to novel and seemingly radical concepts of the existence of miniscule disease-causing "germs." The resultant microbiological developments of the 18th and 19th centuries introduced more confusion than clarity regarding disease etiology at first. It was not until the early 1900s that bacteriology and parasitology cemented their reputations as providers of reasonable explanations about disease causation [1]. Before their acceptance, these theories existed, as Nancy Tomes has argued, as unappreciated challenges to long-standing beliefs about disease origin and transmission, making the complete acceptance of the new theories unlikely for multiple decades. In clinical settings, the ill and affected by disease, as today, cared more about viability of treatments and less about unproven shifts in scientific thought.
But changing disease theories also caused the public to go into a cleaning frenzy because microscopic organisms were being claimed as disease vectors. Desperate boards of health and greedy product swindlers publicized antiseptic products, promising cleanliness would end maladies. In a similar manner, today's accessibility of information increases patient preparedness and understanding of clinical diseases in some cases, yet in others causes heightened anxiety about physician treatment and patient outcome. Where perceived knowledge exists, so does over-confidence.
The majority of yellow fever cases occur in warm, underdeveloped areas like the burgeoning American South in the late 1800s and today in some African and South American regions, where the mosquitoes breed in stagnant water collections. Of those bitten and infected by Aedes aegypti mosquitoes carrying the flavivirus, a third ultimately perish. The others experience harsh fever conditions (elevated body temperatures, body aches, dehydration) and bouts with vomito negro, the coagulated blood expulsions often associated with the malady. At its worse, the disease causes the infected person to relapse several days after the initial fever symptoms disappear, resulting in an intensified illness and potentially organ failure and death. Gruesome and unpredictable, yellow jack remains a particularly feared disease, even though a vaccine is available. In the 19th century, when disease etiology was dominated by confusion and consternation, a lack of hope regarding yellow fever containment and treatment intensified the disease's ravaging.
Historians have frequently commented on the many economic and political consequences of yellow jack in the South, as quarantine and sanitation measures halted commerce and challenged trade relationships, seeking to maintain and purify oft-affected business centers and homes. Others have commented on its role in facilitating, by creating a necessity for, early public health movements across the South. Within this historical and social discourse exists an opening to introduce a new angle from which to view the epidemics of the 19th century: the patient's perspective of failing efforts to halt the disease's wrath.
tArgets oF crItIcIsm
Understanding the characteristics and effects of layperson distrust begins with an explanation of the rationale behind the public's focus on certain entities. During the hysteria of fever epidemics in the American South, two groups received the majority of attention from the public critics: boards of health and medical workers. These groups bore censure from the public for a variety of reasons, including untimeliness of response to outbreaks, unexplained disease preventive measures, and inefficacy to stop the malady. The public often linked these groups together as contributors to failed fever prevention and treatment, damaging the reputations of those in the medical community.
Boards of health emerged across the South asynchronously. The year 1793 introduced the first yellow fever epidemic to the United States, and the founding of health advisory boards closely followed or delayed slightly, perhaps until the threat of yellow fever was more realized. Consequently, boards of health spawned in 1800 (Norfolk), 1804 (New Orleans), 1808 (Charleston), and 1841 (Natchez and Mobile) [2]. Though founded to manage all diseases, including cholera, malaria, tuberculosis, and other common diseases of the time, the boards dealt primarily with yellow fever and the onslaught of epidemics across the region.
Like the disease itself, the boards' activities fluctuated seasonally. In most circumstances, they functioned as advisory boards and fund-channeling councils, working with city council groups to manage a town's sanitation and quarantine. In New Orleans, as an example, the Board of Health served as an administrative group composed of scientists and physicians that organized donations for Howard Association doctors and nurses, taxed citizens for sanitation measures, appointed health wardens, and established timelines for caution regarding outbreaks. In that role, the New Orleans board appropriated $2,500 for the Howard Association from $15,000 donated from citizens and pushed health wardens to "strictly inspect the streets, alleys, yards, privies, open lots and squares, of their respective Wards," during the 1853 epidemic [3]. However, these sanitation measures and ward checks took place only during epidemic months, causing some outspoken citizens to question the reliability of the boards as financial and resource allocators. Not until later would more frequent meetings take place, and by then, the inefficacies had created many skeptics.
The efforts by the boards of health to stop yellow fever appeared infrequent and unsuccessful to the public, and its dissatis- faction with the boards surfaced in newspapers and pamphlets. Utilizing tax dollars for their failed sanitation and quarantine projects, the Louisiana Board of Health made itself an easy target for dissidents. A cartoon posted in the Daily Item ( Figure 1) mocked the handling of fever cases in New Orleans by depicting sufferers from non-fever maladies, like a toothache and a broken arm, as yellow fever cases according to the Board of Health. By doing so, the newspaper vocalized its disapproval of the Louisiana Board of Health's handling of fever cases and the medical knowledge of the board's employed physicians. Another cartoon from the same newspaper frankly addressed the public opinion toward the Board of Health. The image showed a boot labeled "Public Opinion" kicking a man whose pants are labeled as the Board of Public Health ( Figure 2). Positing the Board of Public Health as a group worthy of castigation, laypeople became increasingly insolent toward medical officials in South.
Laypeople were not alone in their expressions of distaste for the boards of health in newspapers. Oftentimes, caregivers used newspapers to spread concern about the outbreaks to the public or suggest new treatments. Physicians and scientists also clashed regarding large-scale disease prevention, using newspapers as an avenue for volatile discourse. In New Orleans, drama ensued from a dispute between a local physician and the Board of Health regarding sanitation. Dr. J.S. McFarlane, an opponent of the health board, publicly proclaimed the actions and intentions of the Board of Health worthless. In a letter to the editor of the Daily Delta, McFarlane fumed, "Could any but a blind idiot ... fail to perceive that from the first moment the Board of Health went into operation, and commenced in its lustrations, that the mortality in our city has increased in a duplicate ratio" [4]. McFarlane did not include statistics with his claim, but the message was clear: the Board of Health had done little to weather the onslaught of fever in 1853. That year, despite sanitation and quarantine measures, 11,000 New Orleanais died -nearly 10 percent of the total population [5]. McFarlane continued his blasting of the Board of Health by referring to it as the "so-called Board of Health," calling attention to his view of the board as a waste of tax money and mental energy. Anger like his was widespread, and considering the board's actions trifling, he concluded, "[the Board of Health] can do no earthly good, but to stir up mud in the streets, and furnish a daily record of the dead, which, from the tone of the public prints, nobody believes" [3]. Occurring in a public medium, in this case a newspaper, the volatile clash of opinion -rather than a polite postulating of fever treatment typical of newspaper writings by physicians -helped fuel cynicism toward the Board of Health and chariness toward fever prevention and treatment measures. Yellow fever medical workers, peers of McFarlane, experienced reproach from the public similar to that the boards of health experienced. In comparison to the criticism of the boards of health, the censure directed at caregivers was often, though not always, more indirect. Unsure of science but positive of mortality, laypeople's reflections during the epidemics trended toward despondency and, incidentally, humor. Realizing -or in the case of humor, escaping -the realities of fever mortality gave patients and laypeople autonomy to demonstrate how they felt about the dilemma of yellow fever and the well-documented public attempts to quell its wrath.
Direct criticisms of medicine called attention to the flaws of science at the time and expressed disdain for its occasional successes. To be fair, not all sanitation and quarantine measures were total failures. Sanitation measures in fact helped remove some of the sewage issues in trade towns like New Orleans and Memphis, which had provided excellent breeding ground for mosquitoes. 1 But for most of the 19th century, the mosquito theory of contagion did not exist, and, once it surfaced, it remained highly controversial.
Following the 1878 epidemic, a Kentucky physician by the name of J.P. Dromgoole gathered fever methodology and perspectives from a wide variety of professionals and sufferers. The publication, titled Heroes, Honors, and Horrors, remains a legacy to the memory of the epidemic and the pain it caused. Further, it serves as a unique and illustrative source for under-standing the fever epidemics from the perspective of the caregiver while simultaneously providing insight into the patient experience. In an article, "Yellow Fever -Medical Mockeries," an anonymous survivor of Memphis recounted his experience with the illness and his evaluation of medicine's role in his recovery and, more broadly, future prevention. Midway through his commentary about fever as neither infectious nor contagious, he ranted, "Has medical science since the history of epidemics ever stayed a plague?" [6]. In the context of regional medical expertise toward tropical maladies, his stance matches the futility expressed by caregivers in the South in response to their inability to slow or treat yellow fever. Further, as Dromgoole's text aimed to publicize the successes (heroes) and failures (horrors) of the 1878 epidemic in order to learn from them, the Memphis survivor picked a high profile medium in which to offer such a scathing remark.
His words also communicated a sense of desperation toward fever measures. An excerpt from his article sheds light onto how lay people viewed the epidemics and ineffective measures to stop them: "We were uncleanly, saith the scientist, yearning to organize a force. We know nothing about it, saith the humane, painstaking nurse and experienced practitioner who has survived many epidemics, and so say we all down here ... It may be they will stumble on a preventive or a cure, but I pray you put not your trust wholly in them" [7].
When writing the article, he obviously knew of the city measures to organize work forces to clean streets and fumigate sickly homes, typical preparations for the fever season and response to the initial outbreaks.
He also must have known about the impassioned, circular medical discourse that sought to resolve the disease by producing scholarship for its treatment yet failed to actually determine a best method. Surely, he suggested, one of these large-scale approaches to refining disease theory might eventually lead to a cure. But if it failed to, he would not be surprised. The health officials had little to calm the onslaught of disease in the city, so why place faith in them? The Memphis survivor did not believe the boards of health and caregivers, though well-intentioned, deserved the public's trust. As a voice for laypeople and a direct commentator on the existing measures for fever prevention and treatment, this anonymous author opened a door to understanding some patient experiences amid medical futility.
Though direct criticisms of medicine like the survivor's existed, indirect denunciations were more typical. Such indirect commentary may be more important to understanding the layperson perspective than direct criticism because it assumes the reader's predisposed awareness of the author's point. Instead of responding to the science of the time, of which much is known and can be described, indirect patient commentary on yellow fever's unchecked raging reveals the mind-frame of a common person during an uncommon and epidemicridden medical time. This commentary had two forms: 1) sad dwellings on the sorrow of the epidemic; and 2) light-hearted escapes from the harshness of epidemic suffering. An analysis of layperson responses helps extend the futility experienced by medical workers to the patient, further illuminating the dilemma of disease in the yellow feverbeleaguered American South.
Despondency, an expression of sadness and vulnerability, affected the states of mind of laypeople in the South during epidemic times. A poem written by a New York Sun writer upon venturing to the South, for example, evoked this emotion. The poet visited Grenada, Mississippi, a place not far from the Mississippi River known in the late 19th century for its railroads and, until 1878, its freedom from the distresses of yellow fever [7]. Writing a poem, the visitor highlighted the despondency communicated to him by Mississippians struck by the new-tothem disease. The emotions elicited by the poem contribute to an understanding of how broad the issue of yellow fever had become by the 1870s and also how little patients expected from those in the medical field.
In the first stanza, the poet noted the unbiased nature of the disease in regard to age, wealth, or purity. In fact, the disease's high infection of people of various groups was perhaps the most distressing factor. Other diseases of the time, cholera for one, were linked to poverty [8]. Writing, "Beware, beware, oh! Men, you helpless lot -Though ye be pure as snow, it pardons not!" the poet reflects the ubiquity of fever infection [9]. The lines also reflect the unpredictability of treatment or recovery, making Southerners "helpless" to avoid the ravages of the disease.
A later stanza [9] reiterated the notion of gloom and reticence toward medicine. In this case, the poet described the fever as a plague no one on earth has the ability to stop, and only when God's will allows, will the sick be given the release of death. The lines are best read continuously: Naught here on earth can stop its awful bane, No tearful plea can give the doomed release; Terrible angel of supremest pain, Until God wills it, it can never cease. Beware, oh! Mortals, for in light or gloom, Before ye yawns a grim expectant tomb The line, "Naught here on earth can stop its awful bane," points to the inability of medical professionals and laypeople alike to limit the disease and its spread, prompting the poet to give the only advice available: "beware." Learned of frequent and inescapable fever outbreaks, the poet communicated the attitude of laypeople in Mississippi during the 1878 epidemic as fearful of infection and hopeless of prevention.
The New York Sun writer's solemn tone toward fever reality starkly contrasted the other dominant attitude: humor. The healing power of humor has been documented in recent years and received much acclaim. Because yellow fever patients and laypeople used humor as a reflective tool, and not a curative one, it has interesting consequences [10,11,12]. I contend that laypeople used humor to distract themselves from the gravity of the epidemics and mask their uncertainty toward available fever treatments and the medical professionals seeking to benefit the region.
Fever humor, from the few sources available, utilized tall tales and exaggerated observations for comical effect. Though hilarious, the jokes have deeper meaning. Contained within the wit lay an understated analysis of the epidemic situation. Take for example, an anecdotal story from the 1853 outbreak in New Orleans. That year, one count listed 8,000 people dead of a total permanent population of 159,000 [13]. Referring to the amazing death toll, a citizen offered a droll description of quarantine measures and mortality, in tandem. Saying, "as soon as a man arrived on one of the steamboats, the office of the Board of Health immediately took his name and entered it in their books as deceased, to save all the trouble in calling upon him again," the storyteller slights the Board of Health as an overwhelmed manager of disease. If the tale held true, the board was so inundated with fever cases that it was forced to shortcut normal procedure. An exaggeration, the storyteller's point remained undisguised: Death was abundant and the board was unfit to control it or even keep up with the unceasing deaths. Other jokes commented similarly, claiming companions would try to beat the fever to death by eating breakfast in the graveyard or getting measured for a coffin upon checking into a hotel [14,15]. Through the instances of humor during fever occurrences, the lighter side of the fever epidemic appeared and offered backhanded commentary on the inadequacies of the medical community.
Medical boards and treatment failures received the majority of attention from laypeople, but not all of it. For its polarizing effect on the public, medical quackery con-stituted another avenue through which the laity commented on its plight. Most commonly, the existence of quacks enhanced the criticisms of laypeople by allowing them to more generously discount all who participated in medicinal, and not homemade, cures. Able to apply blanket criticism to medical professionals and quacks alike, laypeople blended the two together, further disparaging yellow fever medical efforts.
Though of unclear linguistic origins, the term "quack" referred to cunning laypeople or unskilled -at least fanatical -scientists posing as beneficent salespeople or concerned physicians to spread their aid or, simply, pocket cash. Excelling in the arts of publicity, confidence, and persuasion and gaudily dressed to verify their stable finances for cautious onlookers, quacks in the 19th century used then-novel technologies to promise health and happiness [16]. These swindlers targeted the curious, the frantic, and the worried, hoping to spin a convincing enough tale to close a sale. As historian James Harvey Young nicely summarizes in his article "Device Quackery in America," quacks relied on impressiveness and public unease toward orthodox medicine to coax potential buyers [17]. Southerners in the 19th century were both susceptible to these quacks and cognizant of their tricks.
As the industrial revolution loomed, laypeople became interested in -and easily convinced of -novel scientific developments, even those marginalized by scientists. Promising universal health benefits, quacks advertised handsomely priced gadgets ranging from stretching devices to increase body height and compressing devices to reshape breasts to magnetic belts that could improve health, posture, comb hair, press clothes, and promote a luxuriant mustache, all in one [17]. For yellow fever sufferers, quacks offered a variety of treatments ranging from laxatives to help purge the malady from the body and disease preventives. Newspapers carried advertisements for these products, especially in the most fever-ridden cities. In the New Orleans Item, a vegetable compound known as Cascarets was advertised, proclaiming, "If you want to be safe against the scourge ... use a mild laxative that will make your bowels strong and healthy, and keep them pure and clean, protected against any and all epidemic diseases." Other goods promised to "keep the body free from poisonous germs," like Londonerry Lithia Water. These products offered laypeople alternative routes for recovery than provided by boards of health whose large-scale preventives and equally fruitless (and varying) treatments offered by physicians and nurses had filled public discourse [18]. Quack products functioned as a personalized medicine, and the recurrence of advertisements for quack products in newspapers indicates quacks' reasonably successful business and distribution.
If they were occasionally susceptible to quack products, Southerners were not completely blind to their guise. In fact, some laypeople rejected the products either because they were designed by outsiders who had never been to the region or the swindlers were seen as profiteers from the problems of the South. The Picayune observed these themes once saying, "Hundreds of quacks in different parts of the country are trying to introduce their medicines in yellow fever here ... Let them come to New Orleans and try it on themselves" [19]. Not entirely clear, the use of quacks in the Picayune could have either referred to actual quacks or regional outsiders attempting to help cure the disease. Dromgoole's Heroes includes in it some articles for fever prevention written by physicians from Ohio, Kentucky, and elsewhere, suggesting some North-South discourse regarding fever treatment existed. Yellow fever epidemics did in fact make national news, but Southern doctors believed they were the most adept at handling Southern maladies, following their reliance on the notion of a Southern medical distinctiveness, thus laypeople and medical professionals may have considered outsiders as quacks. Regardless, quackery had made an impact on Southern disease culture for its litany of advertisements in local newspapers. Whatever the success or failure of quacks to sell their products, we will see that during the times of utmost anxiety, in particular epi-demic years, the laity used knowledge of quackery to denigrate medicine by making both seem ludicrous in their attempts to quell the fever.
A sAtIrIcAL noveL And Its ImPLIcAtIons
Perhaps the most telling depiction of health professionals during the epidemic years in the Mississippi Valley was a satirical novel, Doctor Dispachemquic: A Story of the Great Southern Plague of 1878, by James Dugan. Written in New Orleans immediately following the worst epidemic to hit the city, the locality of Doctor Dispachemquic matchlessly exemplifies the seeds of disapproval for yellow fever medicine. The novel articulated popular views of irresponsible and incapable medical workers for public enjoyment. Now Dugan's work exists as a one-of-a-kind testimony to laypeople's view of unpreventable, unavoidable, and untreatable disease.
Positioned against newspaper clippings and excerpts relating the themes of skeptics, this cohesive and thorough work becomes an excellent historical indicator of patient distrust, deserving close analysis and consideration. The depth of sarcasm and cynicism toward disease management by officials and medical workers in the text far exceeds other sources, making it a telling reflection on the consequences of failing medicine. Further, it is only fitting that New Orleans, the most yellow fever-maligned city in the South, would produce a novel to satirize those determined to remove the disease. The failures of the local officials to eradicate yellow fever tested the patience and trust of laypeople, resulting in a mostly humorous depiction of medicine's efforts to squash the disease.
The social commentary provided by Dugan, at times outright and at others more embedded, hinges on the main character, Doctor Dispachemquic. The protagonist's semblance to notorious medical quacks suggests the public's acceptance of medical failure as commonplace, as well as a blending of physicians and quacks in popular culture.
Described as a young man wearing a profusion of jewelry -a depiction likely reflective of quackery -and a recent graduate of the "cold-water medical institution in a distant state," Dispachemquic's youth and non-Southern origin made him unheedingly stubborn, headstrong, and offensive to readers. Enhancing his repulsiveness to locals, Doctor Dispachemquic considered himself a great man surrounded by mediocrity and ignorance. When it came to science, Dispachemquic's skills went unsurpassed -at least in his own opinion. So unquestionable was his greatness that according to the author, "If ten thousand died under his treatment, it was because they were beyond the reach of science to save, and not from any defect in his system of practice" [20]. Selfconfidence on such an impressive scale allowed Dugan to create a physician nearly as egotistical as imagination allowed, thereby generating great comedy as readers compared Dispachemquic to doctors they perhaps encountered.
Moreover, his self-confidence and medical greatness exceeded the average, according to Dugan's characterization, and approached the "all-curing" promises of quacks. In reality, as shown throughout the text, Dispachemquic seldom cured a patient through his methods in the same way a quack's laxative likely served no medicophysiological purpose. The bridging of physician and quack to create Doctor Dispachemquic relied on already recognized popular ideas about medicine and hyperbolized them to appeal to readers, desperate to unload their frustrations. The satirical text then became an outlet for hyperbolized stereotypes of caregivers in order to lighten to mood and enjoy shared humor in the routine inadequacy of medical care.
The description of Dispachemquic as an obstinate doctor likely owed its roots to the public opinion of physicians at the time as unhelpful and unchanging. Besides proving no viable method for consistently curing fever patients, caregivers frequently submitted articles to journals and newspapers to compare yellow fever treatments with others, yet many remained entrenched in their own methods despite the discourse. From the patient perspective, this made scientific debates appear trivial, as seen through Dispachemquic's oft-excessive consideration for science during patient visits.
With Dispachemquic and his companion, Dr. deKwarantenus, Dugan derisively depicted physicians as more interested in science than the patient's well-being. When called to the house of an ill, obviously yellow fever-ridden patient being treated by his wife, Drs. Dispachemquic and deKwarantenus immediately halted her tending so they could consider the best course of action for the man. Instead of quickly and decisively settling the issue of treatment, as might be helpful and encouraging to the patient, the physicians painstakingly and circularly analyzed and reanalyzed the case of the patient. Meanwhile, the couple waited impatiently for a decision to be made, causing much anxiety to the wife who wanted only to place a damp rag on the forehead of the man [21]. In the end, the physicians departed the residence without taking action and offered no recuperative advice, needing more time to make a decision. Such detachment from the patient aligns with trends seen in caregiver occupational distresses, where confusion over the best treatment caused trepidation and concern for sentient healers. For patients, this may have appeared as uncaring. As Dugan noted once, "To his patients and nurses, and especially those whom the Association placed under his care and charge, [Dispachemquic] could be as obstinate and presuming as a pet pig; but in dealing with those whom he regarded as already established ... he was as plastic and sycophantic as a vagrant cur" [22]. As a representation of hyperbolic physicians, Dispachemquic and deKwarantenus symbolized a common perception of doctors as people concerned solely with appearing knowledgeable about science. Seen from the patient perspective according to the interpretation of James Dugan, the push for scientific understanding was seen as a sometimes overly frivolous motive, ignorant to the true needs of the ill.
As Dugan focused on slating yellow fever medicine, he spared few. The Howard Association was another primary target. The benevolent association channeled donated funds to fever sufferers by organizing physicians and nurses to epidemic cities. In the fictional world Dugan created, the Association had recruited Dr. Dispachemquic to New Orleans. The distinction was made mischievously by the author. Writing that Dispachemquic was employed by the "Association," Dugan allowed the reader to use his or her imagination and knowledge of public controversy to conclude the Howards' identity. Dugan also criticized private charities that collected and distributed provisions to the sick. Offering kindness, encouragement, food, clothing, and medicine to those in need, these groups sought to alleviate the stresses of rampant illness by providing for those who could not for themselves. Intentions of the groups, like the Ladies Good Samaritans or the Christian Flower Mission and Theological Association of New Orleans, notwithstanding, the author introduced a beautiful and wealthy young woman to facilitate to the reader his contempt for the seemingly self-serving and righteous societies. Traveling door-to-door, presenting and offering flowers to the ill to ease the pain of disease, the girl follows Dugan's model for beneficent groups as an overly loquacious, self-interested, and condescending Christian, unnecessarily eager to help the "poor and needy" [23]. Slights of beneficent groups in this way clue the reader to Dugan's disdain toward all groups related to fever containment, especially those with muddled intentions.
A long student of yellow fever and practitioner of its remedies, Dr. Kancurum, a final physician character in the novel, functioned as Dugan's voice against Howards and other groups. Skilled in his treatment of patients, Kancurum demanded respect from others, but dished out little to Howards and, in particular, the nurses they brought in. Howard nurses earned significantly more money at 3 dollars per day than through other employment, thus making nursing an attractive field. Such good pay and high demand brought in many would-be nurses, and Kancurum quickly dismissed many. "I have dismissed so many worthless nurses since this epidemic began," he stated in regard to his high frequent discharging of aids, "that the Association is beginning to think, I am rather hard to please ..." [24]. Kancurum disdained the lack of training for nurses (none was required in New Orleans) and had no patience for ineffective or obstinate nurses. With pay for nurses being so good, he found himself faced with unhelpful and self-interested women seeking to help little and earn much. Dugan's jabs make New Orleans seem as though it cut corners to circulate nursing aids to the dying ill at an unexpected cost to medical reputation. The source shows then how the public skeptically viewed the expertise, and perhaps even greed, of some medical workers during fever epidemics.
More subtle representations of medical foolery existed in Doctor Dispachemquic. The names of characters, for instance, directed criticism toward doctors and their work. The most obvious allusion to medicine through a character name is Dr. Dispachemquic himself, known for dispatchingpatients-quickly. Said to make more widows and orphans than the battles around Richmond, Dispachemquic's reputation for sprinting through patients earned him the reputation of an exceedingly, perhaps overly, quick worker. In this case, the name fits the man and his mission: the egotistical doctor who sought personal gain and glory from a career in medicine, seeking to see as many patients as possible per day, and proclaiming his methods to be at the forefront of nascent medical practice.
The name of his colleague Dr. deKwarantenus, meanwhile, refers to quarantine measures against germs and fomites, as well as his academic focus on germs and their spread. Giving him the title "Fellow of Border Rhelf and Professor of the Theory of Germs, Spores, and Fomites," the author used DeKwarantenus to represent a faction of scientists supportive of quarantine, despite its commercial drawbacks and inability to stay the fever. Characterized this way, Dugan used the physicians, names and otherwise, to unabashedly cast a negative light on physicians as selfish and indulgent people, cold to the realities of patient suffering. To affirm this, the author describes both their businesses as booming and, despite having lost many patients, Dr. Dispachemquic as "the happy possessor of a clear conscience." 2 For the trends of emotional detachment and selfish indulgence of physicians as profiteering quacks quick to move from a dying patient, James Dugan's work signifies a widespread unhappiness with the help provided by medicine at the time.
Between all the satire, the book manages an occasionally sentimental tone. At its core, the novel serves as form of comedic relief for an emotionally dilapidated region. Constantly at the whim of caregiver suggestions and surrounded by familial and communal mortality, a book satirizing the causes of anxiety likely offered some consolation to its readers, united in distress. A secondary layer deals with the encouraging sendoff Dugan gives to readers in the final chapter. After summing up the characters' lives -Dr. Dispachemquic and Dr. deKwarantenus went on to live happy, rich, and respected lives as professionals and scientists of the greatest type -Dugan settles in to a more serious topic: moving forward with yellow fever. His words say it all: But time will heal the hearts now lacerated with grief, and in a little while, even those who are now most inconsolable will smile again, and learn to pass almost unnoticed, the articles which are now covered with tears, in memory of those who have only preceded them [25].
After a long, satirical novel of this type, these words of kindness, patience, and understanding refreshed the reader and re-contextualized the book in the greater scope of disease. For Dugan and his peers, fever hurt, and having an imperturbable perspective and a light heart may help to look past the flaws of imperfect scientists.
For its expression of comedy toward the ineffectiveness of professional healers during the 1878 epidemic, Dugan's Doctor Dispachemquic satirized the state of medicine in the South while offering practical advice for those recovering from the year's terrors. As the publisher noted, the book "remorselessly depicts the follies and crimes of a certain class of people and so-called officials," allowing the reader to emerge not only "well entertained but greatly benefitted" [26]. Those goals were indeed achieved. In total, Doctor Dispachemquic illustrates the variety of criticisms the public made toward the boards of health, medical scientists, and workers and supports the notion that the South made communal effort to overcome, or at least persevere through, yellow fever outbreaks.
concLusIons And outLook
Yellow fever tested the patience of laypeople in the American South, resulting in expressions of distrust and criticism toward caregivers and caregiving organizations during epidemic times. Local boards of health, physicians, and nurses were targeted by these vocal laypeople in despondent and comical ways in efforts to manage the horrors caused by the disease. Direct and indirect, mild and scathing, the expressions of distrust covered the spectrum of elicited emotions and intended effects, all contributing to a notion of skepticism toward the failed efforts of officials to manage and treat yellow fever. Some wit and camaraderie slightly soothed the struggles of these skeptics, but more generally, people disliked the bleak outlook of fever prevention and management during the epidemic years of the 19th century. Contributing to an understanding of the regional effect on yellow fever and the reaction of patients to ex-pressed medical futility from caregivers, this notion of distrust helps elucidate the scope of yellow fever's impact on the people of the South on an emotional basis. | 2018-04-03T05:18:30.803Z | 2013-12-01T00:00:00.000 | {
"year": 2013,
"sha1": "57321ea5dc1c8e42db7cebf2e4bb6a4bbf01c52b",
"oa_license": "CCBYNC",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "57321ea5dc1c8e42db7cebf2e4bb6a4bbf01c52b",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
62150242 | pes2o/s2orc | v3-fos-license | Distributed Russian Tier-2 – RDIG in Simulation and Analysis of Alice Data From LHC
On the threshold of LHC data there were intensive test and upgrade of GRID application software for all LHC experiments at the top of the modern LCG middleware (gLite). The update of such software for ALICE experiment at LHC, AliEn[1] had provided stable and secure operation of sites developing LHC data. The activity of Russian RDIG (Russian Data Intensive GRID) computer federation which is the distributed Tier-2 centre are devoted to simulation and analysis of LHC data in accordance with the ALICE computing model [2]. Eight sites of this federation interesting in ALICE activity upgrade their middle ware in accordance with requirements of ALICE computing what ensured success of MC production and end-user analysis activity at all eight sites. The result of occupancy and efficiency of each site in the time of LHC operation will be presented in the report. The outline the results of CPU and disk space usage at RDIG sites for the data simulation and analysis of first LHC data from the exposition of ALICE detector [3] will be presented as well. There will be presented also the information about usage of parallel analysis facility based on PROOF [4].
Introduction
In the period of steady LHC operation there is permanent and high usage of GRID in the massive RAW processing and MC production. There is also quiet successful usage of GRID in the end user analysis. The positive results of GRID usage could be owed to many applications developed in the early years at ALICE Grid environment. ALICE has made an effort to consolidate all of these applications in a coherent set of monitoring and control tools. The intensive usage of distributed Tier-2 -RDIG computing resources with a stable operation of its has been realized means of this control and monitoring tools application also. What improvements of ALICE workload have been achieved in the preparation time before the first LHC data? First of all from the side of middleware development: 1. Migration of middleware services and working nodes to 64 bit versions of gLite3.2 [5], tuned under SL5 2. New features of VO ((Virtual Organization) boxes under support of gLite3.2 [6] 3. Migration from LCG-CE [7] to CREAM-CE [8] Iimprovement and new features developed by AliEn; 1. Automatic storage elements discovery 2. Update of data management under xrootd[9], 3. Workload management based at Job Agent(JA) schema 4. Implementation of job and file quotas for end-user analysis.
Results of RDIG -Tier2 operation will be shown in this report at the example of MC simulation and analysis data from detector exposition in pp beams of LHC. The status of GRID services in time of LHC data development and analysis at RDIG sites has to be adequate to these tasks. What we had in that time: a) All 8 sites are processing ALICE jobs under CREAM-CE only, from beginning of 2010. b) All these sites have been updated to the last AliEn version -2.18. c) RDIG ALICE Federation managed of SE under last xrootd version. It is seen that ALICE have used at the time of first LHC operation more RDIG CPU (40%) than it has been pledged in 2010 (28%).
DISK space availability
How intensively has been used the disk space data in the same time? Data of the disk space is presented in figure 2 as MonaLisa [10]
Network Operation
The network operation in LHC run is presented in figure 3(input traffic) and figure 4(output traffic). It is interesting to understand what number of processing events has been contributed by RDIG sites to whole statistics of ALICE. This information could be extracted from data presented in figure 5.
Figure 5 ALICE Report of computing usage in 6 months of first LHC operation
These data are data of different ALICE national federation contribution to processing and done events developed, simulated and analyzed by whole ALICE. It is seen from these data that 1128162 from 12988113 (8.7%) events have processing and 744895 from 7566281(9.8%) events have been done successfully by RDIG sites, only.
The share of these events from April to October between different RDIG sites is presented in figure 6 and percentage contribution of these sites one can find in figure 7. There is shown in these figures the different RDIG sites contribution from CREAM CE as well from LCG CE.
Data Analysing
It is interesting to understand what part of total running jobs has been analyzed at RDIG sites? The answer to this question can be found with help of data presented at figures 8 and figure 9. The contribution of all analyzed DONE jobs to all DONE jobs by whole ALICE is equal 5251232/8620577 = 61%. So from successfully completed 744895 events at RDIG sites there have been not less than 450000 events of analyzed jobs. Not less than 250 users of ALICE were analyzing data ay RDIG sites, including users from Russia and Joint Institute for Nuclear Research. There have been considered above the analysis done in batch. The usage of PROOF facility let ALICE users to do analysis interactively. And this facility was and is using extensively in ALICE for the data analysis.
PROOF
The new type ALICE Analysis Facilities, called AAF is a distributed PROOF cluster used for interactive parallel data processing. There is combined ROOT`s package PROOF, with settings XROOTD (ALICE SE), which is responsible for working with data, where the Packman ensure the timeliness software across a cluster. This new type of PROOF cluster was setup at JINR in Dubna and is called JRAF (JINR Russia Analysis Facilities), see Table 2. Figure 10 presented aggregated network traffic at this PROOF cluster.
In order to analyze data, it is necessary to have the data on the storage space PROOF cluster. In the case of AAF is storage directly to a local drive of each computer in the cluster. It was necessary to ensure copy data from Alien to the cluster catalog. To accomplish this task it was necessary to create so-called dataset, which is basically a list of files. Data are divided into two groups: a) official (real data, data from Monte-Carlo simulation), b) user datasets (datasets created by users). | 2019-02-14T14:17:08.124Z | 2011-12-23T00:00:00.000 | {
"year": 2011,
"sha1": "6dc74e438a31377ee42a4ee05ef17e48bd3f40b1",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1742-6596/331/7/072066",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "a002fc55a8f48b80858ef8496557f49dd9a20126",
"s2fieldsofstudy": [
"Computer Science",
"Physics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
254526620 | pes2o/s2orc | v3-fos-license | Physical imaging parameter variation drives domain shift
Statistical learning algorithms strongly rely on an oversimplified assumption for optimal performance, that is, source (training) and target (testing) data are independent and identically distributed. Variation in human tissue, physician labeling and physical imaging parameters (PIPs) in the generative process, yield medical image datasets with statistics that render this central assumption false. When deploying models, new examples are often out of distribution with respect to training data, thus, training robust dependable and predictive models is still a challenge in medical imaging with significant accuracy drops common for deployed models. This statistical variation between training and testing data is referred to as domain shift (DS).To the best of our knowledge we provide the first empirical evidence that variation in PIPs between test and train medical image datasets is a significant driver of DS and model generalization error is correlated with this variance. We show significant covariate shift occurs due to a selection bias in sampling from a small area of PIP space for both inter and intra-hospital regimes. In order to show this, we control for population shift, prevalence shift, data selection biases and annotation biases to investigate the sole effect of the physical generation process on model generalization for a proxy task of age group estimation on a combined 44 k image mammogram dataset collected from five hospitals.We hypothesize that training data should be sampled evenly from PIP space to produce the most robust models and hope this study provides motivation to retain medical image generation metadata that is almost always discarded or redacted in open source datasets. This metadata measured with standard international units can provide a universal regularizing anchor between distributions generated across the world for all current and future imaging modalities.
Supervised training of deep neural networks has proven successful for learning representations that are task specific. An important example of this is localisation and classification of tumors in mammograms 1 . In this context, neural network based models have the potential to outperform traditional Computer-Aided Detection (CAD) software that relies on handcrafted features 2 in the metrics of sensitivity and specificity. Deployment of such models in real clinical settings can aid prognosis for patients with breast cancer 3 where false positive results are common and patients often unnecessarily take part in a medical biopsy that carries risk as well as potentially psychological distress 4 . These models can also reduce routine workload for radiologists where nationwide screening takes place as well as enable screening to take place in countries where there is a scarcity of trained radiologists. Despite in-house testing success across research groups and even in industry, a ubiquitous pitfall with such models is a significant drop in model performance when released into the real world 5,6 . In the medical field such a lack of robustness is especially concerning as it translates into patient risk.
To further explore this issue of model robustness we must define some terms; 'train set' and 'test set' refer to the set of images an algorithm is trained and tested on. The terms 'source' and 'target' domain are more general and refer to groups of images. We could generate a train and test set from a source domain. A domain label is a vendor or hospital ID in this study. This generalization issue is one of the core challenges in the field of machine learning (ML) at the time of writing and is widely accepted to be caused by the phenomenon of domain shift (DS) i.e the image statistics of a new test set are different from what the model was trained on, clinically, training a lesion classifier on images from hospital A but using the model to make clinical inference for images from hospital B. This kind of test set is termed ' out of distribution' (O.O.D). In representation learning terminology the learnt features in the models internal representation are not general enough to be applicable to new O.O.D datasets 6 . Unlike other dataset generation factors, physical imaging parameters (PIPs), namely mechanical configuration, www.nature.com/scientificreports/ calibration, vendor and acquisition protocol within and between data collection sites are variable and directly measurable as well as routinely logged, but rarely used beyond calibrations (See Table 1 for examples). Optimization of PIPs in digital mammography necessitates maximization of the image signal-to-noise ratio (SNR), while simultaneously minimizing patient X-ray dose 7 . By gathering mammography images and their respective PIPs from multiple hospitals we will explore how they correlate with this issue of performance drop in deep neural networks.
To understand how PIP variation shifts image statistics we must look into the physics of projectional radiography which is well established. The generation of radiographic images can be described with an extension of Beer-Lambert's law to non uniform solids for a flat detector taking into account the inverse square law 8 : with S(E) the spectrum of the incident X-rays, E 0 the X ray energy at the tissue entrance point,E max the maximum energy of the spectrum, r the path through the tissue,θ the angle of the X-ray beam from the normal to the plane and η(E) the measurement noise 9 . I(i, j, E) is the contribution to the X-ray intensity at point i, j from the part of the energy spectrum E. µ(E, i, j) is the tissue attenuation at depth r and energy E for a beam that ends at point i, j on the detector (See Fig. 1).The ML community focused on overcoming the domain shift issue in medical imaging tasks tend to focus on x(i, j) and give the process G less attention despite it being well established symbolically. The parameters of the process can be rearranged as tissue dependent (denoted by a) and physical imaging (denoted by p) parameters G = G(a, p) . In a particular investigation of breast b the process becomes G = G(a b , p b ) as depicted in Fig. 1. In practice, p b is freely available but normally discarded. Our work focused on the effects of p b variation on model generalization.
For this large scale study we trained a Resnet50 11 convolutional neural network (CNN) for a task of age group classification from mammograms in a supervised manner. The two age groups were defined as under and over 58 years of age at the time of imaging (as 58 was the mean patient age in our dataset). We trained the models on Table 1. Table of PIPs available for images from the three Hungarian hospitals. These physical parameters are measured automatically by the machine at the time of imaging. There were many more parameters available for each dataset but they were not all universally named between hospitals so were dropped to avoid false comparisons. Each parameter has a direct or indirect relation to a term in Eq. (1). Some of these parameters are derived from others thus should hold some redundant information, this further justifies the PCA space analysis, see Figure 1. Physical description of the X-ray imaging generative process G(a, p) where a is the tissue of interest and p are the PIPs used for imaging of that tissue. The generated intensities (pixel values at positions (i, j)) of image x(i, j) of breast b are strongly dependent not only on the human tissue a b imaged but also the values of the PIPs ( p b ) used for that particular breast. This framework extends to all anatomy. Technicians attempt to correct for variations in these images but still the variation is present 10 .This process is non-invertible.
Related work
Due to its centrality in ML, domain generalization has well-established theoretical bounds 12,13 . Learning within the risk minimisation framework can be defined as: where h is a binary classifier, x is an image and y is its corresponding label, these images and labels are by definition from the source domain. ℓ is the loss function and E is the expectation over the entire dataset. X is the source domain distribution of images and Y is the source domain distribution of labels. The goal of learning is to optimize for a model h * . We can define a model error for a source ǫ S (h)(training set) and target ǫ T (h) (possible future test samples outside of the training set) as: where D S is the source domain distribution and D T the target. It is proven in 12 that, for perfectly labeled datasets; where d 1 is a measure of divergence of the domains. This bound states that, the larger d 1 is for a given problem setting, the larger the worst case scenario will be for the classifier generalization to the target domain. During medical imaging, physical imaging parameters (PIPs) vary within and between various vendors. These parameters are often recorded and used for image normalization so mammograms are visually consistent for physicians to read. It has been shown that variation in scanners can be subtle yet significantly affect model generalization 10 . Domain labels are still learn-able within multi domain image sets even after state-of-the-art image pre-processing and normalization, meaning residual signature domain information is difficult to remove from images.
In an effort to mitigate generalization error many domain adaptation (DA) methods have been formulated. These can broadly be split into three approaches; dataset alignment, dataset enlargement and representation alignment.
Dataset alignment: In 14 the authors propose a CycleGAN-based DA method for breast cancer classification. They use CycleGAN 15 to transform whole-slide mammogram test sets to match the style of the train set. These methods all aim to align the domain statistics between train x S , y S ∼ D S and test x T , y T ∼ D T sets but are only possible when the test sets are available at training time. This can also be described as the transformation of x T images to "look" as if they were sampled from D S originally, treating the true image generation variation as a learnable non-linear transformation.
Dataset enlargement BigAug 16 models degrade an average of 11% (Dice score change) from source to target domain, substantially better than conventional augmentation for medical segmentation tasks. In 17 the authors create new ultrasound physics inspired augmentations for segmentation and classification tasks. In 18 the authors use MRI physics augmentations for deep learning MRI reconstruction. These methods all aim to expand the source domain distribution to encompass target domain statistics.
Representation alignment These methods treat domain shift as a "style" change and force model internal representations to be agnostic to these changes. Domain Adversarial Neural Networks (DANN) 19 leverage the idea of a gradient reversal on a secondary task that is to predict domain labels in order to force a shared feature representation to contain no domain specific information. In medical imaging 20 , propose a DANN-based multiconnected adversarial network for brain lesion segmentation. Similarly, in 21 the authors work on learning MR acquisition invariant representations with the use of a Siamese loss function enforcing variation between scanners to be minimal while the variation between tissues to be maintained. In 22 the authors separate content and "style" with a Fourier transform to perform Federated Learning with shared styles without sharing the content. An important note and challenge for all of these approaches is that it is still not well established that variation in PIPs only drives high level "style" changes.
Data
The datasets used were collected from three Hungarian hospitals. Two open source datasets [23][24][25][26][27][28] were also used for methodological validation. The years of collection of the Hungarian datasets overlap (see Fig. 2) as well as their mean histograms of gray level values. We defined PIPs (see Table 1) as the list of physical imaging parameters available as metadata. These parameters were only available for the three Hungarian datasets. Each image has its own PIPs. We resolve the PIPs into groups related to known variables in the physics of the X-ray image generation process. These PIPs are in SI units and so are comparable across all vendors worldwide. All standard DICOM files contain this metadata 29 . The two other datasets did not include PIPs. The proxy task of age group classification was used to ensure very high ground truth label accuracy thus avoiding annotation bias. All datasets underwent the same routine mammogram image pre-processing pipeline. Overexposed images were removed automatically with the condition of the mean pixel value of the given image being over 150. The remaining images were first binarized to create a mask based on a threshold pixel value of 50 [26][27][28] . All images underwent the same pre-processing pipeline prior to experiments. Row 1: Example images*. Row 2: Age distributions for each dataset with 20 bins, x axis is age group. Row 3: Mean gray level histogram for each dataset after pre-processing on a scale of (0-255).*Data usage statement*-The images were chosen as the most "average-looking" so as to best reflect the broad dataset they were selected from. These images that have no distinctive signature features of interest or medical relevance, this includes breast implants and surgical markings. In order to anonymize images we downsized the 3000 px × 4000 px DICOM images to 244 px × 244 px resolution. Full description of datasets vendors and years of collection can be found (see Table 2). www.nature.com/scientificreports/ ; pixels with intensities over 50 out of a possible 255 were set to 1 and under 50 were set to 0, and then the largest binarized connected area (always the tissue for our images) was kept. This removed all unwanted image markings in the mask. The original, unbinarized image was then multiplied by the binarized mask creating a version of the original image that was cleaned of markings. The cleaned images then underwent the Contrast Limited Adaptive Histogram Equalization (CLAHE) 30 algorithm. Images with breast tissue on the right side were then reflected along the y axis of the image so breast tissue was always set to the left side. See Fig. 2 row 1 for examples. Finally the images were downsampled to 244 × 244 pixels and pixel intensities were normalized to values between − 1 and 1. Downsizing images does often reduce model performance for medical imaging tasks with CNNs as there are features that are lost in this downsizing process 31 . However, in this study all images are downsized from the same initial dimensions (3000 px × 4000 px) so it is fair to assume the relative reduction in accuracy between each hospital should remain the same. 244 pixels per dimension is not far from the optimal range stated in 31 .
The baseline networks were Image-net 32 pre-trained Resnets. They were trained with lr = 5 × 10 −4 and batch size 32 on 2 Quadro RTX 45GB Nvidia GPU machines. All model training was implemented in PyTorch (1.10) 33 . Training sets contained 1000 images per age group from hospital A, this was chosen as a compromise of training times and size of generalization gap, we performed tests to test this and understand how the generalisation gap changes as training sets varied (See Supplementary Fig. S1. for analysis on training set size and generalization gap). Test sets were 150 images per age group for other hospitals. These test sets were randomly sampled within pre-defined groups based on PIP values see Fig. 3. As the task was binary classification, all test and train sets contained an equal number of samples from each age category.
In this study we control for the known drivers of domain shift to understand how independent variation in PIPs translate to generalization error for deep neural networks. To control for population shift 34 we take images from the same country and ethnic group 35 . To control for selection bias 36 we make sure each data set is gathered with the same criterion, namely: the national mammography screening program in Hungary 37 , there is also supplementary open source data used for methodological validation. To control for annotation shift 34 we train our models on the proxy task of age group classification where we have perfect ground truth labels. The measured domain shift between the datasets can then be attributed almost purely to the variation in PIPs for each image.
Age prediction from mammograms: Breast density is a measurement of the ratio between radiodense epithelium and stroma to radiolucent fatty tissue. Dense tissue has generally been associated with younger age and premenopausal status, with the assumption that breast density gradually decreases after menopause 38 . These visual features give grounds for the potential for age prediction from mammogram images. This potential was realized in 39 where the authors achieved a validation mean average error (MAE) of 8 years and was proposed as a tool for filling in missing tabular data.
The networks were trained with the stochastic gradient descent (SGD) optimizer without augmentations with a cross entropy loss function; where y is the true age group label and P is the model prediction. Non-numerical PIP data was removed. Data from each hospital was combined and normalized with the standardscaler() sklearn function. Principal www.nature.com/scientificreports/ component analysis (PCA) with two principal components was performed on the tabular data for visualization purposes. A control experiment was run to see if the age groups were predictable from the PIPs with a random forest classifier. The model accuracy was 56.75% ± 2.19% (95% confidence interval) indicating PIPs were largely independent from the patient age group and thus age group classification was an appropriate proxy task for the evaluation of PIP values on model robustness (see Fig. 5d for visualisation of homogeneous mixture of points). Figure 3 illustrates the correlation found between sampled test sets position in PIP space and the increase in model generalization error (drop in age group classification accuracy). The "PIP space Euclidean distance from training centroid" is the distance in the PCA projection in PIP space between each hospital PCA PIP centroid and the A PCA PIP centroid (Fig. 3b). These are the results of a experiment where 1000 Images were sampled randomly from hospital A for training. The tests sets are taken from A (images not used in the training set), hospital B (sampled completely randomly), the C hospital which was deliberately separated into 3 test set sampling areas to illustrate that even within one hospital, generalisation error increases as a function of PIP space distance from the training PIP centroid. The results are the mean of 5 tests sets generated for each point with the error of one standard deviation for a given point. Figure 4 illustrates the learning for the task over 50 www.nature.com/scientificreports/ epocs A train manages to overfit. Domain in medical imaging is treated as a set of non-ordinal discrete labels usually referring to hospitals or vendors but we provided evidence that domain has some continuity due to its dependence on PIPs which are themselves continuous values. This bridging space is common for all existing mammography imaging devices worldwide and is more granular than a simple domain or hospital label. There is an equivalent PIP space for other modalities: MRI 40 ,Ultrasound 41 . In MRI some examples of PIPs would be Repetition Time, Time to Echo and magnetic field strength. For Ultrasound some examples would be frequency, dynamic range and incidence angle. In principle a PIP is any parameter that can be varied in G for any modality and may have variable effects on DS (Fig. 5).
In the radial experiment ( Fig. 6) a more granular approach was taken; test sets were sampled from "ring like" regions at a constant distance from the center ring where all the training examples were taken from. This allowed for better sampling as a function of distance in PCA space. Accuracies as a function of PIP distance from the training set were almost overlapping (excluding the point from the A training set at a distance of 4) meaning PIPs have some predictive power for preemptively assessing generalization error for a new hospital. Domain labels are useless for this prediction.
Discussion
The variation of PIPs do not largely show in image histograms post CLAHE yet this pre-processing tool is not enough to eliminate the subtle changes in image statistics as a function of PIP. This is concordant with literature 10 The PIP space PCA allows for feature ranking. Figure 7 outlines further exploration into the feature importance of each PIP in its ability to describe DS. We see many PIPs are of a similar importance further emphasising the need to use some kind of full PIP space or manifold to provide good distance measures that we have shown correlate with DS. Due to this even importance of PIPs, no obvious PIP seems particularly dominant for explaining DS. Entrance dose, relative exposure and body part thickness are correlated and contribute strongly to PC1. Detector primary angle, detector temperature and X-ray current are somewhat correlated and strongly contribute to PC2. The Fig. 7 subplot shows that to explain the full variance of PIPs we may need to take into account more dimensions for our PIP distance measure to be used as a predictive model for DS but even with the first 2 eigenvectors from the PCA we observe PIP variation is a driver of DS.
The proxy task allows for independent analysis of PIP effects on model performance but the domain shift will vary based on the task at hand. Medical tasks are the tasks of real interest so it is important to see if this effect is present for them. We hypothesise it would be as both medical tasks and age detection rely or learning a representation from features of the training sets. With a relatively low frequency of positive cases in the nationwide screening, the number of positive (benign or malignant mass or calcification) images we have access to are limited. Within this image subset we do not have a very large number of biopsy verified images such as in DDSM or CBIS. Conversely we only have access to PIPs for our internal datasets not DDSM or CBIS. We would like to run these experiments on the task of classifying benign vs malignant masses but do not have enough www.nature.com/scientificreports/ reliable data where both PIPs and verified classes are available. If we did have access to this data we would still not be able to rule out variance in radiologist experience between hospitals as well as other annotation and collection biases. We believe that variation of PIPs is one of many drivers of DS. These drivers are only separable with large controlled studies. The accuracy drop itself is seen in many studies which are medically relevant tasks 1 for example where training for lesion localisation and classification training on DDSM and testing in INBreast sets. We also cannot say how big a part the PIP driven domain shift will be in more clinically relevant tasks but we believe that the way CNNs extract features between our task and medical tasks is fundamentally the same so there is no reason to think PIP variation has no part to play in the domain shift observed in medically relevant tasks. Currently further investigation into this is limited by the issue of data availability. We explored leveraging these PIPs with a Domain Adversarial Neural Network (DANN) 19 architecture with the domain prediction changed to a regression task of PIP prediction however, we did not see any improvements over state of the art methods described in the related work section. As G is non-invertable, images generated cannot be fully separated from their PIPs with any mathematical transformations or learnt approximation. Concretely, one cannot gain information present in an MRI image with a CycleGAN transformed CT scan or vice versa 43 . Despite this knowledge, integrating PIPs into conditional CycleGANs 15 or Pseudo-physical augmentations 17 may improve model generalization despite not being fully rigorous and physically aware methods but rather "physically inspired" solutions.
Despite further work required to separate and fully categorise the contributions to domain shift 34 that effect model generalization in medical imaging tasks the results we present in 3 could be a useful tool for predicting the worst case generalization scenario when we have full knowledge of PIPs in the training set as well as full knowledge of PIPs from the hospitals or vendors where the model is deployed. Concretely, if we trained our model on data from hospital A we could go to hospital B and C and give an approximation for the reduction in accuracy of our model with knowledge of the distance in PCA space between hospitals generated data and the training data. This type of worst case scenario would be possible to deploy today if PIP data is preserved at sites. www.nature.com/scientificreports/ Adding this quantifiable error into model predictions could give clinicians more confidence in model outputs. This could be paired with other tools such as visualisations of variations of concepts that change model predictions for a given image 44 or any other interpretability aids.
Our results lead us to believe that a more homogeneous sampling process of images in PIP space may provide better generalization to unseen images as these may lie inside this "evenly tiled" PIP space if the training set is from a diverse enough set of vendors. In the case of models trained on data from one vendor, hospitals may chose which models to use based on their mean distance from the training sets used for different models, for example if hospital C was presented with two models, one trained on A and one trained on B it would be advisable to pick the model trained on data from B see Fig. 3. To extend and transfer this knowledge to the area of federated learning; if we train on data from multiple hospitals in a federated regime there may be large unbalanced clusters in the PIP space and we may end up with sub-optimal generalization power as models are bias to learn well in these PIP manifolds where there are many examples but not outside. Further work could be done to explore this hypothesis and may aid the advancement of federated learning as a regime to enhance generalization. It may well be that "more data" does not necessarily mean better model generalization for medical imaging tasks, however, this is speculative.
Conclusion
We have presented a unique study concerning the effect of physical image generation parameters on domain shift as well as provided evidence that the bias sampling of images with particular PIPs causes significant covariate shift of up to 10% accuracy for our proxy task. We believe this bias is still present in real medical tasks that are at the center of this field's interest and where this technology aims to succeed in deployment.
We found the PIP PCA space to be a helpful projection to understand and predict domain shift for unseen medical images at a more granular level than domain labels. These labels are physically concrete and can be related to the image generation process which is universal and well understood across all imaging modalities. Federated Learning will inevitably increase robustness of models as data will naturally be sampled more globally from PIP space in the training process. A homogenous sampling from PIP space may be more important than an even sampling between hospitals when attempting to optimize for the most robust models. This could be a direction for future research.
A more general point we would like to make is we suggest groups are more mindful about conservation of original non-private metadata for publicly released datasets. We commonly observe destruction of metadata in open source datasets but more effort should be made to retain and organize seemingly useless metadata as future research could leverage it. We were fortunate to be able to study this effect directly because in the Hungarian datasets the anonymization focused only on redacting the personal data and PIPs were not deleted.
PIP information may also be useful for algorithms that are less label dependent. PIPs are free labels describing the generative process of a given image so may be leveraged in unsupervised and self-supervised algorithms.
We also hypothesize that representations that are agnostic to PIPs would generalize better to images generated from a different set of PIPs and that leveraging PIP metadata may assist future models to be more robust and therefore more reliable when assisting physicians with important decisions concerning patients in the real world.
Data availability
The CBIS-DDSM dataset is available online at http://marathon.csee.usf.edu/Mammography/Database. html. The CMMD dataset is available online at https://wiki.cancerimagingarchive.net/pages/viewpage. action?pageId=70230508 The datasets acquired from Hungarian hospitals was used with a special license therefore is not publicly available, however the authors can supply data upon reasonable request and permission from the hospitals. All code is available at: https://github.com/csabaiBio/PIP-variation-drives-DS. | 2022-12-11T16:20:27.636Z | 2022-12-09T00:00:00.000 | {
"year": 2022,
"sha1": "5199f691342ea6a114e23f56db79f660b672877b",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "75a9481d8a3d6ac3b73be8f0fc38904d89b02a7d",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
16345191 | pes2o/s2orc | v3-fos-license | Spin beyond Standard Model: Theory
I use spin as a guide through the labyrinth of possibilities and ideas that go beyond the established understanding of the fundamental interactions.
INTRODUCTION
What is required to define spin are Lorentz-covariant free particles and quantum mechanics. On the other hand, causality (from demanding a Lorentz invariant S-matrix) and the cluster decomposition principle (the requirement that distant experiments yield unrelated results which in relativistic theories naturally leads to the concept of quantum fields) are not necessary. Thus, spin can be considered even outside the framework of a quantum field theory (QFT) -but not conversely -and can serve as a very general organizing principle for the many types of physics beyond the Standard Model (SM) that have been suggested 1 , including extra dimensions, strings, and M-theory.
The complete classification of unitary representations of the inhomogeneous Lorentz group according to Wigner [4] is recalled in Table 1. The little group is defined as the subgroup leaving some conveniently chosen "standard" four-momentum, k µ , unchanged and gives rise to spin and helicity. Thus, for a massive particle the possible spin states are obtained from the SO(3) Lie algebra and are consequently identical to those familiar from non-relativistic quantum mechanics. The little group for massless particles is the Euclidean group in two dimensions, ISO(2) (generated by two translations and one rotation), and (being non-compact) permits, in general, a continuous parameter ("continuous spin"). One assumes (basically on phenomenological grounds) that physical states are non-trivially represented only with respect to the compact SO(2) = U (1) subgroup, and identifies its "charges" with the helicity, h, of the massless particle. The trivial U (1) Lie algebra would allow arbitrary values of h, but the topology of SO(3, 1) = SL(2, C)/Z 2 can be shown to be that of the doubly connected space R 3 × S 3 /Z 2 , so that the requirement of a globally defined wave function restricts h to integer and half-integer values. Still, Lorentz invariance requires consideration of the entire ISO(2) little group. I will return to this point in the discussion of states with h ≥ 1. The spin 0 category includes the (would-be) Nambu-Goldstone bosons of spontaneously broken continuous symmetries, such as Higgs bosons, familons (broken family symmetry), Majorons (broken lepton number conservation), or axions (broken Peccei-Quinn symmetries). Furthermore, there may be radions (graviscalar components of the metric tensor in models with extra spatial dimensions), dilatons (e.g., in string theories), moduli (scalars with flat potential in supersymmetry), inflatons, scalar leptoquarks, and sfermions (scalar superpartners). Spin 1/2 states beyond the SM could be due to a fourth family, right-handed neutrinos, other exotic states (e.g., those needed to cancel anomalies in models with new gauge symmetries), techniquarks from dynamical (strong) symmetry breaking models, or X-inos (fermionic superpartners other than gravitinos). Spin 1 states could be extra Z ′ or W ′ bosons or vector leptoquarks (e.g., from Grand Unified Theories). Examples for spin 2 particles are the graviton (predicted by string and Mtheory) and its Kaluza-Klein excitations (in models with extra dimensions). A massless particle with spin > 2 is not expected to give rise to a long-range force, but the towers of massive string excitations include states with arbitrarily high spins. On the other hand, a massless particle (or massive particle after spontaneous supersymmetry breaking) of spin 3/2 would uniquely point to the gravitino of local supersymmetry (supergravity). Results from Higgs searches at run II of the Tevatron [6]. The meaning of the lines and bands is as in the right panel of Figure 1. Notice, that the right-hand plot is from a slightly larger data set.
SPIN 0
The unique fundamental scalar within the SM is the yet to be discovered Higgs boson.
SPIN 1/2
One way to address the hierarchy problem is to avoid fundamental scalars altogether. In the original technicolor (TC) idea [7,8] (techni)fermions condense through non-Abelian gauge interactions, break electroweak symmetry (EWS), and generate the masses for the W and Z bosons. It is less straightforward to obtain the masses for the SM fermions and extended technicolor gauge interactions are needed. In general, this leads in turn to large flavor changing neutral current effects in conflict with observation. This can be cured by decelerating the running of the TC coupling (walking TC) so as to effectively decouple the extended TC gauge bosons. Another complication is the large m t value and one considers models of top quark condensation (topcolor). Models in which the EWS breaks solely due to topcolor predict too large an m t so that one arrives at hybrid models (topcolor assisted TC). One also has to introduce an extra U (1) ′ gauge symmetry to prevent condensation of the much lighter bottom quarks. This is one of many examples in which a model of EWS breaking predicts a specific massive Z ′ boson to solve some model building problem. Thus, Z ′ diagnostics may be an additional analyzing tool for the new physics that breaks EWS. However, models of TC are generally in conflict with the S parameter ( Figure 3). Incidentally, the S parameter also constrains a fourth fermion generation and rules it out if mass degenerate (S = 2/3π). The non-degenerate case is also disfavored and only marginally consistent for specific mass values. Another way to avoid fundamental scalars is to assume that the Higgs field is composite [9]. A modern reincarnation of this idea is Little Higgs Theory [10] with a fundamental scale Λ of 5 to 10 TeV, and where the Higgs is lighter since it appears as a pseudo-Goldstone boson. In these models the quadratically divergent contributions to M H are then postponed by one loop order (in some models by two orders). One can also assume that quarks and leptons are composite. This introduces effective contact interactions with an effective scale constrained by LEP 2 to satisfy Λ O(10 TeV).
In a QFT, a particle with h = ±1 can produce a long-range (1/r 2 ) force only if it is coupled to a conserved vector current. This can be traced to the ISO(2) little group mentioned in the introduction and implies the concept of gauge invariance [11]. Extra neutral gauge bosons (Z ′ ) can arise from extra U (1) ′ gauge symmetries as contained, e.g., in the E 6 unification group, E 6 → SO(10) ×U (1) → SU (5) ×U (1) 2 , or in left-right symmetric models, SU (2) L × SU (2) R ×U (1) → SU (2) L ×U (1) 2 . The Z ′ could also be a Kaluza-Klein (KK) excitation (sequential Z ′ ), the techni-ρ, or the topcolor-Z ′ mentioned above. The Z ′ mass, M Z ′ , is constrained to be greater than 1305 GeV in the sequential case (DELPHI) and M Z ′ > 630 to 891 GeV for E 6 scenarios (DØ), while electroweak (EW) data limit the mixing angle with the ordinary Z to < 0.01. Likewise, mass and mixing of an extra W ′ are limited, respectively, to > 1 TeV (DØ) and < 0.12 (OPAL).
SPIN 2
In a QFT, 1/r 2 -forces are produced by an h = ±2 particle only if it is coupled to the conserved energy-momentum tensor implying invariance under general coordinate transformations [11]. The ultraviolet completion of gravity requires its embedding into structures like string or M-theory whose growing number of known vacua led to the concept of the string landscape [12] and a revival of anthropic reasoning (the multiverse).
In addition to the graviton one can also allow SM fields to propagate in the factorized EDs. E.g., TeV-scale string compactification [17] with compactification radius R (7 TeV) −1 (from LEP 2) implies gauge fields propagate in the bulk. Gauge-Higgs Unification [18] protects the EW scale by defining the Higgs as a higher-dimensional gauge field component. Universal Extra Dimensions (UEDs) [19] allow all SM particles in the bulk. Grand Unified Theories (GUTs) in EDs with R −1 ≈ M U (the gauge coupling unification scale) can solve many problems of conventional supersymmetric GUTs [20]. There are also models with SM fields propagating in warped EDs (but not the Higgs in order to protect the exponential suppression of the EW scale). These can be viewed as dual descriptions of walking TC by virtue of the AdS correspondence with conformal field theories [21]. No fully realistic models exist, but it is interesting that EWS breaking by boundary conditions of gauge fields in warped EDs yield higgsless models [22].
In a QFT, an h = ±3/2 particle can produce a 1/r 2 -force only if it is coupled supersymmetrically. Conversely, supersymmetry [23] is the only possibility to extend the Poincaré algebra non-trivially and its non-renormalization theorems offer the most elegant solution to the hierarchy problem. Moreover, in models of supersymmetric unification a large m t correctly predict EWS breaking (a fact that was known [24,25] even before m t was measured) and these are further supported by the approximate unification of gauge couplings at M U ∼ 10 −16 GeV κ 4 in the minimal supersymmetric standard model (MSSM). Finally, the MSSM predicts M h 0 135 GeV for the lighter of the two CP-even Higgses, in remarkable agreement with Figure 3 (identifying h 0 with H). The additional assumption of R-parity conservation (sufficient to forbid proton decay by dimension 4 operators) implies that the lightest supersymmetric particle is stable offering an explanation for the observed dark matter (a similar statement applies to UED models). | 2009-03-11T18:02:50.000Z | 2009-01-29T00:00:00.000 | {
"year": 2009,
"sha1": "083fc73cbf7176385a6fe62b08fe869c75b5b5be",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0901.4763v1.pdf",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "02e428636233fac87eaad43f2f04c8f2e3fe2d4a",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
118610798 | pes2o/s2orc | v3-fos-license | Radiation of non-relativistic particle on a conducting sphere and a string of spheres
The radiation arising under uniform motion of non-relativistic charged particle by (or through) perfectly conducting sphere is considered. The rigorous results are obtained using the method of images known from electrostatics.
Introduction
The radiation emitted by a charged particle crossing the boundary between two media with different electrodynamic properties is known as transition radiation (TR), whereas for the particle traveling near the boundary of the spatially localized target without crossing the emitted radiation is called as diffraction radiation (DR). DR and TR of a charge on a perfectly conducting sphere (as well as on a periodic string of the spheres) are studied in the present paper.
Various aspects of DR and TR on spherical targets had been considered in numerous papers, see, e.g., [1]- [6]. However, these papers either deal with some special cases or present the result in rather complicated form. Here we propose the simple and economic method for computation of the radiation characteristics (spectral-angular density as well as polarization, if needed) based on the following idea.
One of the ways to describe these types of radiation is the application of the boundary conditions to the Maxwell equations solutions for the field of the moving particle in two media. It becomes evident that the boundary conditions could be satisfied only after addition the solution of free Maxwell equations that corresponds to the radiation field, see, e.g. [7].
The conditions on the boundary between vacuum and ideal conductor could be satisfied in some cases via introduction of one or more fictitious charges along with the real charged particle; this approach to electrostatic problems is known as the method of images, see, e.g., [8]. The method of images had been used in the pioneering paper [9] where TR on a metal plane had been predicted. The method of images had been used also in [10] for consideration of TR under passage of the particle through the center of the ideally conducting sphere in dipole approximation.
DR on sphere
Consider the real charge e 0 passing near the grounded conducting sphere of the radius R with the constant velocity v 0 ≪ c, see Fig. 1. Its image of the magnitude has to be placed at the point with coordinates So, while the incident particle moves uniformly, its "image" will move accelerated, The radiation produced by non-uniform motion of the fictitious charge will be described by the well-known formula [11] where k is the wave vector of the radiated wave, |k| = ω/c, and (it could be easily seen that it is applicable to the case of time-varying charge e(t) as well as to the case of the constant one). Note that the method of images for the isolated (in contrast to grounded) sphere requires introducing another fictitious charge of the magnitude −e(t) resting at the center of the sphere. However, the last equation shows that such rest charge does not produce any radiation. Substitution of (1), (2), (3) into (5) gives the following integrals: (7) can be easily performed numerically, that leads to the spectral-angular density of diffraction radiation in the form where the typical shape of the angular distribution Φ DR (θ, ϕ, ω) is presented in Fig. 2; for different values of parameters it does not vary too much. Approximate analytical result can be obtained after neglecting the second term in the exponents in (6), (7) (that is valid for low radiation frequencies, ω ≪ cb/R 2 ) as well as the third term (that is always valid for non-relativistic case): where K 0 and K 1 are modified Bessel functions of the third kind.
Integration of (8) over radiation angles leads to the radiation spectrum presented in Fig. 3. For illustrative purposes, we choose the parameters v 0 = 0.1c, R = 20 nm, b = R + 0, for which the DR intensity maximum will lie in the visible spectrum. Note, however, that in this frequency domain the properties of the sphere material can be far from that of a perfect conductor. Particularly, plasma oscillations can be important. On the other side, the perfect conductor approximation is valid for the frequencies less than the inverse relaxation time τ −1 for the electrons in the metal. For instance, τ −1 = 5 · 10 −13 s −1 for copper [12], so the results obtained surely could be applied up to THz and far infrared range. The applicability of the results to higher frequencies needs further investigations.
DR on the string of spheres
Now consider the motion of a charge e 0 along the periodic string of N ≫ 1 spheres. The mutual influence of the fictitious charges induced in the neighboring spheres can be neglected in the case of small impact parameters, b → R (when the radiation intensity is high), for the string period large enough, a 5R. Then the interference of the radiation produced on the subsequent spheres leads to the simple formula for the spectral-angular density of DR: where the delta-function means well-known Smith-Purcell condition [13].
The spectrum for the string period a small enough consists of separated bands, see Fig. 4, a. The bands overlap each other under increase of the string period a gradually forming the spectrum of DR on a single sphere (multiplied by the total number of the spheres N), see Fig. 4, b, c, d.
TR on the sphere
TR arises under b < R, when the incident charge e 0 crosses the sphere. In this case both real and fictitious charges vanish while crossing the sphere (and then appear again) that complicates the formulae describing the radiation. For the convenience of the further interpretation, let us separate the contributions into the vector I (5) from the real particle and its image on that is on the incoming part of the particle's trajectory, , and on the interval t 0 ≤ t < ∞, that is on the outgoing part of the particle's trajectory,
Numerical integration leads to the spectral-angular density of transition radiation in the form
where the function Φ T R (θ, ϕ, ω) is presented for some particular cases as the directional diagram in Fig. 5. We see very sophisticate shapes in contrast to DR case. This is due to interference of the radiation emitted by the real charge and its image while crossing two boundaries between vacuum and conductor. Let us trace out the origin of this interference picture in the case of high radiation frequency Rω/v 0 ≫ 1, when the transverse size of the particle's Coulomb field Fourier component is small and the curvature of the conductor surface is negligible. The elementary contributions from both real and fictitious charges into the radiation under each crossing of the metal surface have similar shape with axis of symmetry along their velocity. So, for the real particle the axis of symmetry is directed along the z axis, see Fig. 6, a, where the directional diagram computed with the vector I (real.in) (11) or I (real.out) (14) instead of complete vector I is presented. The interference of two such contributions from the enter point of the charge into the sphere and from the exit one leads to the directional diagram in the Fig. 6, b.
On the other hand, the axis of symmetry of the elementary contribution from the fictitious charge for b = R/ √ 2 (that is for the incidence under 45 degrees) will be directed along the x axis (Fig. 6, c, which is computed for the vector I (image.in) (12), (13) or I (image.out) (15), (16)). The interference of two such contributions from two points of crossing the sphere leads to the picture in the Fig. 6, d.
Finally, the interference of the contributions from the real charge ( Fig. 6, b) and its image (Fig. 6, d) leads to the final picture of the radiation emission (Fig. 5, d).
TR spectral density for b = R/ √ 2 is presented in Fig. 7.
Conclusion
The radiation resulting from the interaction of non-relativistic particle with perfectly conducting sphere is considered.
The method of images allows the precise description of the radiation in this case. The integration of the resulting formulae can be easily performed numerically that permits to compute the spectral-angular density of the diffraction and transition radiation for an arbitrary impact parameter. The angular distribution of the transition radiation shows a wide diversity of shapes for different values of the radiation frequency and other parameters. This effect is the result of the interference of the waves emitted by the particle and its image at the moment when the boundary between the conductor and vacuum is being crossed by the particle.
The results are valid only for the non-relativistic particles due to the nature of the method of images: the sphere on which the potential of two charges (the real one and its image) is equal to zero exists only for spherically symmetric Coulomb potentials, not for the relativistically compressed ones. However, when the characteristic transverse size of the incident particles Coulomb field v 0 /ω becomes much less than the sphere radius R and the impact parameter is not close to R, one can neglect the spheres curvature. In this case our results for TR could be applied also for relativistic particles.
Our results are valid at least up to THz and far infrared domain. Their applicability for higher frequencies, where the spheres material properties are far from ones of the perfect conductor, needs further investigations.
Note that the trajectory of the particle interacting with a sphere or even a string of spheres remains rectilinear with a high degree of accuracy providing that the particle is heavy, like a proton or a nucleus. One can minimize the perturbation of the particles trajectory via using the spheres made of thin foil or drilling a fine channel through the spheres for the particles flight.
Acknowledgements
The work was supported in part by the grant of Russian Science Foundation (project 15-12-10019). | 2017-07-08T18:22:58.000Z | 2016-10-22T00:00:00.000 | {
"year": 2016,
"sha1": "22caff7ffeccc793722a1b7c1c30ccd4b508c367",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1610.07085",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "22caff7ffeccc793722a1b7c1c30ccd4b508c367",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
209257348 | pes2o/s2orc | v3-fos-license | Posterior lamellar versus bilamellar tarsal rotation surgery for trachomatous trichiasis: Long-term outcomes from a randomised controlled trial
Summary Background We re-examined the participants of a clinical trial four years after enrolment to identify which of the two most commonly used eyelid surgery procedures to treat the blinding stage of trachoma (trachomatous trichiasis, TT), the posterior Lamellar Tarsal Rotation (PLTR) and Billamelar Tarsal Rotation (BLTR), gives better results in the long-term. Methods A randomised, controlled, single masked clinical trial was done in Ethiopia. At baseline, adults (aged >18 years with upper lid unoperated TT were recruited from a community-based screening. Participants were randomly assigned (1:1), to either BLTR or PLTR surgery, stratified by surgeon. At 4 years an independent assessor masked to allocation examined the trial participants’ eyes using the same procedures as for the baseline and earlier follow-ups. The primary outcome was the proportion of individuals who had recurrence (postoperative TT, PTT) at the 4-year examination, or a history of repeat surgery in the 4-year period. The intervention effect was estimated by logistic regression, controlled for surgeon as a fixed effect in the model. The trial is registered with the Pan African Clinical Trials Registry (number PACTR201401000743135). Findings 1000 participants with TT were enrolled, randomly assigned, and treated (501 in the BLTR group and 499 in the PLTR group) between Feb 13, 2014, and May 31, 2014. At year 4, 943 (94.3%) participants were re-examined (471, PLTR; 472, BLTR) and included in the primary outcome analysis. PTT had developed in 169/943 (17•9%) study eyes, among which 129 (76•3%) had minor trichiasis (≤5 lashes touching the eye). PTT was significantly more frequent at 4-year in the BLTR arm (105/472 [22•2%]) than the PLTR arm (64/471 [13•6%]), adjusted OR 1•82 (95% CI, 1•29–2•56); p = 0•0006, with 8•6% (95%CI 3•8–13•5) risk difference. Interpretation The PLTR surgical procedure had superior long-term outcomes to the BLTR with significantly lower risk of PTT supporting the current WHO guideline that the PLTR should be the procedure of choice for training new surgeons in the programmatic management of TT.
Bilamellar Tarsal Rotation (BLTR) and Posterior Lamellar Tarsal Rotation (PLTR) surgical procedures performed by ophthalmologists in a teaching hospital in Ethiopia was conducted on 153 patients in 2002, which found no evidence of a difference in outcome after three months. We conducted a randomised, controlled, single masked clinical trial between February and May 2015 to determine the relative effectiveness of the PLTR and BLTR Ethiopia in larger sample (10 0 0 patients) in a programmatic setting in Ethiopia. The 12-month results showed that the PLTR was superior to BLTR giving a substantially lower trichiasis recurrence rate by one year and fewer intra and immediate post-operative complications. There is no data on the long-term outcome of these two surgical procedures from a head to head comparison trial.
Added value of this study
The one-year results of our trial led to a shift in international guidance from BLTR being the treatment of choice to treat trachomatous trichiasis to a preference for new trainees to be taught PLTR. However, there was much international interest to see the long-term outcome of the PLTR and BLTR. Our trial participants were examined four years after randomised intervention to assess the long-term outcome of these two procedures and to ascertain whether the superiority of PLTR would be sustained beyond one year. The results showed that the PLTR surgical procedure had still superior long-term outcomes to the BLTR with significantly lower risk of recurrent trichiasis four years after surgery.
Implications of all the available evidence
The available evidences support the current WHO guideline that the PLTR should be the procedure of choice for training new surgeons in the programmatic management of TT.
Introduction
Trachomatous Trichiasis (TT), the blinding stage of trachoma, is mainly treated with corrective eyelid surgery [ 1 , 2 ]. Many surgical procedures have been tried for the management of TT [3] . However, the two most commonly used surgical procedures are the Posterior Lamellar Tarsal Rotation (PLTR) and the Bilamellar Tarsal Rotation (BLTR) surgeries [2] . The type of surgical procedure is thought to be one of the major determinants of outcome of TT surgery [4][5][6][7][8] . Poor surgical outcomes pose a major challenge for surgical programmes worldwide. Trichiasis typically recurs in around 20% of patients within a year, and about 10% develop eyelid contour abnormality (ECA) [9][10][11][12] . Empirical data indicate that poor surgical outcomes deter patients from accepting trichiasis surgery [13] , possibly contributing to the recent decline in surgical uptake in some trachoma control programmes [14] .
Four years ago, we conducted a randomised, controlled, single masked clinical trial to compare the relative effectiveness of the PLTR and BLTR procedures. One year after surgery we found that the cumulative rate of recurrent trichiasis (here after postoperative trachomatous trichiasis, PTT) was more frequent in the BLTR group than in the PLTR group with a 9.5% risk difference [9] . Following this, international guidance on the surgical treatment of choice was updated, shifting away from BLTR being the treatment of choice to a preference for new trainees to be taught PLTR [15] .
There are about 3 million un-operated cases of TT globally [16] , requiring surgery using the safest and most successful procedure. However, there are currently no long-term data directly comparing these two surgical procedures. Some studies have reported that the rate of PTT may increase from about 20% at 1-year to as much as 60% at 3 years after surgery [ 4 , 5 , 11 , 12 , 17-19 ].
In this long-term follow-up of a randomised controlled surgical trial, we followed and examined trial participants four years after enrolment to investigate the long-term outcomes of BLTR and PLTR surgery, and to ascertain whether the superiority of the PLTR outcome was sustained beyond one year.
Study design and participants
The trial methods have been previously described in detail [ 9 , 20 ]. In summary, a randomised, controlled, single masked clinical trial was conducted in Ethiopia between Feb 13, 2014, andApril 30, 2015. Adults with TT defined as one or more eyelashes touching the eye or evidence of epilation, identified from a communitybased screening in districts of West Gojam Zone, Amhara Region, Ethiopia were examined for eligibility. People with trichiasis due to other causes, recurrent trichiasis after previous surgery, hypertension, pregnancy, and those under 18 years were excluded. Those eligible and consented to participate following a written informed consent in Amharic were enrolled. This report adhered to standard CONSORT guidelines.
Randomisation and masking
Participants were randomly assigned (1:1) to either PLTR or BLTR surgery. Randomisation was stratified by surgeon and sequences were computer-generated by an independent statistician with random block sizes of 4 or 6. Allocations were concealed in sequentially numbered, sealed, opaque envelopes. Examiners who were responsible for clinical observations at baseline and followups were masked to allocation. The surgery was performed by six experienced nurse/health officer trichiasis surgeons. These were already trained, certified, and regularly performing PLTR surgery. They were trained rigorously on the BLTR procedure using the WHO trichiasis surgery training manual [2] , and were then restandardised on both surgical procedures after six-months of regular practice.
Procedures
Participant eyes were examined (EH) at baseline prior to randomisation using 2 • 5 × binocular loupes and torch, and graded using the detailed World Health Organisation (WHO) Follicles Papillae Cicatricae (FPC) Grading System [21] .. The number, location and type of trichiasis lashes, corneal scarring, and tarsal conjunctival scarring and inflammation were graded and recorded. Presenting distance logMAR (Logarithm of the Minimum Angle of Resolution) visual acuity was measured using PeekAcuity software on a Smartphone in a dark room [22] . Four standardized high-resolution digital photographs of trichiasis, cornea, and tarsal conjunctiva were taken. After the randomisation, during the surgeries, intraoperative and immediate postoperative observations were made to measure incision length, height and regularity by three trained nurses. Number of scissor cuts made to make an adequate dissection medially and laterally. The number, spacing and tension of the mattress sutures were recorded.
Participants were re-examined at 10-days, 6-months, and 12months postoperatively, following the same assessment procedures as per baseline. The only additional elements were assessment for granuloma, level of eyelid correction, and post-operative eyelid contour abnormalities (ECA). ECA were graded according to the PRET trial methodology [23] , and grouped for analysis: (1) clinically non-significant ECA, which included mild ECA; and (2) clinically significant ECA, which included moderate-to-severe ECA.
The four-year follow-up was approved by the Ethiopian National Health Research Ethics Review Committee, the London School of Hygiene & Tropical Medicine Ethics Committee, Emory University Institutional Review Board, and the Ethiopian Food, Medicine and Healthcare Administration and Controls Authority. The study was conducted in compliance with the Declaration of Helsinki and International Conference on Harmonisation-Good Clinical Practice.
Trial participants were re-contacted and invited to attend a four-year follow-up assessment at a local health facility. Those who were not able to come to the health facilities were examined in their homes. Reasons for loss to follow-up were identified and documented. Written informed consent in Amharic had been obtained at baseline, before the initial enrolment from participants. The participants were re-consented at the four-year follow-up. If a participant was unable to read and write, the information sheet and consent form were read to them and their consent recorded by thumbprint.
Participants were asked about any repeat surgery, epilation in the last 6-month, and satisfaction with their surgical outcome. They were examined following the same procedure as outlined above. Outcome assessment was conducted by an independent examiner (BA) who was masked to the intervention allocation and who had no prior involvement in randomisation allocation, outcome assessment, and data analysis. The four-year examiner received rigorous training and was standardised with the baseline and 12-month outcome assessor (EH). They had very strong agreement for the primary outcome ( k = 0.98).
Outcomes
Prior to the start of the 4-year follow-up, a single primary end point was prespecified in the approved protocol. The primary end point was postoperative TT at 4-years analysed as the proportion of individuals who developed one or more lashes touching the eye or clinical evidence of epilation at the 4-year examination, or a history of repeat surgery in the 4-year period. The secondary analysis of the primary outcome measure was cumulative PTT defined as the proportion of individuals who had developed PTT by 4-years , defined as one or more lashes touching the eye or clinical evidence of epilation at all follow-ups (10 day, 6-and 12-months, and 4-year), or a history of repeat surgery in the 4-year period.
Secondary outcome measures included: PTT difference by surgeon and baseline disease severity; under correction; eyelid contour abnormality prevalence and regression at 4-year, corneal opacity and vision changes; effect of PTT on corneal opacity and vision changes, factors influencing long term outcomes; and patientreported outcomes.
Statistical analysis
Sample size determination was described in the 12-month results paper [9] . Data were double-entered into Access 13 (Microsoft) and transferred to Stata 14 (StataCorp) for analysis. For (0 • 0) * 2 participants had repeat surgery between baseline and 12-month (one from each arm). a One missing value in the PLTR arm. b Five missing values (one in the PLTR, four in the BLTR).
participants who had bilateral surgery, the same randomly designated eye used for the 12-month analysis was used for the 4year follow-up analysis (i.e. one eye only per participant included in analysis).
A modified intention-to-treat analysis was performed (modified meaning that participants who died during follow-up or were not seen at follow up for another reason were excluded). Otherwise all trial participants were analysed in the groups they were originally randomised and included in the analysis if they were seen at the 4-year follow-up (for the primary analysis) and at least at one follow-up time point (for the secondary analysis of the primary outcome). The effect of the intervention on primary outcome and binary secondary outcomes (cumulative PTT difference, PTT by baseline disease severity, under correction) was analysed using logistic regression to estimate the odds ratio (OR) and 95% CI. All comparisons between the two surgical procedures were controlled for surgeon as a fixed effect in the model, to account for the stratified randomisation. Vision and corneal opacity changes between baseline and 4-years were categorised as worse, same, better. Effect of the intervention on ordered categorical secondary outcomes (changes in visual acuity and corneal opacity, and patient-reported outcomes) were analysed using ordinal logistic regression. Intervention effects on ECA (categorical variable) prevalence and regression at 4-year were analysed using multinomial logistic regression to estimate relative risk ratio (RRR) and 95% CI. A non-prespecified sign test was used to analyse if ECA regression between 6-and 12month, and 12-month and 4-year is statistically significant in all study participants. In order to identify potential predictors of PTT at 4 years, first a univariable logistic regression was performed using PTT at 4 years as an outcome and factors with possible association with PTT (covariates) as exposures and was done separately for each intervention arm, before including all covariates that were associated with the outcome with p < 0.2 into a multivariable model. Likelihood ratio test was used to decide on the covariates that should be included in the final multivariable model to determine the best fitting predictive model of risk factors for PTT at 4 years. The trial was registered on the Pan African Clinical Trials Registry (PACTR2014010 0 0743135) and overseen by independent data and safety monitoring committee.
Role of the funding source
The funder of this long-term follow-up had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Results
Between Feb 13, 2014, and May 31, 2014, 5168 people were examined for eligibility, of whom 4166 were ineligible and 1002 were eligible among which two ( < 1%) declined surgery. Thus, 10 0 0 trichiasis cases consented, were enrolled, and randomly assigned: 501 in the BLTR group and 499 in the PLTR group), Fig. 1 Baseline demographic and clinical characteristics of participants seen at 4-year were balanced between the two groups, table 1 . The mean age was 46 years and the majority were women (76 • 2%). Trichiasis severity and phenotypes, tarsal conjunctival inflammations, vision, and corneal opacity were comparable between the two groups. At baseline, major TT ( > 5 lashes touching the eye) was present in similar proportions of the two arms (46% in PLTR, 47% in BLTR), and 76% of the trichiasis lashes were corneal in both arms. Higher proportion of cases in both arms had metaplastic only lashes (44 • 4% in PLTR, and 41 • 3% in BLTR).
At the 4-year follow-up, PTT was observed in 169/943 (17 • 9%) study eyes, among which 129 (76 • 3%) had minor trichiasis (1 -5 lashes), and 23 (13 • 6%) had previously received repeat surgery in the study eye between baseline and 4-year follow-up. PTT The secondary outcomes, all adjusted for surgeon effect, are shown in table 2 . The risk of PTT between surgeons ranged between 11 • 1% and 17 • 7% in the PLTR group, and 19 • 2% and 28 • 1% in the BLTR group. There was more under-correction in the BLTR group than the PLTR group (15 • 6% vs 20 • 8%). Most of the undercorrection tends to be peripheral in both the PLTR (65/73 [89 • 0%]) and BLTR (87/98 [88 • 8%]) surgeries. PLTR surgery had a lower risk of PTT in cases with baseline major TT than the BLTR surgery (17 • 8% vs 31 • 5%), and performs better across all severity of entropion. Participants were also asked about their satisfaction on the surgery. Comparable proportion of participants in both treatment arms reported satisfaction with the effect of the surgery on the trichiasis (93 • 0% PLTR and 92 • 6% BLTR), and the cosmetic appearance of the operated eyelid (96 • 2% for both surgeries). There was no evidence of a difference in logMAR visual acuity score changes ( p = 0 • 26), and corneal opacity grade changes ( p = 0 • 92) from baseline to 4-year between the two intervention arms.
There was strong evidence that major trichiasis, conjunctival scar severity, and any under correction at any location measured at immediate post-op during the baseline surgery independently predicted PTT 4-year after PLTR surgery. Increased number of peripheral dissections with scissors intraoperatively in the PLTR surgery at baseline had a long-term protective effect on postoperative TT. In the BLTR group, there was strong evidence that major trichiasis, mixed trichiasis lash location, and central under correction at immediate post-op at baseline independently predicted PTT 4-year after BLTR surgery, table 4 .
Demographic and clinical factors (age, gender, trichiasis severity, entropion severity, conjunctival scarring, surgeon effect, number of scissor cuts during dissection, number of suture notes, suture knot symmetry, suture tension irregularity) that may predict ECA regression at 4-year were analysed by intervention group and for all participants (data not provided) and there was not strong evidence found of an association of any of these with ECA regression.
Discussion
It is possible that outcomes may change a few years after TT surgery. However, the data from this long-term follow-up indicate PLTR remains superior to the BLTR with a significantly lower risk of postoperative trichiasis both cross-sectionally at and cumulatively by 4-years after trichiasis surgery and is consistent with the oneyear results we have previously reported [9] . Moreover, PLTR still had a lower risk of PTT across all severity of trichiasis and entro- pion groups, and had lower risk of under-correction as was found at 1-year. These data support the WHO recommendation that new surgeons should be trained on the PLTR procedure for the programmatic management of TT [15] . We have discussed in detail in the 1-year report why these outcome differences between the PLTR and the BLTR might have occurred [9] . We believe the PLTR procedure provides a greater, more stable outward rotation of the distal portion of the eyelid. The major factors predicting long-term outcomes were also similar to those reported at 1-year which included preoperative disease severity and surgical factors such as peripheral dissection and under-correction [20] . Encouragingly, making adequate peripheral dissection still had a long-term protective effect on PTT in PLTR surgery indicating that it can be prevented with quality surgery and can be addressed easily during surgical trainings and supportive supervision. Special attention should also be provided for cases with advanced disease which should be operated by the most experienced surgeon available in the programme using the PLTR surgical procedure.
ECA has been a major issue for surgical programmes. It is cosmetically disfiguring, posing probably a greater concern than PTT for patients and surgeons. The good news is ECA, regressed between 6-and 12-months, and 1-and 4-years after surgery in about 73% and 66% of the cases in PLTR and BLTR surgeries respectively. However, about 50% of the clinically significant ECA cases at 12month in both procedures remained un-changed. The regression seen in clinically significant ECAs between 6 and 12-month in the PLTR arm (15 • 0%) was much less than we have found in one of our recent trials which used the PLTR (40 • 6%) [10] . Rather the regression at 4-year in this trial (48 • 3%) was comparable to the regression at 12-month in the earlier trial. These results suggest that those with clinically significant disfiguring ECA need to be addressed surgically. However, in most trachoma endemic settings, neither highly skilled personnel which can correct ECA, nor a standard surgical procedure that can be used to correct ECA are available.
The strengths and limitations of this trial with regard to its design have been discussed in detail elsewhere [ 9 , 20 ]. We managed to follow 94% of the trial participants 4 years after enrolment. Trichiasis surgeons operated in this trial received rigorous training and standardisation. Risk of unmasking posed a potential design limitation at the 6-and 12-month follow-ups in relation to possibility of visible skin scar from the BLTR surgery, which was addressed with independent photographic grading. However, this was not an issue in this long-term follow-up as no eyelid skin scar would be visible 4-years after surgery. Another potential limitation could have been unmasking of the outcome assessor as the randomisation code has been broken for earlier analyses. However, the outcome assessment was done by an independent assessor masked to allocation of intervention who had no involvement either in randomisation or data analysis.
Overall, there is strong evidence that PLTR remains superior to BLTR with reduced long-term risk of postoperative trichiasis supporting the current WHO guideline that the PLTR should be the procedure of choice for training new surgeons in the programmatic management of TT. Surgical programmes need to provide attention in improving outcomes and establish a system to comprehensively manage cases with poor surgical outcomes. The majority of PTT cases had five or less metaplastic lashes indicating that most of these cases can be treated with less invasive non-surgical methods. A relatively simple surgical procedure that can be used in trachoma endemic settings is needed to address un-resolving clinically significant ECA cases. Note : Analysis is done using logistic regression model. Factors with possible association with postoperative TT were tested in univariable analysis, and those with p < 0 • 2 were included in the initial model. Then, likelihood ratio test was used to decide on variables to be included in the final multivariable model. Surgeon was included in the multivariable model regardless of significance level in both PLTR and BLTR but results not presented as there was no effect. Central undercorreciton was not included in the multivariable model for the PLTR to avoid collinearity with central undercorrection. In the PLTR, all listed in the table except age, and central undercorrection were included in the initial model. Then Lash location was excluded from the final model after likelihood ratio test. In the BLTR, all listed in the table except undercorrection at any part of the eyelid were included in the initial model. Then tarsal conjunctival scarring was excluded from the final model after likelihood ratio test. PTT = Postoperative Trachomatous Trichiasis.
Data sharing statement
The Amhara Regional Health Bureau Ethics Committee requires that all data sharing requests are reviewed and approved by them before data can be shared. Data is available to any researcher under reasonable request. To facilitate the data access process please contact ethics@lshtm.ac.uk.
Funding
This four year follow-up was funded by the Coalition for Operational Research on Neglected Tropical Diseases (COR-NTD), at the Task Force for Global Health (NTD-SC 129D). The initial trial was funded by The Wellcome Trust (Grant Number 098,481/Z/12/Z).
Declaration of Competing Interest
We declare no competing interests. | 2019-11-07T15:04:29.629Z | 2019-11-01T00:00:00.000 | {
"year": 2019,
"sha1": "b23c1adfa27b3e15627a433cd6c264d805623984",
"oa_license": "CCBY",
"oa_url": "http://www.thelancet.com/article/S2589537019302019/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d48732484da64c22d9d60afbf2683341ef6304a7",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
264603648 | pes2o/s2orc | v3-fos-license | Plasticity-related gene 5: A novel surface autoantigen in paraneoplastic cerebellar degeneration
Paraneoplastic cerebellar degeneration (PCD) is one of the most frequent paraneoplastic syndromes affecting the CNS. It is associated with antibodies targeting intracellular neuronal antigens (Hu, Yo, Ri, CV2/CRMP5), which are not thought to be directly pathogenic, and surface antigens (DNER, mGluR1, VGCC), which are potentially pathogenic.1,2 However, in many patients the immunologic target remains unidentified, resulting in diagnostic and therapeutic challenges. We report a patient with PCD and a squamous cell lung carcinoma with antibodies to a novel neuronal surface antigen, plasticity-related gene 5 (PRG5).
Paraneoplastic cerebellar degeneration (PCD) is one of the most frequent paraneoplastic syndromes affecting the CNS. It is associated with antibodies targeting intracellular neuronal antigens (Hu, Yo, Ri, CV2/ CRMP5), which are not thought to be directly pathogenic, and surface antigens (DNER, mGluR1, VGCC), which are potentially pathogenic. 1,2 However, in many patients the immunologic target remains unidentified, resulting in diagnostic and therapeutic challenges. We report a patient with PCD and a squamous cell lung carcinoma with antibodies to a novel neuronal surface antigen, plasticityrelated gene 5 (PRG5).
Case description. A 72-year-old man presented with a 2-week history of progressive dizziness, vomiting, oscillopsia, dysarthria, and ataxia. Five months earlier he had been diagnosed with a poorly differentiated squamous cell lung carcinoma, which was radically resected. On examination, he had a symmetrical horizontal gaze-evoked nystagmus and saccadic pursuits. He had severe cerebellar dysarthria and trunk and limb ataxia and was unable to walk.
CT of the thorax showed a recurrence of the tumor in the mediastinum. MRI of the brain was normal. CSF examination showed pleocytosis (25 white blood cells/mL), elevated protein (1.05 g/L), normal glucose, and normal cytologic findings. All tests for classical paraneoplastic and published surface antibodies were negative (supplementary data at Neurology.org/nn). The patient received irradiation to the mediastinum after which his neurologic syndrome stabilized. He died 4 months after onset of disease.
Results. Immunohistochemistry (IHC) of rat brain slices showed neuropil staining of cerebellum and hippocampus with patient serum and CSF but not control serum and CSF ( figure 1A). Both serum and CSF labeled the surface of rat hippocampal neurons and stained the tips of dendritic spines (figure 1B). Using immunoprecipitation of whole rat brain lysate with patient serum and CSF followed by mass spectrometry analysis (as described in reference 3), we identified PRG5 as the autoantigen. PRG5 is a transmembrane protein enriched in plastic areas of the adult brain and involved in neurite outgrowth and the formation of dendritic spines. 4,5 To validate the antigen, serum and CSF were used in cell-based assays (CBAs) for PRG1, 3, and 5. Both samples showed strong reactivity to PRG5, weak reactivity to PRG1, and no reactivity to PRG3 ( figure 1C). PRG5 was also recognized under nonpermeabilizing conditions ( figure 1D). Anti-PRG5 titers were .1:200,000 (serum and CSF in CBA). We did not find additional patients with anti-PRG5 antibodies among 214 patient sera (98 with cerebellar ataxia, 95 with neuropil staining on IHC, and 21 with PCD and small cell lung carcinoma) and 137 control sera (supplementary data).
The patient serum recognized a conformational epitope on PRG5 as it immunoprecipitated PRG5 from lysate, but it no longer recognized the epitope when denatured on Western blot (figure 1E, figure e-1). Using chimeric proteins of PRG3 and PRG5, this epitope was mapped to the second and third extracellular loop (figure 1F). When staining with patient serum depleted of both PRG1 and PRG5 antibodies, specific labeling of dendritic spine tips was diminished (figure 1, G and H).
To study the effects of anti-PRG5 antibodies, mature cultured hippocampal neurons were treated for 4 days with purified healthy control IgGs or patient IgGs. No changes occurred in dendritic spine number, spine morphology, synapse number, or dendritic branching (data not shown). However, 24-hour treatment of hippocampal neurons with patient IgGs (not healthy control IgGs) induced internalization of green fluorescent protein (GFP)-tagged PRG5 ( figure 1I). Internalized PRG5-GFP accumulated within EEA1-positive early endosomes in the soma of the neurons ( figure 1J).
Discussion and conclusion. We have identified a patient with PCD and autoantibodies to a novel extracellular conformational epitope on PRG5, a transmembrane protein expressed in the hippocampus and cerebellum and involved in dendritic spine formation. 4,5 Its expression in the cerebellum is in line with the phenotype of our patient. Similar to what has been shown for anti-NMDA receptor Identification and characterization of plasticity-related gene 5 (PRG5) as a neuronal surface autoantigen (A) Immunohistochemistry of adult rat cerebellum using patient CSF (top panel) and control CSF (bottom panel). The patient's CSF stains the neuropil of the cerebellar molecular layer and Purkinje cell cytoplasm. Scale bars: 100 mm in the overview, 20 mm in the magnification. (B) Immunocytochemistry of cultured rat hippocampal neurons (18 days in vitro [DIV]). The top panel shows a permeabilized staining with patient CSF (green) and anti-PSD95 (red) to mark the post synapse. The patient CSF labels the tips of both mature and immature dendritic spines. Arrows indicate colocalization between the patient CSF and PSD95 in the tips of mature dendritic spines. The bottom panel shows a neuron surface labeled with patient CSF (green) followed by permeabilized staining with anti-MAP2 (red) to mark the dendrites. The patient CSF recognizes an extracellular epitope located along the dendrites. Scale bars: 20 mm. (C) HeLa cells expressing PRG1, 3, or 5 tagged with green fluorescent protein (GFP) (green) were permeabilized and stained with patient or healthy control serum (red). The patient serum strongly recognizes PRG5 and to a lesser extent PRG1. Scale bars: 10 mm. (D) HeLa cells expressing PRG1 or 5 tagged with GFP (green) were surface stained with patient serum (red). The patient serum strongly recognizes an extracellular epitope on PRG5. Scale bars: 10 mm. (E) Immunoprecipitation (IP) of GFP-tagged PRG1, 3, and 5 using patient or healthy control (HC) serum. The sample was run on SDS-PAGE and subsequently stained with anti-GFP. The patient serum, but not HC serum, strongly pulls down PRG5 and to a lesser extent PRG1. Bands visible in the control blot at 50 kDa are background bands representing the IgG heavy chain. (F) Schematic representation of PRG5 (based on reference 4). GFP-tagged chimeric proteins (green) of PRG3 (schematic purple) and PRG5 (schematic green) expressed in human embryonic kidney cells and stained with patient serum (red). The patient serum only recognizes chimera 3, containing the second and third extracellular loop of PRG5. Scale bars: 10 mm. (G) Permeabilized immunofluorescent staining of rat hippocampal neurons (18 DIV) with anti-PSD95 (red) and serum (green) depleted of PRG1 and 5 antibodies or GFP as a control. The specific labeling of dendritic spine tips is diminished if the serum is depleted of PRG1 and 5 antibodies. Scale bars: 5 mm. (H) Quantification of depletion. Bars represent the number of enrichments in dendritic spine tips per 20-mm dendrite. N 5 18 cells/condition. Error bars 5 SEM; p 5 0.0301 (Mann-Whitney test). (I) Rat hippocampal neurons (20 DIV) transfected with PRG5-GFP treated for 24 hours with purified patient IgGs or healthy control IgGs (10 ng/mL). Cells were acid washed to remove all protein from the cell surface and stained to visualize human IgG (red) and anti-EEA1 (blue) to mark early endosomes. Upon incubation with patient IgGs, PRG5-GFP is internalized from the dendritic spine tips and moves to early endosomes. The arrows indicate triple colocalization of PRG5-GFP, IgGs, and EEA1. Scale bars: 5 mm. (J) Percentage of EEA1-positive early endosomes containing PRG5-GFP in the soma of hippocampal neurons. N 5 17 cells/condition. Error bars 5 SEM; p 5 0.007 (Mann-Whitney test). antibodies, 6 PRG5 is internalized and accumulates in early endosomes after treatment with patient IgGs. Because the localization of PRG5 in the plasma membrane is critical for its role in spine formation, 4 its internalization could lead to neuronal dysfunction and neurologic symptoms.
Incubation of cultured hippocampal neurons with patient IgGs did not result in morphologic changes, possibly due to the biological redundancy of other PRG proteins. Cerebellar neurons may lack this redundancy; therefore, cerebellar slice cultures might be a better model system to study the pathologic effects of patient IgGs. Unfortunately, lack of serum sample precludes these experiments.
Although only 1 patient with PCD and PRG5 antibodies has been identified, we encourage physicians to consider this antigen, as extracellular antibodies are potentially pathogenic and the symptoms can improve or stabilize with aggressive treatment. | 2016-05-12T22:15:10.714Z | 2015-09-24T00:00:00.000 | {
"year": 2015,
"sha1": "d965c74a9f41efd7086719ed550ed546b29ba7cf",
"oa_license": "CCBYNCND",
"oa_url": "https://nn.neurology.org/content/nnn/2/5/e156.full.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d965c74a9f41efd7086719ed550ed546b29ba7cf",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
116892653 | pes2o/s2orc | v3-fos-license | Radial Velocity Studies of Southern Close Binary Stars.I
Radial-velocity measurements and sine-curve fits to the orbital velocity variations are presented for nine contact binaries, V1464 Aql, V759 Cen, DE Oct, MW Pav, BQ Phe, EL Aqr, SX Crv, VZ Lib, GR Vir; for the first five among these, our observations are the first available radial velocity data. Among three remaining radial velocity variables, CE Hyi is a known visual binary, while CL Cet and V1084 Sco are suspected to be multiple systems where the contact binary is spectrally dominated by its companion (which itself is a binary in V1084 Sco). Five additional variables, V872 Ara, BD Cap, HIP 69300, BX Ind, V388 Pav, are of unknown type, but most are pulsating stars; we give their mean radial velocities and Vsini.
INTRODUCTION
The origins of this paper are related to those of the series of radial velocity studies of short period binaries currently conducted at the David Dunlap Observatory (DDO papers 1 -10) (Lu & Rucinski 1999;Rucinski & Lu 1999;Rucinski, Lu, & Mochnacki 2000;Lu et al. 2001;Rucinski et al. 2001Rucinski et al. , 2002Rucinski et al. , 2003Pych et al. 2004;Rucinski et al. 2005). Both authors realized in the 1990's that with availability of good Hipparcos parallaxes (ESA 1997), the limiting factor in gathering spatial velocities of contact binaries would be radial velocities (RV). While the DDO studies have since succeeded in obtaining RV data for now over one hundred of northern binaries, for many reasons the data presented in this paper are so far the only effort for the southern binaries. We present these results because chances of continuation of these observations is basically nil: the telescope has been retired, and the remaining ESO telescopes are assigned for technically more demanding tasks.
The observations reported in this paper have been collected on 4 nights of August 8 -11, 1998. To optimize on the returns from such a short survey, the 17 targets were selected to be a mixture of contact binaries possibly offering reasonable orbital solutions with a selection of variables suspected to be contact binaries (Duerbeck 1997). The next paper will contain similar results for Spring southern targets.
In this paper, we attempt to stay close to the format of the DDO series. In particular, we use the same data extraction procedures through the Broadening Function (BF) approach, as described in the DDO interim summary paper Rucinski (2002a, DDO-7). Of 17 stars discussed in this paper, four contact binaries (EL Aqr, SX Crv, VZ Lib, GR Vir) in the meantime have been observed during the DDO program resulting in good RV orbits; we include these systems here to report the Southern observations as a check of consistency. The remaining stars have been observed by us for radial-velocity variations for the first time. We have derived the radial velocities in the same way as described in the DDO papers; see the DDO-7 paper for a discussion of the broadening-function technique used in the derivation of the radial-velocity orbit parameters: the amplitudes, K i , the center-of-mass velocity, V 0 , and the time-of-primary-eclipse epoch, T 0 . The primary radial velocity standard used to determine the BF's as well as to find radial velocities was 6 Cet (F5V) assumed to have the velocity of +14.9 km s −1 (Simbad). This was the only sufficiently well observed standard which could be used as the BF template, but appeared to serve well for the whole range of the spectral types from mid-A to mid-G; the disparity of the spectral types manifested itself mostly in the broadening function intensities which would not normalize to unity as expected for perfect spectral matches.
We describe our results in the context of the existing photometric data from the literature and the Hipparcos project. We also utilize the mean (B − V ) color indexes taken from the Tycho-2 catalog (Høg et al. 2000) and the photometric estimates of the spectral types using the relations published by Bessell (1979). The spectral types are taken uniformly from the 5 volumes of the Michigan Catalogue of HD Stars (Houk et al. 1975(Houk et al. -1999; from now on called HDH. Because the high incidence of companions to contact binary stars (Pribulla & Rucinski 2006), we checked all stars for possible membership in visual systems using the Washington Double Star Catalog (WDS) 2 . DE Oct and CE Hyi have been identified as members of already known visual binaries. VZ Lib is a previously recognized , DDO-4) spectroscopic triple system.
The observations were carried with the ESO 1.52-m telescope at ESO La Silla, equipped with a Boller & Chivens Cassegrain spectrograph. Holographic grating No. 32 (2400 lines/mm) was used in combination with Loral CCD No. 39 (2048× 2048. The slit with was set to 220 µm. The broadening functions were extracted from the wavelength region of 401.6 -499.8 nm. Thus, compared with the DDO results based on the Mg I triplet at 518.4 nm, with the window of about 25 -30 nm, the ESO spectra are longer and more blue. They also have a lower resolution: While for the DDO spectra, the broadening functions have a resolution of typically σ ≃ 13 − 18 km s −1 , the spectra described here have the BF resolution of typically σ ≃ 23 − 27 km s −1 . Stellar exposure times ranged between 10 and 20 min, depending on the brightness of the object; each exposure was followed by the exposure of a He-Ar spectrum. Spectrum extraction and wavelength calibration was carried out using the ESO MIDAS software system 3 . This paper is structured in a way similar to the DDO series. We comment on individual contact binary systems in Section 2. Three additional variables which may be members of multiple systems with a dominating third component are described in Section 3, while additional five stars of mostly unknown types are described in Section 4. In each section, the stars appear in the constellation order. The individual measurements are listed in Tables 1 and 2, while Table 3 gives parameters of spectroscopic orbits derived for 9 binary stars discussed in Section 2. The broadening functions for selected phases of the 9 contact binaries are shown in Figure 1, while radial velocity orbital solutions for these stars are shown in Figure 2. We show the broadening functions for single stars or binaries without orbital solutions in Figure 3. The conclusions of the paper are summarized in Section 5.
V1464 Aql
The variable star V1464 Aql (HIP 97600, HD 187438) was suggested to be a contact binary by Duerbeck (1997) (hereinafter D97) on the basis of the low-amplitude Hipparcos light curve. We found that the star is indeed a close binary and our six observations confirm that the period is two times longer than the original Hipparcos period and is equal to 0.697822 days. The SB1 orbit is shown in Figure 2 while the orbital elements are given in Table 3, together with the remaining binary stars. The RV data are in Table 2, together with the oher stars showing single-lined spectra.
A search for the presence of the signature of the secondary star in the BF was unsuccessful. The visible component shows a relatively large broadening of its spectral lines with V sin i = 94 ± 4 km s −1 (Figure 1). The star is bright, V max = 8.6. The Tycho-2 color index B − V = 0.237 suggests, with no reddening, the spectral type of A8 -F0, much later than given in HDH, A2 V.
The Hipparcos parallax of 7.16 ± 1.26 mas suggests M V = 2.9 at the light maximum, which corresponds to F1/2V rather than early A. We do not have sufficiently many spectral standards to attempt our own classification, but the general appearance of our spectra for V1464 Aql indeed supports an early F spectral type.
EL Aqr
EL Aqr (HIP 117317) was the subject of a previous DDO study (Rucinski et al. 2001, DDO-5) where the orbital coverage was good, but the reported final elements had a larger scatter than for most of the DDO systems, probably because of typically large zenith distances and of a relative faintness of the system at V max = 10.35. The B − V given in DDO-5 was incorrect; the value of B − V = 0.47 better agrees with the DDO spectral type of F3V, but still suggests some amount of interstellar reddening. The spectral type is not available in HDH. For more information about the system, please consult DDO-5.
Our 15 observations are concentrated in the first half of the orbit. They confirm the DDO results, but the K 2 semi-amplitude is significantly smaller. This may be an indication of an insufficient spectral resolution, although the peaks in the BF's are quite well separated ( Figure 1). The primary eclipse prediction of T 0 from the DDO observations served well the new observations and was adopted here without a change.
V759 Cen
The bright contact binary V759 Cen (HIP 69256, HD 123732) was discovered by Bond (1970). In spite of its brightness of V max ≃ 7.45, it has not been much observed since then with only sporadic photometric observations for eclipse timing. The color b − y = 0.39 and the spectral type F8 (Bond 1970) or G0V (HDH) and the period of 0.394 days suggest a typical contact binary.
The binary was observed 6 times within our program and these were its first RV observations. The phase distribution of the observations was far from optimal so that the orbital elements must be treated as preliminary. As can be seen in Figure 1, the spectral resolution was insufficient for this binary which is probably visible at a low inclination angle.
The Cracow database consulted in April 2006 provided an ephemeris used for our observations, as given in Table 3. Our bootstrap estimated errors are very large because of the insufficient number of observations. The mass ratio is probably close to q ≃ 0.2.
SX Crv
The very interesting and important contact binary SX Crv (HIP 61825, HD 110139) with the currently smallest known mass ratio of q ≃ 0.07 (Rucinski et al. 2001, DDO-5) was observed 8 times. The BF's show the faint peak of the secondary component quite well although the DDO-5 elements are definitely better established as they were based on 49 best observations selected from among 96 available ones. As may be expected for a lower spectral resolution, the current observations give a smaller K 2 , but the center of mass velocity appears to be also different; the latter effect may be due to the uneven phase distribution of the observations.
For more information about SX Crv, please consult DDO-5. Note the incorrect value of B − V in that paper which should be 0.44. The HDH spectral type is F7V.
VZ Lib
The spectroscopically triple system VZ Lib (HIP 76050) was analyzed in Lu et al. (2001, DDO-4). The current observations show all three components, but the lower spectral resolution results in a much stronger merging of the primary and tertiary peaks in the BF. As observed for other binaries, the value of the semi-amplitude K 2 is again smaller for the current observations. For more information about VZ Lib, please consult DDO-4.
In DDO-4, a continuous change of the radial velocity of the companion in four seasons was noted. The current observations with the mean value V 3 = −36.7 ± 2.9 at JD 2,451,034 (see Table 2 for individual observations) confirm the V 3 variability within the combined span of 4 years very well. The "kink" in V 3 visible in Figure 5 in DDO-4 is apparently real so that the orbital period of the triple system is probably quite short, of the order of a few years. It may be necessary to look for systematic changes in the center of mass data for the binary VZ Lib itself to confirm its motion. The fact that no obvious changes of V 0 have been noted so far suggests that the third component is probably much less massive than the binary.
DE Oct
DE Oct (HIP 100187, HD 191803) was observed spectroscopically for the first time within this program. D97 had suggested that this is a contact binary with the orbital period twice as long as the Hipparcos discovery period, 2 × P = 0.5555922 days. With this period, our 3 observations cannot properly define an orbit and no estimates of element uncertainties could be determined. However, we can exclude applicability to our observations of the Hipparcos conjunction time at T 0 = 2, 448, 500.157; the new value of T 0 is given in Table 3. The BF's are poorly resolved so that the measured velocities, particularly of the secondary component, are very tentative. DE Oct is a visual binary with the angular separation of 22.9 arcsec at the position angle 129 • and the magnitude difference between the visual components of 2.85 (WDS 20194-7608). The secondary was far enough not to be included in the spectrograph slit.
The Tycho-2 color index B − V = 0.319 suggests a spectral type F1/2V, while the HDH spectral type is A9IV. The star is relatively bright, V max = 9.15.
MW Pav
MW Pav (HIP 102508, HD 197070) was observed within our program for the first time and was the best observed star of this series with 18 observations defining a good radial velocity orbit. The spectral signatures are well separated in the BF's, although one must take into account the warning signs from the other binaries that K 2 might be systematically underestimated at the available resolution. We assumed the value of the period from the Hipparcos results.
MW Pav is a well known southern contact binary with V max = 8.80, B − V = 0.33 (Tycho-2), the spectral type F3IV/V (HDH) and a relatively long orbital period of 0.795 day. It was discovered by Eggen (1968) and initially designated as BV 894. A light curve solution was presented by Lapasset (1980). The secondary eclipse seemed to be total, so that evaluation of the mass ratio appeared to be possible. However, q phot = 0.122 ± 0.003, disagrees with our spectroscopic determination, q sp = 0.228 ± 0.008, even if we consider a possibility of a probable systematic underestimate of K 2 by (at most) 10%. Our spectroscopic observation should permit a combined solution of the parameters of this binary.
BQ Phe
BQ Phe (HIP 2005, HD 2145) was suggested by D97 to be a contact binary with the period twice longer than given by the Hipparcos discovery observations, 2 × P = 0.437 days. We confirm that BQ Phe is a contact binary, but with only 4 observations our orbital solution is indicative rather than definitive and the formal errors are very large. We assumed both the T 0 and the double Hipparcos period (see Table 3).
GR Vir
GR Vir (HIP 72138) was analyzed for radial velocity variations by Rucinski & Lu (1999, DDO-2) where a good orbital solution was presented. With only 5 new observations we can only say that we fully confirm the DDO-2 solution. We assumed both the T 0 and the period from the DDO-2 results.
For more information about GR Vir, please consult DDO-2. As for other systems observed before, we see that our value of K 2 is slightly lower than that observed at DDO.
CL Cet
CL Cet (HIP 2274, HD 2554) was suggested in D97 to be a contact binary with a period twice longer than the Hipparcos discovery result, 2 × P = 0.6216 days. The star has V max = 9.9 and the Tycho-2 color index B − V = 0.313; the latter agrees with the spectral type of F2V (HDH).
Our spectroscopic observations do not have sufficient resolution to analyze apparent changes in the single-peaked, wide broadening function ( Figure 3). It is possible that the binary signature is masked by a relatively rapidly rotating companion with V sin i = 135 ± 8 km s −1 . The single peak in the BF has the velocity V = −18.9 ± 1.2 km s −1 . However, a significant shift by 10 km s −1 from the average was observed for the last of our four observations. The case for a complex blending of three components in this system is the weakest one among the three cases discussed here; the star may be in fact a pulsating one.
CE Hyi
CE Hyi (HIP 7682, HD 10270) is an other case suggested to be a contact binary by D97. Again, the orbital period suggested was 2 × P = 0.4408 days.
The star is known as a visual double star WDS 01389-5835 (HU 1553) with the angular separation of 1.9 arcsec at the position angle of 10 • and a small magnitude difference of only 0.24. Our 3 observations show very clearly that the spectrum is dominated by a slowly rotating companion, while the close, low-inclination, contact binary is visible only in the base of the combined BF profile. Hipparcos and Tycho photometry of individual components shows that it is the fainter star (B) which is the photometric variable and thus the contact binary.
The comparable light contribution of both components to the combined spectrum is visible in the BF where the sharp-lined star shows the peak with V sin i which is un-measurably low, below the spectral resolution of our observations, while the contact binary light is distributed in the velocity domain within ±200 km s −1 (Figure 3). The radial velocity of the slowly rotating companion is V 3 = 9.00 ± 0.33 km s −1 .
The observed V max ≃ 8.3 is for the combined light of both visual components. The Tycho-2 catalog gives V A = 9.08, V B = 9.29 and (B − V ) A = 0.333, (B − V ) B = 0.497, respectively. The Simbad database gives B − V = 0.49 and F5V for CE Hyi. The spectral type is after HDH.
V1084 Sco
V1084 Sco (HIP 86294, HD 159705) was suggested by D97 to be a contact binary with the period twice the Hipparcos period, 2 × P = 0.3003 days.
We have only 3 observations which show that system is a complex one: It appears to be a quadruple system consisting of a detached binary giving two sharp peaks in the BF (see Figure 3), and of a slightly fainter contact binary responsible for the short-period photometric variability. The contact binary -because of the stronger line broadening -is just barely detectable at the base of the BF. The radial velocities of the sharp-line binary components (designated as "3" and "4" in Table 2) varied during the 3 days of observations between −19 and −31 km s −1 for the stronger component and +77 and +83 km s −1 for the fainter component. Thus, the detached binary must be also relatively compact, but our observations were insufficient to determine any parameters of the radial velocity orbit. The star was included in the major radial velocity survey of Nordstrom et al. (2004) where it appears with the average radial velocity of +21.3 km s −1 .
This star is a very interesting object for further studies, particularly if the mutual period of revolution of the two binaries turns out moderately short to be observable. The star is relatively bright, V max = 9.0, while the color and the spectral type given in Simbad are late, B − V = 0.76 and G6V (HDH). The Tycho-2 catalog is in agreement with B − V = 0.73.
V872 Ara
This star, at that time identified as HIP 81650 (HD 149989) was suspected in D97 to be a contact binary with the orbital period of 0.8532 days. Very little can be said on the basis of its light variations which are very small (0.02 mag.). Three observations obtained here show a wide, rotationally broadened profile with the average V sin i = 142 ± 6 km s −1 . The mean velocity is constant at +42.1 ± 2.4 km s −1 , but the variation between +37 and +45 km s −1 is larger than the measurement error of about ±1.2 km s −1 so that some small variability may be present.
Our results are fully consistent with the recent study of de Cat et al. (2006) which explains the variability of V872 Ara by γ Dor-type pulsations with the originally suggested period of 0.42658 days. The measured value of V sin i = 134 ± 3 km s −1 is consistent within the combined errors with our estimate. We refer the reader to the paper of de Cat et al. (2006) for more information on this star. The spectral type is A8/F0V (HDH).
BD Cap
BD Cap (HIP 99365, HD 191301) was suggested by D97 to be a contact binary with the period twice as long as the one given by the Hipparcos project, 2 × P = 0.3204 days. Our three spectra show a very broad BF with V sin i = 133 ± 10 km s −1 . The mean velocity is practically constant at +9.7 ± 1.0 km s −1 . We cannot say more about this star except we note that it was included in the catalog of suspected and confirmed δ Sct pulsating stars (Rodrigez et al. 2000) as well as in the survey of spatial velocities of nearby stars (Nordstrom et al. 2004). The spectral type is A9III (HDH).
Anon Cen = HIP 69300
HIP 69300 (HD 123720) was an other suggestion of D97 to be a contact binary. Our two observations substantially differ in radial velocity of the star, −94.6 and −25.9 km s −1 , but the broadening profile has the same V sin i = 116 ± 7 km s −1 . The star does not have an entry in the General Catalog of Variable Stars 4 and no variable star name has been assigned to it yet, but it is definitely a radial velocity variable. The spectral type is A4V (HDH).
BX Ind
BX Ind (HIP 108741, HD 208999), an other candidate of D97, appears to be a slowly rotating star. Our seven observations all show a BF peak consistent with no rotation. Some small radial velocity changes within −32 and −20 km s −1 appear to be present with the mean value −27.6 ± 1.7 km s −1 . This is definitely not a close binary star. It is listed in the Catalog of Delta Scuti stars of Rodrigez et al. (2000). The spectral type is F2V (HDH).
V388 Pav
We have only two observations of V388 Pav (HIP 103803, HD 199434), another candidate of D97. The radial velocity may be constant at the mean of +5.6± 1.2 km s −1 , while the BF's indicate a mild broadening of V sin i = 45 ± 7 km s −1 . It is not a close binary star. It is listed in the Catalog of Delta Scuti stars of Rodrigez et al. (2000). The spectral type is F5II (HDH).
SUMMARY
This program of radial velocity measurements of known and suspected southern contact binary stars was performed to fill the growing disparity in the available RV data for northern and southern hemispheres. With only four successive nights, the program could not achieve the same goals as the current David Dunlap Observatory survey. Still some useful results have been obtained for 17 targets of the Fall southern sky.
We have confirmed the suggestion of Duerbeck (1997) (D97) that V1464 Aql, DE Oct, BQ Phe are contact binaries and obtained the first preliminary orbital data for these systems; V1464 Aql is a single-lined binary (SB1) while the rest are double-lined systems. We obtained the first radial velocity orbital data (SB2) for the well known southern systems V759 Cen and MW Pav. We confirmed the David Dunlap Observatory results for the double-lined binaries EL Aqr, SX Crv, VZ Lib, GR Vir, but we noticed that in all these systems the secondary star semi-amplitude K 2 is by a few percent smaller than observed at DDO which may be a result of the lower spectral resolution.
Three systems could not be analyzed because of the presence of companions. In the case of CE Hyi, a visual companion had been known, but we see spectral signatures of a binary companion in V1084 Sco (so that the system is a quadruple one) and suspect a presence of a companion in CL Cet. We are not able to say much about other variables suggested in D97: V872 Ara, BD Cap, HIP 69300, BX Ind, V388 Pav, but most appear to be pulsating stars and have been included in catalogs of such objects; we give their mean radial velocities and V sin i.
Thanks are due to George Conidis for participation in reductions of the data used in this paper. Thanks are also due to the reviewer, Dr. Vakhtang S. Tamazian, for a few very pointed suggestions on the improvement of the paper. Support from the Natural Sciences and Engineering Council of Canada to SMR is acknowledged with gratitude. The research made use of the SIMBAD database, operated at the CDS, Strasbourg, France and accessible through the Canadian Astronomy Data Centre, which is operated by the Herzberg Institute of Astrophysics, National Research Council of Canada. This research made also use of the Washington Double Star (WDS) Catalog maintained at the U.S. Naval Observatory and the General Catalog of Variable Stars maintained at the Sternberg Astronomical Institute, Moscow, Russia.
Captions to figures: Fig. 1.-Broadening functions for 9 contact binary systems discussed in Section 2. The orbital phase is given in the right side of each panel. The last panel gives the BF representing the nominal resolution of the method. -The 3 first panels show broadening functions for multiple systems discussed in Section 3, while the next 5 panels show the respective functions for radial velocity variables of mostly unknown type, as discussed in Section 4. Note that "strengths or "intensities" of the BF's have different vertical scales for different stars. This is because the BFs depend not only on the geometric (rotational) broadening, but also on how well spectral types of the template and of the program star match. For a perfect fit, the integral of the BF should give unity. The Y-axis units correspond to the BF sampling at 12.5 km s −1 per point. Note. -The table gives the radial velocities RV i and associated weights W i for observations of 8 stars described in Section 2. The velocities are expressed in km s −1 . The weights W i were used in the orbital solutions and can take values of 1.0, 0.5 or 0; the zero weight observations may be eventually used in more extensive modeling of broadening functions. Note. -The column DDO gives the number of the DDO paper or "new" for this paper. n obs and n used give the number of available and used RV measurements, respectively. The radial velocity parameters V 0 , K 1 and K 2 and their rms errors are in km s −1 . T 0 is the heliocentric Julian Day of the superior conjunction (eclipse). The period P is in days. The assumed and fixed quantities are in square brackets. | 2019-04-14T01:46:18.875Z | 2006-06-21T00:00:00.000 | {
"year": 2006,
"sha1": "04172feea9a82f04bfec98297fbfccf6d45865b3",
"oa_license": null,
"oa_url": null,
"oa_status": "CLOSED",
"pdf_src": "Arxiv",
"pdf_hash": "04172feea9a82f04bfec98297fbfccf6d45865b3",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
13928254 | pes2o/s2orc | v3-fos-license | Exploring cross-language statistical machine translation for closely related South Slavic languages
This work investigates the use of cross-language resources for statistical machine translation (SMT) between English and two closely related South Slavic languages, namely Croatian and Serbian. The goal is to explore the effects of translating from and into one language using an SMT system trained on another. For translation into English, a loss due to cross-translation is about 13% of BLEU and for the other translation direction about 15%. The performance decrease for both languages in both translation directions is mainly due to lexical divergences. Several language adaptation methods are explored, and it is shown that very simple lexical transformations already can yield a small improvement, and that the most promising adaptation method is using a Croatian-Serbian SMT system trained on a very small corpus.
Introduction
Statistical machine translation has become widely used over the last decade -open source tools such as Moses (Koehn et al., 2007) make it possible to build translation systems for any language pair within days, or even hours. However, the prerequisite is that appropriate bilingual training data is available, which is actually one of the most severe limitations of the statistical approach -large resources are only available for a few language pairs and domains. Therefore exploiting language closeness can be very convenient if there are no appropriate corpora containing the desired language, but it is possible to acquire corpora containing a closely related one. Croatian and Serbian are very close languages, and both 1 are under-1 as well as other South Slavic languages resourced in terms of free/open-source language resources and tools, especially in terms of parallel bilingual corpora. On the other hand, Croatian has recently become the third official South Slavic language in the EU 2 , and Serbian 3 is the official language of a candidate member state. Therefore investigating cross-language translation for these two languages can be considered very useful.
Both languages belong to the South-Western Slavic branch. As Slavic languages, they have a free word order and are highly inflected. Although they exhibit a large overlap in vocabulary and a strong morphosyntactic similarity so that the speakers can understand each other without difficulties, there is a number of small, but notable and frequently occurring differences between them.
In this paper, we investigate the impact of these differences on cross-language translation. The main questions are: • How much will the translation performance decrease if a Serbian-English SMT system is used for translation from and into Croatian? (and the other way round) • What are the possibilities for diminishing this performance decrease?
Related work
First publications dealing with statistical machine translation systems for Serbian-English (Popović et al., 2005) and for Croatian-English (Ljubešić et al., 2010) are reporting results of first steps on small bilingual corpora. Recent work on Croatian-English pair describes building a parallel corpus in the tourism domain by automatic web harvesting (Esplà-Gomis et al., 2014) and results of a SMT system built on this parallel corpus which yielded significant improvement (10% 2 together with Slovenian and Bulgarian 3 together with Bosnian and Montenegrin BLEU) over the Google baseline in the tourism domain (Toral et al., 2014). A rule-based Apertium system (Peradin et al., 2014) has been recently developed for translation from and into Slovenian (also closely related language, but more distant). Techniques simpler than general SMT such as character-level translation have been investigated for translation between various close language pairs, where for the South Slavic group the Bulgarian-Macedonian pair has been explored (Nakov and Tiedemann, 2012). Characterbased translation has also been used for translating between Bosnian and Macedonian in order to build pivot translation systems from and into English (Tiedemann, 2012).
Developing POS taggers and lemmatizers for Croatian and Serbian and using Croatian models on Serbian data has been explored in (Agić et al., 2013).
To the best of our knowledge, a systematic investigation of cross-language translation systems involving Croatian and Serbian, thereby exploiting benefits from the language closeness and analyzing problems induced by language differences has not been carried out yet.
2 Language characteristics 2.1 General characteristics Croatian and Serbian, as Slavic languages, have a very rich inflectional morphology for all word classes. There are six distinct cases affecting not only common nouns but also proper nouns as well as pronouns, adjectives and some numbers. Some nouns and adjectives have two distinct plural forms depending on the number (less than five or not). There are also three genders for the nouns, pronouns, adjectives and some numbers leading to differences between the cases and also between the verb participles for past tense and passive voice.
As for verbs, person and many tenses are expressed by the suffix, and the subject pronoun (e.g. I, we, it) is often omitted (similarly as in Spanish and Italian). In addition, negation of three quite important verbs, "biti" (to be, auxiliary verb for past tense, conditional and passive voice), "imati" (to have) and "ht(j)eti" (to want, auxiliary verb for the future tense), is formed by adding the negative particle to the verb as a prefix.
As for syntax, both languages have a quite free word order, and there are no articles.
Differences
The main differences between the languages are illustrated by examples in Table 1.
The largest differences between the two languages are in the vocabulary. Months have Slavicderived names in Croatian whereas Serbian uses standard set of international Latin-derived names. A number of other words are also completely different (1), and a lot of words differ only by one or two letters (2). In addition, Croatian language does not transcribe foreign names and words, whereas phonetical transcriptions are usual in Serbian although original writing is allowed too (3).
Apart from lexical differences, there are also structural differences mainly concerning verbs. After modal verbs such as "morati" (to have to) or "moći" (can) (4), the infinitive is prescribed in Croatian ("moram raditi"), whereas the construction with particle "da" (that/to) and present tense ("moram da radim") is preferred in Serbian. An inspection of the Croatian and Serbian web corpora 4 (Ljubešić and Klubička., 2014) shows the prescription being followed by identifying 1286 vs. 29 occurrences of the two phrases in the Croatian and 40 vs. 322 occurrences in the Serbian corpus. It is important to note that the queried corpora consist of texts from the Croatian and Serbian top-level web domain and that the results in discriminating between Croatian and Serbian language applied to these corpora are not used at this point.
The mentioned difference partly extends to the future tense (5), which is formed in a similar manner to English, using present of the verb "ht(j)eti" as auxiliary verb. The infinitive is formally required in both variants, however, when "da"+present is used instead, it can additionally express the subject's will or intention to perform the action. This form is frequent in Serbian ("já cu da radim"), whereas in Croatian only the infinitive form is used ("jaću raditi"). This is, again, followed by corpus evidence with 0 vs. 71 occurrences of the phrases in the Croatian corpus and 13 vs. 22 occurrences in the Serbian corpus. Another difference regarding future tense exists when the the auxiliary and main verb are reversed (5b): in Croatian the final "i" of the infinitive is removed ("raditću"), whereas in Serbian the main and the auxiliary verb merge into a single word ("radiću"). future tense a) jaću raditi jaću da radim I will work b) raditću radiću I will work 6) "trebati" = should a) trebam raditi treba da radim I should work trebaš raditi treba da radiš you should work = need b) trebam posao treba mi posao I need a job Petar treba knjige Petru trebaju knjige Petar needs books Table 1: Examples of main differences between Croatian and Serbian.
Corpus evidence follows this as well with 611 vs. 9 occurrences in the Croatian corpus and 4 vs. 103 occurrences in the Serbian one. A very important difference concerns the verb "trebati" (to need, should) (6). In Croatian, the verb takes the tense according to the subject and it is transitive as in English. In Serbian, when it means "should" (6a) it is impersonal followed by "da" and the present of the main verb ("treba da radim"). When it means "to need" (6b), the verb is conjugated according to the needed object ("treba" (job), "trebaju" (books)), and the subject which needs something (I, Petar) is an indirect grammatical object in dative case ("meni", "Petru").
Apart from the described differences, there is also a difference in scripts: Croatian uses only the Latin alphabet whereas Serbian uses both Latin and Cyrillic scripts 5 . However, this poses no problem regarding corpora because a Cyrillic Serbian 5 During the compilation process of the Serbian web corpus (Ljubešić and Klubička., 2014), 16.7% of retrieved text was written in the Cyrillic script. text can be easily transliterated into Latin. The idea of Figure 1 is to illustrate the closeness and the differences between the two close languages of interest by numbers: overlapping of word level and character level n-grams for n = 1, ...6 in training, development and test corpora together is presented via the F-score. In order to give a better insight, overlaps with English are calculated as well. It can be seen that the Croatian-Serbian overlap on character level is very high, and still rather high on the word level. Character overlaps with English are below the Croatian-Serbian overlap on the word level, whereas the word level overlaps with English are very low.
Translation experiments
In order to explore effects of the described language differences on cross-language SMT, four translation systems have been built: Croatian→English, Serbian→English, English→Croatian and English→Serbian. For the sake of brevity and clarity, we will use the terms "corresponding source/output" when the test language is same as the language used for training, and "other source/output" when the cross-language translation is performed. For translation into English, the translation outputs of the other source text and its adapted variants are compared to the translation output of the corresponding source test with respect to the English reference. For translation from English, the other translation output and its adapted versions are compared to the corresponding output with respect to the corresponding reference. The investigated adaptation methods are described in the next section.
Language adaptation methods
The following methods were investigated for adaptation of the test set in the other language: • lexical conversion of the most frequent words (conv); The most frequent 6 different words together with simple morphological variations are replaced by the words in the corresponding language. This method is simple and fast, however it is very basic and also requires knowledge of the involved languages to be set up. It can be seen as a very first step towards the use of a rule-based Croatian-Serbian system.
• Croatian-Serbian translation system trained on three thousand parallel sentences (3k); 6 occurring ≥ 1000 times in the training corpus This method does not require any language knowledge, and a small bilingual corpus is often not very difficult to acquire. It is even not very difficult to create it manually from a monolingual corpus by translating it, although in that case the language knowledge is needed.
• Croatian-Serbian translation system trained on the large parallel corpus (200k); This method is interesting in order to see the upper limits of the adaptation, however it is not realistic -if a large in-domain corpus is available in both languages, there is no need for cross-language translation, but pivoting or synthetic corpora can be used.
The language adaptation is performed in the following way: for translation into English, the other language test set is first preprocesssed, i.e. converted or translated into the corresponding language, and then translated. For the other translation direction, the English test is translated into the other language and then converted/translated into the corresponding one.
In addition, training a system using the converted corpus has also been investigated for all translation directions.
Experimental set-up
The enhanced version 7 of the SEtimes corpus (Tyers and Alperen, 2010) is used for translation experiments. The corpus is based on the content published on the SETimes.com news portal which publishes "news and views from Southeast Europe" in ten languages: Bulgarian, Bosnian, Greek, English, Croatian, Macedonian, Romanian, Albanian and Serbian. We used the parallel trilingual Croatian-English-Serbian part of the corpus. The detailed corpus statistic is shown in Table 2. The Croatian language is further referred to as hr, Serbian as sr and English as en.
The translation system used is the phrase-based Moses system (Koehn et al., 2007). The evaluation metrics used for assessment of the translations are the BLEU score (Papineni et al., 2002) and the F-score, which also takes recall into account and generally better correlates with human rankings which has been shown in (Melamed et al., 2003) and confirmed in (Popović, 2011
Croatian↔Serbian language adaptation
This section presents the results of conversion and translation between Croatian and Serbian in order to better understand advantages and disadvantages of each of the adaptation methods. The effects of each method on translation into and from English will be reported in the next section. Table 3 shows the BLEU and F-scores as well as the percentage of running OOVs for each adaptation method. If no adaptation is performed (first row), the word level scores are about 40%, CHARF score is close to 75% , and a large number of OOVs is present -13% of running words are unseen. A large portion of these words differ only by one or two characters, and for a standard SMT system there is no difference between such words and completely distinct ones.
The conv method, i.e. simple replacement of a set of words, already makes the text more close: it reduces the number of OOVs by 3-5% and improves the scores by 3%. The best results are obtained, as it can be expected, by 200k adaptation, i.e. translation using the large Croatian-Serbian training corpus; the amount of OOVs in the adapted text is comparable with the text in the corresponding language (presented in Table 2). The 3k translation system, being the most suitable for "realword" tasks and improving significantly the text in the other language (almost 10% reduction of OOVs and 13% increase of scores) seems to be the most promising adaptation method.
Croatian/Serbian↔English translation
The translation results into and from English are presented in Table 4. It can be seen that the BLEU/WORDF loss induced by cross-language translation is about 12-13% for translation into English and about 13-15% for the other direction. The effects of language adaptation methods are similar for all translation directions: the simple lexical conversion conv slightly improves the translation outputs, and the best option is to use the 200k translation system. The small training corpus achieves, of course, less improvement than the large corpus. On the other hand, taking into account the significant improvement over the original of the text of the other language (about 9%) and the advantages of the method discussed in Sections 3.1 and 5.1, this performance difference is actually not too large. Future work should explore techniques for improvement of such systems.
Last two rows in each table represent the results of the additional experiment, namely using the converted other language corpus for training. However, the results do not outperform those obtained by (much faster) conversion of the source/output, meaning that there is no need for retraining the translation system -it is sufficient to adapt only the test source/output. The examples are given only for translation into English, and the effects for the other translation direction can be observed implicitly. Generally, the main source of errors are OOV words, but structural differences also cause problems.
Translation examples
For the first sentence (1), the conv method is sufficient for obtaining a perfect cross-translation output: the obstacles are three OOV words, all of them being frequent and thus converted. The outputs obtained by 3k and 200k methods as well as the output for the corresponding language are exactly the same and therefore not presented.
The second sentence (2) is more complex: it contains three OOV words, two of which are not frequent and thus not adapted by conv, and one future tense i.e. a structural difference. The OOV words do not only generate lexical errors (untranslated words) but also incorrect word order ("from 17 dječjih kazališta"). The conv method is able to repair only the month name, whereas other errors induced by language differences 8 are still present. The 3k translation system resolves one more OOV word ("theater") together with its position, as well as the future tense problem, but the third OOV word "children's" is still untranslated and in the wrong position. This error is fixed only when 200k translation system is used, since the word occurs in the large corpus but not in the small one. It should be noted that the word is, though, an OOV only due to the one single letter and probably could be dealt with by character-based techniques (Nakov and Tiedemann, 2012) which should be investigated in future work.
Conclusions
In this work, we have examined the possibilities for using a statistical machine translation system built on one language and English for translation from and into another closely related language. Our experiments on Croatian and Serbian showed that the loss by cross-translation is about 13% of BLEU for translation into English and 15% for translation from English.
We have systematically investigated several methods for language adaptation. It is shown that even a simple lexical conversion of limited number of words yields improvements of about 2% BLEU, and the Croatian-Serbian translation system trained on three thousand sentences yields a large improvement of about 6-9%. The best results are obtained when the translation system built on the large corpus is used; however, it should be taken into account that such scenario is not realistic.
We believe that the use of a small parallel corpus is a very promising method for language adaptation and that the future work should concentrate in improving such systems, for example by character-based techniques. We also believe that a rule-based Croatian-Serbian system could be useful for adaptation, since the translation performance has been improved already by applying a very simple lexical transfer rule. Both approaches will be investigated in the framework of the ABU-MATRAN project 9 .
Depending on the availability of resources and tools, we plan to examine texts in other related languages such as Slovenian, Macedonian and Bulgarian (the last already being part of ongoing work in the framework of the QTLEAP project 10 ), and also to do further investigations on the Croatian-Serbian language pair. hr Subotica in Serbia will be will host the 16th International Festival from 17 dječjih kazališta to 23 svibnja. hr-sr.conv Subotica in Serbia will be will host the 16th International Festival from 17 dječjih kazališta to 23 May. hr-sr.3k Subotica in Serbia will host the 16th International Theatre Festival from 17 dječjih to 23 May. hr-sr.200k Subotica in Serbia will host the 16th International Children's Theatre Festival from 17 to 23 May. sr The Serbian town of Subotica will host the 16th edition of the International Children's Theatre Festival from 17 to 23 May. Table 5: Two examples of cross-translation of Croatian source sentence into English using Serbian→English translation system. | 2015-07-06T21:03:06.000Z | 2014-10-01T00:00:00.000 | {
"year": 2014,
"sha1": "2c49297c32571b60c60f5d5644296a3d1243c89a",
"oa_license": null,
"oa_url": "http://www.aclweb.org/anthology/W/W14/W14-4210.pdf",
"oa_status": "GREEN",
"pdf_src": "ACL",
"pdf_hash": "2c49297c32571b60c60f5d5644296a3d1243c89a",
"s2fieldsofstudy": [
"Computer Science",
"Linguistics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
234861006 | pes2o/s2orc | v3-fos-license | DEMONSTRATIVE REFERENCES OF THIS AND THAT IN ENGLISH SPEECHES BY CHINESE COLLEGE STUDENTS
Demonstrative reference is a common means of cohesion in discourse. Based on Cohesion Theory by Halliday and Hasan, the paper is aimed at exploring the use of this and that as a means of deixis and reference in college English speeches from the perspective of discourse cohesion to achieve more effective analysis and appreciation of English speeches.
INTRODUCTION
In the concept of discourse, cohesion is the indispensable means of generating ideas, and reference is an important way to embody this means. Halliday and Hassan, representatives of the functional linguistic school, subdivided anaphora into personal reference, deictic reference and comparative reference (1976: 4). Anaphora is the connection point between one linguistic element and another. Both the explicit reference point in English and the implicit reference point in Chinese can be found in the text context respectively. In the process of interpreting and translating a text, the reference relationship between language components is an important basis for the interpreter to decode text information. According to Halliday & Hasan (1976), the interpretation of a word cannot be obtained from the word itself, whereas the answer must be found from the object meant by the word, which leads to anaphora. Anaphora provides people with the information needed to understand the relevant language components. Deictic reference is an important means of discourse cohesion. Grammarians have done a lot of research on English demonstrative pronouns this and that in the past, but most of them are confined to the traditional syntactic field. With the development and change of language, traditional syntactic research has been unable to answer some linguistic phenomena of this and that demonstrative references. With the development of modern linguistics, the study of this and that has turned to the discourse field involving pragmatics, semantics and cognition. Hou Guojin (2002) reviewed the pragmatics of this; Deng Xiaoling (2004) explored the pragmatic similarities and differences between Chinese and English deictic words "zhe" and "this"; Yang Huiwen (2012) conducted a textual translation equivalence study on the Chinese translation of this and that; David Peeters, Aslı Özyürek (2016) reconsidered spatial demonstratives this and that from a social and multimodal approach. In sum, this and that as demonstrative references have been covered from many perspectives, with few studies on the use of this and that in speech discourses. College Students' English speeches can fully demonstrate the speech skills and English proficiency of contemporary Chinese college students. This paper intends to explore the deictic reference of this and that as deictic and referential devices in college English speeches by Chinese college students.
RESEARCH DESIGN Research questions
This study aims to analyze the frequency of this and that as determiners and pronouns, and that as conjunctions in English speeches by Chinese college students, together with their contextual analysis in College English speeches. As demonstrative pronouns, this and that can be used as a subject, an object, a predicative, etc., to refer back or avoid repetition; as determiners, they can be followed by various common nouns, that is, individual nouns, collective nouns, material nouns and abstract nouns. What are their specific usages in college English speeches when used as demonstrative pronouns and determiners? How is the correlation manifested in students' speeches accordingly?
Research corpus
The corpus of this paper is a total of 40 college English speeches by the winners of Chinese National English speech contests. Students choose different topics to give speeches. The content involves politics, economy, society, campus, personal life and so on. The brief information of 40 college English speeches is shown in Table 1. As can be seen from Table 1, the average length of each speech is less than 200 words, whereas each speech has to be powerful and persuasive in diction. The second person pronouns of the 40 speakers were used the least and the first person pronouns were used the most. This is because the second person often produces a hint that separates the speaker from the listener. The first person is convenient for the expression of subjective psychology, which is highly infectious and easy to express emotions.
Research method
In this paper, qualitative and quantitative analysis methods are used. Qualitative research in this paper refers to cohesive devices, with each speech carefully being analyzed to identify and confirm the language environment in which this and that are used as determiners and pronouns respectively in 40 speeches. In this paper, the quantitative research is accurate with Corpus tool and SPSS17.0 software being used to count the frequency of this and that, and to analyze the correlation of the same usage of this and that, for the accuracy and objectivity of data analysis. According to the classification of referential means, the software annotation mainly refines the usage of this and that as qualifier and pronoun, and that as conjunction respectively. After the framework being set in the software, each speech is coded and the targets are counted.
RESULTS AND DISCUSSION
As for pronouns, this and that can function as a subject, an object and a predicative in a sentence; that can also appear at the beginning of attributive clauses as relative pronouns; as determiners, they can be followed by single nouns, collective nouns, material nouns and abstract nouns; as conjunctions, that can be located at the beginning of subject clauses, predicative clauses, object clauses and appositive clauses. Table 2 shows the frequency of this and that in 40 College English speeches. As can be seen from Table 2, this is mainly used as a subject in the speech, whereas that is mainly used as a relative pronoun to lead a clause; this and that are mainly followed by individual nouns when they are used as determiners; that as conjunctions, leading objective clauses the most, while the predicative clauses and subject clause the least. As for pronouns, that is used more frequently than this; while as determiners, that is used less frequently. When used in a text, this and that together with the referent can play an important role in the text structure cohesion, the anaphora being essential in the context of the discourse. The deictic usage of this and that means that their referents and their meanings in sentences can be confirmed only under the explicit contexts. Without the specific context, their referential contents or information will be ambiguous. Table 3 presents the language environment of this and that in 40 English speeches by Chinese college students. In Table 3, six pairs of related samples being presented in the T-test, with this as a subject, an object and that as a subject, an object respectively, because P-value is greater than 0.05, there is no significant correlation.
However, when this and that being followed by individual nouns, P-value is less than 0.01, indicating a significant correlation. Table 4 partial correlation analysis is used to analyze the correlation between this and that as determiners used after individual nouns, collective nouns, material nouns and abstract nouns respectively. Table 4 shows the variables that may affect the linear correlation between two variables. When the total amount of this and that as determiners being controlled, whether there is any correlation between these nouns following this and that, it can be seen from the table that the correlation between this followed by individual nouns and collective nouns is stronger than the correlation between this followed by material nouns and abstract nouns. As for the use of nouns used after that, there is no correlation between collective nouns and individual nouns, material nouns and collective nouns, abstract nouns and collective nouns. Table 5 partial correlation analysis shows the correlation between this and that as pronouns and their use as subjects, objects and predicatives respectively. It can be seen from Table 5 that the correlation between this as a subject and as an object is the strongest, while the correlation between this as an object and as a predicative is the weakest; that as a subject and as a relative pronoun being the strongest, and that as an object and a predicative being the weakest. Table 6 presents the correlation between that as a conjunction and its leading a subjective clause, an objective clause, a predicative clause and an appositive clause respectively. It can be seen from Table 6 that when that is used as a conjunction, the correlation between its leading an objective clause and an appositive clause is the greatest, while the correlation between its leading a subjective clause and an appositive clause is the weakest.
CONCLUSION
Being a powerful means of communication, speeches are demanding in language use. In speech contests, students express their opinions and display their speech skills. Speech contests are supposed not only to be novel in content but unique in conception, thus grasping the audience from the beginning to the end, moving the audience, conquering the audience and leaving a deep impression on the audience. The frequency analysis of this and that as deixis and pronouns, as well as the correlation in the language environment, will help Chinese college students better master the writing skills of speeches and improve their appreciation of public speaking so as to hopefully avoid vagueness and ambiguity in communication. | 2021-05-21T16:56:43.888Z | 2021-04-17T00:00:00.000 | {
"year": 2021,
"sha1": "3b75f2f7e00f9f0e73ee2cf491cb0a5f29819645",
"oa_license": "CCBY",
"oa_url": "https://esciencepress.net/journals/index.php/JSAS/article/download/3331/1788",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "5622d99a43391dcebfa710c333d8ba32b4e7119f",
"s2fieldsofstudy": [
"Education",
"Linguistics"
],
"extfieldsofstudy": [
"Sociology"
]
} |
8970433 | pes2o/s2orc | v3-fos-license | Parity Types, Cycle Structures and Autotopisms of Latin Squares
The parity type of a Latin square is defined in terms of the numbers of even and odd rows and columns. It is related to an Alon-Tarsi-like conjecture that applies to Latin squares of odd order. Parity types are used to derive upper bounds for the size of autotopy groups. A fast algorithm for finding the autotopy group of a Latin square, based on the cycle decomposition of its rows, is presented.
Introduction
For a given positive integer n let [n] denote the set {1, . . . , n}. A Latin square of order n is an n × n array of numbers in [n] so that each number appears exactly once in each row and column. Let LS(n) be the set of Latin squares of order n and let S n be the symmetric group of permutations of [n]. The group S 3 n = S n × S n × S n acts on LS(n) by isotopism. An isotopism is a triple (α, β, γ) ∈ S 3 n that acts on LS(n), so that α permutes the rows, β permutes the columns and γ permutes the symbols of a given Latin square. Two Latin squares are called isotopic if there is an isotopism that transforms one to the other. Being isotopic is an equivalence relation and the set of Latin squares that are isotopic to each other is an isotopy class. Let I n = S 3 n be the group of isotopisms on LS(n), I n (L) the isotopy class of L, and A(L) = {Θ ∈ I n |Θ(L) = L} the autotopy group of L (also denoted in the literature by Is(L) [12], Atp(L) [2,15], and Atop(L) [13]), whose members are called autotopisms of L. We have that |I n | = (n!) 3 and |I n (L)| = (n!) 3 / |A(L)|. The rows and columns of a Latin square can be viewed as permutations of the ordered set (1, 2, . . . , n). A reduced (also called normalized ) Latin square is a Latin square whose first row and first column are the identity permutation. Let RLS(n) be the set of reduced Latin squares of order n. For L ∈ RLS(n) let I ′ n,L = {Θ ∈ I n |Θ(L) ∈ RLS(n)} and let I ′ n (L) = I n (L) ∩ RLS(n). Any Latin square can be transformed by isotopism to a reduced Latin square by permuting the rows to form the identity permutation in the first column and then permuting the columns 2, . . . , n to form the identity permutation in the first row. Thus, each reduced Latin square is isotopic to n!(n − 1)! distinct Latin squares in this way. It follows that |I n (L)| / |I ′ n (L)| = n!(n − 1)! and hence I ′ n,L = n · n! and |I ′ n (L)| = n · n!/ |A(L)| (1.1) For these and other facts about isotopisms and autotopisms the reader is referred to Dénes and A. D. Keedwell's book [4] and to the works of Janssen [10], McKay and Wanless [12] and McKay, Meynert and Myrvold [11] among many others.
The sign (or parity) function sgn : S n → {−1, 1} is defined so that sgn(σ) = 1 if and only if σ is an even permutation. The parity of a Latin square L, denoted par(L), is the product of the signs of all its rows and columns. A Latin square is even (resp. odd ) if its parity is 1 (resp. -1). Let ELS(n) (resp. OLS(n)) be the set of even (resp. odd) Latin squares of order n. Let RELS(n) and ROLS(n) be the corresponding sets of reduced Latin squares. If n is odd |ELS(n)| = |OLS(n)|. If n is even, a known conjecture of Alon and Tarsi [1] states that |ELS(n)| = |OLS(n)|. It was also conjectured in [15] that |RELS(n)| = |ROLS(n)| for all n (for even n this is equivalent to the Alon-Tarsi conjecture). For further results regarding even and odd Latin squares, the reader is referred to the works of Janssen [10], Drisko [5,6], Zappa [17], Glynn [8] and Stones and Wanless [15]. This work is mainly concerned with the relation between parities of Latin squares and of their rows and columns and autotopy groups. In Section 2 an expression for the members of I ′ n,L is derived and the notions of parity type and parity class are introduced. These lead to expressions for the numbers of even and odd reduced Latin squares of odd order in an isotopy class. These expressions, although stated and proved differently coincide with results of Stones and Wanless [15]. In Section 3 an extension of the Alon-Tarsi Latin square conjecture is discussed and a related conjecture on parity types is presented. In Sections 4 and 5 bounds are obtained for the order of the autotopy group A(L). In Section 4 the bounds are obtained using parities of the rows and columns of L and in Section 5 the bounds are derived from the cycle structure of the rows of L. Convention: Throughout the manuscript, when viewing rows and columns of a Latin square as permutations in S n , the following convention will hold: the number i appearing in the jth place of the row (column) σ of L signifies that σ(i) = j. By this convention, in order to transform a column (or row) σ to the identity permutation, we have to apply the permutation σ −1 to the rows (resp. columns) of L (see Lemma 1(i)). It is also assumed that permutations are applied from right to left.
Parity types and parity classes
A permutation acting on a Latin square acts on another permutation (a row or a column) in two different ways, as described by the following lemma. Recall the convention of this paper that when i appears in the jth place of a row (or column) π it means that π(i) = j. Lemma 1. Let L be a Latin square of order n.
(i) The result of permuting the rows (resp. columns) of L, by α ∈ S n , on any column (resp. row) π is απ.
(ii) The result of permuting the symbols of L, by α ∈ S n , on any row or column π is πα −1 .
Proof. (i) Let π be a row or a column. Suppose π(i) = j and α(j) = k then i appears in the jth place of π, and after applying α, the number in the jth place of π (i in this case) moves to the kth place. Thus the resulting permutation takes i to k, so it is απ.
(ii) Denote by α(π) the permutation obtained by applying α to the each symbol In the row (or column) π. Suppose π(i) = j and α(i) = k. Since i is in the jth place of π, k is in the th place of α(π). Thus α(π)(k) = j. It follows that α(π) = πα −1 .
The following proposition appears in a different form as part of the proof of Theorem 2.1 in [15]. It describes the n · n! elements in the set I ′ n,L .
When n is even, applying an isotopism preserves the parity of a Latin square ( [5], Lemma 1), and thus the parity of all the Latin squares in an isotopy class is the same. When n is odd an isotopy class contains an equal number of even and odd Latin squares. This is not the case when only reduced Latin squares are concerned. In order to determine the number of even and odd reduced Latin squares in an isotopy class the following definition will be useful: the electronic journal of combinatorics (), # Definition 1. The parity type of a Latin square L of order n is defined to be (k, m), for 0 k, m n/2, if k (respectively m) of its rows (resp. columns) have one sign and the remaining n − k rows (resp. n − m columns) have the opposite sign. The parity class (k, m) consists of all Latin squares of parity type (k, m).
Proposition 2. A parity class is the union of isotopy classes.
Proof. Since an isotopy either changes the signs of all the rows (columns) or leaves them all unchanged, the parity type of two isotopic Latin squares is equal.
Thus, it is valid to refer to the parity type of an isotopy class, as parity type of any of its members.
The following theorem appears in a different form in [15] (Theorem 2.1). The proof here is different and was obtained independently. Theorem 1. Suppose n is odd and the parity type of L ∈ RLS(n) is (k, m), then Proof. Suppose k and m are either both even or both odd (first row in (2.1). Since n is odd, n − k and n − m are either both odd or both even respectively. Let I = {i 1 , . . . , i k } be the set of indices of the k rows of L having the same sign and let J = {j 1 , . . . , j m } be the set of indices of the m columns of L having the same sign. Suppose L is even. Since k and m have the same parity, the rows indexed by I and the columns indexed by J must have the same sign and the rest of the rows and columns all have the opposite sign. Now, let Θ = (α, απ j σ −1 α −1 (1) , απ j ) ∈ I ′ n,L (see Proposition 1) be such that Θ(L) is even. By Lemma 2, σ α −1 (1) and π j must have the same sign. So, either α −1 (1) ∈ I and j ∈ J or α −1 (1) ∈ [n] \ I and j ∈ [n] \ J. Since the number of permutations α such that α −1 (1) ∈ I (resp. α −1 (1) ∈ I) is (k/n)n! = k(n − 1)! (resp. (n − k)(n − 1)!), the number of isotopies Θ such that Θ(L) is even is (km + (n − k)(n − m))(n − 1)! and thus the number of distinct reduced even Latin squares that are isotopic to L is [km the electronic journal of combinatorics (), # Hence, The other case is proved in an analogous manner and is left to the reader.
Remark 3. Note that if k = m = 0 all the squares in I ′ n (L) are even.
On even and odd Latin squares of odd order
The following extension to all n of the Alon-Tarsi Latin square conjecture [1] appears in [15]: When n is even Conjecture 1 is equivalent to the Alon-Tarsi conjecture. When n is odd it is not known whether Conjecture 1 is equivalent to another extension of the Alon-Tarsi conjecture by Zappa [17]. Let N n (k, m) be the number of reduced Latin squares of order n in the parity class (k, m). We have the following corollary of Theorem 1: Proof. Since a parity class is the union of isotopy classes (Proposition 2), the result follows from Theorem 1 by summing over all the isotopy classes contained in a given parity class.
Corollary 1 yields an equivalent version of Conjecture 1 for odd n: the electronic journal of combinatorics (), # Proof. By Corollary 1, if k ≡ m (mod 2) the difference between the number of even reduced Latin squares and the number of odd reduced Latin squares in the parity class (k, m) is and when k ≡ m (mod 2), this difference is − (n−2k)(n−2m) . Summing over all parity classes, Assuming Conjecture 1 statement (3.1) follows.
This calls for a study of the values N n (k, m). When n = 5 only the parity classes (0, 0) and (1, 1) are nonempty. When n = 6 the only empty classes are (0, 2) and (2, 0). When n = 7 all parity classes are nonempty, as is the case for n = 8, 9, 10 and 11 (Tables 1-5). Thus, it seems reasonable to conjecture that for all n 7 and all k, m n/2 the parity class (k, m) is nonempty. Furthermore, an estimate for the values N n (k, m) is desired. It is conjectured in [3] that the distribution of the permutation that maps between two randomly chosen rows of a randomly chosen Latin square of order n, among all derangements, approaches uniformity as n → ∞ (a derangement is a permutation without fixed points). Based on this, it seems reasonable to conjecture that for large enough n the distribution of the parities of the rows and columns of a random Latin square will be close to the random distribution. This should also hold when restricting to reduced Latin squares, namely, if the parities of the rows and columns were chosen randomly, the proportion of reduced Latin squares of parity class (k, m) among all reduced Latin squares would be rs n k n m /2 2n , where r = 2 (resp. s = 2) if k < n/2 (resp. m < n/2) and 1 otherwise. (since k, m n/2). Thus, rs is always 4 if n is odd.
Conjecture 2 is supported by the data in the following tables. Table 1 counts of N n (k, m), as calculated for 100,000 randomly selected reduced Latin squares (using the Jacobson-Matthews method [9]), for each of n = 8, 9, 10 and 11, along with the expected values calculated using (3.3) and (3.4) for this sample size (similar simulations were performed for n = 12, . . . , 20 and close values were observed as well for estimated and observed counts). The counts for the classes (k, m) and (m, k), when k = m, were added, since the corresponding classes have equal size. Note that 0 as an estimated value does not mean that the class is expected to be empty, but that for the given sample size no member of the class is expected to be observed. Lemma 2.4 in [15] states that almost all Latin squares belong to parity classes (k, m) such that k, m n/63, and thus it can be viewed as a partial result for Conjecture 2.
Thus, the sum in (3.6) is 0 and the result follows.
The conjecture presented in Proposition 4 is supported by Theorem 3.2 in [15] which states that |RELS(n)| and |ROLS(n)| share a large divisor which grows exponentially as n → ∞. The parity type of a square provides some information about its autotopy group. Browning, Stones and Wanless ([2], Lemma 4.1) showed that for a prime order p there is exactly one isotopism class that contains Latin squares L for which p 2 divides |A(L)|, namely, the isotopism class containing the Cayley table of Z p . The following corollary of Theorem 1 extends that result: Proof. Statements (i) and (iii) follow from the above mentioned result in [2] and the fact that the parity type of a Cayley table of odd order is (0, 0). Without loss of generality we may assume that L is reduced (otherwise we take a reduced Latin square in the isotopy class of L and use the fact that isotopy preserves the parity type). Setting n = p we look at one of the expressions for I ′ p (L) ∩ RELS(p) in (2.1). If just one of k and m is nonzero, then p divides the numerator but p 2 does not. Thus, p must divide I ′ p (L) and by (1.1) the highest power of p that may divide |A(L)| is 1. This proves (ii).
For the proof of the next theorem the following definition will be useful: . Since the same argument applies by looking at the second component of the parity set, the inequality (4.1) follows. Let Σ = {α 1 , . . . , α ( n k ) } be a set of permutations that, when taken as the first component of corresponding n k isotopies on L, will yield n k distinct reduced Latin squares with distinct first parity set component. Now, let α ∈ Σ and suppose that Θ 1 = (α, απ j σ −1 α −1 (1) , απ j ) ∈ I ′ n,L and Θ 2 = (α, απ l σ −1 For simplicity, denote σ = σ α −1 (1) . Let β = σπ −1 l π j σ and γ = π −1 l π j = σ −1 βσ. We have that Φ = (1, β, σ −1 βσ) ∈ A(L). While applying Φ to L, after permuting the columns by β, the first row is β (by Lemma 1(i), since it was originally the identity permutation) and we have to apply again β on the symbols in order to transform the first row back into the identity permutation (Lemma 1(ii)). Thus γ = β and we have Φ = (1, β, β). Suppose β(1) = r. After permuting the columns by β, the original first column (the identity permutation) moves to the rth position. Then when applying β on the symbols The rth column becomes β −1 (Lemma 1(ii)), and since we assumed that (1, β, β) ∈ A(L) we must have that β −1 = π r . But β −1 = γ −1 = π −1 j π l . Thus π l = π j π r . If we assume that no column of L is the product of two other columns, we must have that Θ 1 (L) = Θ 2 (L). We conclude that for any α ∈ Σ the different n isotopies obtained by taking the n columns of L yield n distinct squares that are isotopic to L. Since for different elements of Σ the squares obtained have parity sets with distinct first component, |I ′ n (L)| n · n k . By
Computing A(L) by cycle structures
Every permutation α ∈ S n can be decomposed into a unique (up to order) product of disjoint cycles. Cycle structures of permutations were considered in the context of Latin the electronic journal of combinatorics (), # squares in different aspects. Cavenagh, Greenhill and Wanless [3] considered the cycle structure of the permutation that transforms one row of a Latin square to another row. Other works [7,14] considered the cycle structure of the permutations α, β and γ in an isotopism Θ = (α, β, γ), in order to derive information on Latin squares for which Θ is an autotopism. Here the cycle structure of the rows of a Latin square L are considered in order to derive information on A(L).
Theorem 6. The following is an algorithm for finding A(L) for a given reduced Latin square L of order n.
(1) Compute the cycle structures of the rows of L, viewed as permutations in S n . Let C 1 , . . . , C s be the distinct cycle structures of the rows, sorted in some well-defined way (see [7]) and let λ = (λ 1 , λ 2 , . . . , λ s ) be a partition of n where each λ i is the number of rows with cycle structure C i .
(2) For each row σ k , compute the cycle structures of the set of permutations {σ −1 k σ i } n i=1 . Let C k 1 , . . . , C k t be the cycle structures of these permutations, sorted as in (1), and let λ k be the partition of n corresponding to these cycle structures.
Proof. The condition (*) follows directly from Lemma 3. Since α is a bijection the condition in step (3) must hold.
Remark 7. Condition (*) implies that α(k) = 1 since the identity permutation has its own unique cycle structure.
the electronic journal of combinatorics (), # 11. It performed faster than the algorithm by McKay, Meynert and Myrvold [11], based on "nauty", but slower than an improved version of "nauty" using vertex invariants (such as the "train" [16]).
The algorithm of Theorem 6 can be massively sped up by constructing the permutations α in Step (4), in parts, corresponding to the cycle structures C k i in Step (3). After constructing each part we can perform Step (4) on the rows of L that are permuted by the part of α already constructed. This may rule out most of the candidates, or produce candidates α while saving the time needed to consider all the parts corresponding to the cycle structures C k i . This idea is illustrated in the following example.
the electronic journal of combinatorics (), # | 2012-10-03T19:29:39.000Z | 2012-03-01T00:00:00.000 | {
"year": 2012,
"sha1": "059bf14a2074d05d7a0319f91620f74542645712",
"oa_license": null,
"oa_url": "https://www.combinatorics.org/ojs/index.php/eljc/article/download/v19i3p10/pdf",
"oa_status": "GOLD",
"pdf_src": "Arxiv",
"pdf_hash": "059bf14a2074d05d7a0319f91620f74542645712",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics",
"Computer Science"
]
} |
193577096 | pes2o/s2orc | v3-fos-license | The Reception of Harold Pinter’s Plays in Slovenia between 1999 and 2014
Harold Pinter started his career with a conspicuous lack of success. He faced negative critical reviews of his early works, but his typical style eventually opened doors to new worlds in modern drama. On Slovene stages, Pinter’s plays also received a similarly modest welcome. The audience as well as the reviewers found his long pauses, silences and incoherent dialogue insufficiently engaging. One of the main reasons for this could have been their unfamiliarity with Pinter’s style, which eventually acquired its own adjective – ‘Pinteresque’. With time, Pinter’s popularity increased more rapidly on the world stages than in Slovenia, and today this playwright is not a stranger to the Slovene theatre. This article deals with Pinter on Slovene stages as well as the popular and critical reception of his plays. The period before 1999 was thoroughly analysed by Darja Hribar, while this study is the first to focus on the decade and a half following.
The Reception of Harold Pinter's Plays in Slovenia between 1999 and 2014 1 Introduction
According to Peter Raby (2009, 1), statistical data shows that, in the present international world of drama, Pinter is one of its best known as well as most widely performed playwrights. His public activity is by no means limited to drama and theatre. Rather, "the horizon of his literary, cultural and political projects stretches far beyond the borders of his homeland, as well as beyond theatrical and literary world" (Onič 2012, 5). Richard cave agrees with this and even reaches beyond this statement by adding that "one is astonished at the sheer range and variety of endeavour to which [Pinter] has brought a focused and profound commitment" (2009,123).
Apart from his literary heritage, Pinter is most widely known for his activism in defending the human rights of politically oppressed nations and individuals. He frequently stated his world views through his work -articles and letters, poems, plays and radio plays (Derbyshire 2009, 269). Andrew Goodspeed explains that Pinter significantly objected to the "falsity of politicians' analyses and justification of the pain their policies inflict" (2012,54) and that his political beliefs and criticism had appeared continuously in his previous public speeches or interviews; however, his Nobel lecture in 2005 gave him a wider international audience than the usual limited crowd. Therefore, comments suggesting that Pinter had won the award for his political activity and not for his artistic contribution are not surprising (Onič 2007, 117).
Fairly early in Pinter's career, his literary and theatrical excellence as well as his political impact started to move beyond the English-speaking world. His plays first appeared in Slovene theatres in the late 1960s and early 1970s. Initially, critical material and literary research results on this topic were scarce, but the situation has changed recently. Darja Hribar (1999) conducted the first major study of the Slovene translations of Pinter's plays and collected a body of critical reviews referring either to the original plays or to their Slovene theatre productions, thus initiating Pinter Studies in Slovenia. This article gathers the results of a similar research that focused on the theatre productions of Pinter's plays and critical response to them during the 15-year period following Hribar's study.
Pinter on Slovene Stages before 1999
Harold Pinter wrote his first play, The Room, in 1957, and it was staged in London in the same year. Slovenia got its first production of a Pinter play a decade later with The Homecoming in 1967, and Slovene theatres have continued to stage Pinter ever since. Altogether, there were 13 productions of 6 Pinter's plays in Slovenia before the turn to the 21st century. These were Homecoming, The Caretaker, Old Times, The Birthday Party, Betrayal and Ashes to Ashes.
The plays Old Times (Njega dni, 1974/ Stari časi, 1982, 1987 and The Birthday Party (Zabava za rojstni dan, 1979(Zabava za rojstni dan, , 1991(Zabava za rojstni dan, , 1997 were produced three times each. Moreover, the 1979 staging of the latter saw 43 performances, which is the highest number of any single Pinter production on a Slovene stage to date. The next play on the list of repetitions per production is Homecoming (Vrnitev, 1967), with 37 performances, another result that has never been repeated since. The only number of performances in a single theatrical season that surpasses both previously stated figures is 68, which is the number of repetitions of the two 1979 productions of The Birthday Party and Betrayal (Prevara) together. If we compare these figures to those of either traditionally more popular genres (various comedies, etc.) or plays by the then more established playwrights that could receive three or four times as many repetitions, we must agree with Onič (2008b) that Pinter cannot be considered popular with a wide theatre audience of the time. On the other hand, theatre practitioners, who are still today challenged by Pinter's texts, have tended to return to them. Good examples of this are Zvone Šedlbauer, who directed four productions of Pinter in Slovenia, and Miran Herzog, who directed two. In terms of the number of performances, Betrayal comes first with 50 performances altogether. Šedlbauer's 1995 production even ran for two seasons. The only other production achieving this number was Ashes to Ashes (V prah se povrneš) in 1998. It must be noted that some data regarding early performances are not available.
Based on the Slovene theatre production of Pinter's plays before 1999 (Hribar 1999), 1 one can conclude that the playwright was a breath of fresh air for Slovene repertories. The audience's expectations as well as their curiosity were high, the actors and directors were inspired, but most reviewers felt confused. The first Pinter play staged in Slovenia, Homecoming, premiered on 27 October 1967. The reviewer(s) looked for comical aspects of the play and wondered at the reasons for Pinter's writing the piece. Much praise was given to the acting team by all reviewers, but the play itself was marked as an "unnecessary part of the repertory" 2 (vidmar, 1967;cited in Hribar 1999, 206-7), which suggests that even renowned critics like Josip vidmar failed to recognize the value and power of Pinter's text. Some tried to determine the meaning of the play by comparing "human intellect with human primitivism" (Novak, 1967;cited in Hribar 1999, 207) or "finding the truth of human rhapsodically inspired life and non-symphonically regulated existence" (Predan, 1967;cited in Hribar 1999, 208). We can see from these quotations that the reviewers tried to comment on the meaning of the play and sought its comic perspective. This shows the lack of understanding of the playwright's style as well as the adopted misconception that his plays are traditional plot-oriented pieces.
The situation changed slightly with the second Pinter play staged in Slovenia, which was The Caretaker. 3 The theatre programme for its 1970 production offered an extensive study on the playwright and his style. Obviously, the intention was to bring both closer to the public, yet there were still no openly supportive reviews of the performances. Similarly to the response to Homecoming, which had been reviewed as unnecessary and not at all enriching for the repertory, one of the reactions to The Caretaker was that the "director could have, without causing any harm, shortened that typical but tiring repetition of certain phrases" (Onič 2004, 92;see Javornik 1970, 10). According to Hribar (1999, 212-14), the first positive reviews followed the Slovene premiere of Old Times in 1974, which was labelled as a piece of a new dramatic style. In addition, many typical Pinteresque elements were recognized and highlighted in the reviews. Pinter's psychological word play was recognized, and his mastery of interfering with subtextual speech was praised. 1 The majority of data in this section of the article were acquired from Hribar's doctoral dissertation (1999). She commented on the performances of Pinter in Slovenia to 1999 and analysed the reviewer's response. In the same year, Mirko Mrčela (1999) The Birthday Party, which premiered in Ljubljana in 1979, was presented as timeless and independent of any specific setting. As Hribar (1999, 219) reports, the reviewers were still trying to present and explain Pinter's stylistic features to the audience to get a better response from the audience. For example, they interpreted Pinter's dialogue, explained the relevance of information, the allegorical message and the grotesque features. The reviews were increasingly positive, or at least not too harsh. The actors were still getting more praise than the author, but the attention gradually shifted to the characteristics of Pinter's writing style, and the reviewers began to compare it to Beckett and Ionesco, usually referring to its absurdist features (Hribar 1999, 220). One of the first reviewers to use Pinter's name as an adjective 'pinterjevska' to describe the production of a play was Dimitrij Rupel, whose review appeared in Teleks (cited in Hribar 1999, 219). Betrayal was premiered in the same year, and the reviewer pointed out that much attention was given to the existential relationships between the characters through the typical Pinter dialogue. Silence was recognized as a tool of communication, and Pinter was again compared to Beckett and also to Kafka (Hribar 1999, 220).
The 1982 Old Times was a student production at the Academy of Theatre, Radio, Film and Television. Hribar mentions a review by Franc vurnik focusing on Pinter's ability to present an ordinary relationship from a different point of view: "caught in their memories, Pinter's characters somehow always seem to end up in a never ending circle of solitude" (cited in Hribar 1999, 220-21). The reviewers were clearly starting to focus more on Pinter's style and not so much on the performance. The third Slovene production of Old Times introduced a new Slovene version of the adjective Pinteresque, i.e., 'pinterjansko'.4 Although most of the reviewers found Pinter's plays fascinating, and some even called him "a master of playwriting", some still commented that Pinter's plays did not deserve a place on stage, since they could have worked just as well on the radio (Hribar 1999, 222). The dialogues and monologues were critiqued as being boring and too long (Lah 1987, cited in Hribar 1999. At the time of the 1990 production of The Caretaker, the London theatre audience was trying to see Pinter's plays from the political angle. Because of his involvement in political activism, his plays were no longer compared only to Beckett and Kafka; but Michael Billington in The Guardian (1994) also compares him to other artists like Michael Miles or Jack Benny (cited in Hribar 1999, 222-23). After The Birthday Party in 1991 and Betrayal in 1995, the reviews of the 1997 Homecoming took a slight turn towards suggesting how the play should have been understood, highlighting what a difficult task it was for the actors and producers to "capture the intended atmosphere" (Pezdir, 1997;cited in Hribar 1999, 227). The actors were praised for doing a wonderful job (Jež, 1997cited in Hribar 1999, and better reception of the play was attributed to a more mature society, which had, however, still not done away with the taboos "to the extent to make prostitutes, grotesque family relationships and scary unexpected characters part of their acceptable vision of everyday life" (Šuklje, 1997;cited in Hribar 1999, 228). Notably, the reviewers found that the dated translation influenced the quality of the play.
The Birthday Party produced in 1997 was labelled comedy of menace, 5 and the theatre programme dealt extensively with salient characteristics of Pinter's discourse. It praised the playwright's ability 4 There is probably no particular reason for the varied adjective. The Slovene word formation allows pinterjanski and pinterjevski, as well as the calque pinteresken. 5 The expression was coined by Irving Wardle (1958, 28), referring to a rather unusual amalgamation of the comical and the seriously threatening in drama. Although not originally created with Pinter's plays in mind, the concept was more and more frequently used to denote his early plays, most notably The Caretaker, The Dumb Waiter and The Birthday Party.
to use the power of words to create tension, mood and mystery in the play. Pauses and silences earned a place in the positively oriented review by Primož Jesenko (Hribar 1999, 228-29). The response to Ashes to Ashes in 1998 was similarly appreciative of the play and explanatory in terms of possible interpretations. This was the last Pinter produced in Slovenia before the turn to the 21st century.
Pinter on Slovene Stages after 1999
There were fourteen productions of Pinter plays on Slovene stages in the period from 1999 to 2014, which is one more than in the period from 1967 to 1999. A quick comparison suggests that, with almost the same number of premieres in both periods, where the earlier period is more than twice as long as the one after 1999, the frequency of performances has more than doubled in the last decade and a half. This may have to do with the fact that Pinter became an established author, and staging his plays was no longer an experiment, let alone a risk. Apart from his literary activity and a series of prestigious literary awards -some of which came just before the turn of the century -his political engagement contributed to his growing fame. Let us now proceed to particular productions and a selection of critical reviews.
The most frequently staged play after 1999 was The Dumb Waiter, which saw five different productions. This play is also the one with the most variants of title translations within this time frame: Futrlift, Strežni jašek, Jašek and Mutasti natakar, which in back-translation mean, respectively, fodderlift (fodder as animal food; both parts of the compound are dialectal/substandard language use and so is the entire translation 6 ), serving shaft (meaning food elevator connecting dining room and kitchen when they are on different floors), shaft (similar to the previous meaning but less specific, so without the concept of serving attached to it, it hints at a mining shaft or a drain shaft, which allows rich interpretations), and dumb waiter (literal meaning with no reference to food elevator but alluding to several possible readings of the original; a detailed explanation of this translation problem is given by Onič (2011b) in the theatre programme for the 2011 Ptuj production). In terms of the number of productions, The Dumb Waiter is followed by The Birthday Party and Celebration, each of which was staged twice.
The reviewers of the 2000 production of The Dumb Waiter mainly focused on description of and commentary on Pinter's style and on the theatre of the absurd, assuming that the audience would still be unfamiliar with it. Therefore, the reviews included other representative authors of the absurdist movement and set out to popularize the genre of the comedy of menace. The reader cannot avoid the impression that, in a way, the reviewers wanted to comfort the audience and reassure them that the play was supposed to be understood in the way they experienced it; "the performance managed to present Pinter's typical scary and mysterious atmosphere with the hint of humour" (Svetej 2001, 20).
In 2001, The Lover was staged in Ljubljana. The reviewers praised the actors and their acting skills, while, surprisingly, modest praise was given to the playwright. It is possible that applauding Pinter might seem, to them, to be stating the obvious and thus unnecessary. The reviewers seemed obliged to convey their own interpretation of the play or parts of it in the review as an explanation of the complex interpersonal relationships: "[I]maginary lover functions as a symptom of the corroded couple's relationship" (Jesenko 2001, 11). Moreover, a comment by the same reviewer suggests a mysterious hidden meaning for the play: "The play […] gives the impression of an analytical, disciplined in terms of acting, and Pinteresquely bizarre commentary on the stability of the respectable bourgeois marriage, protected by seemingly strong walls" (Jesenko 2001, 11). With this comment, Jesenko points out that everyone has some skeletons in the closet and includes his perception of that observation.
With the production of The Birthday Party in 2002, we witnessed a significant turn in the reviews. The reviewers tended not to dabble in interpretation of the play. They started allowing the mystery of the 'Pinteresque' to speak for itself, encouraging the audience to create their own interpretations. This particular production of The Birthday Party was recognized as a modern classic, since it was placed in modern times, and also as a "well-thought-over mix of genres, atmospheres and relationships" (Jurca Tadel 2003, 265). The reviews suggested that the play was not simple and transparent, but interesting enough for those seeking a theatrical thrill. The readers of the reviews were faced with the options of finding the mystery in the comedy of menace either challenging or too complicated. This production of The Birthday Party had 20 performances, was seen by 3674 playgoers, and also received an award at the most important Slovene theatre event, The Borštnik Festival.
We noticed that a significant rise in publicity accompanied the 2003 production of Celebration. This was a new play by Pinter written in 1999, and some reviewers expressed doubt whether it would live up to expectations; these were probably higher, since Pinter was no longer an obscure author, and the audience already had a certain knowledge of his style. However, all the reviewers commented that Pinter's plays were still as good as forty years ago. The production was a success; it was performed 25 times and seen by 5592 playgoers. The reviewers still commented on Pinter's style, but not in an ambiguous way. His plays in general were acknowledged as quality pieces of theatre, and, in the case of Celebration, it was the performance that was scrutinized and reported as lacking in quality. The reviewers also praised the high quality translation by Alja Predan, which respected and preserved much of Pinter's style, and recognized this fact as significant in securing a good production.
In terms of Pinter on stage, 2004 was a prolific year. Slovenia saw three productions: Remembrance of Things Past, The Dumb Waiter and The Birthday Party. The recognition of Pinter and his style remained unquestioned. He was again noted as a representative of the theatre of the absurd and was often compared to Beckett and Ionesco. The reviewers brought forward the details that could be identified as the 'Pinteresque' essence of the production (or the lack of it, for that matter): "[S]ome scenes were just not scary enough, and towards the end the tension started to drop" (Golob 2004, 11). The evolution of the public response continued in 2006 when the reviews began to embrace the fact that Pinter's plays are applicable at any time and any place, since two performances of The Dumb Waiter were set in the present. The producers and reviewers started to appreciate the depth and the variety of dimensions of Pinter. Marjana Ravnjak commented that the play was a "successful presentation of the present relationship between society and politics, and a constant struggle of lie versus truth, and art versus life in the intense performance" (2006).
The reviewers of the 2008 production of Homecoming observed how 'Pinteresque' elements of the performance influenced the audience: "The absence of actors' feelings was interesting for the audience" (Jurca Tadel 2008, 305). The latter seemed ready to accept Pinter's plays as amusing and entertaining, not only challenging. The production was rated highly, lasting through 16 performances. In the public response it was observed that individual interpretation becomes important, which, along with the choice of scenery, acting skills and other elements also highlights the value of high quality translation. All new translations were duly noted, and their effect in each production was analysed. Moreover, as observed in the response to Pinter's plays in the following years, the reviewers started to expect performances to thoroughly express the specifics of Pinter's style. For example, in commenting on The Dumb Waiter in 2011, Rak found that the performance lacked the "magnetic field to create the 'Pinteresque' absence of sense" (2011a, 15). Reviews were more focused on the directing and sought the reasons for what could be seen as flaws in the performance within the production and not in the original text of the play. For instance, the cultural difference seemed to remain an obstacle when presenting British humour. Rak (2011a, 15) believed that it was impossible to put such a typical British play with British characteristics on a Slovene stage and integrate it into Slovene culture, despite the promising new translation. The reviewers also identified producers who favoured Pinter's plays and had picked up on Pinter's attitude when staging his plays. Rak, for example, comments on the director/producer who staged Old Times in 2011: "[Peter] Boštjančič and Pinter have one thing in common: they do not care about what other people think.
[…] Boštjančič is only interested in those 20 or 30 people who take the time in the evening to see Old Times" (Rak, 2011b, 16). Moreover, this reviewer has no problem declaring a performance successful even if it does not attract a huge audience and also claims this as a major step forward in Slovene reception of and response to Harold Pinter. Pinter's plays are not aimed at the masses, so a small, solid audience is a reliable indication of good reception.
The importance of quality directing is brought to our attention in the reviews of the production of Betrayal that was staged in 2011. The production had no official director, and that was the main reproach noted among the reviewers. Tadel stated that "being without a director caused some flaws in interpretation; although the team included only good actors" (Tadel 2011, 27). It is also likely that the reviewers commented on this, because staging a play without a director is more the exception than common practice on the Slovene theatre scene.
The production of Celebration in Kranj in 2013 received great publicity. In addition to praising Pinter, reviewers observed that the performance succeeded in leaving the audience with more questions than answers, and that it implemented other expected elements of Pinter's style. Additionally, they found the play amusing and educational at the same time, including humorous inserts. Štaudohar comments that this play was a "reflection of real life attracting the viewers in a very interesting way: by saying that by attending this performance you will find out that you are not the only one with a screwed up life" (2013). The awareness that we are still able to identify with the relationships and situations on stage, even though decades have passed since these plays were written, suggests that they are timeless.
Although Pinter's masterpieces eventually spoke for themselves and made him famous, this study found that several factors influenced the popularity and reception of Pinter and his work in Slovenia. Undoubtedly, among these factors are the Nobel Prize in Literature in 2005 and his death in 2008. Judging by the number of Pinter plays produced in Slovenia shortly after those two events, we were unable to prove their influence at first. It was anticipated that these two events would have caused an increase in producing Pinter's plays after 2005 and after 2008, nevertheless; the final results of the analysis showed no major increase in productions of Pinter's plays staged in Slovenia. However, we found that Pinter's Nobel Prize as well as his death each triggered a major media response, which significantly influenced the reception of his plays and their popularity. The reviews, for example, were more positively oriented, giving praise to Pinter and his work; they were longer and also more glowing, regardless of whether the performance of the play was good or bad. Moreover, events were organised that were solely dedicated to Pinter, for example, the project Pinter Abroad: Other Stages Other Rooms in 2011 (see Onič 2012). The vast development in technology and easier media accessibility could be considered reasons behind the larger number of articles and news previews referring to Pinter. They were longer and more frequent in many cases, since the amount of text in the electronic media is not directly connected to material cost, as in print, and is often less limiting in this respect.
In the conclusion of this subdivision, some statistical data comparing the periods before and after 1999 will be given. There were six Pinter plays performed in Slovenia before 1999 and eight after 1999. In the 31 years before 1999, there were 13 performances of Pinter's plays in Slovenia altogether. After 1999 this number increased to 14 in the last 15 years, which is a significant increase. The Pinter plays with the most productions in Slovenia before 2014 are The Dumb Waiter (5 productions) and The Birthday Party (5 productions), followed by Old Times (4 productions), Homecoming (3 productions) and Betrayal (3 productions). If we list the performances according to number, we can see that Betrayal was performed most often, followed by The Birthday Party and Homecoming. All this considered, it could be claimed that Pinter's popularity in Slovenia has increased since 1999.
Pinter in the Slovene Intercultural Context
This brief overview of Harold Pinter's plays on the Slovene stage shows that introducing this playwright to the Slovene audience was a lengthy process. If lack of insight into Pinter's significant style represented a reception obstacle to the British public and critics, it comes as no surprise that the process took longer in the culturally different and considerably closed communist/postcommunist Slovene environment. Gradually, the public began to appreciate Pinter's dialogue, with utterances full of recurrences, interruptions, hesitations, incomplete syntax, silences, pauses and other typical features of Pinter's style. 7 The reviews of productions mounted after 1999 (and particularly after the Nobel Prize Award in 2005) began to treat his plays, without exception, as masterpieces. Thus, the reviewers moved from predominantly critiquing Pinter's style to mainly explaining it in their reviews, sometimes including their own interpretation of the play; eventually they reached the stage of praising the plays and commenting on the translations and the performances. Nowadays, the reviewers are familiar with Pinter's style; they embrace the fact that the interpretation should be left to each individual and mark a performance as marked as good if it raises more questions than answers; yet, his plays are still occasionally seen as complex and hard to understand. In the reviews before 1999, it often happened that the actors and the performance received praise, but the play itself was marked as worthless for the repertory or insufficiently engaging for the audience. contrary to this, in the reviews after 1999 and particularly after 2005, the plays are always much appreciated, regardless of the performance quality.
Apart from the cultural differences that undoubtedly exist between the British and the Slovene literary and theatre spaces, the Slovene audience also faces the fact of translation. The critics as well as the academic researchers have recently identified this as one factor that often hampered the audience's perception of this new stream in drama. Particularly in the post-1999 commentaries, the translation quality is frequently recognized as crucial for the reception of the plays, while in early reviews this issue is almost never addressed. Hribar and Onič even report attempts in early translations to 'mend' the play in an explanatory way 'so that the reviewers and the audience would understand it'. Such translators, obviously, failed to realize that Pinter had already mastered and "captured all the feelings, psychological states, moods, intentions and secrets with the power of words" (2011,13). The translators must overcome the fear of possible misinterpretation, since, according to Pinter himself, this does not exist.
It is well known today that translating Pinter represents a particular challenge, since the translator must know the playwright's style well in order to be able to transfer its effects into the target context. Any unawareness in the phase of reading the original, or interventions or superficiality in the translation process may affect the interpretive potential of the text. As Onič (2005a) suggests in his study about preserving register in translation, any unawareness in the phase of reading the original, or interventions or superficiality in the translation process may affect the interpretive potential of the text. According to Meta Grosman (1997, 26), the translator can only translate the meaning s/he created in his/her own reading of the original, while all other potential meanings are inevitably lost. It would, of course, be naïve to expect that a good translation can secure a quality production by itself, but it is obviously vital. Possibly, under the influence of reviews referring to this issue as well as academic studies addressing the importance of a good translation, four new translations of Pinter's plays appeared after 1999: The Birthday Party in 2002 by Zdravko Duša, Homecoming, staged in 2008, by Darja Dominkuš andThe Dumb Waiter, staged in 2011, by Tomaž Onič; the new play, Celebration, was translated by Alja Predan. 8 Judging by the reviews of the plays with new translations, these were better, fresh and improved.
Many reviewers used to believe that Pinter's plays were hard to understand and made no sense because the original scenery was placed in Britain, and Slovenes were therefore unable to identify with the concept of a big city or British humour. 9 In the more recent performances of Pinter's plays in Slovenia, the settings are placed in the present, or possibly even an indefinite time, and are independent of place. Their ability to function in any place and any time indicates that the cultural gap has diminished, probably owing to globalization, modernization and media accessibility.
Conclusion
The reviewers seemed to be expecting more from the productions of Pinter's plays, so the experience of performance companies with staging Pinter as well as the knowledge of his style have become the norm rather than a bonus. In the reviews and with the audience, Pinter's reputation remains intact, independently of the performance quality. The latter mostly coincides with positive media response, which is evident from the fact that successful performances receive many awards, attract a lot of media attention and consequently also numerous playgoers and replays.
comparing the reception of Pinter before 1999 and after, we can claim that a notable rise in awareness regarding his works and thus his popularity exists in this century. A few new, fresh and improved translations have brought Harold Pinter's plays closer to the public. In addition, there are several Slovene equivalents for the term 'Pinteresque' in the current vocabulary, and their use has become more frequent, which is another indication that Pinter's style has been successfully integrated into the Slovene cultural awareness. The number of reviews with definitions of Pinter's 8 Entire translations are available either in book format, Duša (2006); or in the relevant Theatre Programmes: Dominkuš (2008), Predan (2003), Onič (2011a). 9 A study on humour in Pinter and its translation into Slovene was published by Onič (2003). It includes multiple examples from The Caretaker. style and suggested interpretations has decreased. Instead, suggestions for individual initiative to find personalized interpretations of the plays has emerged, with a strong awareness that a good Pinter performance should give more questions than answers. As Darja Hribar puts it, "Forcing us to find our own interpretations of his characters' behaviour and reactions, the dramatist exerts trust in our judgements and capabilities, thus allowing different interpretations of human condition -an aim which is the core of Pinter's view on complexity of life" (Hribar 2004, 206). By comparing the situation before and after 1999, we can observe that there has been a considerable increase in Pinter plays staged per year since 1999. The constant presence of Pinter's plays on the Slovene theatre scene since 1967 proves that this playwright and his style have been successfully integrated in the Slovene theatre repertory. The awareness that we still can identify with the relationships and situations on Pinter's stage, even though decades have passed since some of these plays were written, suggests that Pinter's work is timeless and can become permanently harmonized into a non-British background. | 2018-12-27T00:39:19.428Z | 2016-12-16T00:00:00.000 | {
"year": 2016,
"sha1": "00d74e6e498798b335d2366a597f642f06d11878",
"oa_license": "CCBYSA",
"oa_url": "https://revije.ff.uni-lj.si/elope/article/download/7060/6835",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "00d74e6e498798b335d2366a597f642f06d11878",
"s2fieldsofstudy": [
"Art",
"History"
],
"extfieldsofstudy": [
"Sociology"
]
} |
24478610 | pes2o/s2orc | v3-fos-license | Alpha-blockade and vasodilatation induced by nipradilol, arotinolol and labetalol in pithed rats.
In pithed rats two recently-introduced beta-blockers, nipradilol and arotinolol, as well as labetalol shifted the pressor dose-response curve for phenylephrine to the right. Labetalol and arotinolol did not modify the pressor dose-response curve for clonidine, while nipradilol induced a definite rightward shift. These results indicate that labetalol and arotinolol are selective alpha 1-blockers, while nipradilol is a non-selective one. In addition, all the three beta-blockers produced complex changes in the blood pressure in pithed rats. A fall of the diastolic blood pressure induced by labetalol and nipradilol was preceded by a slight rise, while arotinolol produced a fall at lower doses and a rise at higher ones. The hypotension by labetalol was abolished after propranolol, while the hypertension was suppressed by prazosin, indicating that labetalol has an intrinsic beta- and alpha 1-sympathomimetic effect. The hypertension and the hypotension produced by nipradilol and arotinolol persisted even in the presence of propranolol and prazosin or propranolol and yohimbine.
Numerous ,3-blockers are now being used in ischemic heart diseases, arterial hyper tension and cardiac arrhythmias. Among them labetalol is unique in that it possesses both ;3 and a-blocking activities (1 , 2). According to Blakely and Summers (3), it is a more potent blocker of a, -adrenoceptors than of a2 -receptors. in the pithed rat. As the pithed rat pre paration was a suitable model for exploring the vasodilatory action of drugs, we also examined the vasodilatatory action of these compounds.
To insure a good oxygenation of the arterial blood, oxygen gas was blown to the inspiratory tube. For drug injections, the right femoral vein was cannulated with a polyethylene tubing. The left carotid artery was cannulated with a polyethylene tubing (PE 50) filled with heparin (200 U/ml) saline solution, and the blood pressure was measured via a pressure transducer (Gould P-50) connected to a carrier amplifier (San-ei 1236). Heart rate was monitored with a card iotachometer (San-ei 2130) triggered by the arterial pressure pulse. Both parameters were recorded on a linear recorder (Watanabe Mark V). Pithing of the rats was made according to the method of Shipley and Tilden (7). After injection of atropine (1 mg/ kg, i.v.), bilateral vagus nervi were cut. A trocar (O.D.=2.5 mm, 11 cm long) was introduced through the left orbit into the spinal cord (C6) after ligation of the right carotid artery. Using the trocar as a guide, a pithing rod (O.D.=1.5 mm, 19 cm long) was inserted to destroy the spinal cord com pletely. During the experiment, body temper ature was maintained at 37°C using a thermostatically controlled heating pad. After a 30 min period of equilibration, during which the cardiovascular parameters were allowed to stabilize, experiments were performed.
2. Studies on the selectivity of the a blocking action: The pressor responses to phenylephrine and clonidine were determined from the changes in the diastolic blood pressure. Phenylephline was administered as a single injection, while clonidine was administered in a cumulative fashion. In experiments in which the effects of labetalol, nipradilol and arotinolol on the pressor effects of phenylephrine and clonidine were tested, the pressor responses to the agonists were determined 5 min after administration of the three ,3-blockers. All the experiments were performed in the presence of propranolol (1 mg/kg). The drugs were injected via the polyethylene cannula inserted into the right femoral vein without flushing. Volumes of single injection of drugs ranged 0.01-0.1 ml per 100 g.
3. Statistics: Results were expressed as the mean±S.E. The data were evaluated using Student's t-test and P-values less than 0.05 were regarded as significant.
Results
1. Studies on the vasodilatatory effects: The initial mean values of the diastolic blood pressure and the heart rate of the pithed rats were 48.8±1.2 mmHg and 265±12 beats/min, respectively (n=15). The three l3-blockers with a-blocking action produced complex changes in the blood pressure when administered to the pithed rats in a cumulative fashion. Labetalol and nipradilol induced an initial slight rise of the blood pressure followed by a fall, while arotinolol produced a fall of the diastolic blood pressure at lower doses and a rise at higher doses (Fig. 1 ). The hypotensive effect of labetalol was abolished after treatment of the preparation with propranolol (Fig. 2), while the initial hypertension was suppressed by prazosin (Fig. 3). Both the hyper and hypotensive effects of nipradilol and aro tinolol persisted even in the presence of a and F3-blockers. Figure 4 depicts the effects of treatment of the preparation with prazosin or yohimbine on the hypertension produced by arotinolol in the presence of propranolol.
2. Studies on the selectivity of the a blocking action: To evaluate the selectivity of the a-blocking effects, log dose-response curves for the pressor effect of phenylephrine (a, -agonist) and clonidine (a2-agonist) obtained in the presence of the three 13 blockers, labetalol, nipradilol and arotinolol, were compared with those obtained in the absence of the ;3-blockers (Fig. 5). All the three compounds used produced a shift to the right of the pressor dose-response curves for phenylephrine, while the shift to the right of the pressor dose-response curves for clonidine was observed only with nipradilol (Fiq. 5). Discussion a1-adrenoceptor.
Blakely and Summers (3) have also reported that the a2-blocking effect of labetalol is extremely weaker than the a,-blocking effect. Arotinolol behaved in a similar manner to labetalol, while nipradilol produced an inhibition of the pressor effects not only of phenylephrine but also of clonidine. Thus, it may be concluded that arotinolol is a (3-blocker with a,-blocking activity, while nipradilol does not have such a selective a-blocking action. pig atria and in the heart-lung preparation supported by a donor dog (9) and vasodilation in the perfused femoral artery (10) and hindlimb of dogs (11). In our present study conducted in the pithed rat, labetalol induced a hypotensive response and tachycardia, which were abolished after treatment of the preparation with propranolol. Thus, the existence of the intrinsic sympathomimetic activity was further substantiated.
The hypo tension by labetalol was always preceded by a transient phase of hypertension. This hyper tensive phase, which became manifest after propranolol, was suppressed by prazosin, indicating the participation of the at stimulation in the observed rise of the blood pressure.
Nipradilol produced a biphasic blood pressure response. Arotinolol produced a hypotension at lower doses and a hyper tension at higher doses. Both the hyper and hypotensive effects of these two compounds (arotinolol and nipradilol) persisted even in the presence of propranolol and prazosin or propranolol and yohimbine, indicating that these two substances have direct vaso constrictor and vasodilator effects of unknown mechanisms. | 2018-04-03T00:52:26.943Z | 1985-01-01T00:00:00.000 | {
"year": 1985,
"sha1": "e422cb403719bd231d8da9088536681bf9e9a15e",
"oa_license": null,
"oa_url": "https://www.jstage.jst.go.jp/article/jphs1951/39/4/39_4_481/_pdf",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "e7cd6d8763bea03896c9b8272ced91901e40b25d",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
252393263 | pes2o/s2orc | v3-fos-license | Business‑Civil Society Collaborations in South Korea: A Multi‑Stage Pattern Matching Study
In this study, we use an empirical example to demonstrate how a multi-stage pattern matching process can inform and substantiate the construction of partial least squares (PLS) models and the subsequent interpretation of and theorizing from the findings. We document the research process underlying our empirical investigations of business – civil society collaborations in South Korea. The four-step process we outline in this paper can be used to ensure the meaningfulness of the structural model as well as to maximize the use of PLS for theorizing. This methodological advancement is particularly helpful in situations when literature reference points exist, but further contextual information may add nuances to prevalent knowledge. The findings from the qualitative flexible pattern matching part of the study prompted us to conduct a multi-group analysis. The resulting path changes in the base model led to the identification of four partnering strategies for business-CSO collaborations: (1) partnering for visibility; (2) partnering for compliance; (3) partnering for responsibility outsourcing; and (4) partnering for value co-creation.
Introduction
This paper uses an empirical example to demonstrate how a multi-stage patternmatching process can inform and substantiate the construction of partial least squares (PLS) models and the subsequent interpretation of and theorizing from the findings.We document the research process underlying our empirical investigations of business -civil society collaborations in South Korea.We first outline why and how a multiple-pattern matching process can enhance the use of partial least squares structural equation modeling (PLS-SEM).We then provide a framing for our empirical study and discuss how the study contributes to the international business and international management literature.PLS-SEM as a method and SmartPLS as a software tool (Ringle et al. 2015) have rapidly gained popularity over the past decade.This is partially due to PLS-SEM's suitability for predicting and theorizing, and its ability to deal with complex models, estimate formative constructs, and handle smaller sample sizes, among other features (Richter, Cepeda, et al. 2016;Ringle et al. 2020).However, despite the method's high level of sophistication, its relative flexibility and the user-friendliness of the software have generated some unintended consequences (Zeng et al. 2021).Although PLS-SEM was originally designed to facilitate theory development through the exploration of data (Richter Sinkovics, et al. 2016;Ringle et al. 2020), researchers still need to bear in mind that an inductive approach in quantitative analysis requires sound conceptualization and operationalization and an adequate execution of data collection (Sinkovics 2018;Trochim 1989;Zeng et al. 2021;Wible and Sedgley, 1999).Therefore, it is important to focus more attention on methods and techniques that can be applied at the beginning of the research process to pave the way for generating sound PLS models and to enhance the benefits of theorizing with PLS (cf., Sinkovics 2016Sinkovics , 2018;;Sinkovics et al. 2021b).
In this study, we draw on the pattern-matching framework to demonstrate how the use of qualitative techniques can strengthen the conceptualization and theorybuilding aspects of PLS.The overall pattern-matching process can be divided into three stages: partial, flexible, and full pattern matching (see Appendix Fig. 5 for an overview).By bringing together all three stages, our study demonstrates how they inform each other (see Fig. 1), specifically, how the first two stages support the development of a PLS model and the subsequent theorizing based on the empirical findings from 215 firm responses to a survey.
Multinational enterprises (MNEs) frequently interact with sociopolitical stakeholders such as civil society organizations (CSOs) across their home as well as host countries (Sun et al. 2021).These interactions are considered an aspect of MNEs' non-market strategies, and they contribute to MNEs' competitiveness by reducing challenges associated with social, political, and institutional contexts (Mellahi et al. 2016).A growing number of international business studies have highlighted non-market strategies as an integral part of MNEs' overall international business strategy (Boddewyn and Doh 2011;Cuervo-Cazurra et al. 2014;Doh et al. 2015Doh et al. , 2017;;Kobrin 2015).Lucea and Doh (2012) propose that if MNEs are to design non-market strategies that appropriately fit their non-market context, they need to pay attention to four sociopolitical dimensions, namely, stakeholders, issues, networks, and geography.Therefore, there is a need to match what we know about these dimensions in frequently explored research settings such as the United States and Europe to knowledge generated in less frequently explored settings such as South Korea and other Asian and African geographies (cf., Doh et al. 2015).An additional factor that adds urgency to these explorations is the United Nations' (UN 2015) stance on the importance of cross-sector partnerships for the attainment of the sustainable development goals (Bäckstrand 2006).In this paper, we define business-CSO collaboration as "a system of formalized cooperation between several institutions [involving at least one firm and one CSO], based on a legally contracted or informal agreement, links within cooperative activities and jointly adopted plans" (Wyrwa 2018, p. 123).
We chose South Korea as our research context because of the highly influential role that CSOs play in the country's political and business environment.Understanding this context could help foreign MNEs in South Korea reduce institutional distance and design better non-market strategies.For instance, Kim et al. (2013) highlight a comment by a South Korean Corporate Social Responsibility (CSR) manager: "we get too much political influence on CSR.I think this is typical in Korea … so businesses are not free to do what they think they should do anymore.Businesses have to pay attention to these pressures (from CSOs)" (p.2584).As exemplified by this quotation, an important characteristic of South Korean CSOs is their active and direct participation in politics, both at an individual and group level.Many even become politicians themselves, and as a group they have been involved in the conception and running of past administrations.
Observed paƩerns in survey data
PaƩern match and theorizing based on results (path coefficients, R 2 , effect sizes, etc.) Step 1 Step 2 leads to Step 3 Step 4 informs informs
Figure. 1
The multi-stage pattern matching process underlying this study.Source: adapted from (Sinkovics 2018) Therefore, on a conceptual level, our study contributes to the international business and international management literature by furthering understanding of how the factors that drive the formation of business-CSO collaborations influence firms' collaborative behavior and ultimately the outcomes of collaboration in the South Korean context.The findings hold important implications for MNEs aiming to expand into the South Korean market given the extent to which CSOs actively shape the sociopolitical environment.Further, an understanding of how South Korean firms engage with CSOs in their home country will aid future theorizing about their collaborative behavior with CSOs in host countries.
Partial and Flexible Pattern Matching to Pave the Way for Structural Model Specification
Appendix Fig. 5 provides an overview of the three stages of pattern matching.Partial pattern matching is completed either in the theoretical realm, where the researcher works with the literature to identify initial theoretical patterns, or in the observational realm, where the researcher starts with empirical observations to identify theoretical patterns (Bouncken et al. 2021;Shah and Corley 2006;Sinkovics 2018).
Flexible pattern matching builds on partial pattern matching (Sinkovics 2018), either within the same study or in a subsequent study.Initial theoretical patterns are deduced from the literature or a previous inductive study and matched to observed patterns in empirical data.Therefore, flexible pattern matching combines a deductive and an inductive component.In other words, it seeks to identify matches and mismatches between initial expected patterns based on the literature and observed patterns that emerge from the empirical data while simultaneously allowing new patterns to emerge from the data (Bouncken et al. 2021;Sinkovics 2018).The third stage of pattern matching is full pattern matching.This aims to determine which alternative theory best explains an empirical observation.Structural equation modelling arguably represents the highest level of full pattern matching to date because it involves pattern matches at the structural level as well as the measurement level (cf., Hair et al. 2017;Sinkovics 2018).
Figure 1 demonstrates how we used the different stages of pattern matching to construct our structural (and later, measurement) model.Step 1 involved a systematic literature review (see Appendix Table 5 for the protocol).The aim was to obtain an understanding of what is known in the literature about business-CSO collaborations and what theories are commonly used to underpin investigations on this phenomenon.We conducted this literature review with the aim to derive an initial framework -that is, a collection of theoretical dimensions as a starting point for our structural model specification.In step 2, we conducted interviews and applied a flexible pattern-matching analysis technique to check the relevance of the theoretical dimensions that we had identified from the literature to our study context.This was necessary since most studies we identified were conducted in the United States and Europe, and we needed to ascertain that those insights were relevant for our South Korean context.Flexible pattern matching further allowed us to explore whether there were any theoretical dimensions that the literature had not yet uncovered but that were important in this context.Step 3 then entailed the finalizing of our structural model and the hypotheses.
Appendix Table 7 lists the main theoretical dimensions, the corresponding operationalizations, and the expected theoretical patterns that we identified from the literature review.The three main dimensions correspond to the three phases of business-CSO collaborations: (1) formation, (2) implementation, and (3) outcomes (e.g., Selsky and Parker 2005).The formation phase refers to the factors that drive organizations to collaborate with CSOs.Prior studies suggest two main drivers (e.g., Dahan et al. 2010;Weber et al. 2017): The pressure from external stakeholders to collaborate with CSOs and the desire to gain access to the resources of CSOs.These drivers are linked to two main theories in the literature -stakeholder theory and the resource-based view.
The implementation phase of business-CSO collaborations may be linked to the collaborative behavior displayed by partners to achieve their shared objectives (Heckman and Guskey 1998).Our review of the literature uncovered three main concepts: interorganizational connectivity, shared resources, and trust in CSOs' competence and good intentions (Jiang et al. 2015;Rivera-Santos and Rufín 2010;Weber et al. 2017).Interorganizational connectivity refers to "the communication and interaction mechanisms and relational structures that support the back-and-forth flow of knowledge and ideas" (Sinkovics et al. 2019, p. 132) in collaborations.The resource-sharing dimension includes the sharing of knowledge, capabilities, materials, human resources, and social capital.Lastly, the trust dimension can be broken down into "goodwill trust," that is, trust that the partner wishes to contribute to the collaboration, and "competence trust," that is, trust that the partner has the capability to contribute to the collaboration (Jiang et al. 2015;Lui and Ngo 2004).Finally, the outcomes phase of business-CSO collaborations encompasses two performance dimensions: business and social performance of the collaboration.
The fourth column in Appendix Table 7 offers sample comments to demonstrate observed patterns in the interview data (see Appendix Table 6 for an overview of firm characteristics in our sample).The last column in Appendix Table 7 provides the outcome of the flexible pattern match, that is, how the observations from the qualitative data compare against prior research results or theoretical expectations derived from the literature.The matches, mismatches, and emerging aspects provide a quality check for the meaningfulness of the structural model and form the basis for our hypotheses.Therefore, the hypotheses formulated below, in step 3 of the multi-stage pattern matching process (see Fig. 1), are partly grounded in the literature (step 1 in Fig. 1 corresponding to columns 1-3 in Appendix Table 7) and partly the result of theorizing based on the qualitative data analysis via flexible pattern matching (step 2 in Fig. 1 corresponding to columns 4-5 in Appendix Table 7).Figure 2 provides the structural model and the hypotheses.
The Relationship Between Stakeholder Pressure and Collaborative Behavior in Business-CSO Collaborations
Although our findings largely match the predictions from the literature regarding the fundamental relationships between stakeholder pressure and collaboration with CSOs (see Appendix Table 7), we found some differences in intensity stemming from the South Korean context.Specifically, whereas stakeholder pressure is expected to lead to a certain degree of connectivity between the firm and the CSO in most institutional contexts (c.f.Pagell et al. 2010), this relationship is likely to be much stronger in South Korea.This is a consequence of targeted government initiatives and the prominent and influential role of CSOs (Kim et al., 2013).Therefore, we hypothesized a positive relationship between stakeholder pressure and investment in interorganizational connectivity with the CSO.
Hypothesis 1a: A positive association exists between stakeholder pressure and investment in interorganizational connectivity
Our findings also corroborate evidence from other contexts that government-driven business-CSO collaborations may lead to resource sharing.This is because both the firm and the CSO may be required to sign written commitments to contribute resources to a given project (cf., Pratt Miles 2013).Consequently, the CSO may register a complaint if the firm does not share resources to the extent stipulated in the agreement, which in turn may lead to strict penalties (Tripsas et al. 1995).Hence, firms may be obliged to share resources with a CSO because of the government's intervention.If a business-CSO collaboration is formed owing to consumer or client pressure, the parties may form joint teams to address the underlying issue as well as to perform public communication activities.Such activities may include joint promotional events, marketing campaigns, publications, and media briefings (Shumate and O'Connor, 2010).We found evidence in our qualitative data to support these expected patterns in the South Korean context.Further, we found that the resource-sharing level is higher when the competence level of the CSO is high.
Figure. 2 Conceptual Model
Hypothesis 1b: A positive association exists between stakeholder pressure and investment in resource sharing with the CSO However, stakeholder pressure to collaborate with CSOs may lead to low levels of goodwill trust in CSOs (cf., Jiang et al. 2015).This occurs because a firm may feel too concerned about the motives of a CSO to initiate collaboration, for example, a CSO's intention to use information received through collaboration to fuel future criticism.Further, when collaboration is forced, a firm does not tend to have the opportunity to perform due diligence with respect to the CSO's ability to contribute to the project outcome (cf., Rivera-Santos and Rufín 2010).We found evidence of this in our qualitative data.Specifically, when there is government pressure to collaborate with a specific CSO, firms do not feel they are in a position to refuse the collaboration, despite having little opportunity to gauge the CSO's competence.This may lead to problems during the collaboration and reduced general trust in CSOs in future collaborations.Hypothesis 1c: A negative association exists between stakeholder pressure and the level of trust in CSO.
The Relationship Between Accessing CSO Resources as a Driver for Collaboration and Firms' Collaborative Behavior
We also found support for the general proposition that when a firm collaborates with a CSO to access its resources, it is more likely to build interorganizational connectivity.This is because building interorganizational connectivity allows partner organizations to understand the depth and breadth of each other's resources (Ferreras-Méndez et al. 2015) as well as identify who holds the required knowledge in each organization.In general, CSOs possess knowledge about local communities or population segments on the fringes of the mainstream market.CSOs are also likely to have network ties with local community leaders (Dahan et al. 2010;Sinkovics et al. 2014).Further, in developing countries, CSOs are often hired by the government to design and deliver social projects on its behalf (Barr et al. 2005), and thus CSOs may possess high levels of social capital within government organizations (Den Hond et al. 2015).This can be leveraged by a firm to obtain regulatory approval and a "social license to operate" (Wilburn and Wilburn 2011).Since tacit knowledge is difficult to codify, it is mostly transferred through personal relationships (Nonaka 1994).Therefore, firms are required to implement communication strategies that facilitate information exchange between the two groups of employees.
Hypothesis 2a: A positive association exists between firms' desire to access CSO resources and to invest in building interorganizational connectivity
We also found evidence to support that when a firm is driven to collaboration by the prospect of accessing CSO resources, it is more likely to implement strategies and routines that foster resource exchange.For instance, joint integrated teams allow partner organizations to become closer and access each other's tacit knowledge (Lam 1997).Further, the presence of joint teams on public platforms and at public events signals a high degree of integration between CSO and firm, resulting in a better reputation.Our findings also indicate that co-location, such as shared office space, can further facilitate this process.Hypothesis 2b: A positive association exists between firms' desire to access CSO resources and firms' desire to share their own resources in return When a firm initiates a collaboration with a CSO to access the CSO's resources, it is assumed that the firm is already aware of the potential of the CSO's resources and is anticipating synergies between the two organizations' resources.As opposed to situations in which the collaboration is stakeholder driven, a firm is more likely to undertake due diligence when the collaboration is motivated by a desire to access CSO resources.As a consequence, the firm will have a better understanding of the CSO's level of expertise and capacity to achieve the overall objectives of the proposed collaboration (Rondinelli and London 2003;Seitanidi and Crane 2009).Therefore, when a collaboration is formed with a CSO to access the CSO's resources without any pressure from external stakeholders, the role of trust in the CSO is likely to be more important for the collaboration (cf., Jiang et al. 2015) than when the collaboration is stakeholder driven.Hypothesis 2c: A positive association exists between firms' desire to access CSO resources and firms' level of trust in the CSO.
The Relationship Between Collaborative Behavior and the Outcomes of Business-CSO Collaborations
In business-CSO collaborations, interorganizational connectivity is likely to play a bigger role than in the usual equity alliances.This is because of differences between the organizations in terms of culture, values, routines, performance measurement, leadership style, decision-making processes, and goals (Quélin et al. 2017).The findings from our interviews suggest that frequent and welldesigned communication helps the collaborating partners to understand existing differences and fosters conflict resolution.Further, interorganizational connectivity also fosters a better understanding of the underlying issues that the project aims to address.Several respondents highlighted the importance of two-way communication and knowledge exchange to co-create solutions that were not only related to the company's core business but also had significant societal implications.Examples include product design for the visually impaired or the design of environmentally friendly products.Therefore, interorganizational connectivity is expected to lead to positive business as well as social performance outcomes.
Hypothesis 3a: A positive association exists between investment in interorganizational connectivity and business outcomes in business-CSO collaborations.Hypothesis 3b: A positive association exists between investment in interorganizational connectivity and social outcomes in business-CSO collaborations.
In the context of equity alliances, scholars have empirically confirmed a positive association between resource sharing and collaboration outcomes (Weber et al. 2017).Similarly, resource sharing in business-CSO collaborations is expected to lead to enhanced collaboration outcomes.CSOs often lack the financial resources required for their projects (Hale and Mauzerall 2004).Therefore, sharing financial and other resources with CSOs can be expected to lead to better social outcomes.Our interview data indicate that resource sharing may come in different shapes and sizes, including applying for government funds and co-designing products for an underserved population segment.Hypothesis 4a: A positive association exists between a firm's resource sharing with its CSO partner and business outcomes in the business-CSO collaboration.Hypothesis 4b: A positive association exists between a firm's resource sharing with its CSO partner and social outcomes in the business-CSO collaboration.
Based on the findings from the literature review, it is expected that a high level of trust in the CSO will lead to cost reductions in terms of legal, monitoring, and negotiation costs (Lui and Ngo 2004;Zaheer et al. 1998).In business-CSO collaborations, differences in objectives (Mars and Lounsbury 2009) can lead to delays in agreeing on the overall objectives of the collaboration and drafting its terms and conditions.Further, a high level of trust between partners is expected to facilitate discussion of social and business issues, including those external to the project.Moreover, when a firm's trust in its CSO partner is high, it is more likely to draw on the partner in the product development process.Research and development-related information and product pipelines involve sensitive information that firms tend to protect (Dahan et al. 2010).While our interview data support the link between trust and the business outcomes of the collaboration, our findings suggest that trust in the CSO partner may be less relevant regarding social outcomes.Nevertheless, we hypothesize a positive relationship between trust and social outcomes to further test the theory in the second part of the study.
CSO Dominance and Firms' Level of Standard Adoption
Two dimensions emerged during the flexible pattern matching process that seem to have an influence on how the relationships between our theoretical concepts play out.These are the level of dominance of a CSO within the industry and the level of standard adoption by the firm.The dominance of a CSO within the industry may be seen as a proxy for its experience and reputation.For example, one of our interviewees stated, "If a CSO is not dominant in a particular sector or domain [environmental or labor issue], it implies that they do not bring sufficient experience and 1 3 knowledge to a particular collaborative project.When we collaborated with CSOs that were less influential in the past, we faced a lot of difficulties…".
Conversely, the level of standard adoption by the partnering firm may be regarded as a proxy for the firm's internal resources and capabilities that enables the firm to learn about and tackle social and environmental issues.The implementation of standards or other voluntary frameworks generally requires firms to develop processes, routines, and capabilities to learn about and address aspects of the targeted social and environmental issues.This in turn is expected to increase the level of value co-creation in the collaboration.The following comment from one of our interviews represents a case where there is no value co-creation: "We are not able to check every single step taken by the CSO in our project because of our resource constraints.As a small firm, we don't have the time or human resources for that.We also don't have much knowledge about what is required to fulfil these social and environmental standards.So, we just have faith in our CSO partner.We trust in what they are saying and what they are doing … and that the way they are carrying out the project will contribute to the community."Examples of standards include the ISO 14000 suite covering environmental management and ISO 26000 Social Responsibility.
These two theoretical dimensions that emerged from the interview data can be used to create four scenarios: (1) Low standard adoption/Low CSO dominance, (2) High standard adoption/Low CSO dominance, (3) Low standard adoption/High CSO dominance, and (4) High standard adoption/High CSO dominance.Based on the qualitative data, we expected that the paths from the original model would change under these different conditions.However, we did not have sufficient data points to formulate hypotheses pertaining to how exactly the paths might be expected to differ across the four scenarios.Therefore, we drew on PLS-SEM's suitability for exploration and conducted a multi-group analysis to examine whether -and, if yes, how -the paths changed across the four scenarios.
Full Pattern Matching with PLS-SEM
Step 4 in our multi-stage pattern matching approach encompasses the measurement model specification, model estimation, and results evaluation stages of the PLS-SEM analysis process (Ringle et al. 2020).We collected survey data between November 2018 and February 2019.We drew on the databases of two intermediary organisations -the CSR Forum and the CSR Academy in Seoul -to identify potential respondents (in total, 530 companies).The survey was sent to CSR managers by email.We also provided the option to complete the survey over the phone or offline (cf., Dillman et al. 2014).The survey was originally designed in English and then translated into Korean, which was validated by an accredited language expert.
We received a total of 224 responses (i.e., 42% response rate), out of which 215 were complete and valid.Table 1 provides an overview of the descriptive statistics.This distribution was in line with data from the National Statistical Office (2015), which shows that small and medium-sized enterprises in Korea account for 99.9% of the total number of Korean firms.More than half of the respondents (60.5%) considered their firms' collaborations with CSOs to be either extremely important (31.2%) or very important (29.3%)(see Table 2).
Measures
All variables were measured using seven-point Likert scales, with 1 corresponding to strong disagreement and 7 to strong agreement.Stakeholder pressure was measured based on items involving a range of stakeholders, including (1) government, (2) supply chain, (3) CSOs, and (4) industry/trade associations (Van Huijstee and Glasbergen 2010).In most Asian countries, including South Korea, owing to their high power-distant culture, if a firm is "asked" or "invited" to collaborate by an organization with a presence in the regulatory, social, or business environment (most CSOs, supply chain collaborators, and government organizations fall into this category), this usually implies pressure.Access to CSO resources was measured by using items adapted from Dahan et al. (2010), comprising CSOs' (1) social capital, such as networks and contacts, and (2) knowledge resources.The propensity of firms to build interorganizational connectivity was measured using items adapted from Jamali et al. (2011).Interorganizational connectivity was categorized into (1) project-specific, and (2) relation-specific interorganizational connectivity.The propensity of firms to share resources was measured using items adapted from Jiang et al. (2015), including (1) human resources, (2) knowledge resources, and (3) social capital in collaborations.The extent of trust in CSOs was adapted from Lui and Ngo (2004).
The measurement items captured both (1) goodwill trust and (2) competence trust.Our dependent variables were the outcomes of business-CSO collaborations.The measures of social performance were adapted from Hansen and Spitzeck (2011), whereas the measures of business performance were adapted from Steckel and Simons (1992).We also controlled for firm size and past alliance experience.Firm size was measured in terms of the number of employees.Two dummy variables were created -"0" (or small-sized firms) if the number of employees was less than 500 and "1" (or large-sized firms) if the number of employees was ≥ 500.Similarly, "0" was created if firms had no past alliance experiences with CSOs and "1" if they had past alliance experiences with CSOs.Since our control variables were categorical, the effect of each dummy variable was analyzed via a bootstrapping procedure with a resample of 4,999 (Henseler et al. 2016).Results indicated that control variables did not exhibit any significant effect on dependent variables.
Common Method Bias
We employed the marker variable technique to address the potential issue of common method bias (Rönkkö and Ylitalo 2011).We chose the level of information technology to use as a marker variable because it was not theoretically correlated with any constructs in our research model.The mean correlation coefficient value for the marker item was 0.035, indicating an insignificant influence of common method bias.We further included this marker as a control variable (i.e., additional exogenous variable predicting each endogenous construct) in our PLS model (Rönkkö and Ylitalo 2011).We compared the results of the marker model with those of our baseline model.Since we noted minimal changes to the estimates for the path coefficients had occurred and all significant effects remained significant, we concluded that common method bias was not a concern in this research.
Measurement Model Assessment
First, we assessed item reliability using Cronbach's alpha and composite reliability (CR) for each construct.CR is considered a more suitable measure of reliability for the PLS-SEM method (Hair et al. 2018).All constructs had CR and alpha values above 0.7, confirming a high level of internal consistency reliability (see Appendix Table 8).Second, discriminant validity was assessed based on the cross-loading criterion suggested by Fornell and Larcker (1981).Third, the standardized root mean square residual (SRMR) is the only estimated model-fit criterion in PLS path modelling (Henseler et al. 2016;Hu and Bentler 1998).We found an SRMR value of 0.055, which indicated a good model fit for PLS-SEM analysis (Henseler et al. 2016).Overall, it may be concluded that we had a reliable and valid measurement model (Table 3).Last, we evaluated endogeneity according to the systematic procedure proposed by Hult et al. (2018).The Gaussian copula approach (Park and Gupta 2012) to endogeneity testing was inapplicable owing to criteria violations; thus, we pursued endogeneity using the control variable approach in PLS-SEM (Hult et al. 2018).We applied two control variables, firm size and past alliance experience, in the model.According to Arndt and Sternberg (2000), firm size is an important factor influencing corporate behavior such as collaboration.Since small firms tend to lack resources, they are likely to engage in collaboration.Path coefficients between the control variables and the endogenous variable are explained in Appendix Table 9.The findings confirm that the control variables had no significant effect on the dependent variables.Therefore, we can conclude that this research had no endogeneity issues, and the findings confirm that the PLS-SEM model was robust.
Confirmatory Tetrad Analysis
We applied confirmatory tetrad analysis (CTA-PLS), widely considered an appropriate approach, to determine whether the latent construct was reflective or formative (Hair et al. 2017).This statistical measure is based on an assessment of construct indicators.The latent construct is considered reflective when all the tetrad values are non-significant (Hair et al. 2017).Appendix Table 10 provides the CTA-PLS results, indicating that none of the tetrads displayed a statistically significant difference from 0, which confirms the reflective nature of the constructs (Gudergan et al. 2008).
Assessment of the Structural Model
In accordance with Hair et al. (2013), the structural model was evaluated with R 2 , corresponding t-values, effect sizes (f 2 ) and predictive relevance (Q 2 ).First, we assessed the effect sizes, which signify the strength of relationship among variables (Appendix Table 11).Second, we performed a blindfolding procedure to evaluate the predictive relevance of the path model (Hair et al., 2018).We evaluated the predictive relevance of the latent constructs by adopting cross-validated redundancy Q 2 and cross-validated communality Q 2 (Fornell and Cha 1994).The Q 2 values of all latent constructs were greater than zero, indicating the predictive relevance of the model (Appendix Table 11).
Appendix Table 12 and Fig. 3 provide an overview of the outcomes of hypothesis testing.With the exception of H1c indicating a positive association between stakeholder pressure and trust in CSOs, and H5b indicating a positive association between trust in CSOs and the social performance of the collaboration, we found support for all hypotheses in the main model.We discuss the results in more detail in the Discussion and Conclusions section.
Multi-Group Analysis
Appendix Table 13 provides an overview of the measurement invariance testing.We used a MICOM (measurement invariance of composite models) procedure involving a three-step approach: (1) configural invariance, (2) compositional invariance, and (3) equality of composite mean values and variances (Henseler et al. 2016).After completing the MICOM procedure, we conducted multi-group analysis across the four scenarios that emerged from the flexible pattern-matching aspect of the study based on high/low combinations of the focal firms' standard adoption and high/low combinations of the CSOs' dominance within the industry.Splitting the data along these dimensions resulted in sub-groups of 36, 38, 43 and 35 observations.All sub-groups exceeded the acceptable minimum sample size of 34 (Hair et al. 2018).The level of adoption of standards and CSO dominance were measured on a seven-point Likert scale (1 = Not at all, 7 = Extremely), and a cut-off value of ≥ 4 (median) was used to determine groups.
Figure 4 provides a graphical overview of the findings, while Appendix Table 14 presents the standardized coefficients and t-values.Table 4 summarizes R 2 values and the effect sizes for the full model as well as for the four models for each scenario.
The findings tell a compelling story demonstrating the power of SmartPLS for theorizing and pattern matching at the highest level (cf., Liu et al. 2020;Sinkovics, 2018;Sinkovics et al., 2021b).We discuss the findings and their implications in the next section.
Discussion and Conclusions
In this study, we employed a multi-stage pattern-matching process to identify from the literature the main dimensions that shape the outcomes of business-CSO collaborations and examine to what extent and how they apply in our specific study context, South Korea.The underpinning rationale for combining qualitative and quantitative methods in this study was to support the process of theorizing in stages, and specifically to ensure the meaningfulness of the structural model as well as to maximize the use of PLS for theorizing.This methodological advancement is particularly helpful in situations for which literature reference points exist but further contextual information may be required to add nuances to prevalent knowledge.The findings from the qualitative flexible pattern-matching stage of the study revealed that although the general theoretical concepts and the relationships between these concepts may largely hold in the South Korean context, they may mask a more complex and nuanced story.This can be largely explained by the unique history and strong role of civil society in South Korea resulting in a high level of trust of citizens in CSOs (cf., Bae and Kim 2013; Willis 2020).Additionally, two dimensions emerged from the qualitative interviews that prompted us to explore them further with a multi-group analysis in SmartPLS.Without the flexible pattern-matching aspect of the study, we would not have identified these dimensions and the study would have lost some of its richness.The level of CSO dominance within the industry and the level of standard adoption of the collaborating firm provided us with four scenarios that in turn led to the identification of four partnering strategies in our data.In the remainder of this section, we describe and discuss the findings.We conclude by outlining the implications of the findings for MNEs wishing to enter the South Korean market and providing some suggestions for future research.
Although we hypothesized a negative association between stakeholder pressure and firms' trust in CSOs, this relationship was consistently insignificant in the main model, as well as across the four scenarios.The other consistent result across all models (main model and models 1-4) was the non-significant relationship between trust in CSOs and the social outcomes of collaboration.Additionally, we found deviations from the main model in scenarios 1 and 2 (see Fig. 4).In scenario 1 (Low standard adoption/Low CSO dominance), only five paths were significant.The focal firm's desire to access CSO resources had a positive association with interorganizational connectivity and resource sharing.Interorganizational connectivity in this scenario only had a significant impact on the social outcome of collaboration, whereas resource sharing had a significant impact on both business and social outcomes.In scenario 2 (High standard adoption/Low CSO dominance), neither stakeholder pressure nor firms' desire to access CSO resources had a significant impact on trust in CSOs.Additionally, trust in CSOs in this scenario only had an impact on the business outcome of the collaboration.The effect sizes provided in Table 4 helped us theorize about the implications of these results.
In general, (see full model and models 2-3 in Table 4), stakeholder pressure to collaborate was less effective in driving firms' collaborative behavior in a business-CSO partnership than firms' motivation to access CSO resources.Further, in the presence of stakeholder pressure to collaborate, the importance of trust in CSOs seemed to be crowded out by the importance of satisfying the stakeholder that exercised the pressure.This was especially relevant in scenarios 2 and 4 where the focal firms had a high level of standard adoption and, by extension, a higher visibility to stakeholders.In these two scenarios, stakeholder pressure had a medium effect on interorganizational connectivity and a small effect on resource sharing.Investment in interorganizational connectivity would be necessary to demonstrate sufficiently to stakeholders that they fulfil their responsibilities in the partnership and thus comply with societal expectations, which would be required to obtain or retain their social license to operate (cf., Prno and Slocombe 2012;Wilburn and Wilburn 2011).
However, in scenario 3, the partnering CSO had low dominance in the industry, whereas in scenario 4 partner CSOs had high dominance.The dominance of CSOs seemed to influence the perceived trust of the focal firm in their CSO partner's competence and goodwill and in turn the extent to which there was an attempt to co-create value.In scenario 4, trust in the CSO partner had a strong effect on the business outcome of the collaboration, whereas in scenario 2 it only had a medium effect.Further, in scenario 4, firms' motivation to access CSO resources seemed to be more substantial than in scenario 2. This is evidenced by the significant path between the desire to access CSO resources and the level of trust in CSOs, as well as the medium effect of firms' motivation to access CSO resources on their propensity to share their resources with the CSO.In scenario 2, this effect was small.This implies that focal firms in scenario 2 used their partnerships with CSOs strategically, yet not for co-creating solutions.Based on these insights, we can label the partnering strategy in scenario 4 "partnering for co-creation" and the partnering strategy in scenario 2 "partnering for compliance." In scenarios 1 and 3, firms had a low level of standard adoption.However, in scenario 1, firms partnered with non-dominant CSOs, whereas in scenario 3, firms partnered with dominant CSOs.Again, the role of trust in a business-CSO collaboration appeared to play a more important role when partnering with dominant CSOs.In scenario 3 (Low standard adoption/High CSO dominance), firms' trust in their CSO partners had a strong effect on the business outcomes of the collaboration.Interestingly, only interorganizational connectivity had a noteworthy, even if small, effect on the social outcome of the collaboration, despite the fact that the path from resource sharing to the social outcome was significant.Firms in this scenario had little knowledge and experience with social and environmental issues.Therefore, they relied heavily on the CSO partner to implement the project.The strong effect of trust on business outcomes combined with the small effect of interorganizational connectivity and resource sharing on both business and social outcomes supports this proposition.This raises a question about how much learning takes place in the partnering firm, because the strategy seems to be to outsource the social or environmental project.Therefore, we label the partnering strategy in scenario 3 "partnering for responsibility outsourcing." Lastly, in scenario 1 (low standard adoption/low CSO dominance) stakeholder pressure was not relevant for any of the collaborative behaviors.This indicates that firms in this scenario tended to fly under the radar of stakeholders.They likely collaborated with CSOs to gain more visibility and capture the attention of stakeholders.This interpretation is underpinned by the lack of importance of trust in their CSO partners.Further, interorganizational connectivity only had a significant impact on social performance, and not on business performance.Hence, while these firms wished to understand what CSOs were doing, they did not interfere.They most likely needed this information for communication and public relation purposes.Therefore, we label the partnering strategy in scenario 1 "partnering for visibility." These findings reinforce results from other contexts -that stakeholder pressure on its own is insufficient to achieve meaningful and lasting outcomes (cf., Sinkovics et al. 2016).Further, even in an environment where CSOs abound and the government has initiatives in place to facilitate interaction between firms and CSOs, the right match is difficult to achieve.Our four scenarios indicate the existence of four partnering strategies that firms adopt, depending on the degree of competence match between firms and their CSO partners.MNEs wishing to enter South Korea are advised to seek collaboration with CSOs to enhance their legitimacy.
However, the differences across the four scenarios imply that MNEs may need to adjust their partnering strategy depending on the extent of their existing knowledge relevant to the project they seek to establish.Partnering with dominant CSOs seems to produce the highest business and social benefits when both parties have complementary resources and the project is both societally relevant and connected to the firm's core business.Therefore, MNEs are advised to search for such complementarities to maximize the benefits from the collaboration.However, if MNEs are not able to secure collaboration with dominant CSOs, they are advised to undertake due diligence because less dominant CSOs may struggle with high staff turnover, which can cause disruptions in a project, or they may not possess the right combination of capabilities needed for value co-creation.Therefore, if MNEs cannot secure collaboration with dominant CSOs despite having complementary resources, it may be more beneficial to adopt a "partnering for visibility strategy," focusing on a project that is in the competence domain of a less dominant CSO and outside the immediate competence domain of the firm.This way, the MNE can gain visibility vis-à-vis stakeholders but is not tempted to take too much control of the project.The MNE may thus focus on building interorganizational connectivity to learn about the issue and utilize the project duration to gauge whether future collaborations and mutual skill upgrading is feasible.
Future research will need to uncover additional factors that contribute to enhancing the social outcomes of business-CSO collaborations.Although interorganizational connectivity and resource sharing seem to have a significant impact on social outcomes, the R 2 values derived in this research indicate additional important dimensions that were not part of this study.Future research will also need to control in more detail for the breadth and the depth of collaboration projects (Sinkovics et al. 2021c(Sinkovics et al. , 2021d) ) and examine the micro-foundations of progressing toward value co-creation at the highest level where the project is related to the core business of the firm and simultaneously has a meaningful societal impact (Sinkovics et al. 2015(Sinkovics et al. , 2021a)).Further, future research is needed to explore details of the dark side of stakeholder pressure (cf., Sinkovics et al. 2016).Government intervention and policy are needed to safeguard against the harmful opportunism of the private sector as well as to guide and incentivize desired behavior (Hamilton 2022;Hofstetter et al. 2021;Sinkovics et al. 2021d).However, interventions are not without unintended consequences, and we need to learn more about how to create safeguards to recognize and remedy such consequences early in the implementation process.
Appendix
See Appendix Fig. 5 See Appendix Tables 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 TheoreƟcal Realm TheoreƟcal Realm TS = ("business-CSO partnership" OR "business-CSO alliance" OR "business-CSO collaboration" OR "firm-CSO partnership" OR "firm-CSO alliance" OR "firm-CSO collaboration" OR "corporate-CSO partnership" OR "corporate-CSO alliance" OR "corporate-CSO collaboration" OR "business-NGO partnership" OR "business-NGO alliance" OR "business-NGO collaboration" OR "firm-NGO partnership" OR "firm-NGO alliance" OR "firm-NGO collaboration" OR "corporate-NGO partnership" OR "corporate-NGO alliance" OR "corporate-NGO collaboration" OR "business-community partnership" OR "business-community alliance" OR "businesscommunity collaboration" OR "firm-community partnership" OR "firm-community alliance" OR "firm-community collaboration" OR "corporate-community partnership" OR "corporate-community alliance" OR "corporate-community collaboration" OR "business-civil society partnership" OR "business-civil society alliance" OR "business-civil society collaboration" OR "firm-civil society partnership" OR "firm-civil society alliance" OR "firm-civil society collaboration" OR "corporatecivil society partnership" OR "corporate-civil society alliance" OR "corporate-civil society collaboration" OR "business-nonprofit partnership" OR "business-nonprofit alliance" OR "business-nonprofit collaboration" OR "firm-nonprofit partnership" OR "firm-nonprofit alliance" OR "firm-nonprofit collaboration" OR "corporatenonprofit partnership" OR "corporate-nonprofit alliance" OR "corporate-nonprofit collaboration" OR "multi-stakeholder partnership" OR "multi-stakeholder alliance" OR "multi-stakeholder collaboration" OR "multi-stakeholder initiative" OR "multistakeholder intervention" OR "nonmarket strategy" OR "nontraditional partnership" OR "non-traditional alliance" OR "non-traditional collaboration" OR "non-traditional partnership" OR "non-traditional alliance" OR "non-traditional collaboration" OR "public-private partnership" OR "PPP" OR "multi-stakeholder approach" OR "cross sector partnership" OR "cross sector collaboration" OR "cross sector alliance" OR "cross sector initiative" OR "cross-sector collaboration" OR "cross-sector partnership" OR "cross-sector alliance" OR "cross-sector initiative" OR "inter-sectoral partnership" OR "inter-sectoral alliance" OR "inter-sectoral collaboration" OR "inter-sectoral initiative" OR multi-sector partnership" OR "multi-sector alliance" OR "multi-sector collaboration" OR "multi-sector initiative" OR "green partnership" OR "green alliance" OR "green collaboration" OR "environmental partnership " OR "environmental alliance" OR "environmental collaboration" OR "sustainability partnership" OR "sustainability alliance" OR "sustainability collaboration" OR "social partnership" OR "social alliance" OR "social collaboration" OR "non-equity partnership" OR "non-equity alliance" OR "non-equity collaboration" OR "third sector partnership" OR "third sector alliance" OR "third sector collaboration" OR "third-sector partnership" OR "third-sector alliance" OR "third-sector collaboration" OR "base of the pyramid partnership" OR "base of the pyramid alliance" OR "base of the pyramid collaboration" OR "BOP partnership" OR "BOP alliance OR "BOP collaboration")
2010)
A few years ago, we developed a collaboration with an NGO as part of a government project.We were asked by the government to collaborate with the NGO so we had to do it without actually checking its background, experience and capability.We invested a good amount of financial resources, but we didn't know where they spent it or how much actual progress we made on that project.They didn't update us about the progress and at the end of the collaboration they submitted the receipts for the costs occurred, but it was really difficult to believe them.Had we collaborated without any pressure from the government, we would have done a thorough background check on the NGO before collaborating Some NGOs put pressure on us to support their campaigns or activities, which may or may not suit our company's missions and aims.For example, last month, there was a vegan festival which was held in Seoul and we received thousands of calls for collaboration from CSOs working for animal welfare Last year, we participated in an event called 'CSO matching day', which was organized by governmental organizations in Seoul.We found one CSO that matches our business values and developed a collaboration.As we collaborated through this governmental organization's event, there were quite strict requirements regarding our collaboration.We had to regularly update them with information about the progress of the collaborative project.We were also asked to make a joint presentation with the CSO in front of government officials in the final stage.During the presentation, we had to report how much we spend on collaborative projects and [give] details of our collaborative performance as well Each year, the government gives an award to companies for ethical business operation and if the company receives it, their reputation among consumers can be increased." When a firm collaborates with a CSO due to stakeholder pressure: (1) There is a high level of connectivity among partners when firms are held accountable for the outcomes of the project ( ity overnight.This could really improve our reputation.Also, reporters from media houses are interested in our social activities with CSOs and they write articles about it even if we don't ask them to do so.So, getting pictured together with the CSOs is very important for our reputation CSOs know exactly where people in need are and with their help, we can organize CSR programs efficiently.They have expertise in social issues.In order to get access to their resources, we try to establish regular interaction methods with the CSO.For instance, we developed the Senior Parcel Delivery, which creates new job opportunities for senior citizens.This is a novel Sustainable Development Goal model that addresses the issue of job creation and poverty.Senior Parcel Delivery has expanded to incorporate 1,400 seniors around 170 bases nationwide so far.In addition to senior citizens, other socially underprivileged classes such as low-income families are also actively joining the program.For this program, we made a huge effort to build relationships with CSOs, because CSOs had the contact lists of senior citizens and people from underprivileged classes that we needed to complete the project.We had regular meetings with the partner CSO and organized corporate training for the CSO employees as well When a firm forms a collaboration with a CSO to access the CSO's resources: (1) There is frequent or continuous two-way communication to absorb the CSO's expertise and knowledge (2) There is a high level of resource sharing during collaboration to utilize CSO's resources effectively (3) Partnering firms whose core business is connected to the social/environmental project try to collaborate with CSOs that have a high level of dominance/good reputation in the industry.These firms, if they have the resources, invest in thoroughly vetting the capabilities of the CSO (4) The dominance of a CSO in an industry is less important in terms of inducing trust in a CSO in a collaboration when the partnering firm has a high level of expertise already or when the collaboration is not related to the firm's core business (i.e.volunteering, altruistic giving, CSR projects etc.) (2015). 1 = "strongly disagree", 7 = "strongly agree" (alpha = 0.864, CR = 0.898, AVE = 0.598) (1) You made active contributions to the project in terms of investing your time Business outcomes -adapted from Steckel and Simons (1992). 1 = "strongly disagree", 7 = "strongly agree" (alpha = 0.825, CR = 0.873, AVE = 0.535) (1)
Figure. 3
Figure. 3 Result of hypothesis test
Figure. 4
Figure. 4 Results of the multi-group analysis articles) Selection of articles that are published in ABS ranked journals only (economics or international relations or management or geography or public administration or environmental sciences or social sciences interdisciplinary or business finance or water resources or urban studies or environmental studies or business or agriculture multidisciplinary or multidisciplinary sciences or green sustainable science technology or planning development or political science or area studies or sociology) Relevance screening (198 articles) Comprehensive screening of relevant articles by reading title, abstract and introduction '5' = Relevant; '4' = Of some relevance; '3' = Relevance not clear a priori; '2' = Less relevant or unclear nature of research work; '1 You made active contributions to the project in terms of investing your You made active contributions to the project in terms of sharing your You made active contributions to the project in terms of sharing your human You made active contributions to the project in terms of offering the CSO access to your networks and con-
Table 1
Descriptive analysis of sample firms
Table 2
General characteristics of collaborative projects
Table 3
Discriminant validity (Heterotrait-Monotrait ratio) Propensity of firms to build inter-organizational trust 0.170 0.790 0.114 0.307 0.459 0.167 .35 are considered small, medium and large at the structural level.We indicate this in normal font (small), italics (medium) and bold (large)
Table 5
Systematic literature analysis protocol
Table 7
Flexible pattern matching between expected and observed patterns
Table 7 (
We use conflict minerals in order to make electronic products.Buyers such as Walmart or Bestbuy ask us to report where conflict minerals come from and how they were collected.Wecollect tin from Indonesia and collaborate with local CSOs to supervise how mine is operated and what can be the problem in this particular area.It is difficult to know what kinds of standard they have in Indonesia and we want to minimize any problems we might face.Therefore, we try to collaborate with local CSOs that have specialized knowledge and credibility in this particular sector.We try to increase our engagement with them so that we can learn from them To identify partner CSOs for collaborations, we compare various CSOs first and try to look into what kinds of expertise they have and how dominant they are in the particular sector and if their expertise is aligned with the aim of our social projects "If a CSO is not dominant in a particular sector or domain [environmental or labor issue], it implies that they do not bring sufficient experience and knowledge to a particular collaborative project.When we collaborated with CSOs that were less influential in the past, we faced a lot of difficulties."When we choose NGO collaborators, we are looking to engage in open innovation.We examine the collaborator's capability to transform an idea into a marketable product, their communication capabilities, and how well they can collaborate with others and what kinds of social or environmental value they can create in society.So, we examine if they have a clear vision for creation of social value in the community and how they try to solve the social issues and if it can be realized, how much profit we can make from the collaborations and if they have built good networks within the industry, we examine them from various perspectives Our interactions and collaborations with CSOs can increase the visibility of our social commitments in the public.These days, most people use social media to share information fast, our company's interactions with the CSO are revealed easily.For example, if someone captures pictures or videos of our employees (wearing t-shirts with our company's logo) planting trees with NGO's employees or removing plastics together from the beaches and shares it on Facebook, Instagram or Twitter, we might gain high popular-
Table 7 (
We have collaborated with local NGOs to help people with visual impairments.We communicate very often with these NGOs by using all the channels we have.Bycommunicating with NGOs frequently, we have come to realize the needs of the final consumers i.e. what sort of functions they would need from our product.These NGOs work regularly for people with disabilities and hospitals and thus hold great knowledge about the requirements of people with visual impairments.The frequent and extensive connections with NGOs helped us develop an understanding of the requirements, and finally to develop our product.Following this collaboration, we have developed a new mobile phone handset for people with visual impairments.And this has been a success.The collaborations with the NGOs have helped us identify new market opportunity and develop our products accordingly and ultimately increased sales For this collaboration [vegan festival], we had to meet the NGOs quite often to identify what kind of expectation they had from us.By building connections and through extensive talks and discussions with them, we came to understand exactly what they wanted We usually focus on environmental CSR programs and have been planting trees together with local NGOs and organizing events to spread awareness to protect our environment.For this project, our partner NGO had previous experience and knowledge.Hence, we developed a joint team and it helped employees to get closer to the NGO's employees and access their expertise efficiently.And also, if media reports tomorrow show that our company and an NGO planted hundreds of trees together, this would be well received by our customers and the regulatory authorities As CSOs have good connections in the local community, we share human resources and work together with CSOs for social projects.For example, we organize events every month, where children with disabilities in the community have a chance to learn how to read and write.For those events, our employees and employees from CSOs gather, discuss social issues and work together to resolve them.As we know that CSO has expertise in completing social projects, we tend to share our human capital and engage in more interactions with them.Because this will allow our employees to better understand how to solve social issues and will be beneficial for our future CSR projects
Table 9
Path coefficients between control variables and endogenous variables | 2022-09-21T15:16:30.092Z | 2022-08-01T00:00:00.000 | {
"year": 2022,
"sha1": "25005231e4c52e3591bebd1f0ee8b77e49b18991",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s11575-022-00476-z.pdf",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "f0e15ee7011793567e746e721a92114b50e9dbe1",
"s2fieldsofstudy": [
"Business",
"Political Science"
],
"extfieldsofstudy": []
} |
251262953 | pes2o/s2orc | v3-fos-license | A Study of English Translation Theory Based on Multivariate Statistical Analysis of Random Matrix
With the continuous advancement of articial intelligence in natural language processing technology, machine translation based on machine learning technology has been fully transformed from traditional machine translation methods to neural network machine translation methods. In particular, the tremendous development of large families has made data-driven a reality. With deep learning as the research and design background, the neural network structure of random matrix multivariate statistical analysis is designed according to the language characteristics of English. e model was tested on a Chinese-English panning model, and the optimal model fused with a Bleu value of 39.53.emodel results were applied to a real system to achieve language detection, multidirectional language translation, and manual correction of results to be able to learn long dependencies and overcome the limitations of recurrent neural networks to translate long sentences more uently.
Introduction
e basic model of machine translation system is a natural language processing system, and its basic principle is the principle of element synthesis [1][2][3]. e process of machine translation can be simply understood as three stages: rst, the source language text is decomposed into basic constituents such as words, phrases, and grammatical structures, and the original text is analyzed. Morphological analysis, syntactic analysis, and semantic analysis are used to form the internal representation of the source language, and then translation using the target language eventually produces the internal representation of the source language by transforming and ordering at the structural level using the compound rules of the target language [4]. erefore, to improve the accuracy of machine translation, the processing of natural language is taken as the core problem of machine translation system research. e rapid development of arti cial intelligence technology in natural language processing technology and the rapid development of the Internet, has led to a boom in machine translation research, which has become a hot topic not only in the eld of scienti c and technological research, but also in the keen research of linguists all over the world [5,6].
Although the statistical machine translation system is also developing and improving, its statistical machine translation which ignores the grammar rules is highly dependent on the large-scale corpus and relies on statistical data for disambiguation and translation selection, while the translation e ect ultimately depends on the probabilistic model and the coverage capacity of the corpus, [7] so the probabilistic matching results of words and phrases do not achieve the accuracy of the nal translated sentences and the translation e ect is poor. In order to improve the translation e ect, neural network technology was introduced into machine translation [8]. A neural network is an arti cial system for intelligent information processing developed by exploring models that simulate the function of the human brain's nervous system with functions such as learning, association, memory, and pattern recognition [9].
Research has shown that although recurrent neural networks are good at using global information, due to the exibility, complexity, and diversity of natural language, they are unable to remember content that is too far ahead or too far behind, i.e., the interval between relevant information and the current predicted position is widened and recurrent neural networks become unable to learn how to connect information, a problem known as "long-term dependency." is is known as the "long-term dependency" problem [10]. [11] Long Short Term Memory Neural Network (LSTM), a variant of recurrent neural network, is able to learn long dependencies and overcome the limitations of recurrent neural networks, resulting in more fluent translation of long sentences. e attention mechanism (AM)-based neural machine translation can dynamically obtain the source language word information related to the generated words during decoding, and obtain word alignment information [12]. [13] For example, when generating "I", the word "I" is dynamically calculated to be the most relevant, rather than the other words. Neural machine translation is rapidly replacing statistical machine translation as the mainstream machine translation technology in academia and industry [14].
Neural Network Structure for Chinese-English Translation
Unlike traditional statistical machine translation, which uses multiple modules to complete the translation task, neural machine translation uses an encoder-decoder architecture [15][16][17][18][19]. e encoder-decoder architecture was first proposed by Kalchbrenner and Blunsom at the University of Oxford in 2013.
Basic Network Structure of Chinese-English Translation.
e encoder-decoder model is that encoder converts the input language sequence into a fixed length intermediate vector [20,21]. Figure 1 shows the overall structure of the encoder-decoder model for translating from Chinese to English. e decoder then obtains the output Chinese-English sequence based on the fixed vector C.
Among them, the length of the parallel utterances in the training set is not unique and the words in the utterances will have sequence information; a recurrent neural network (RNN) is used in order to obtain a robust translation system; the structure of which is shown in Figure 2.
As you can see from the expansion diagram, the current moment is influenced by the previous moment. is feature is very suitable for language sequences, because no matter if it is Chinese-English, English, or linguistic, the meaning of the words in the sentence will depend on the context. e RNN not only reads the sequence information from x 0 to xRNN and not only reads the sequence information from front to back, but also from back to front, i.e., from x n to read information, so as to ensure the integrity of the input sequence information.
e backpropagation process can lead to gradient explosion or gradient disappearance, which can cause oscillation in the parameter update of the model learning, while gradient disappearance can make the learning very slow and lead to ineffective learning in the end.
Improvement of the Underlying Network Structure.
is is to improve the translation effect. e improved model structure is shown in Figure 3.
In the case of the translation of English I am a student into Chinese Saya pelajar, when parsing the Chinese English word, so that after the learning of the neural network, it will generate the vector C 2 .
As shown in Figure 4, the stronger the relationship between the source language words corresponding to each target word, i.e., the higher the weight, the darker the colour of the connecting line.
e EOS represents the end-ofsentence marker for each utterance, and also calculates the weight, because some words end when the sentence is read, and need to be fed back to the neural network to tell the model which word is going to end when the target language is generated. e body dynamic vector C i , is calculated as follows: where T x represents the length of the input sentence, i.e., the length of the English sentence, plus the stop character, T x here being 5.
Here, the f function represents encoder's transformation function for the input English word. a ij is the weight value, which is continuously optimized during the learning process of the neural network. erefore, c 1 � g(a 11 * h 1 , a 12 * h 2 , a 13 * h 3 , a 14 * h 4 , a 15 * h 5 ), and the g function is the weighted summation.
Improvement of the Decoder Hidden Layer Unit.
Since the introduction of mathematical methods that simulate actual human neural networks, people have slowly got used to referring to such artificial neural networks directly as neural networks. Neural networks have a wide and attractive prospect in the fields of system identification, pattern recognition, and intelligent control, etc. Especially in intelligent control, people are particularly interested in the self-learning function of neural networks and regard this important feature of neural networks as one of the key keys to solve the difficult problem of controller adaptability in automatic control. e sequence information is read from front to back, and from back to front, i.e., from reading information, thus ensuring the integrity of the input sequence information.
We use the encoder-decoder neural network structure to conduct experiments on Chinese-English machine translation. After the source and target language texts are pretrained with word vectors. e specific structure of the co-decoder's implicit unit is shown in Figure 5.
In the Chinese-English translation experiments in the computation of decoder decoding, an improvement was made to the hidden layer unit in it: that is, the GRU module is composed of the prehidden layer state St-i [22,23].
Structural Design of Chinese-English Translations
Based on the above analysis, an integrated translation system can be designed, which is based on file configuration and can complete the whole translation system. e architecture of the Chinese-English machine translation system is shown in Figure 6. Before training starts, a parallel training corpus is loaded and the model is pretrained with word vectors to obtain word vectors for the source and target languages, respectively [24,25].
Google Translate Example Analysis
Language is flexible and diverse, and many words and expressions are related to specific contexts and can sometimes be difficult to understand. e first two words "how much" and the last two words "how much" mean exactly the opposite. In the following, I will analyse the application of Google's neural network translation in technical English with examples.
Example 1.
Where the separation of the two light components and the two recombinant components takes place, column means column; but in the context of chemical texts, this refers specifically to the distillation column, where the components are separated by distillation, and cannot be translated as column. Compared to human translators, the AI translation results are not only readable and fluent when dealing with complex, specialised or technical passages, but also still intolerably flawed in terms of correctness as a basic requirement [26,27].
Example 2.
A Chinese technology company with 80,000 employees is banned by the Trump administration from dealing with American firms. Google translation follows the original language directly and uses passive sentences, which is not difficult for the target readers to understand, but the manual translation changes passive to active, which makes the expression more in line with the habits of the target audience and more authentic. Human translation: genetically modified plants can be cultivated to possess enhanced stress behaviour. Stress behavior is a dynamic response completed in a short time, not an adaptive inertia behavior. e source text uses the passive voice, but the Google translation avoids the rigidity of the expression by not using the word "to be," but by translating the subsequent purpose clause "to possess improved stress behaviour." is is not clear what is meant by "to possess improved stress behaviour," which creates ambiguity and leads to confusion in understanding. e translation in the human translation adds "its" to the object of reference, translating it as to have improved stress behaviour, which means the transgenic plant, avoiding ambiguity [25].
Example 3. Momentum is building for American regulators to catch up with Europe in promoting "biosimilars," which are generic approximations of patented drugs. Google Translate: US regulators are building for American regulators to catch up with Europe in promoting "biosimilars," which are generic approximations of patented drugs.
Human translation: US regulators are catching up with Europe in the promotion of generic approximations of patented drugs, "biosimilars." Momentum is variously understood to mean, momentum, impetus, etc., and this case is meant to indicate the efforts made by US regulators, so momentum is more appropriate. e translation of approximations in this sentence is inaccurate. Google does not affect the understanding of the meaning of the original text too much, but there is a certain difference with the Chinese language expression. However, there is a certain gap with the Chinese language, especially when translating long and difficult sentences, and due to technical limitations, the translation may only achieve formal equivalence and basic sentence fluency.
Example 4.
Science has cracked the genetic information code. Google Translate: Science has cracked the genetic information code.
Human translation: Science has cracked the genetic information code. Code has multiple semantic meanings, such as code, password, encoding, and codex. e deciphering of the genetic code in biology is a great milestone in the history of biology; genetic engineering technology is not a code, and machine translation cannot identify the exact meaning of words in a particular application context.
Test of Chinese-English and Indo-English Translation
Models. In the experiments, two seeds were selected for model training in each translation direction, and then the model with the better round was selected for ensemble, i.e. fusion model, at the end of training, and the specific experimental results are shown in Tables 1, 2, and 3. (1) Model test results of the better round of two seeds (2) e result of the better model after model fusion, with the best value in bold. e table shows that the BLEU of the optimal model after fusion can reach 37.99.
(1) e model test results for the two seeded better rounds are shown in Tables 4 and 5. (2) e results of the better model after model fusion are shown in Table 6.
e table shows that the two models are equally effective, with a BLEU of 34.44.. Table 9.
e table shows that the optimal model after fusion has a high BLEU of 39.53.
Competitive Translation
Reviews. If the translation model is not available in the corresponding language direction, it cannot be evaluated, and it is necessary to check e evaluation method is to use our test data to crawl the online translation results of Baidu, Microsoft, and Google on the cluster, and then perform BLEU evaluation on the crawled data. e evaluation data from the experiment are shown below. Only the data of English and Indonesian translation are available here, while the online translation results of Chinese and Indian translation are not available for evaluation, as shown in Tables 10 and 11. is project is embedded in the system, and after the model has been tested, the results of the competing translations are calculated and then saved as one file output in the same language translation direction. e data obtained are also much clearer and more convincing.
Conclusions
Artificial intelligence has greatly facilitated the development of translation services, allowing people to quickly understand the general content of languages other than their native language without having to rely to some extent on human translation. e whole training and testing results are very simple, just start loading our collated bilingual material cat algorithm, run the script and then no longer need excessive human cost to test the BLEU value of each model in real time, and finally the results can be obtained. However, due to the differences between the source and target languages, it is found that the translation results of the above-mentioned technical English translation examples have some semantic deviations from the original sentences, and there is a gap with the human translation, and machine translation still cannot achieve perfect translation. erefore, for the time being, Data Availability e dataset used in this paper are available from the corresponding author upon request. | 2022-08-03T15:08:11.209Z | 2022-07-31T00:00:00.000 | {
"year": 2022,
"sha1": "53e53b51c364e5f174f01b6152f17330bc49f57d",
"oa_license": "CCBY",
"oa_url": "https://downloads.hindawi.com/journals/mpe/2022/3077453.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "0d1c0ab2221faa70aa6468d1b2085df240334fd0",
"s2fieldsofstudy": [
"Computer Science",
"Linguistics"
],
"extfieldsofstudy": []
} |
117600489 | pes2o/s2orc | v3-fos-license | Optimization of Electric System for Offshore Wind Farm Based on Lightweight Substation
The electrical system of offshore wind farms is the key to transmit the power from wind turbines to onshore power grid. With the increase of the offshore wind farms capacity and the distance from shore, the traditional centralized offshore substations is facing the challenges of capacity and construction. To solve the above problems, we propose the lightweight substation and an improved k-medoids clustering method to find lightweight substation locations, in addition, we establish the life cycle cost model of electrical system. The example of a real offshore wind farm is taken as an optimization object. The final result shows that the lightweight substation can decrease the economic costs of the offshore wind farm electrical system.
Introduction
In recent years, offshore wind power has developed rapidly. Currently, the capacity of large-scale offshore wind farms (OWFs) has reached up to 2.9GW and the location of the offshore wind farm off the coast can be 80 km. With the increasing power capacity of future OWFs, the optimization of the electrical system of offshore wind farms becomes more important.
Some experts have made relevant research on the electrical system of offshore wind farms. Literature [1] used the fuzzy C-means algorithm (FCM algorithm) to partition the large-scale offshore wind farm, and chose the cluster centroid of each sub-area as the installation location of substations. Literature [2] studied the transmission system between Penghu substation and Taiwan land-based substation, and considered its charging current, overload, over voltage and so on. Literature [3] took economic analysis into the redundant configuration of the main transformer based on the whole life cycle, and provided the theoretical basis for the optimal selection of the offshore wind farm transmission system. In the electrical system of offshore wind farm, the offshore substations undertake the task of converting voltage with optimal efficiency [4]. In the above study, the electrical design of offshore wind farms is applicable for centralized offshore substations. Centralized substations have the same functions as onshore substations, but its construction cost is high, and with the increase of wind farms 2 1234567890 ''"" capacity and the distance from shore, the installation and construction of substations are facing huge challenges.
In response to the above problems, this paper introduces a new offshore substation technologylightweight substation. Compared with the centralized offshore substation, the substation has the decisive advantages of miniaturization and lightweight weight, which is the trend of offshore wind farm.
It is important to find the best position of offshore substations during the optimization of electrical systems. In [5], a clustering algorithm based on attribute threshold is proposed to cluster wind turbines in wind farms to determine the cluster centroid and serve as the location of offshore substations. In [6], the influence of the offshore substations location on the cost of the electrical system was considered, and the economic cost is optimized by the genetic algorithm. However, none of the above studies apply to lightweight substations. The location of lightweight substation is more flexible. It could be built not only in wind farm cluster center but in the location of wind turbine, and reduce the construction difficulty of offshore substation infrastructure.
Based on the above analysis, this paper builds a Life Cycle Cost (LCC) model applied for the offshore wind farm electrical system and proposes an improved k-medoids method suitable for site selection of lightweight substations. In addition, this paper adopts single parent genetic algorithm to optimize the topology structure of electrical system.
Cutting-edge technology of offshore wind farm----Lightweight-Substation
With the rapid development of offshore wind power, the fan capacity and the scale of wind farm are continuously increasing. In addition, the size and weight of substations are also increasing and the top weight exceed the single hosting capacity of vessels. These factors have bright great challenges to substation construction. In order to solve such matter, Siemens, ABB, DONG and other companies have started to design the lightweight substation for larger offshore wind farm. The structure is shown in figure 1 Different from traditional centralized substations, lightweight substations are single layer structures, and employ only one large capacity transformer (traditional offshore substations often equipped with multiple transformers). The problem of reliability reduction can be solved by connecting the lightweight substations with sea cables to each other. The traditional electric system optimization model will no longer be suitable for lightweight substation, and it is necessary to establish a new electric system model for offshore wind farms.
LCC model based on centralized maritime substation
In the optimization of electric system of offshore wind farms, the centralized offshore substation location will often be selected at the cluster centroid, or selected based on the designer's experience [1,4], so it's difficult to get the optimized results. In summary, the LCC model based on the centralized substation is shown as follows: Where, initial investment C I includes the purchase cost and installation cost of submarine cable(C Icable ), the purchase cost and installation cost of transformer (C Isub ) and the substation construction cost (C lsubf ) etc. C F is power loss, C o is loss of operation, C M is the operation cost of cable and transformer, C D is the recovery cost (The value of the recycled material is considered to be offset by the cost of recycling, that is, C D =0); P ossum is the capacity of the substation and P WT is the total installed capacity of the wind farm; I imax is the maximum continuous load current through the sea cable i, I i0 is the long-term ampacity; K i is the total correction coefficient of the long-term ampacity; S imin is the minimum cross-section meeting the requirement of the short-circuit thermal stability, I i∞ is the steady-state short-circuit current, t i is the short circuit time, C ir is the coefficient of thermal stability, X sub and Y sub are the abscissa and ordinate of centralized substation respectively, D sub the area outside the fan in wind farm. P V.sum is the discount factor of the annual investment, P V is the discount factor, expressed as follows: r and t are the discount rate and useful life respectively. The annual operating loss of the electrical system includes the submarine cable power loss during operation and the transformer operating loss [7], and its expression is: Where: I j is the root mean square current through the high-voltage submarine cable, the expression is shown in Equation 6, I e is the current generated by the fan at a certain wind speed, and t e is the duration of the current, T is 8760 hours. c is the online price of offshore wind power, R j is the resistance value, t j is the total running time, P 0u is the transformer no-load loss, t 0u is the transformer's annual running time, P ku is its load loss, ρ u is the transformer load factor, τ u is the year Average maximum load hours.
Maintenance cost is divided into two parts, one part is the fault maintenance of submarine cable, and the other part is the transformer maintenance costs, so the repair costs are as follows in expression: In the formula: k j is the annual failure frequency of the cable, c j is the cost of each maintenance; k u is the annual failure rate of the transformer, and c u is the cost of each maintenance.
LCC model suitable for lightweight offshore substation
Different from the centralized offshore substation, lightweight substations can share a foundation with the wind turbine. An offshore wind farm can use multiple lightweight substations instead of a traditional centralized substation Therefore, an offshore wind farm can be divided into several subareas according to the number of lightweight substations. In addition, according to the reliability requirements of lightweight substations, it is also possible to interconnect several lightweight substations to improve the electrical system redundancy.
Where, the initial investment C I includes C Icabl , C Isub , the purchase and installation cost of spare submarine cables needed for the interconnection between lightweight substations(C IHsp ), and the reconstruction costs of wind fan foundation occupied by lightweight substations; X WTi and Y WTi are the abscissa and ordinate of the fan shared base with lightweight substation j respectively.
Substations can be spared to each other by interconnection, and it reduces the possibility of all the turbines shut down in the emergent situation. High-voltage submarine cable and alternate submarine into "E" shaped arrangement, shown in Figure 3. The solid circle in the figure represent a lightweight substation, and the rectangle represents the fan group connected to each lightweight substation. Table 1: Table 1 The outage of wind turbines and its probability in a district Power failure in a certain area.
All-stop Partial-stop non-stop Probability P T +(1-P T )P H P s (1-P T )P H (1-P s )P (1-P T )P H (1-P s )(1-P bi ) +(1-P T )(1-P H ) Loss power nP ti n i P 0 In the table, P T is the fault probability of the transformer, P H is the failure probability of high voltage submarine cable, P s is the corresponding probability of standby submarine fault, P bi is the probability of medium voltage submarine cable failure, n is the number of stoppages, P ti is the average fan power. Therefore, we can get the expected annual power outage(E) of the wind farm in this area according to the table. The expression of annual power loss is as follows: Where, T F is the power failure time.
Lightweight Substation Location Based on Improved k-medoids Clustering Algorithm
The lightweight substation location is related to the fan coordinates and it is in a discrete range. The kmedoids clustering algorithm can be used to find a cluster centroid from the current discrete population, and the Euclidean distance from the cluster centroid to all other points is shortest. Therefore, the clustering algorithm meets the requirements of selecting the lightweight substation location in wind farms. It is worth noting that the sample of this paper includes all the wind turbines location and onshore substation location. Medium voltage submarine cables are used between the fans and the offshore substations, and high voltage submarines are used between the offshore substations and the onshore substations. Due to the wide price variance between those two submarines, it will result in the large error if consider only the Euclidean distance. Therefore, it is necessary to improve the algorithm, and the specific model is as follows: Where: D H is the total length of the high-voltage submarine cable; D M is the total length of the medium-voltage submarine cable; d i s2land is the distance from a lightweight substation to the onshore substation; d ij WT2s is the distance from the wind turbine to the lightweight substation. k 1 and k 2 are the price factors of medium and high voltage submarines, and both factors are based on the lightweight substation capacity and the number of fans in the fan string or fan ring connected to it. k 1 is the average unit price of all medium-voltage submarines, and k 2 is the price of the high-voltage submarine cable; D T is the product of all distances and the corresponding price coefficients, and the k-medoids algorithm flow chart shown in Figure 4
Topology Optimization Based on Partheno -Genetic Algorithm
After the site selection of substation, we adopt single parent genetic algorithm to optimize the topology. The single parent genetic algorithm does not use the crossover operator which used in the traditional genetic algorithm, and all the genetic operations are completed in a single individual. The algorithm has the merits of simple genetic operation and high calculation efficiency. In addition, it has much less strict requirements for the individual diversity of the initial population, and has no problem of "premature convergence". Typical wind farms electrical system wiring mainly includes radial and ring: ①about the radial structure, this paper adopts the single-parent genetic algorithm and minimum spanning tree to optimize the topology. ②about the ring structure, this paper combines the single-parent genetic algorithm with the multi-traveling salesman problem to get the optimal ring topology of the electrical system.
Case introduction
In this paper, an offshore wind farm with the capacity of 360 MW is taken as an example. There are 100 fans with a capacity of 3.6MW, and the fans location is shown in Figure 5. The fault rate of submarine cables is 0.03 times/ (km a), the cost of repairing high voltage submarine cables is 5 million yuan, and that of medium voltage submarine cables is 1 million. The fault rate of the main transformer is 0.01. The wind farm service life is 25 years and the annual utilization hours of the fan is 2600 hours. For the traditional station model, we use the method proposed in Reference 8 to find the cluster centroid as the best location of centralized substation and optimize the electrical system topology by the method proposed in the previous section. The results are shown in Figure 6 and Figure 7, Black solid line express high voltage transmission submarine cable.
The optimization results of the lightweight station mode are shown in Figure 8 and Figure 9. Black dashed line is the standby submarine cable. | 2019-04-16T13:28:58.073Z | 2018-10-11T00:00:00.000 | {
"year": 2018,
"sha1": "c787a19cac87a89fa2364abe9b6d359d498b8800",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1755-1315/186/5/012023",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "5de4afa9d946aa11b8c3fc81ef5e79c2f970d3e6",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
234080190 | pes2o/s2orc | v3-fos-license | Development of Personal Income Taxation Achievement by Practicing Skills within the Class in the Course of Tax Law
This academic article presents the principles of teaching and learning to promote proactive learning or active learning. Action by focusing on learning is important, and also using the proactive learning management process. The use of learning through the use of practicing skills within the class. The goal of teaching and learning management that focuses on the students is to manage learners to create new knowledge through self-thinking process, allowing the learner to learn by doing, and to understand. It can bring knowledge to integrate into daily life and have qualifications for the goals of education that want learners to be smart, good and happy. Good teaching is a dynamic movement, constantly changing, both in organizing activities and teaching method which consists of creating atmosphere, technique, style, content, and things that are not too far from the students and planning of teaching and learning systematically.
Introduction
National Education Act 1999, Section 22 states that the provision of education must be based on the principle that all learners have the ability to learn and develop themselves, and that learners are of the most importance in academic learning. On the legal side must have good knowledge, skills and attitudes in the legal profession. At present, the business industry has expanded more. Taxation is more important. Coupled with the economic situation and the trade of the country has evolved rapidly. The importance of taxation information requires accuracy, certainty and reliability. All educational institutions would like to have education management to be effective and meet the goals according to the vision mission of each place by establishing strategies for developing the process for selecting learners with potential. The teaching and learning management according to the National Education Act 1999 has focused on the learners most important. All learners have the ability to learn and develop themselves. In the management of education, it is necessary to encourage learners to develop themselves to their full potential. Therefore, in the management of teaching, there must be a variety of forms to meet the needs of individual learners. The instructor must give the learner the greatest care. By studying the teaching methods and learning methods of learners in new ways to be used in solving problems or developing learners. With a focus on teaching instructors to integrate research processes or integrate them into learning development [1] From observing teaching and learning in the subject of tax law on personal income tax for the 4th year students in the Bachelor of Laws program, it was found that students could not distinguish the type of tax in the personal income tax and calculation error. The main item relating to the calculation of personal income tax must be able to classify the individual income and deduct the deductible correctly. From the teaching results of the students in the aforementioned law programs Found that students studying tax law only 20 percent of which can calculate personal income tax correctly, the other 80 percent do not understand the principles of personal income tax calculation. Relying solely on memorization and do not understand about calculating personal income tax. When analyzing the main causes of teaching and learning, it was found that the students' failure to prepare the exercises by themselves. Caused by the students do not understand it. Because the students will not seek help with classmates in matters they do not yet understand. Fear of friends to insult and it is also important to note that once each day of study is complete, students will hurry home. It may cause students to incorrectly calculate personal income tax, which is a reason for confusion. If it still going on or has not been resolved it will cause an impact on academic achievement will lead to a bad attitude towards continuing to study tax law. Therefore, the instructor decided to solve the problem for students to have the skills to calculate the personal income tax correctly. Using the personal income tax calculation skill form to develop students to know the process of calculating personal income tax on personal income tax in Tax Law. The Bachelor of Laws program recognizes the great importance and necessity of tax law. This is a course that focuses on the importance of students to increase their potential to be efficient and effective by studying the principles of taxation in respect of income tax from the Revenue Code, for example personal income tax. Corporate income tax VAT Specific business tax And stamp duty. Which will enable learners to apply their knowledge and ability to apply in daily life and professional practice in order to develop the potential of the lecturers of the Bachelor of Laws to international. This is in line with [2] research on the expectation of Bachelor of Law students in teaching English for lawyers The key findings of this study are that students have high overall and individual expectations for teaching English for Law. In particular, the ISSN: 00333077 1592 www.psychologyandeducation.net instructors' teaching expectations were at the highest level. And the instructors should understand and recognize the importance of students' basics in English. For the achievement of teaching and learning that will make the students learn real benefits. Teaching this subject in tax law the instructor has organized a student-centered instruction. In order for learners to create new knowledge through the self-thinking process. Which allows the learner to learn by doing understand and can bring knowledge to integrate into daily life. Have qualifications for the goals of education that want learners to be smart, good and happy. Instructors therefore take into account various issues in organizing teaching and learning activities. Individual differences of learners focusing on the needs of the learners improving the quality of life of the learners activities to be interesting does not make the students feel bored kindness to learners challenging the learner to know recognizing the right time for learners to learn creating an atmosphere or situation for learners to learn by doing. Support and promote learning teaching aim learners understanding learner background not sticking to one particular method teaching good is dynamic, that is, the movement is constantly changing in terms of activities. Atmosphere creation Form, content, techniques, methods of teaching things that are not too far from the learner And teaching and teaching planning systematically by definition of academic achievement (Achievement) at [3] has defined the achievement as it is learning according to a predetermined plan. Arising from the teaching and learning process during a particular period of time which find [4] meaning that the measure of academic achievement is a measure of academic success. Or measure the learning experience that students gain from teaching and learning. It is measured according to the aim of teaching or to measure the achievement of the education and training programs. This is consistent with [5] defining academic achievement as attributes, including knowledge. The competence of the individual as a result of the teaching or the mass experience that the individual gains from teaching Causing individuals to change behavior in different areas of brain function Which is intended to check the level of brain competence of a person that learns what And what kind of ability. As well as the consequences of learning, training or experiences in school, at home and in other environments, including feelings, values and ethics are the result of practice. From this definition Academic achievement is the learning outcome according to the course Acquired according to the measurement and evaluation principles covering both cognitive and cognitive aspects emotional and emotional or spiritual aspects and the skill, practice, or character skill set by the instructor at a particular time. In general, the measure of academic achievement will measure knowledge and abilities according to the subject matter studied which is mostly cognitive or cognitive Most of the tools used to measure the test are called the achievement test. The objective is to know that the learner after the teaching and learning process will have the knowledge level. So that instructors can find a way to improve, correct, develop and encourage learners to develop their knowledge to their full potential. But to create a quality test the instructor must have knowledge of the nature of the test. Planning, creating principles, creating a selection of test types to suit the content; and use of exam results to improve and summarize grades
Meaning of achievement test
The word the achievement test has many interested people to study the meaning and say it. Achievement tests mean that the achievement test is a test used to measure the brain behavior of the learners having knowledge and ability in the subject they have learned. Or how much has been trained? [1] [6] said that the learning achievement test was a test that measures knowledge, skills, and academic ability that the learner has learned how accomplished the stated purpose. So it can be concluded that Academic achievement test is a test used to measure knowledge and skills learned from the past or in the present state of the individual. However [7] said achievement means success, fluency. Proficiency in the use of skills or the application of knowledge, the achievement refers to the knowledge or skills that arise from learning in the subjects that have been learned. This is obtained from the results of the instructor test or the person responsible for teaching or both. From the above, it can be concluded that academic achievement is the learning outcome according to the course which has been applied in accordance with the measurement and evaluation principles covering both knowledge thought or cognition emotional and emotional or psychological aspects and the skill, practice, or skill set by the instructor during the period of instruction.
Types of tests
There are two types of achievement tests: instructor made tests and standardized tests. Both of which ask for the same content. It was asked what the learners had received from the teaching which could group 6 types of behaviors: knowledge, memory, comprehension, application, analysis, synthesis and assessment. [5] 1) Instructor-created quizzes are self-created quizzes to test students in a classroom. For [8], the test was classified into 3 types as follows: 1) The oral version is a test that relies on individual questioning. It works well if a small number of people take the exam. Because it takes a lot of time can ask carefully because they can interact with each other 2) The written answer is a test that is changed from the oral exam. Due to the large number of people who take the exam and the number is limited, it can be divided into 2 types: (1) Essay or subjective is an exam that allows respondents to independently compile their own words in expressing attitudes, feelings and thoughts under a given subject. Is an exam that can Good measure of synthesis behavior but there is a downside to the rating. Which may not be accurate making it difficult to have multiple choice.
(2) Answer limit it is an exam that has correct answers under limited conditions. This type of test is divided into 4 types: correct, wrong, additive, matching and multiple choice.
3) Practice exercise is a test in which the test taker demonstrates the behavior by actual action or practice, such as a musical test, mechanic, physical education, etc.
Fig.2 Types of tests
It can be concluded that the academic achievement test can be divided into two types: standardized test. Which is built by experts in content and educational measurement there is a good quality search. The other type is instructor-created quizzes. For use in classroom testing in designing, testing, measuring academic achievement, vocabulary for communication. The instructor has chosen a researcher-built practice test to measure vocabulary ability to be used in oral and written communication and choose a written quiz that limits answers by selecting answers from the given options. In creating quizzes to cover content and measure behaviors that are appropriate to the content. Developing the table of specifications should be created to guide the building, just like building a house drawing. Also known as a Test blueprint, the curriculum analyzes the curriculum topic. And learning objectives and desired behaviors to be measured. Creating a Tables Analysis Course starts with creating a 2D table, ie vertical, the behaviors that need to be measured including memory, understanding, application, analysis, synthesis and valuation the horizontal section is for the topic, content or learning objective. Which depends on the subject matter and / or objectives. Then determine the weight of the content. Consider the importance of the content which may define the weight in percent Along with determining the behaviors that need to be measured and set the importance By considering the learning purpose along with the content Finally, the test to be measured, such as correct, wrong, match, add-on, choice or subjective, etc.
The benefits of skill exercises
The benefits and importance of skill exercises. [9] 1. Make students better understand the lessons. 2. Make the instructors know the understanding of students towards learning. 3. Helps instructors improve teaching content and activities in each lesson. 4. Help children to learn better according to their abilities. 5. Train students to have confidence and be able to assess their work. 6. Train students to work sequentially with responsibility for the assignments. Where [10] said that the benefits of the skill training are as follows: 1. As an additional or supplement to textbooks for skills training 2. Helps to enhance the use of skills. 3. It helps with individual differences. 4. Helps to strengthen the lasting skills by practicing immediately after learning the subject, repeat many times. The practice should focus on the subject that was practiced. In the section [11], the benefits of skill training are described as follows. 1. It is a device to help reduce the burden of instructors. 2. Help students to improve their skills. 3. It helps with individual differences. Causing students to achieve great mental results 4. Helps to strengthen skills in durability 5. It is a tool to measure grades after learning the lesson. 6. Help students to review lessons by themselves. 7. Help instructors look into problems. Of students clear 8. Allowing students to practice fully In addition to what was learned in the lesson 9. Help learners see their own progress. However, [12] mentioned the benefits of the skill training as follows: 1. To enhance skills Practice exercises are tools that assist students in practicing their skills, but they require encouragement and attention from their instructors. 2. Helps in the differences between people. Because students have the ability Different language Allowing students to do exercises that are appropriate for their abilities will help students achieve more mental achievement. Therefore, exercise is not a workbook. The more the instructor gives to the students. Chapter by chapter or page by page. 4.2 Focus only on wrong matters 5. The exercises used will be a tool to measure learning achievement after each lesson. 6. Exercises that are created as a booklet, students can be kept for use as a guide. For further self-review. 7. Having students complete the exercises helps instructors to clearly see the student's hot spots or problems. This will help instructors to improve and solve the problem promptly. 8. Exercises held other than those contained in the textbook. It will help students to practice fully. 9. The training has already been printed. This will help instructors save both labor and time. In order to always be prepared to create a practice. For the learners, there is no need to spend time copying the practice from textbooks or blackboards. This gives them more time and opportunities to practice various skills. 10. The exercise style saves money. Because the publication is a certain format. It is not necessary to invest less than to use the printing method on wax paper every time. And see their progress in a systematic and orderly manner.
The meaning of Practicing
The person who proposed this meaning is [13], who has proposed the meaning of Practicing as follows. Practicing is the repeated act of behavior that has been learned in order to develop a skill or expertise or a habit, for example when learning mathematical principles or theories. Science already it has to be practiced by doing the exercises, etc. and in the practice should follow the following principles. 1. Begin training when students are ready first. 2. Should take into account the difficulty of the things to be practiced. 3. Define the training period appropriately, not too long or too long. 4. If there is a lot of practice must be divided into periods 5. Students should know the progress in the practice.
The good nature of a good skill exercise
By creating a practice exercise to train analytical skills to achieve that objective requires a variety of forms of training. This must be consistent and appropriate for the skills to be practiced. The concept of characteristics and patterns of good skill training which [14] has proposed, the characteristics of a good practice should look like this: 1. Relate to the story that has been learned 2. Suitable for the class level or the age of the students 3. There is a short statement to understand. 4. Use the right time. 5. There are interesting and challenging things to show your abilities. 6. Should have instructions for use. 7. There are limited choices and free answers.
8. If it is a practice that requires students to study by themselves the exercise should be multiple forms. However, [15] has proposed the following characteristics of a good training model. 1. It should be created to practice what it is teaching. Not a test that students what did you learn? 2. Should be about the structure of what is taught alone. 3. What you practice is what students have already seen. 4. Should be a short message. 5. Do not use too many words. 6. Should be a practice that encourages the child to respond to the desires. Where [16] said that a good practice should consist of 1. Contain content that matches the purpose 2. Activities suitable for age level or the ability of students 3. The illustrations are well laid out. 4. Spend time appropriate for training. 5. Challenge students' abilities and be able to practice by themselves. However, [17] said that the instructor or the person creating the exercise should adhere to the following characteristics of good practice. 1. Good exercises should be clear in both instructions and how to do them. The instructions or examples showing the actions used should not be too long. Because it will make it difficult to understand. It should be adjusted to be easy and suitable for the user so that students can study by themselves if desired. 2. Good exercises should be meaningful to learners and meet the purpose of the practice. Little investment, can be used for a long time and be up to date on a regular basis. 3. The language and images used in the exercises should be appropriate for the age and knowledge base of the learners. 4. Good exercises should be separate, individual topics should not be too long. But there should be a variety of activities to stimulate students' interest. And not boring in doing and to practice any skill until mastery. 5. Good exercises should include both the fixed-answer and the free answer. Choosing to use words, texts or pictures in exercises. It should be something that students are familiar with and meet their interests. This is so that the exercises are created to be enjoyable and pleasurable for the user, which is in line with the learning principle that students tend to learn quickly in satisfying actions. 6. Good exercises should give students the opportunity to study by themselves, to learn, to research, to collect common things that they have seen or used themselves. This will enable the students to understand the subject better and will be able to apply that knowledge in their daily life properly and properly. And see that what students practice is meaningful and beneficial to students forever in the future. 7. Good exercises should cater to different students' individual differences. They differ in many areas, such as needs, attention, readiness, intelligence level. And experience, etc. Therefore, each exercise should be sufficiently prepared and available at all levels, from easy to medium to relatively difficult. So that both students are good, middle and weak, can do according to ability. This is for every student to be successful in completing the exercise. 8. Good exercises should be able to excite students' attention from the cover to the last.
www.psychologyandeducation.net 9. Good exercises should also be exercises that can be assessed and classified by students. From studying the characteristics of good training enough to conclude that good exercises should include psychology, proper linguistics, fun practice, awakening interest and meaning in life. Suitable for age and ability and may be able to study by themselves, which the report can be considered as a guideline for preparing the exercise to suit the age Student abilities several elements match the content. Appropriate to the age, time, ability, attention and problem of the learners.
A good practice component
The components of a good exercise should include the following documents. [14] 1. Practice Manual It is an important document for the practice of what it is used for and how to use it. Used for home use, repairing, which should consist of 1.1 Components of the exercise how many sets are indicated in this exercise? What and whether there are other components, such as a test or an assessment record.
1.2 What instructors or students must prepare It will tell instructors or students to prepare in advance.
1.3 Purpose of using exercises 1.4 The steps for using each item are listed in order of use and may be written in the form of teaching or lesson plans to be more clear. In the importance of the exercise, good exercises are essential to improve learners' skills and are an integral part of the Primary Education Curriculum, Year 2521 (Revised Edition 2533) [18] in line with [19] said that exercises are very necessary. Instructors must provide appropriate exercises in order to practice after having learned the content from the textbooks to have extensive knowledge. It can be considered that exercises is one of the teaching materials. Which instructors are able to use in teaching activities very well and helping instructors to teach success and in line with [17] also said that Work book is a teaching medium that is prepared. This is to enable learners to study, understand and practice until they come up with correct concepts and skills in a particular subject. Moreover, the exercises are also used to indicate that the learner or the user of the exercise has knowledge of the lesson and is able to how much of that knowledge can be used Learners have strengths that should be promoted or have weaknesses that need to be improved, where and how good and complete exercises. Therefore, it may be used as a substitute for a diagnostic test for assessing student progress. Exercises can be considered an important tool that all instructors use to examine their cognition and develop students' skills in a range of subjects. Appropriate exercises for each age student contribute to student success. Creating pride that brings joy and pleasure in doing. There is an opportunity to use imagination to develop creativity. On the body side, it will lead to the development of the muscles and the senses from the practice of writing hand gestures, hand gestures.
Personal income tax
Personal income tax is tax levied on individuals. Or from a special taxation unit as required by law and generate income according to the specified criteria usually stored annually Income generated in any year income earners are obliged to present themselves on the required tax return by January to March of the following year. For some income earners, the law also requires the filing of the form. Tax payable at half year on actual income in the first half of the year. In order to relieve the tax burden that must be paid and in some cases the law requires that the payer acts withholding tax on part of the income paid. So that there is a gradual tax payment while there is income arising as well. Persons liable to personal income tax are those whose income arising during the past year with One of the following (Section 56 and Section 57) as follows: 1. Natural person. 2. Non-juristic partnership. 3. Those who died during the tax year. 4. The estate that has not yet divided.
Fig.3 Persons liable to personal income tax
Assessable income is income of the following categories. 1. Income derived from employment, whether in the form of salary, wage, per diem, bonus, bounty, gratuity, pension, house rent allowance, monetary value of rent-free residence provided by an employer, payment of debt liability of an employee made by an employer, or any money, property or benefit derived from employment 2. Income derived from a post or from performance of work, whether in the form of fee, commission, discount, subsidy, meeting allowance, gratuity, bonus, house rent allowance, monetary value of rent-free residence provided by a payer of income, payment of debt liability of a taxpayer made by a payer of income, or any money, property or benefit derived from a post or from performance of work, whether such post or performance of work is permanent or temporary. 3. Fee of goodwill, copyright or any other rights, annuity or annual payment of income derived from a will, any other juristic act, or court decision. 4. Income that is: Persons liable to personal income tax Natural person Non-juristic partnership Those who died during the tax year The estate that has not yet divided www.psychologyandeducation.net (a) Interest on a bond, deposit, debenture, bill, loan whether with or without security, the part of interest on loan after deduction of withholding tax under the law governing petroleum income tax, or the difference between the redemption value and the selling price of a bill or a debt instrument issued by a company or juristic partnership or by any other juristic person and sold for the first time at a price below its redemption value. Such income also includes income assimilated to interest, benefit or other consideration derived from the provision of a loan or from a debt-claim of every kind whether with or without security. (b) Dividend, share of profits or any other gain derived from a company or juristic partnership, a mutual fund or a financial institution established under a specific law in Thailand for the purpose of providing a loan in order to promote agriculture, commerce or industry; the part of dividend or share of profits after deduction of withholding tax under the law governing petroleum income tax. For the purpose of income calculation under paragraph 1, if a lawful child who is a minor derives income and the marital status of the parents exists throughout the tax year, the income of the child shall be treated as income of the father. However, if the marital status of the parents does not exist throughout tax year, the income of the child shall be treated as income of the parent who exercises parental power, or of the father if both parents jointly exercise parental power. The provisions of paragraph 2 shall apply mutatis mutandis to an adopted child who is a minor deriving income. (c) Bonus paid to a shareholder or partner of a company or juristic partnership; (d) A decrease of the capital holdings in a company or juristic partnership which does not exceed the total amount of profits and reserves; (e) An increase of capital holdings in a company or juristic partnership that is determined from the total amount of profits or reserves; (f) A benefit derived from the amalgamation, acquisition or dissolution of a company or juristic partnership and having the monetary value which exceeds the capital; (g) Gains derived from transfer of partnership holdings or shares, debentures, bonds, or bills or debt instruments issued by a company or juristic partnership or by any other juristic person. 5. Money or any other gain derived from: (a) Rent of property, (b) Breach of a hire-purchase contract, (c) Breach of an installment sale contract, where the seller regains the property sold without paying back the money or gains already received. 6. Income from liberal professions, namely, laws, arts of healing, engineering, architecture, accounting, fine arts or other liberal professions as prescribed by a Royal Decree; 7. Income derived from a contract of work where the contractor has to provide essential materials besides tools; 8. Income from business, commerce, agriculture, industry, transport or any other activity not specified in (1) -(7).
Related research
It is very important for instructors to acquire knowledge in creating training exercises that will train students' skills to create such exercises as highly effective and suitable for the students. With [20] doing research on Using the skill training On account recording according to the dual accounting system In the financial accounting course For students in the Accounting Division, Vocational Certificate 1, Lanna Polytechnic College of Technology, Chiang Mai, found that students had higher academic achievement in line with the [21] conducting research on Using a skill set to solve the problem of students' lack of skills in recording trade in general ledger Of vocational students in Vocational Year 1/7 of Rayong Commercial School, the first semester of the 2007 academic year, totaled 12 students, found that the students had higher academic achievement and passed the assessment criteria for everyone Creation and development of a practice for enhancing cost accounting skills 1 for vocational students level 1 in accounting Roi Et Vocational College, 38 people. It was found that 30 students from a total of 38 students, representing 81.08 percent, received an examination score of more than 50 percent, and the results of absenteeism after 35 students passed the test, representing 91.89 percent who did not pass the test. 8.11 percent of people. And [22] conducts research on the development of academic achievement in introductory accounting for 2 subjects, recording transactions in general journals of Vocational 1/4 students using the skill training form, Srithana Commercial College of Technology, Chiang Mai, 20 people found. That students have higher academic achievement And pass all criteria. From the study of related research making it possible to conclude that using the Pre-School and Post-Study Skill Training Set, you can see how you improve your students. And can do exercises correctly.
Conclusion
Using the Pre-School and Post-Study Skill Practice Set, students will be able to learn about the content and practice doing a practice exercise at the same time. Make students better understand the content and be able to practice the skills properly. The skills training appropriate for each age student contributes to the success of the teaching and learning process to the success of the course. | 2021-05-10T00:04:18.855Z | 2021-01-29T00:00:00.000 | {
"year": 2021,
"sha1": "2d5b48c6ae5dc43857fc1633b4db8634001f1b14",
"oa_license": "CCBY",
"oa_url": "http://psychologyandeducation.net/pae/index.php/pae/article/download/951/765",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "5683d220dd895d6da259caef27daecbe401f1fff",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Psychology"
]
} |
220886235 | pes2o/s2orc | v3-fos-license | Linking Theories, Past Practices, and Archaeological Remains of Movement through Ontological Reasoning
: The amount of information available to archaeologists has grown dramatically during the last ten years. The rapid acquisition of observational data and creation of digital data has played a significant role in this “information explosion”. In this paper, we propose new methods for knowledge creation in studies of movement, designed for the present data-rich research context. Using three case studies, we analyze how researchers have identified, conceptualized, and linked the material traces describing various movement processes in a given region. Then, we explain how we construct ontologies that enable us to explicitly relate material elements, identified in the observed landscape, to the knowledge or theory that explains their role and relationships within the movement process. Combining formal pathway systems and informal movement systems through these three case studies, we argue that these systems are not hierarchically integrated, but rather intertwined. We introduce a new heuristic tool, the “track graph”, to record observed material features in a neutral form which can be employed to reconstruct the trajectories of journeys which follow di ff erent movement logics. Finally, we illustrate how the breakdown of implicit conceptual references into explicit, logical chains of reasoning, describing basic entities and their relationships, allows the use of these constituent elements to reconstruct, analyze, and compare movement practices from the bottom up. a paved road and a trackway made by a wild animal near Pugey (France), source L. Nuninger.
Introduction
Over the past two decades, the basis for the creation of archaeological knowledge about past societies archaeology has fundamentally changed. This shift has been driven by multiple factors, notably the acceleration of development-led archaeology [1][2][3] and the integration of large digital observational datasets, particularly those acquired through remote sensing methods for earth observation and monitoring [4,5], the proliferation of scientific methods applied to gather diverse data [6], and the increasing digitization of archaeological data [3,7,8]. Within these broader trends, emerging applications of machine learning methods for the identification of features of archaeological interest in remote sensing derived datasets (e.g., [9,10]) are causing a particularly important change in knowledge production practices. The machine learning methods being applied in many archaeological remote sensing projects take as an input examples of features which represent different types of entities. For example, a project team studying field clearance as a practice might seek to provide examples to an algorithm of the "most important" types of features related to field clearance activities. The research team might provide examples of round cairns, linear cairns, irregular cairns, and field boundary walls, which based on their knowledge at the time were the most relevant feature types. The algorithm learns to identify features similar to these input examples, and when applied it does so quickly and at an extensive scale. This method, in multiplying the identified examples of features of a type similar to the examples used as inputs, rapidly reinforces the dominance of those types of features in the archaeological record. An unintended consequence of this process is that features which were not considered important when designing the machine learning project could be further marginalized within the archaeological record, as the existing state of knowledge of the project team and their research biases are reproduced and reinforced at an unprecedented scale. This problem of "circular reasoning" and "finding more of what you expect to find", posed by the strong reinforcement through machine learning of implicit ideas and knowledge about what physical features represent different activities and processes, has been noted by researchers working in this domain [11][12][13], particularly in the context of the transfer of these algorithms between different environmental or archaeological contexts.
Two key sources which inform the implicit ideas and knowledge base of teams working in this context are the domain-specific literature on the activity or process they aim to study, as well as prior identifications in remote sensing data by expert interpreters of features related to that activity or process. To address part of this problem of strong reinforcement of implicit knowledge, we propose an approach to integrating a re-reading of the rich but challenging written archive of archaeological research on a given topic with the existing information structures used in expert identifications of features in remote sensing data sources, focusing on extensive aerial data sources such as those derived from satellite imagery, aerial photography, and airborne laser scanning (lidar). To illustrate this approach, we explore the process of creating archaeological knowledge about past movement and the material evidence of its impact on the landscape.
Using three case studies drawn from the archaeological research literature (primarily journal articles and books), we analyze how researchers have identified and conceptualized the material traces of various movement processes in different regions. Although the case studies are very different in their descriptive vocabularies and the physical remains they discuss, our analysis shows that all of them combine formal pathway systems, characterized by infrastructure intentionally devoted to enabling travel and movement, and informal movement systems, characterized by diverse features and structures which are co-opted to enable movement. As these two systems are not hierarchically integrated but rather intertwined, we propose to introduce a new concept, the "track graph" that aims to record observed material features in the most neutral form possible, providing a basis for analyzing the combined systems. Of a purely analytical nature, this abstract "track graph" concept can be employed to reconstruct trajectories that follow various types of movement logic.
We outline our approach to constructing ontologies, in this paper's example built around the "track graph", through which we explicitly relate material elements identified in observational data from remote sensing or field-based surveys to a body of knowledge or theory as expressed in the written research, in this paper's example about movement. Movement is a particularly useful topic through which to explore our proposed approach because as a concept it integrates diverse practices, manifested archaeologically through varied physical forms, and it has been approached in quite different ways by researchers working in different contexts. It presents the kind of intellectual complexity, shared by many topics salient to contemporary research on archaeological landscapes, which requires us to grapple with core issues of knowledge creation.
Biases Affecting the Corpus of Physical Features Used as Evidence of Movement
As we noted above, there are two key sources which inform new identifications of features as evidence of past movement: domain literature and past expert identifications in observational datasets (remote sensing) or field observations. Each of these are affected by inter-related methodological, observational, contextual, and semantic biases. These biases are well known, as they impact on most observational datasets created and used by archaeologists, and we summarize the key impacts for studies of movement here.
Methodological biases: The choice of remote sensing technique and the design of the survey affect the probability of detecting features, depending on their specific physical characteristics. Some techniques are better for observing features on the ground surface and others for detecting features below the surface. Some are more suitable for detecting smaller features, while others will only reveal features under specific conditions of lighting, soil humidity, or vegetation cover [14,15]. Survey design choices related to areal coverage and spatial resolution have significant effects on detection probabilities [15,16].
Observational biases: Observers tend to recognize features similar to those they have seen before [11], so an observer's familiarity with features has a significant impact on the frequency of recognition (see, e.g., [12]). Observational bias may lead to false positives as well as false negatives.
Contextual biases: Some features are easily identified and interpreted in isolation, because they have an unequivocal physical expression that cannot be mistaken for something else. However, in most cases we need an understanding of a feature's context in order to correctly identify it [17], taking into account environmental, cultural, and historical context [18]. Certain environmental conditions will enhance or suppress the physical expression of some features, e.g., because they are easily eroded, leading to potential misidentifications. Equally, if contextual information about the known activities in a region at different periods is lacking or incorrectly interpreted, the identification of features may be erroneous [19], a situation complicated by the changing use of physical features over time [20].
Semantic biases: The assignment of a feature to a specific class or type may be a source of disagreement between observers, since the two observers may draw on different classifications and typological systems. Further, even when notionally working within the same system, these two observers might rely on different implicit ideas about the character or function of a feature belonging to each class, using the same term to indicate and imply different things, further complicating the effects of this bias [21]. This problem is only increased when communicating across languages, and it is the basis of many problems associated with cross-linking datasets. Even if we know that crêtes de labour in French are translated as headlands in English, are we sure that these are really the same thing? Do other observers understand these terms in the same way? Further complicating matters, the meaning of these terms can change over time as the research context evolves [19].
Focusing on Observational Biases: How to Recognize a Path
When looking for material and physical traces of movement, both in the field or in the "digital terrain" created from aerial and satellite remote sensing data, all these biases are at play. Our eyes will follow linear features that we immediately "translate" as roads, trails, or trackways. In the field, we can walk along a stretch of an unconstructed path, describing its changing shape, drawing associations with artefacts or features nearby, and recognizing when it merges with the built environment, for example as a causeway. When this walked segment is recognized on digital imagery or terrain models, a process in itself affected by methodological and observational biases, its field-based description is then implicitly extended to the whole linear form observed from the aerial point of view.
This field-based description and its direct aerial-view correlate become the knowledge model used to recognize morphologically similar features in aerial imagery and terrain models as paths. Through this process, particularly when accelerated through machine learning applications, the generalization of field observations through extensive remote sensing datasets will reinforce existing ideas. This interpretive practice does not lend itself to insights into different systems of movement which do not align with our field experience, which is itself embedded in received ideas of what physical features related to movement should look like, ideas developed and reproduced through the literature and affected by semantic and contextual biases. In sum, through our practices of engaging with the literature, field observation, lidar-derived digital terrain models (LDTMs), and complementary remote sensing datasets, we have enhanced our ability to see more of what we already know.
The impact of these inter-related biases can be seen in projects that have generated spatially extensive but semantically narrow datasets. For example, projects working in Central America have identified extensive evidence for major Maya causeways, usually called sakbeh (white, constructed roads) [22]. This type of road is well known from the fieldwork, and its connection to the major cities of the Mayan world is well established [23,24], with key exemplars of the type well mapped and recognized as part of a network. Integration of this existing knowledge base with new observational data from lidar provided an extensive overview of this causeway system [25]. The knowledge gained from the lidar data, however, is only relevant to the major transport network defined by these features. It does not extend our knowledge of different types of movement in this landscape and the traces these might leave.
In another context, in eastern France, we encounter a similar situation. Thousands of features related to paths can be identified in the LDTM, based on ideas of paths drawn from fieldwork and literature. In this area, the physical manifestations of paths appear as a tangle, often lacking clear endpoints because they are only partly preserved or visible in the LDTM ( Figure 1). While we know the location and parts of the course of many more paths as the result of this digital survey work, insight into different types of movement and interconnections between different networks of features that make up this tangle remain elusive, as the impacts of methodological and observational biases result in an incomplete picture. The impact of these inter-related biases can be seen in projects that have generated spatially extensive but semantically narrow datasets. For example, projects working in Central America have identified extensive evidence for major Maya causeways, usually called sakbeh (white, constructed roads) [22]. This type of road is well known from the fieldwork, and its connection to the major cities of the Mayan world is well established [23,24], with key exemplars of the type well mapped and recognized as part of a network. Integration of this existing knowledge base with new observational data from lidar provided an extensive overview of this causeway system [25]. The knowledge gained from the lidar data, however, is only relevant to the major transport network defined by these features. It does not extend our knowledge of different types of movement in this landscape and the traces these might leave.
In another context, in eastern France, we encounter a similar situation. Thousands of features related to paths can be identified in the LDTM, based on ideas of paths drawn from fieldwork and literature. In this area, the physical manifestations of paths appear as a tangle, often lacking clear endpoints because they are only partly preserved or visible in the LDTM ( Figure 1). While we know the location and parts of the course of many more paths as the result of this digital survey work, insight into different types of movement and interconnections between different networks of features that make up this tangle remain elusive, as the impacts of methodological and observational biases result in an incomplete picture. 6.06 to 6.09 • E. Source: Lieppec-C. Fruchart [26]. Beyond the methodological and observational biases that impact our knowledge creation practices when drawing on observational data, we face semantic and contextual biases when developing data models, typologies, and classification systems, as we seek to organize and make sense out of the tangle of features identified and sought in the observational data. While some methodological biases can be accounted for in a straightforward way, for example, data resolution will have an easy to understand impact on our ability to recognize features smaller than a certain size, semantic biases can be more challenging to assess and understand.
Focusing on Semantic Biases: What Words to Use to Describe A Path
While reading many papers addressing the subject of movement we observed that archaeologists, whatever their field of expertise, employ various words to describe pathways, such as path, trail, road, corridor, causeway-canal, or causeway. The terminology adopted, however, is almost never substantiated and this use of implicit definitions makes comparisons between case studies or their integration difficult.
For example, in a cross-cultural Franco-English study the use of implicit definitions is particularly problematic because some terms which are used as translational matches are not true semantic equivalents. The word "chemin" in French is semantically equivalent to the English word "trail", whereas "sentier" would be more like "path". However, the English word "path" is usually directly translated into "chemin" creating a strong ambiguity. Beyond these problems of semantic mistranslation, depending on the socio-cultural context, the terms chosen may imply other meanings beyond that related to movement. The term beh in the Yucatec Maya language is a good example of this situation. According to Keller [22], it has both a literal and metaphorical meaning. It implies the idea of displacement in space but also in calendar time. It also implies the idea of work done and/or completed, as well as the ideas of life course, prosperity, and destiny. These additional meanings pose further problems for translation to a different cultural context.
In 2009, to address this problem of terminology, Timothy Earle [27] suggested a working typology to study paths, trails, and roads within a comparative perspective in order to examine their distribution and function. In his approach, Earle considers paths as "local trodden ways" with an "essentially individual" function, associated with daily tasks and with a low investment in their construction. A path is by nature ephemeral and "largely unrecognizable" in ploughed fields, for example. Paths are associated with local scale "household movements" and with logistical activities and daily use. Trails are more permanent routes "marked by repeated use, by signs such as blazes, cairns, and petroglyphs, and on maps". Contrary to paths, trails are "regional or long-distance routes". Their function is logistical and ceremonial and therefore "trail use is both individual and group oriented". Finally, roads are constructed routes which need labor investment and "chiefdom and state" political integration to be planned. Roads "formalize movement" and their functions are ceremonial and military. They are mostly regional in scale, with segments of long-distance routes sometimes also serving as local routes.
A close reading of this typology raises several problems. Earle, himself, recognized many of them and concluded that through his typology the "examination of variation in routes of movement shows the alternative ways that societies operate and change under contrasting sociopolitical and environmental conditions", and that the static model was not entirely satisfactory. He suggested, "that routine methods should be developed to describe and analyze them [routes of movement]" which would allow for greater diversity to be taken into account", and that the analysis of variation in routes "offers great insight into the essence of human societies and their evolution". Reviewing the difficulties posed by Earle's typology in detail provides a useful mechanism for highlighting broader issues of semantic bias.
Earle's typology implies that under the same very generic terms we will find a set of material and physical traces that are morphologically relatively homogeneous. However, in both his own study and more broadly, we contend that field observations provide a more complex picture from the outset, and the distinction between road and path is not always clear. For example, Hyslop [28] reminds us Information 2020, 11, 338 6 of 34 that for the Inka world the same route can vary from a formally constructed road to a simple path over only a few kilometers. The same type of observation about the variable physical form of a road was made regarding "Caesar's Causeway" in France during excavations carried out along its route [29].
His typology also illustrates the implicit links that can arise between typologies and models of social structures. Earle's typology is embedded in a strong model of social hierarchy which associates forms of movement infrastructure with stratified levels of social complexity. A level of social and political organization is attached to each term and its associated set of morphological evidence, as is a specific geographical scale that determines the structure of the traffic system. Therefore, the definition of a physical feature using one of these terms implicitly gives it the particular function and status associated with that term. This association between a term, function, and status is a common semantic bias problem. By associating morphological entities with specific functions, scales, and social hierarchy levels which are all associated with a single word, we reduce our ability to grasp the social complexity of movement.
Even with full knowledge of the complications and nuance implicit in the words used in the definition of any class or type, and when only using a working typology or informal schema, these problems are present. Further, in a classification system with single, fixed semantic categories, wherein a route can only be classified as a road and not as both a road and a path, the infrastructure created through and enabling movement is thought of as single-purpose, rather than as involved in several movement-related activities. The use of systems or working typologies that assign features to a single class or single type might be appropriate for discussing the components of an idealized communication network that has been planned and organized into hierarchical levels. However, they are less effective for describing the structure that supports the movement processes of individuals or groups of individuals as they renegotiate their journeys depending on their activities, social rules and technical capabilities, and on the landscape's opportunities to move around [30,31]. For example, in applying Earle's working typology, which assigns features to single classes, a team might assign all the diverse constituent features which are recombined ad hoc to form diverse types of paths and are associated with multiple types of movement to the a single class of "path", and in doing this they implicitly reinforce a hierarchical view of systems of movement.
When considering change over time, the assignment to single categories is again problematic because a simple path can become a road and conversely a road can "degenerate" into a path due to lack of maintenance or change of purpose [32]. Returning to the embeddedness of typologies in general models of social structures, we note that behind the distinction between road and path there is usually an evolutionary conception of the world and an implied equation of growth and increase in size or material consumption with progress, with roads being associated with a certain, increased degree of societal complexity [33]. Therefore, as soon as we order the material evidence into classes of pathways, we implicitly include a hierarchy in their function as well as in their spatial and temporal scale. Through these various mechanisms, the act of assigning features to classes as part of the interpretive process and the selection of a specific set of classes to which features can be assigned tends to reinforce broader implicit ideas about how processes and activities in the past took place and delimits the realm of possible understandings of the character of the past societies and landscapes in question.
Things, Words, and Concepts: Methodology
How can we qualify the material and physical traces observed and identified as potentially used for displacement without getting lost in a descriptive typology that compartmentalizes complex reasoning into tidy morphologically homogeneous sets? How can we deal with the "variation of itineraries" based on a set of morphologically heterogeneous material traces, while at the same time giving them meaning? How can we integrate this variation into our approach to identifying and interpreting observational data so as not to reduce our field of knowledge to what we already know how to recognize? Finally, how can we demonstrate the relationships we identify between these observations of heterogeneous material traces, themselves subject to the biases discussed above, and the whole of our archaeological, historical, anthropological and theoretical knowledge?
Attempting to respond to these questions, our approach involves focusing on properties in terms of physical topographic characteristics and basic functions (e.g., depression, alignment, embankment), rather than directly applying the terminology of higher order categories (e.g., hollow ways) and culturally specific categories of objects (e.g., sakbeh in the Maya world: Elevated causeways linking temples, ceremonial centers, or cities). By decomposing the broad concept of "movement" as it is used in archaeology [31,[34][35][36] into more specific constituent concepts and by building explicit links to interpretations of how these concepts are expressed in observational data, we aim to develop a common conceptual framework, the so-called "track graph". This common framework can be used to design and connect various explicit chains of inference related to the analysis of pathways and past movement flows.
The Implications of the Semiotic Triangle for Our Approach
We have seen that the diversity of observed features which enable movement leads either to a multiplication of descriptive terms or to a narrowing of approved terminology, motivated by a classificatory or typological approach. In the first case, the use of overly specific terms makes it difficult to identify connections between entities that effectively serve the same role in systems of movement, while in the second case features with similar morphologies may be grouped together, creating false connections between entities that represent different kinds of movement. In both cases, it is difficult to compare different case studies or even different systems of movement within the same case study.
In order to avoid these dead ends, we have relied on the semiotic triangle designed by C.K. Ogden and I.A. Richards [37]. For these authors, who worked on language, the misunderstanding between two people who use the same term lies in the fact that we often confuse the name (symbol) of an object and this object (referent) itself (the base of the triangle in Figure 2). Starting from the principle that the meaning of a word is determined by the lived experience of the speaker, they reject any idea of canonical or standardized meaning. This is why, in order to clarify the relationship usually implicitly established between a word and an object, they define what they called the domain of reference or thought (the top vertex of the triangle). The field of reference is what we refer to as a concept, that is to say, the body of empirical and theoretical knowledge that allows us to conceive what a thing is-in our case, a material or physical entity dedicated to movement-and to link it to a word that designates and defines it as such-in our case, the terminology associated with pathways in general. Thus, for example, in the Pueblo world, the material traces of a staircase observed carved into a cliff, of deep ruts, and of scattered ceramics on a line, could all be grouped together under the same term "regional trail". In the same way, the material traces of a staircase observed carved into a cliff could be associated with the term "trail" and the term "place" [38]. Therefore, it is the relationship with the concept, or field of reference, that will give meaning to the identification of a material or physical entity as a road, for example.
Based on this theory, our approach therefore consists of analyzing the way in which material characteristics identified in the observed world (terrain, imagery, LDTM, etc.) are related to terms connected to the process of movement (trail, path, road, trackways, holloways, causeways, canals, riverpaths, roads, local trodden pathways, etc.). Our goal is to develop a framework for describing and linking practices of movement, topographic morphology and other material evidence, and the historical and environmental contexts of pathways and their trajectories. This framework should include all the key elements of the domain and explain their relationships in order to reconstruct one or several logical assemblages that constitute a pathway, an intersection, or even a network or a meshwork [31]. Based on this theory, our approach therefore consists of analyzing the way in which material characteristics identified in the observed world (terrain, imagery, LDTM, etc.) are related to terms connected to the process of movement (trail, path, road, trackways, holloways, causeways, canals, riverpaths, roads, local trodden pathways, etc.). Our goal is to develop a framework for describing and linking practices of movement, topographic morphology and other material evidence, and the historical and environmental contexts of pathways and their trajectories. This framework should include all the key elements of the domain and explain their relationships in order to reconstruct one or several logical assemblages that constitute a pathway, an intersection, or even a network or a meshwork [31].
An Ontological Approach to Structuring and Formalizing the Description
In analyzing cross-cultural case studies developed by different research teams, we adopt an ontological approach to address the issue of ambiguity identified above ( Figure 3). This has several advantages. First, it provides a structured, formal way of describing and relating the various terms and knowledge bases used. It also clarifies the structure of the available knowledge. Second, it allows us to consider the various relationships between the terms and concepts used: Hierarchical, topological, temporal, and spatial. In this way, we can create the metadata needed to describe and access meaningful information on movement, and strive towards interoperability, as defined under the FAIR principles [13].
An Ontological Approach to Structuring and Formalizing the Description
In analyzing cross-cultural case studies developed by different research teams, we adopt an ontological approach to address the issue of ambiguity identified above ( Figure 3). This has several advantages. First, it provides a structured, formal way of describing and relating the various terms and knowledge bases used. It also clarifies the structure of the available knowledge. Second, it allows us to consider the various relationships between the terms and concepts used: Hierarchical, topological, temporal, and spatial. In this way, we can create the metadata needed to describe and access meaningful information on movement, and strive towards interoperability, as defined under the FAIR principles [13].
Information 2020, 11, x FOR PEER REVIEW 9 of 35 Figure 3. Schema of the approach adopted by our team.
In practice, when developing their own study, a research community or group describes their chosen question about past movement using terms and ideas drawn from several published studies. The published studies serve as sources from which the research group can assemble a broad "world" of context-specific concepts used to address a single, more global, idea. These concepts are then
Domain ontologies
Track Graph CIDOC-CRM ontology Figure 3. Schema of the approach adopted by our team.
In practice, when developing their own study, a research community or group describes their chosen question about past movement using terms and ideas drawn from several published studies. The published studies serve as sources from which the research group can assemble a broad "world" of context-specific concepts used to address a single, more global, idea. These concepts are then organized, with the aim of connecting what was observed, which we think of as things, or nouns, with how we understand the activities that produce them, which we think of as verbs. To create an operational model of these conceptual domains, we proceeded in three steps: Analyzing articles for the language used to describe landscape features and actions of movement, producing a graph model of the language used, and formalizing this into an ontology. By drawing together and studying the language used in multiple case studies, we attempt to define a conceptual model that is implicitly broadly shared by the research community interested in the question, but which does not rely on a predefined model. Rather, domain and question specific ontological models are developed ad hoc to address specific research agendas.
While we do not seek to encourage the mechanization of the interpretation process, our approach relies on the formalization of concepts and alignment with shared community standards for describing knowledge domains. Therefore, we also make use of existing conceptual reference models such as the CIDOC CRM [39,40], which we aim to cross-map with our ontology's domain-specific entities and properties. This mapping to a shared community standard (currently in progress) aims to increase the usefulness of the specialized domain ontology, as it could be made interoperable with related systems which also make use of the CIDOC CRM.
We underline the creativity, discussion, and reflection emergent through the practice of creating these domain ontologies, which enables us to better understand and map the limitations and potential of our knowledge. In short, the practice of ontology creation and mapping is a heuristic tool that can help us in the process of interpretation. Importantly, the resulting ontology provides a formal framework that can be used to share data which remains flexible, and may be adapted by other research projects to their own contexts.
This heuristic aspect of the ontological approach is particularly useful for the development of a cross-cultural analytical framework. The linking of concepts drawn from the literature on different areas allows a research group to identify points of commonality and differences between concepts of movement as discussed in diverse regions. The way concepts are linked provides a common definition for assemblages of various morphological, temporal, spatial, and functional characteristics, which are not necessarily the same across different geographical areas or even within the whole of the geographic area where they appear.
Formalizing these mappings into a UML schema to visually represent the structure of the knowledge domain precedes further formalization within an ontology editor (such as Protégé: https://protegewiki.stanford.edu/wiki/Main_Page). Using a logical language such as OWL (https://www. w3.org/2001/sw/wiki/OWL) allows for the visualization of concept hierarchies and the development of correspondences with other conceptual references, such as the CIDOC CRM.
To illustrate our approach, we present three case studies from the literature for which we modeled the domain of movement concepts as a UML schema represented as a graph (produced with software such as yEd; https://www.yworks.com/products/yed). Clearly the definitions of the concepts and their relationships may be debated, and it is these points of convergence and divergence of our views on these concepts and relationships relating to movement which become the basis for further study and debate.
Three Case Studies
The analysis of three cross-cultural case studies, used to illustrate our approach, are based on five papers treating regions in Mesoamerica (Caracol, Belize: [24,25]), South America (Bolivian Amazon: [41,42]), and north-western Europe (Beauce, France: [43]) ( Figure 4). All three case studies have recorded material and physical traces linked to movement processes using various sensors and protocols (lidar, remote sensing imagery, field work, historical maps and documents, as well as ethnographic testimony) (Table A1).
further study and debate.
Three Case Studies
The analysis of three cross-cultural case studies, used to illustrate our approach, are based on five papers treating regions in Mesoamerica (Caracol, Belize: [24,25]), South America (Bolivian Amazon: [41,42]), and north-western Europe (Beauce, France: [43]) ( Figure 4). All three case studies have recorded material and physical traces linked to movement processes using various sensors and protocols (lidar, remote sensing imagery, field work, historical maps and documents, as well as ethnographic testimony) (Table A1). In the case of Caracol, we engage with a territory characterized by a readily identified network of built causeways to enable travel between centers, but where the means of mobility from isolated residences (household groups) is less apparent and does not seem to be associated with extensive built features or terrain modifications. The relationship explored here, between sets of features which enable movement between centers, creating connections at a regional scale, and those operating locally, is also explored through the other two case studies which focus on the pre-Columbian period in Bolivia and the Medieval period in north-western Europe. The aim is not a direct comparison of these case studies, but rather to observe in each one how a formal communication network can be articulated with the patterns of movement developed by the population through everyday repetitive activities across the landscape. How do people move over the landscape to join residential centers and resource areas, or to carry out agricultural practices? How do formal road networks influence these daily trajectories? How does the articulation between these two systems produce a structure that imprints its mark on the landscape and influences channels of movement?
Caracol Region (Belize)
In Caracol, numerous pathways were discovered through the analysis of the lidar-derived DTM. The resulting collection of new causeway segments adds to the dataset compiled from previous observations during fieldwork and from satellite imagery based surveys [24]. The lidar based study revealed a large part of the main road system linking the major epicenter of Caracol to other places, called termini, understood as secondary centers which may be residential [25]. The LDTM shows a largely continuous swath of settlement and terracing that exists between the epicenter and these termini. This extensive urbanized landscape area is characterized by small groups of houses organized around small squares, which are integrated into the agrarian landscape, a pattern recognized in other Mayan regions. These settlement units are not directly connected to the epicenter or to the secondary centers by causeways [44]. The density of these areas remains relatively low, estimated at 4 to 10 units per hectare after the lidar survey [26]. This low density and lack of causeway or road infrastructure creates the impression that each small residential group (so-called household groups, or plazuela groups), while forming part of the urban fabric, is relatively isolated from the region's most important residential centers. However, it is plausible that features within the agricultural field system provided strong connecting links between these dwellings and the main residential centers [44]. These are referred to in the Caracol case study as vias: "more informal and shortest roads" [24] which connected household groups to major causeways or joined "important residential groups directly with various non-residential termini". The authors note that in Caracol, "many causeways ( . . . ) are intermixed with agricultural terraces" and that the vias can be primarily identified through a detailed mapping of these terraces. While, "only a half-dozen instances of vias have been formally noted attaching to Caracol's longer causeways," they emphasize that, "detailed mapping of the terraces would undoubtedly turn up other examples" [24]. Since the identified vias appear to provide direct access to the major causeways, the authors suggest that these causeways were partly used for everyday purposes and to facilitate movement and communication of the household groups. This example illustrates how, behind the more obvious and formal network of causeways, there is a meshwork drawn through daily practices of household groups going into the fields, to residential centers, and to other locations. To do so, individuals used various types of features to enable their quotidian routes or trajectories, including formal causeway segments and agricultural terraces features such as narrow pathways. Angela H. Keller's study of Yucatec Mayan language [22] highlights the use of a single root beh connecting terms for diverse types of paths and roads, supporting the argument that multiple physical types of routes were recognized by this community.
We suggest simplifying the description of the Caracol Mayan communication system in the following form ( Figure 5), describing a limited, artificially self-contained system which would have operated within a larger settlement system. Each component of this landscape, described in the UML schema, is connected by a path "network" composed of very formal causeways called sakbehs and more informal paths called vias. Vias may be inter-mixed with agricultural terraces and may be connected to the path network through junctions with the major sakbehs. These causeways connect the epicenter's residential group to other residential termini and in some cases through junctions to non-residential termini (reservoirs, agricultural terraces, or other features). The household groups have no clear position in this case study. It is not obvious how they are connected to their fields (agricultural terraces), to the path network and, by extension, to the residential groups (epicenter and termini), and to the non-residential places such as reservoirs. It is possible that they are connected to the main path network using vias and some part of the sakbeh causeways segments.
Baures and Llanos de Mojos (Bolivian Amazon)
In two chapters of the book "Landscapes of movement", edited by Snead et al. [45], Clark L. Erickson and John H. Walker discuss two case studies in the Bolivian Amazon during the pre-Columbian period [41,42]. Both regions are characterized by a marshy savannah landscape that was topographically transformed over several generations. A major investment of labor was made in order to construct earthworks, not to prevent inundation but to create an expanded and productive wetland. A region composed of savannah, forest areas, and rivers was progressively settled by the Moxos (or Mojos) and transformed into an anthropogenic landscape with settlement mounds, raised fields (farming areas), canals and raised roads/causeways (Figure 6), reservoirs and fish weirs. This anthropogenic landscape is discussed within the model of "landesque capital" by the authors, who propose that this cultural landscape was, "created, used and maintained by small farming communities over hundreds of generations". the epicenter's residential group to other residential termini and in some cases through junctions to non-residential termini (reservoirs, agricultural terraces, or other features). The household groups have no clear position in this case study. It is not obvious how they are connected to their fields (agricultural terraces), to the path network and, by extension, to the residential groups (epicenter and termini), and to the non-residential places such as reservoirs. It is possible that they are connected to the main path network using vias and some part of the sakbeh causeways segments.
Baures and Llanos de Mojos (Bolivian Amazon)
In two chapters of the book "Landscapes of movement", edited by Snead et al. [45], Clark L. Erickson and John H. Walker discuss two case studies in the Bolivian Amazon during the pre-Columbian period [41,42]. Both regions are characterized by a marshy savannah landscape that was topographically transformed over several generations. A major investment of labor was made in order to construct earthworks, not to prevent inundation but to create an expanded and productive wetland. A region composed of savannah, forest areas, and rivers was progressively settled by the Moxos (or Mojos) and transformed into an anthropogenic landscape with settlement mounds, raised fields (farming areas), canals and raised roads/causeways (Figure 6), reservoirs and fish weirs. This anthropogenic landscape is discussed within the model of "landesque capital" by the authors, who propose that this cultural landscape was, "created, used and maintained by small farming communities over hundreds of generations".
People moved through this highly modified landscape either on foot on causeways or by canoe in canals. According to the authors, "because of the intentionality, design, monumentality, and engineering used in their construction, causeways and canals are classified as formal roads rather People moved through this highly modified landscape either on foot on causeways or by canoe in canals. According to the authors, "because of the intentionality, design, monumentality, and engineering used in their construction, causeways and canals are classified as formal roads rather than informal trails or paths" [42] in this context. The case study of the Baures region (N. Bolivia) also discusses a less formal system with shorter causeway-canals produced with minimal planning or labor, and "minor causeway-canals". These shallow canals are interpreted by the authors as "pre-Columbian canoe paths". They are produced by the "repeated paddling, poling, or dragging a large canoe" creating "canal-like depressions" over time. These shallow depressions are used as canoe paths during the wet season and as routes for pedestrian traffic during the dry season. These systems of major and minor causeway-canals appear to play complementary roles in the flow of goods and people over the savannah landscapes to join forested areas, rivers, settlement mounds, raised fields, resource locations, or other causeways "forming physical networks of local and regional scale".
The schema below ( Figure 6) represents the main concepts and relationships used by the authors to describe the complex movement systems of the Bolivian Amazon case studies. Their descriptions reveal two types of pathways, pedestrian and navigable, combined within the same system of movement. The production of pathways takes place through three types of processes. The planned building produced major causeway-canals. Minor causeway-canals were created through the wearing-in of shallow depressions in the earth through repeated journeys in canoes using techniques of paddling, polling, or dragging. These features were used as canoe paths or pedestrian paths seasonally. The building of earthen "bunds" created embankments for agricultural purposes, which were used opportunistically as pedestrian pathways. Referring to the concept of habitus in the sense introduced by Mauss [46] and later promoted by Bourdieu [47,48], the authors suggest that this pre-Columbian movement system was formed by intertwined processes of collective and individual actions, containing features both intentionally constructed and features generated through daily activities, rather than by a political entity's planned projects. The material features connected to movement include diverse types of topographic features, supplementing the network of major causeways, joining settlements, farmed land, and other land use areas and features (forests, fishponds, etc.). The practices of movement by the population at a given period can be read through the assemblages of topographic elements created by that period's activities. These are structured regionally around the main built routes (major causeway-canals), and locally in arrangements related to daily activities.
wearing-in of shallow depressions in the earth through repeated journeys in canoes using techniques of paddling, polling, or dragging. These features were used as canoe paths or pedestrian paths seasonally. The building of earthen "bunds" created embankments for agricultural purposes, which were used opportunistically as pedestrian pathways. Referring to the concept of habitus in the sense introduced by Mauss [46] and later promoted by Bourdieu [47,48], the authors suggest that this pre-Columbian movement system was formed by intertwined processes of collective and individual actions, containing features both intentionally constructed and features generated through daily activities, rather than by a political entity's planned projects. The material features connected to movement include diverse types of topographic features, supplementing the network of major causeways, joining settlements, farmed land, and other land use areas and features (forests, fishponds, etc.). The practices of movement by the population at a given period can be read through the assemblages of topographic elements created by that period's activities. These are structured regionally around the main built routes (major causeway-canals), and locally in arrangements related to daily activities. Figure 6. UML schema of a circulation system in the Bolivian Amazon. Figure 6. UML schema of a circulation system in the Bolivian Amazon.
The Beauce Region (France)
In north-western Europe, in the Beauce region (France), an extensive settled agricultural landscape was maintained during the Medieval and Early Modern periods (12th-17th century AD), as discussed in detail in [13]. Based on the work of Samuel Leturcq [43], this study concentrated on the organization of the openfield landscape, specifically on the distribution of transportation routes across the agricultural areas. While historic maps revealed a dispersed local road network composed of "formal path roads" (Figure 7), in extensive continuous agricultural areas composed of a large number of contiguous small parcels with long strip shapes, no evidence of pathways is apparent which would provide a means for a peasant to reach plots of land enclosed within the field system. Written historical sources state that the borders of some fields can be used as paths by farmers to access their fields [43], providing a likely explanation for how they were reached. In this region, the practice of using field boundaries as paths is further evidenced by the term sommière, which is sometimes connected to the term for a path in the Medieval and Modern terrier, which is a register of lands belonging to a single landowner [43]. In England, in the written estate surveys and 16th-century Elizabethan maps of isolated farms and settlements of the All Souls College estate, the term "balk" is frequently used to refer to the grass access-way between two furlongs [49], a parallel concept. access their fields [43], providing a likely explanation for how they were reached. In this region, the practice of using field boundaries as paths is further evidenced by the term sommière, which is sometimes connected to the term for a path in the Medieval and Modern terrier, which is a register of lands belonging to a single landowner [43]. In England, in the written estate surveys and 16th-century Elizabethan maps of isolated farms and settlements of the All Souls College estate, the term "balk" is frequently used to refer to the grass access-way between two furlongs [49], a parallel concept. This is what we defined as an "access path" in our schema (Figure 8). How were these "access paths" produced and by whom? What archaeological traces did they leave in the present landscape? And how can we identify them? In the dialect language of Toury en Beauce's peasants, the terms sommière, sommier, or têtière refer to an "elevated portion of a field" located at the junction between two fields, otherwise called a headland ridge [50]. This topographic feature, potentially used as a pedestrian pathway by farmers, was not built to enable movement, but rather results from a specific ploughing technique employed in combination with particular social rules. This is what we defined as an "access path" in our schema (Figure 8). How were these "access paths" produced and by whom? What archaeological traces did they leave in the present landscape? And how can we identify them? In the dialect language of Toury en Beauce's peasants, the terms sommière, sommier, or têtière refer to an "elevated portion of a field" located at the junction between two fields, otherwise called a headland ridge [50]. This topographic feature, potentially used as a pedestrian pathway by farmers, was not built to enable movement, but rather results from a specific ploughing technique employed in combination with particular social rules. The schema above (Figure 8) summarizes the process of headland ridge production during the extended period in which farmers were using a style of turnplough or a plough-tail (a type of ardplough) that is held at an angle while ploughing in order to push the earth to one side and then the other [51]. Rows were ploughed in straight lines, and the ploughs were turned at the boundaries of The schema above (Figure 8) summarizes the process of headland ridge production during the extended period in which farmers were using a style of turnplough or a plough-tail (a type of ard-plough) that is held at an angle while ploughing in order to push the earth to one side and then the other [51]. Rows were ploughed in straight lines, and the ploughs were turned at the boundaries of the strip field ("parcel" on the graph, Figure 8). As the land was tilled, the soil was moved and progressively accumulated on the edge of the field. Over the years, these accumulations of soil at the edge of each field developed into raised beds, or headlands, which can still be recognized in Beauce, and in many other regions [49,[52][53][54] (Figure 9). The visible boundaries of the community's cultivated lands are, consequently, materialized by the headland ridges. In cases where the direction of ploughing is regulated by the community, as is common in Medieval north-western Europe, the headland will gradually become a continuous border along numerous contiguous fields, creating connected field borders within a block of fields, or an "aggregate district" [55]. The accumulated evidence suggests that the headland-as-border developed into a structure that facilitated movement between fields and settlements including villages and isolated farms. While when fields were empty farmers could walk across the field, the headland was consistently available for use as a path, providing access when the fields contained growing crops or under other restrictive conditions defined by the village community, without having been intentionally created for this purpose. Even where we may lack textual evidence, we can infer that headland ridges likely supported the movement of farmers elsewhere, for example in England [49], when crossing fields would have been detrimental to crop growth. Combining the formal road system mapped on historical documents with the headlands, we obtain a more complete picture of a flow of movement through the network of rural communities. It is only by bringing together the observational data created through the study of remote sensing data and maps and the knowledge of movement practices derived from the literature that we can start to fully map and understand the actual movement patterns. In our UML schema, the components representing physical features are linked by processes. For example, the farmer can access parcels using an "access path". This could be either by headlands used as a pathway or by another, more formalized path such as a series of roads or routes depicted on a map.
In this case study, knowledge is structured in related classes in the UML schema, which allows us to connect important details, such as the workings of the turn plough, to a more general view of the landscape. Structuring knowledge and expressing it through this type of ontological description allows us to reconnect elements of an analysis to a body of knowledge and to transfer conceptual frameworks between case studies more easily. As a result of this exercise, what we called the "access path" is now better defined and this allows us to better model the local movement network of the farmers. We gain information on its morphology, its relation to social and landscape organization, and its use and temporality, linked to the rules of the agrarian community. The same ontology provides a framework for an informed interpretation of the results of either a visual interpretative survey or machine learning exercise. The abstract entity referred to as the "access path" enables us to integrate into a single functional term diverse physical structures which enable movement such as Combining the formal road system mapped on historical documents with the headlands, we obtain a more complete picture of a flow of movement through the network of rural communities. It is only by bringing together the observational data created through the study of remote sensing data and maps and the knowledge of movement practices derived from the literature that we can start to fully map and understand the actual movement patterns. In our UML schema, the components representing physical features are linked by processes. For example, the farmer can access parcels using an "access path". This could be either by headlands used as a pathway or by another, more formalized path such as a series of roads or routes depicted on a map.
In this case study, knowledge is structured in related classes in the UML schema, which allows us to connect important details, such as the workings of the turn plough, to a more general view of the landscape. Structuring knowledge and expressing it through this type of ontological description allows us to reconnect elements of an analysis to a body of knowledge and to transfer conceptual frameworks between case studies more easily. As a result of this exercise, what we called the "access path" is now better defined and this allows us to better model the local movement network of the farmers. We gain information on its morphology, its relation to social and landscape organization, and its use and temporality, linked to the rules of the agrarian community. The same ontology provides a framework for an informed interpretation of the results of either a visual interpretative survey or machine learning exercise. The abstract entity referred to as the "access path" enables us to integrate into a single functional term diverse physical structures which enable movement such as stairs, terraces, causeways, or minor causeway-canals.
Connecting Case Studies and Finding Commonalities
All the case studies presented here combine formal path systems and informal movement systems. They all illustrate a high degree of heterogeneity among the material and physical features related to movement through the landscape, e.g., built causeways, field boundaries, agricultural terraces, embankments, and formal and informal canals. Further, when there is no impetus or opportunity to construct a path, we see similarities in the way in which the landscape structures co-opted to enable movement are adapted and used. This ad hoc use of landscape features to move around the landscape creates a sort of "open-work fabric of interlaced or knotted cord", or a meshwork [31] of informal paths. Sometimes these activities are regulated by the community, more or less explicitly, either by rules or repeated practices (habitus), and these repeated actions tend to progressively structure the mesh. This meshwork has a footprint in today's tangible world, characterized by material and physical features that reflect the movements of a population in order to carry out tasks, to access particular spaces in the environment, and, especially in the cases discussed here, for agricultural activities.
In each case study, we observe a co-existence of two important motivations, or logics, for movement which use many of the same physical features in the landscape. The first is driven by the need to get from one place to another via an organized and socially recognized circulation network. This is what we refer to as "formal routes" which can be defined "as tangible, physical evidence of a route of travel serving as a means of communication between points or activity areas" [56]. Their morphology can generally be characterized, they are usually constructed to some degree, and they are maintained through a purposeful investment of labor such as cleaning and/or repairing the roadbed or repairing road curbs. This set of formal routes has generally left its imprint in the present tangible landscape and in documents, historical and cartographic, and in oral tradition through songs and narratives. This logic of movement using formal routes is conceptualized as a "path framework system". This system can be materialized by a road, by a series of landscape markers, or by a narrative without associated physical features. It has contextual and temporal properties and it is consciously designed by a group of individuals, in other words by a society, which recognizes it.
The second is influenced primarily by the activities of individuals and motivated by a set of repeated practices, defined and progressively appropriated by a society as a whole (habitus). This set of movements also leaves an imprint in the current tangible landscape, but this imprint is composed of a heterogeneous set of material and physical features, which can vary according to the seasonal context, modes of displacement, or the tools used in the practice of an activity. This type of movement is more difficult to identify because it is not organized as a network with connected places but rather as a meshwork of intersecting routes [31]. Attempting to comprehend these trajectories through the meshwork requires us to mobilize our knowledge of social and economic practices, rules, norms, and customs. This knowledge, in turn, allows us to identify the imprint left in the landscape, especially its residual form in today's tangible landscape and the potential arrangements for articulation with components of the formal route network. This second logic of movement is defined within our approach as the "pathway system". Unlike the "path framework system", the "pathway system" is unconsciously designed by a group of individuals and its temporality is largely disconnected from the temporality of the physical features used to create it.
Finally, we find complex relationships between multiple physical features which are used together to enable movement in all our cases studies. These features may belong to different semantic classes or types, but they may be assembled into a coherent entity, structured according to their use for movement. At the landscape scale, the overall configuration of the features defining and structuring movement must be understood in two radically different but co-existing conceptual frameworks: networks and meshworks. Over the long term, the influences of these two frameworks co-evolve, leading to the complexities experienced in attempting to explain the origin of the features observed in the physical landscape in terms of movement. Thus, for example, a segment of a road, formally identified as a road in a well-defined network, may in fact be based on an agricultural structure used as a passageway, the usage of which prefigures the development of the road. Conversely, a section of road formally created as a road in a planned network may subsequently disappear and be replaced by a simple boundary of fields that is topographically recognizable and still used "unofficially" to move around.
Toward an Abstraction: The Track Graph
While, from a theoretical point of view, it is possible to present this complexity on the basis of well-described and relevant examples, it is much more difficult to envisage methods of analyzing it more globally and systematically on the basis of data sets that are, by definition, heterogeneous. How can these pathways composed of heterogeneous trace elements be identified? How to describe and organize the heterogeneous data on physical features and the information relating to movement in order to be able to analyze these pathways? How can all these pathways that structure the flow of movement in the landscape be identified? Attempting to meet this challenge, we propose the definition of a strictly abstract analytical concept based on graph theory which may be operationalized in terms of spatio-temporal analysis, the "track graph" referred to in the opening section of this paper.
Motivations and Conceptual Workflow
The identifications of the named entities and their relationships are used to define an abstract composite object called a "track graph". This corresponds to a construct that allows us to interpret movement based on the various features observed regardless of their temporality. The interest of this abstract object lies in its ability to structure data and knowledge to describe the system of movement as a set of potentialities, rather than a juxtaposition of incompatible systems, and in the opportunity it provides for a cross-cultural analysis of movement processes, while keeping the logical link between these processes and the specific features being interpreted in each case study. This is particularly relevant when considering how we generate knowledge through repeated observation and construction of hypotheses ( Figure 10). In the hermeneutic spiral [19,57,58], the stage of data analysis is crucial in formulating hypotheses and conceptual models. Pattern recognition and hypothesis formulation are often strongly dependent on specific analytical techniques, such as remote sensing, GIS, network analysis, statistics or simulation modeling, which presuppose a formalized structure for the observed data. Therefore, we propose the track graph as a data structure that will allow for a relatively wide range of analytical approaches, based on graph structures. The track graph collapses all (hypothetical) evidence of movement observed, present and past, into a set of nodes and edges that can be supplied with an unlimited array of attributes. These attributes can be associated to each element of the graph through logical reconstruction via ontologies that describe and structure the body of knowledge according to various world views.
From a methodological point of view, we can subdivide our conceptual workflow into three phases ( Figure 10). The first phase is observation from one or more data sources (fieldwork records, imagery, LDTM, maps, historical documents, etc.). The observers identify archaeological features in the available set of data using a shared or individual observation protocol. While this can induce biases (see part 1), we argue that even if they appear inconsistent in the first instance, these different observations can complement each other within a given study area. In order to identify characteristics, observers also pre-interpret them in relation to a particular problem or question, a more or less defined conceptual framework, and a body of knowledge on which their expertise is based. In doing so, they develop hypotheses and an initial conceptual model to drive their observations and the way they will record and interpret them.
formalized structure for the observed data. Therefore, we propose the track graph as a data structure that will allow for a relatively wide range of analytical approaches, based on graph structures. The track graph collapses all (hypothetical) evidence of movement observed, present and past, into a set of nodes and edges that can be supplied with an unlimited array of attributes. These attributes can be associated to each element of the graph through logical reconstruction via ontologies that describe and structure the body of knowledge according to various world views. From a methodological point of view, we can subdivide our conceptual workflow into three phases ( Figure 10). The first phase is observation from one or more data sources (fieldwork records, imagery, LDTM, maps, historical documents, etc.). The observers identify archaeological features in the available set of data using a shared or individual observation protocol. While this can induce biases (see part 1), we argue that even if they appear inconsistent in the first instance, these different The second phase consists of analyzing these observed datasets, considering the way they have been recognized and interpreted from the point of view of the movement process, in order to propose a schema of the concepts used and the links established by the observers to relate various concepts. This is precisely what we attempted to do by applying an ontological approach to each of the case studies analyzed in this article. This exercise makes it possible to select all the characteristics regardless of their name or morphology and to associate them in a generic domain of "potential path". For instance, agricultural terraces or headland ridges, which are not strictly defined as roads, are nevertheless attached to this generic domain of "potential path". In the cases studied, the features identified all have a spatial reference that is conventionally represented by lines and points. Within the same study area, these heterogeneous layers of information, grouped under the generic domain, can then be combined. Their cartographic generalization allows them to be represented in the form of an abstract geometric model composed of nodes and edges, which constitutes the track graph.
In the third phase, we can explore the logical reconstruction of plausible paths or even movement patterns, for a given time period. This relies on the track graph together with complementary knowledge and data. In the approach developed here, the track graph is a skeleton of plausible tracks, which define the "playing ground" within which we can explore different hypothetical models of movement practices (induced models) and interpretations using, for example, rule-based or agent-based simulation models. Based on several domains of knowledge ("routes", "networks", "trajectories") organized within different ontologies, we can produce various models of movement. Then, the track graph allows us to identify connections between objects that are coming from different observers and methodologies (see the discussion of bias in Section 1) related to these models of movement.
Subsequently, the abstract object represented by the track graph can be used to reconstruct several types of significant networks. These may be recontextualized a posteriori with temporal attributes, material expressions and socio-environmental conceptions of movement, and clearly defined by a new ontology based on spatial relationships within the graph's network and on a set of contextual knowledge. Following this approach, the same set of nodes and edges may be articulated differently depending on the model of mobility chosen. For example, we might compare the articulations produced by a regional transport road or a farmer's trajectory in the course of his daily activities. An edge, corresponding to an archaeological feature with a specific morphology, can then simultaneously be considered as a formal road segment participating in a planned network, or as part of a trajectory defined by the practice of a farmer's routine activity.
While these reconstructions remain hypothetical, they have the benefit of being based on all the archaeological evidence observed but being abstracted from the regionally specific frameworks implicated in their attributes. The link to these attributes is maintained by an identifier and geographical coordinates linking each segment to one or more observations with specific attributes and ontologies attached, thus making it possible to return to the initial data during the evaluation phase of the reconstruction.
The third step of this conceptual workflow is still a work in progress. In the remainder of this article, we will therefore focus in more detail on the concept of the "track graph", its construction, and its articulation with two other concepts coming from our observational ontologies: The "pathways system" and the "path framework system" (discussed in Section 4).
Track Graph Composition
We define the track graph as an abstract object, composed of a set of nodes and edges. This approach follows that of representations where geometric entities representing physical features are abstracted from their descriptive attributes, such as the approach formalized in the context of urban archaeology to explore the complexity of the urban fabric [59]. This permits us to clearly distinguish interpreted properties such as type or class, described in the attributes, from the abstract entity, described through the geometry.
A node may represent a place, such as a city, a marketplace, a single dwelling, a marker in the landscape, or an intersection between paths. Such an intersection could be recognized as a junction or crossroad, a semantically meaningful place, or simply as a crossing of two pathways without any specific semantic meaning, for example, the intersection between an animal trail and a hiking trail. The essential characteristic of a node is that individuals and groups can move between them and through them. In other words, a node in the track graph denotes a potentially meaningful place.
Movement itself takes place on the edges, representing features understood to serve as pathways. These can be assigned various attributes, describing their material manifestations or functions. Through time, intersections and places as well as "pathways" can appear, disappear, and reappear. The track graph records and accumulates each observed feature. This geometric graph is expended as observations are made ( Figure 11). An observation allows the creation of an edge or a set of edges and nodes, and the same edge can correspond to several observations (from various observers, and/or various sources, at different observation times or according to different protocols).
Edges and nodes are given an ID and spatial coordinates that maintain a link to the initial observations which are described by a set of attributes via an ontology. While some of these attributes are purely descriptive, e.g., size, height, width, length, materials, spectral signature, topological attributes (crossing, next to, etc.), and consequently will not change between different ontological schemas, other attributes, e.g., class or type, might change between schemas. This compartmentalized approach allows nodes and edges to have different attribute sets for each ontological schema and each set of observations. For example, a researcher visually interpreting a LDTM could identify and interpret a linear topographic anomaly as a part of a Roman road. The same feature could be identified and interpreted by a Medieval archaeologist as a field boundary or as a potential pathway for farmers. This type of interpretational controversy is quite common and the source of heated debates when a research team is investigating the development of a landscape [60]. Equally, studies combining multiple surveys for a single area highlight the complexities that arise when combining interpretations generated by multiple teams [61]. The track graph allows us to combine these diverse sets of observations, acting as a dynamic representation of observations, evolving as new ones are made. The track graph constitutes a shared abstract canvas or skeleton that can be used to recompose meaningful systems which can be associated with different worldviews and conceptions of movement.
The essential characteristic of a node is that individuals and groups can move between them and through them. In other words, a node in the track graph denotes a potentially meaningful place.
Movement itself takes place on the edges, representing features understood to serve as pathways. These can be assigned various attributes, describing their material manifestations or functions. Through time, intersections and places as well as "pathways" can appear, disappear, and reappear. The track graph records and accumulates each observed feature. This geometric graph is expended as observations are made ( Figure 11). An observation allows the creation of an edge or a set of edges and nodes, and the same edge can correspond to several observations (from various observers, and/or various sources, at different observation times or according to different protocols). Figure 11. The track graph creation process. Figure 11. The track graph creation process.
A single observation, for example a line identified on an aerial image, can also be represented as a set of edges connected by nodes. In this case, the nodes do not have the role of origin or destination but will only link edges to geometrically represent an entity, while some may contribute by representing other observations. This highly abstract structure has several advantages. One is that it allows nodes and edges to be treated as active or inactive. For example, in a set of edges and nodes that represent a Roman road, one of the nodes may also correspond to a junction that allows a farmer to join the road from a terrace edge passageway. In the first case, the node cannot be related to any semantic data, it is a simple graphical convention and is therefore inactive in the graph when it is analyzed as a network. In the second case, the same node is active since it corresponds to an element that has a semantic role, representing a junction of two or more edges.
In the track graph, the network of active places and pathways is only a subset of the total set of realized places and pathways. This graph structure provides a mechanism through which edges and nodes can be activated for modeling movement at specific points in time. A second interest of the track graph which arises from its abstract nature is the possibility to use it as an analytical framework that can be transposed to several transcultural case studies. This allows for the comparison of patterns of paths based on the same structure, while interpreting each of them using the ontology specific to the community under study.
Summing Up and Making Connections: Track Graphs, Pathway Systems, and Path Framework Systems
In this paper, we illustrated our approach to developing formal models of two types of logic which underpin processes related to movement: The "path framework system" and the "pathway system", by analyzing a body of literature about a landscape. Then, we explored how these formal models of movement can be leveraged in the interpretation of data from sources such as aerial imagery, LDTMs, field surveys, and historic maps, and how these observations can be integrated through a "track graph".
To summarize, the "path framework system" and "pathway system" provide complementary models of movement and are intended to be used together. The "path framework system" (Figure 12) is used to describe a path network designed by a society, or at least recognized as such by a group. This includes formal paths which can be used for transport or travel from an origin to a destination. In this conceptual framework, movement is essentially destination-oriented [62]. In developing territorial models, the "path framework system" could be used to model how formal systems of movement contribute to the social, political, or cultural integration of the population. The "track graph" is a network graph in which each entity (node or edge) represents a physical feature in the landscape, and their physical and spatial connections are represented as connections in the graph. The "track graph" links the two models of movement in a single analytical framework provided by its abstract representation of features in the physical landscape. Unlike the "path framework system" and the "pathway system", the "track graph" does not contain any interpretations of what the features represented are used for or when they were in use. It only acts as an abstract support for both conceptual frameworks. Its structure, composed of edges and nodes, is simply a convention to facilitate analysis. In the articulation of the three concepts, the handling of time deserves careful attention. As explained, the "track graph" is a set of nodes and edges that covers the total of all observed and inferred potential paths and places, regardless of their chronological range. In this sense, the "track graph" itself historically and archaeologically atemporal, conflating features from all periods. At a given moment of observation (O1), the "track graph" structure records the totality of the realized or potential paths and can be used for exploring plausible paths, using a specific ontology describing movement behaviors and associated knowledge about a given space-time (Figure 11). New observations may be added to the "track graph", tagged with their observational moment (O2...On), capturing the development of the understanding of the landscape and movement in it through an iterative process of modeling, observation, and interpretation. The temporality of the "track graph" is related to the time at which the observations are made, at which we have arrived at a particular state of knowledge.
Unlike the "track graph", the "pathway system" and the "path framework system" are historically and archaeologically temporal. They are chronologically bounded and their dynamics reflect a dynamic that is historically meaningful. In archaeology, we are interested in analyzing patterns and Figure 12. Theoretical structure of the track graph, pathway system, and path framework system. The "pathway system" is used to describe the trajectory produced by one or more actors through the performance of their activities. This trajectory is essentially informal, although it may include components of the formal road system. In the "pathway system", the movement is activity oriented rather than destination oriented. The movement logic of the "pathway system" could be approached analogously to the wayfaring process, as developed by Ingold [31], but associated with the idea of habitus [47,48], which over the long term produces a collectively created imprint in the landscape.
The "track graph" is a network graph in which each entity (node or edge) represents a physical feature in the landscape, and their physical and spatial connections are represented as connections in the graph. The "track graph" links the two models of movement in a single analytical framework provided by its abstract representation of features in the physical landscape. Unlike the "path framework system" and the "pathway system", the "track graph" does not contain any interpretations of what the features represented are used for or when they were in use. It only acts as an abstract support for both conceptual frameworks. Its structure, composed of edges and nodes, is simply a convention to facilitate analysis.
In the articulation of the three concepts, the handling of time deserves careful attention. As explained, the "track graph" is a set of nodes and edges that covers the total of all observed and inferred potential paths and places, regardless of their chronological range. In this sense, the "track graph" itself historically and archaeologically atemporal, conflating features from all periods. At a given moment of observation (O 1 ), the "track graph" structure records the totality of the realized or potential paths and can be used for exploring plausible paths, using a specific ontology describing movement behaviors and associated knowledge about a given space-time ( Figure 11). New observations may be added to the "track graph", tagged with their observational moment (O 2 . . . O n ), capturing the development of the understanding of the landscape and movement in it through an iterative process of modeling, observation, and interpretation. The temporality of the "track graph" is related to the time at which the observations are made, at which we have arrived at a particular state of knowledge.
Unlike the "track graph", the "pathway system" and the "path framework system" are historically and archaeologically temporal. They are chronologically bounded and their dynamics reflect a dynamic that is historically meaningful. In archaeology, we are interested in analyzing patterns and dynamics within a time frame (T) with defined durations (from t 0 to t n ). Therefore, we can refer to different reconstructions of movement patterns (induced models) as being valid for a particular time frame, defining the nodes and edges that were present at t 0 , and defining which ones become active or inactive until the end date t n . Thus, within our time frame T, we can have reconstructions R 1 , . . . ,R n .
Then, modeling techniques can be used to simulate movement and establish routes within the track graph structure. This is, for example, what happens in an application such as ORBIS (http://orbis.stanford.edu/), where a network of Roman roads and cities is used to explore different travel routes within the Roman Empire. Within this static network, which forms a typical example of a path framework system, routes can be defined as subsets of nodes and edges that are connected for the purpose of a single, individual journey. These routes can be short or long, straight or circuitous, and can be connected to and nested in other routes. Simulations similar to these can explore various manifestations of one or more movement behaviors at a single point in time, for example focusing on understanding the consequences of uncertainties in data attributes [63][64][65] to assess the plausibility of routes.
Next, we consider the role of nodes in the "track graph", and how they are used in modeling the trajectories of journeys made within the framework defined by the "pathway system". The concept of the "pathway system" is rooted in the paradigm of the meshwork, a theoretical framework radically different from that of a network, and consequently any modeling within this meshwork-based paradigm requires a fundamentally different approach [31]. In a meshwork, we focus on the people who move through the landscape and how their practices of movement cause "knots" to emerge. These "knots" are defined as places where their journey's trajectories intertwine (an interweaving of lines in Ingold's language). Visually, these are places where the physical features associated with movement intersect, but which have no semantic meaning for the actors involved in their journeys as they move along these pathways. An example of this kind of knot is the intersection between a paved road and the route taken by a roe or a wild boar moving across the landscape (Figure 13). emerge. These "knots" are defined as places where their journey's trajectories intertwine (an interweaving of lines in Ingold's language). Visually, these are places where the physical features associated with movement intersect, but which have no semantic meaning for the actors involved in their journeys as they move along these pathways. An example of this kind of knot is the intersection between a paved road and the route taken by a roe or a wild boar moving across the landscape ( Figure 13). In contrast to the situation in the network paradigm where each node is a point of connection, the knot of the meshwork paradigm is not, a priori, making a meaningful connection between routes. In order to articulate these two conceptions, the network and the meshwork, in a formal and operational model which can be used in graph-based analyses and calculations, we have chosen to treat these knots as nodes within the "track graph". This is a pragmatic decision taken because when constructing a graph in most current software systems it is a technical necessity that the nodes are present as fixed features, rather than being dynamically generated during the running of an analysis. To reconstruct the plausible paths or the circulation patterns which might have been used by actors moving around the "track graph" at a given moment or period of time, we use attributes to assign a node, acting as an entirely abstract element on the graph, the function of a "connector" (active nodes) or a "knot" (inactive nodes). This approach semantically separates the presence of a node in the graph from its usual function as a connector.
To explore this idea, consider a research exercise in which a team models movement through the landscape at two different moments in time. In the first moment, the junction between a farmers' Figure 13. The intersection between a paved road and a trackway made by a wild animal near Pugey (France), source L. Nuninger. In contrast to the situation in the network paradigm where each node is a point of connection, the knot of the meshwork paradigm is not, a priori, making a meaningful connection between routes. In order to articulate these two conceptions, the network and the meshwork, in a formal and operational model which can be used in graph-based analyses and calculations, we have chosen to treat these knots as nodes within the "track graph". This is a pragmatic decision taken because when constructing a graph in most current software systems it is a technical necessity that the nodes are present as fixed features, rather than being dynamically generated during the running of an analysis. To reconstruct the plausible paths or the circulation patterns which might have been used by actors moving around the "track graph" at a given moment or period of time, we use attributes to assign a node, acting as an entirely abstract element on the graph, the function of a "connector" (active nodes) or a "knot" (inactive nodes). This approach semantically separates the presence of a node in the graph from its usual function as a connector.
To explore this idea, consider a research exercise in which a team models movement through the landscape at two different moments in time. In the first moment, the junction between a farmers' path crossing a road has a specific meaning for farmers and is recognized and marked by a group of farmers as a crossroads. Therefore, it is an active node having the function of a connector in the track graph when running a model simulating movement through the system. In the second moment, the formal road exists, but there is no place marked by a group as an official crossroads. In this second moment, although the direction of the farmer's travel may change, turning from the farm path onto the road, and this may be represented in the simulation as a change of direction between two edges geometrically separated by a node in the "track graph", this node will have an inactive status because it is not acting as a connector because it is not a "crossroads"-a recognized destination or otherwise meaningful feature in the landscape. The assignment of "active connector" or "inactive knot" attributes to nodes on a graph provides a mechanism through which we can attempt to implement network analysis and modeling approaches dependent on graphs within the conceptual model of a meshwork.
Through dynamic simulation, using an agent-based modeling (ABM) for example, a large set of reconstructed individual trajectories can emerge, using various combinations of the track graph elements, which change at each iteration. The trajectories emerging at t i , based on the modeled behavior of the agents, and related to the pathway system, will lead to the creation of a specific set of "pathways" encapsulated in a set of edges and nodes within the track graph. These sets will have an effect on modeled trajectories in the next iteration at t i +1, because the existence of pathways influences the beliefs and knowledge of subsequent groups of agents. This provides a mechanism for the simulation to drive changes in the attributes of nodes and edges in the "track graph" (e.g., "crossroads", "path") from iteration to iteration. Models such as these can be used to validate specific hypotheses, for example if observed formal "path framework systems" could have served other purposes, or if certain movement practices imply or preclude the combined use of path framework and pathway systems.
Some Conclusions and Implications of This Approach for Practices of Archaeological Knowledge Creation in the Contemporary Context
In this paper, we presented an approach to archaeologically studying the diverse expressions in the physical landscape of phenomena, such as movement, through a process of semantic modeling of domain literature and observation-based interpretation using fieldwork and remote sensing data. We focused on movement in three contexts, highlighting the variability in archaeologically recognizable physical evidence for movement and in the concepts and language used to describe movement and its infrastructure. The increasing use of extensive remote sensing datasets and, in particular, the uptake of machine learning to scale up identifications of archaeological features in the landscape motivated the development of this approach, which aims to guard against uncritically reinforcing standardized and strongly codified ideas about how complex phenomena appear in these data.
The impact of the current step-change in the scale of available archaeological data on our approaches to interpretation and recording-our practices of knowledge creation-echoes the impact of the vast increase in the amount of archaeological data generated through fieldwork associated with the expansion of development-led archaeology. While the study of relatively small scale artefactual collections and research-led fieldwork, designed to support the observation of material traces of past human behavior through survey and excavation, dominated much archaeological knowledge creation in the first part of the 20th century, the importance of development-led archaeology increased in the late 20th c., accelerated in the 1990s by the new legislation [66]. This shift in the context of the production of archaeological information, from one primarily constructed around the interests and practices of individual researchers and institutions, and dependent on limited dedicated funding, to one based on the needs and practices of market-driven heritage management, with total funding at a much greater scale, which today continues to increase the scale of archaeological work, led to the proposal and development of new approaches to fieldwork, data collection, and data management [67][68][69]. Adding to the collection of new data from excavation and survey, diverse scientific techniques, from isotope analysis, to micromorphology, to aDNA are increasingly applied in connection with archaeological fieldwork [6,70]. These each generate bodies of observational and metric data as well as interpretations, following their own specific standards.
The myriad problems of strongly codified recording and reporting norms associated with the professionalization of archaeology and the attempts of practitioners to deal with their burgeoning data in the late 20th c. have been discussed at length, particularly in the context of excavation [8,40]. The standardization of survey recording and reporting practice, motivated by its increasing use for creating archaeological inventories, meeting regulatory requirements, and other forms of heritage management [71,72], is similarly recognized as problematic. Despite these widely acknowledged problems, in practice the standardization of recording and reporting for both excavation and survey has increased. This latest step-change is propelled by improved methods and decreasing costs of collecting extensive observational data, notably through remote sensing methods including satellite imagery, UAV-based sensing, and geophysical prospection. While these datasets have been large by archaeological standards for some time, their scale has grown exponentially in recent years. More importantly, while until recently the archaeological use of these data resources has been constrained by the pace of manual interpretation, improvements in machine learning and automated feature detection, notably since the 2010s (see, e.g., [4,5,73]), are speeding their interpretation.
At present (2020), the imperative for integrating this morass of digital information to produce coherent, compelling, data-embedded archaeological narratives is frequently argued in connection to the archaeology's ability to contribute to debates on societal, climate, and environmental issues [3]. This emerging drive to re-articulate archaeological data to address contemporary agendas has implications for the development of new practices of archaeological knowledge creation. There is great potential to re-articulate archival data and synthesis of past research to play an important role in these debates, and to bring to bear the information created through the interpretation of large-scale remote sensing data. However, we must do so thoughtfully. The challenges of re-reading and re-interpreting the records, reports, syntheses, and analyses which emerged from evolving fieldwork practices and contextual understandings, as discussed in the debates over standardization and as illustrated throughout this paper, are substantial. The challenges of the interpretation of remote sensing data, as discussed here, are similarly daunting.
Conclusions
This paper illustrates one example of how we might combine and re-articulate the information, as well as ideas produced through working with observational data and synthesis of reports and research literature. We set out to investigate the influence of the use of context-and observer-specific terminology on the study of past movement processes and pathway patterns based on observed features. Our analysis of the terms and interpretive frameworks used to describe pathway systems in three different case studies revealed a discrepancy between how pathways, recognized through fieldwork and on digital imagery, are recognized and interpreted, and the conceptualization of the actual movement practices involved.
In all three case studies, we noted that observed pathways can be the result of formal construction and movement practices (e.g., processional ways), as well as of prolonged informal movement practices that generate non-constructed features (e.g., canoe paths) or make use of features initially constructed for different purposes (e.g., terraces, embankments, or headlands). Movement itself can take place over all these different features, or even leave no observable trace. While the attention of archaeological observers often focuses on formal pathways, understanding a movement "system" of the past is only possible when we connect observations to knowledge of the different practices of movement and the processes of pathway generation, maintenance, and renewal. To refocus our collective attention, we need to consider the role of observer bias, not just in terms of methodologies employed, available data sources, or individual expertise, but bias rooted in how knowledge is expressed in natural language in specific knowledge domains in different cultural and linguistic contexts.
We demonstrated that by creating an ontology of movement practices based on text analysis we can attempt to disentangle, structure, and clarify some of the semantic biases involved in the practice of identifying and interpreting features in large observational datasets as carried out by archaeologists whose knowledge base is inevitably embedded in the literature representing the current state of disciplinary knowledge. In this context, developing ontologies can serve as a useful heuristic exercise, aiding in understanding the reasoning behind largely implicit frames of reference and inference, and supporting comparing diverse situations. The breakdown of implicit conceptual references into explicit, logical chains of reasoning which describe basic entities and their relationships enables the use of constituent elements to reconstruct, analyze, and compare practices, such as those related to movement, from the bottom up.
Then, we introduced the concept of the track graph as a possible analytical tool for exploring and comparing pathway systems and movement practices at multiple spatial and temporal scales. The track graph is defined as a set of nodes and edges representing all observed features in a study region that are related to movement. Based on graph theory, it offers possibilities for applying well-established analytical approaches such as network analysis and agent-based modeling. At the same time, it allows for a richer description and understanding of observed features related to movement through the conceptualization, modeling, and connecting of informal pathway systems, as exemplified by the concepts of wayfaring and meshworks.
While the practical application of this approach to new datasets is still on the horizon, we tried to illustrate the potential and necessity of synthesizing data from various sources using a formalized, but not standardized, approach based on ontological reasoning and basic graph theoretical concepts. We hope that this paper will provide an impetus for developing these concepts and tools further to meet the challenges posed for archaeological knowledge creation by current remote sensing data collection and interpretation practices.
Medieval and Early Modern Periods-French Beauce Region
Other sources historical text (mid-18th century Jesuit testimonies) historical texts and map: a plan of terriers which dates from 1696, associated to a terrier which is a register of lands belonging to a single landowner (about 350 declarations of farmers exploiting the land in 5500 field parcels) with in addition several other terriers which date from the 16th, 17th and 18th century, and a series of censiers (register for the tax-census payment) which date from 14th and 15th century Formal movement features -"Causeway heights at Caracol range from ground level to some 3 m above the surrounding terrain. In several cases, the sides of hills were cut away to form the causeway." -"Longer intrasite causeways connect the epicenter directly with non-residential causeway termini at distances ranging from 2.5 to 7.3 km from the Caracol epicenter" -2.5 m to max. 12 m wide: "Hatzcap Ceel, an additional 1.9 km east of Cahal Pichik and linked to that site by a 12-m-wide causeway, lies 9.2 km away from the Caracol epicenter" -"Major Causeways are highly visible as tree-lined features flanked on one or both sides by canals filled with dark aquatic vegetation, which stands out against the grass-covered savanna" -"Major Causeways range in width from 1 to 10 m and elevations vary from 0.5 to 3 m tall; Major Canals are comparable in dimensions. Most Major Causeways-Canals are straight and extend up to 7.5 km, although most are several kilometers long. Pedestrians used the elevated causeways and canoe traffic circulated in the adjacent canal(s)" The old road network was highly transformed, first in the 19th century with the construction of well-structured road network, then in the 1950s by land consolidation. From the Middle Ages to the 19th century, it was organised as follows: -A paved road about ten metres wide, called a 'paved path' in 17th century written sources. This route, which undoubtedly dates back to Antiquity, crosses the territory from North to South. It is the Paris-Orléans road, which is essential in the French network (now called "route nationale 20"). -Secondary network of 7 dirt roads roughly in a star pattern, used to link the village of Toury with neighbouring villages, hamlets and isolated farms. Local service network. Today, only about one third of these paths are still in use. | 2020-06-25T09:09:02.678Z | 2020-06-24T00:00:00.000 | {
"year": 2020,
"sha1": "87386803ea5df5aafbfa342b888ac8aa96a3ec99",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2078-2489/11/6/338/pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "f7cf1b9e925ade372a82d375234df5458347f234",
"s2fieldsofstudy": [
"History"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
203991495 | pes2o/s2orc | v3-fos-license | Barkhausen Noise Probes and Modelling: A Review
This review looks at the main types of magnetic Barkhausen noise (BN) probes that have been developed. The aim of this review is to summarize the existing knowledge of magnetic Barkhausen noise probes and the magnetic modelling of them. The BN probes have been the focus of many previous studies, but no sufficient review or conclusions have been made so far. This review focuses on combining information regarding the different types of BN probes and their modelling. The review is divided into two sections; in the first part the different designs and types of Barkhausen noise probes are introduced. The second part of the review deals with the BN probe modelling with various modelling software. Finally, a comparison of the experimental measurements is made and BN sensitivity is discussed.
Introduction
Barkhausen noise (BN) is a non-destructive magnetic measurement method, which has been used in many industrial applications for quality control and as an aid to verify process outcomes. The method is versatile because it is sensitive to changes in both stress state and microstructure. Huang and Qian [1] collected the latest advances in magnetic nondestructive testing and the application of this technique in their recent review article. The BN method has been used widely in grinding burn studies [2] of different ground (transmission) components and inspecting surface characteristics after different heat treatments [3]. Due to its stress sensitivity, BN is also used for detecting stresses in pipelines [4]. However, the BN method is not yet standardized; it is under development. Recently some guidelines for recommended practices have been published [2,5]. Traditional BN measurement systems usually consist of a measurement device, electronics, computer and a sensor or probe consisting of a magnetization coil and pickup coil. In the literature, both terms "sensor" and "probe" are utilized to refer to the magnetization and sensing unit for BN. In this paper, the term probe is used. As Durin and Zapperi [6] pointed out, there are no typical or standard designs, sizes and specifications for conventional BN probes. Deveci [7] stated that the variety of different possible probes for different applications is one of the main advantages of the BN method. In a simplified system for BN measurements, the magnetic circuit is formed by a copper coil wound on a core, which forms an electromagnet that is supplied by alternating current. The alternating magnetic field is then induced by the yoke to the studied surface. As White [8] summarizes, a large amplitude low frequency magnetization field is used to cycle the sample around the hysteresis loop. Cyclic magnetization is generated in the material, and the generated outcome of the irreversible domain wall motion is measured by the pickup coil as Barkhausen noise. The pickup coil detects the domain structure reconfigurations causing the changes in the magnetization as a response to the magnetization field [8]. The BN response expressed as voltage is induced in a pickup winding coil placed around the material sample or on its surface [9] according to Faraday's law of induction [4]. The electromagnetic field that is produced by the magnetizing yoke will decay exponentially into the depth in a direction perpendicular to the surface, due to eddy current damping. The electromagnetic skin depth will be affected by several material and measurement parameters, i.e., the frequency, conductivity and permeability of the material. The skin depth δ is given by the relation where f is the frequency, σ is the conductivity of the material, µ o is the permeability of vacuum and µ r is the relative permeability of the material [10].
Barkhausen noise is not produced uniformly throughout the magnetization cycle but is concentrated in two bursts of activity per cycle near the coercive field, whereas close to positive or negative hysteresis loop saturation the signal is minimal or zero [11]. This means that the BN probe needs to be fixed on the sample surface for at least half a magnetization cycle, when BN measurements are carried out. This traditional method of measuring Barkhausen noise can be considered a stationary BN technique. Rotating measurements are applied for components which are inspected by rotating the component under a stationary probe. For example, transmission components like crankshafts or camshafts can be inspected with either inline, automated or semi-automated systems [12]. Many versatile applications exist for BN measurements and, therefore, most applications require a customized probe design that accommodates the sample geometry. For flat surfaces with static measurements, the coupling between the electromagnetic yoke and the sample is similar in all cases and changes in the BN are due to variations in the sample's magnetic properties. Probe design is crucial for the operation of the whole measurement system. The choice of probe design parameters affect the generation of the magnetic fields that excite BN in the sample under study. In addition, the magnetic behaviour of the probe can be analysed by numerical simulation using the finite element method (FEM) to limit the need for experimental measurements.
Barkhausen Noise Probes
Traditionally, BN signals are measured with a probe or sensor, consisting of a magnetic excitation system, i.e., magnetizing yoke and pickup coil. The probe components are embedded normally in a case made of either aluminium [4] or stainless steel [7], where they are cast with epoxy inside the case [7]. To ensure the wear resistance of the probe, a wolfram carbide insert can be attached to the magnetizing pole pieces [7].
Materials and Geometry: Magnetization Unit
The magnetization unit can be placed in the same embedded case as the pickup unit or it can be external. The external magnetization unit can be used for example in a case of bearing measurements where the pickup is in a stationary location. As Moorthy [13] concluded, external magnetization can be carried out with an open loop solenoid or with a U-shaped or rectangular electromagnetic yoke with rounded areas in the upper parts of the magnetization legs. The C-shaped, U-shaped or rectangular yokes are more common and practical in industrial use than the solenoid. In the U-shape yoke, the copper coil can be wrapped in the upper area of the core. Prabhu Gaunkar et al. [14] studied the effect of core length on the magnetic field and concluded, based on their modelling, that the magnetic field strength decreases with increasing core length (path length from pickup to coil). Thus, in the magnetization unit, the copper coil for magnetizing has also been wrapped around both legs as demonstrated in Fig. 1a, based on Prabhu Gaunkar et al. [14][15][16], or wrapped around each of four magnetizing legs as in White's tetrapole probe [8]. However, for some special cases, for example in the magnetization of a wire sample, Capo-Sanchez et al. [17] reported the use of a solenoid with 700 turns of copper wire. Solenoid magnetization for a bar sample is demonstrated in Fig. 1b.
In some cases, a flux sensing coil or feedback coil for controlling the currents can also be placed in the magnetizing leg, as in the studies of White et al. [8,18] and Vengrinovich and Tsukerman [19]. In White's assembly, the magnetizing coil was wrapped around the upper part (head bow) of the magnetization yoke and a flux sensing coil was wrapped at Fig. 1 Magnetization unit with a copper coil wrapped around both legs, schematic image based on [15] and b wrapped around the sample, schematic image based on [17] the bottom of the leg, as demonstrated in Fig. 2. This position for the flux sensing coil was justified as it best represents the flux entering and leaving the studied sample [18].
The single-head or single-core layout has been introduced in Lo et al. [20], for example. Stupakov [21] utilized a special single yoke measurement set-up where a C-core was used for the magnetizing yoke with the driving copper coil positioned on the head bow and two induction coils wrapped around the legs close to the sample. The induction coil recorded the magnetic flux in the head-sample magnetic circuit. Another single-core layout for special measurements was introduced by Kazanci et al. [22,23], which was a miniaturized fixture for BN measurements for deep drilled holes. The material of the magnetizing yoke core and the pickup yoke core is usually ferrite, which is a ceramic homogeneous material composed of oxides, mainly iron oxide [24]. Ferrites are divided into two categories: Mn-Zn ferrites and Ni-Zn ferrites [24]. As Stupakov [21] concluded, the yoke core is usually chosen to be magnetically softer than the core of the sample in order to allow the magnetic field to penetrate easily and produce magnetization. In some special cases, a magnetizing yoke core made of iron has been studied [13,14,25]. The magnetization yoke in the studies of Moorthy [13] had a solid pure iron core. The permeability values depend on the composition and, in the case of iron, on the impurities, but generally ferrite has lower relative permeability values compared with iron [14]. Another important difference between the different core materials is the saturation magnetization value. A ferrite core yoke yields much lower saturation magnetization (300 mT) than an iron core yoke [25]. The lower saturation magnetization of the ferrite core may be one reason for the differences in the observed pickup signal with different magnetizing yokes seen in the studies of Vértesy et al. [25]. Prabhu Gaunkar et al. [14] verified, by means of modelling, that the magnetic flux densities vary with different core materials. They [14] noticed that, among the materials that they studied, the 78 Permalloy had the highest and the iron core had the second highest magnetic flux density values. Due to the lower saturation magnetization of Permalloy Prabhu Gaunkar et al. [14] utilized the iron core instead to avoid saturation of the core material [14]. Stupakov has also used a core manufactured from laminated Fe-Si transformer steel [26]. Besides the above-mentioned materials used for the core, other alloys also exist e.g., Supermendur [8], with even higher saturation flux densities than iron or ferrites.
Materials and Geometry: Pickup Unit
For pickup units, the simplest form is an air coil where the copper wire is wound in a circular form. However, the air coil has relatively low sensitivity as explained in [27]. Thus, the addition of a soft magnetic ferrite core inside the copper coil leads to better sensitivity of the coil [27]. Prabhu Gaunkar et al. [15] verified with simulations that using a high magnetic permeability core material in the pickup enhances the recorded BN signal although the core material should not be saturated in the measurements. Pickup coil modifications have been studied by many researchers e.g., [17,27]. For example, Stupakov et al. [28] tested many differently modified attached coil structures to observe their effect on the BN response. In their studies [28], two different pickup coil types were introduced: a pancake coil and a cylindrical coil wrapped around the sample, which are shown schematically in Fig. 3a and b. The pancake coil is also referred to as a pickup bobbin coil [26].
Stupakov et al. [28] utilized both an air core and a core made of CNS 12021 steel or Fe-Si transformer steel for the pancake coil. The cylindrical coil had a ferrite core of 25 mm in height. Capo Sanchez et al. [17] also studied a pancake pickup, which was wound around a small cylindrical plastic core with 200 turns. In addition, they [17] studied a pickup wound around the studied sample with 1000 turns of AWG44 copper wire. For the pickup yoke core, ferrite has been a widely used material. Stupakov et al. [26,28] used laminated Fe-Si transformer steel for the pickup core in their studies. In addition, Stupakov [26] used Cu-shielding with a grounded case for the pickup. Prabhu Gaunkar et al. [15] used a nickel-zinc ferrite core for their pickup. Besides the use of different pickup core materials, Stupakov et al. [28] experimentally showed that the responses of different pickup coils with different materials were found to be fairly similar to each other. The main difference was with an air core: the addition of the core material to the air core pickup led to 3-5 times the amplification of the BN responses depending on the core material. The shielding with a copper-grounded probe case decreased the environmental noise compared with open ferrite coils [28].
Pickup Placement
Normally one pickup coil is located between the two magnetizing yoke feet as a separate unit. However, in some applications, a single magnetizing yoke is applied with multiple search pickup coils. In addition, Patel et al. [29] utilized two pickup coils to record BN: one coil to obtain flux density and one H-coil to obtain the magnetic field, whereas the so-called tetrapole probes utilize two electromagnetic yokes and two pickups [19]. Augustyniak et al. [30] used a double-core set-up, where all four coils were wrapped on two perpendicular C-cores. In some specific geometries (rotational symmetry) the pickup coil can be wound around the sample as demonstrated in [8] or in [17,28]. In this case, the flux is measured directly in the sample. However, for the typical BN measurements of nonrotationally symmetric samples, the optimal magnetic field sensing location would be on the sample surface in the region of interest.
Probe Lift-Off and Other Geometrical Issues
Typically probe design is application-specific as most applications require a customized probe design to accommodate the sample geometry. This means that the probe design takes the sample surface geometry into account, and the probe and its leg geometry is carefully considered to minimize the air gap at the surface. As White et al. [18] concluded, even small changes in the magnetizing unit yoke lift-off can produce large changes in magnetic circuit permeability, and thus the probe position dramatically affects the sensitivity of the Barkhausen effect. The air gap between the probe and the target specimen affects the magnetic coupling between the electromagnet's poles and the sample [8]. The air gap significantly reduces the permeability of the circuit formed by the core, sample and air gap, and also reduces flux density B [18]. The geometry of the magnetizing leg and pick-up surface depends on the geometry of the surface of the sample. For flat surfaces, flat-shaped pieces are normally used, while round-shaped ones are used for round surface samples [7]. In addition, the magnetizing pole pieces can be removable and attached so that they can be changed according to the surface to be measured [7]. However, Prabhu Gaunkar et al. [14] pointed out with their modelling studies that a curved magnetizing yoke core tip, with an appropriately calculated arc length, can ensure consistent magnetic flux coupling with varying surface geometries.
The pickup location with only minor lift-off can be verified by a spring-loaded pickup mechanism [26,28,31,32].
The air gaps also have an effect on the pickup sensitivity and different pickups have been demonstrated to exhibit different sensitivity to air gaps or lift-offs. Stupakov et al. [28] noticed that a pickup with a soft magnetic core was more sensitive to air gaps than an FeSi or CSN 12021 steel core. Thus, ensuring the proper contact without air gaps is essential and spring loading should be considered. McNairnay [4] noticed that the lift-off effect scales with the coil diameter and increasing the size of the coil would ensure that the pickup assembly is less sensitive to the air gap or lift-off variations. White et al. [18] performed lift-off studies by placing plastic spacers between the sample and the probe. They [18] observed that the drive current must be increased with lift-off to compensate for the drop in magnetic circuit permeability due to the air gap.
Commercial Probes
Commercial Barkhausen noise system manufacturers offer readily available probes and measurement technology for Barkhausen noise inspections. Stresstech [33] produces probes, central units and readily available automated solutions for Barkhausen noise inspections for grinding burn studies, for example. Fraunhofer Institute IZFP (Germany) produces a 3MA device where Barkhausen noise can be measured along with other magnetic variables [34][35][36]. In addition, the Introscan system by Vengrinovich in Belarus is one Barkhausen noise related measurement system for stress qualitative assessment [19,37]. The QASS company of Germany also manufactures a commercial Barkhausen noise inspection device, the QASS µ magnetic, mainly for hardness inspection utilizing known hardness references [38]. Stresstech is the biggest commercial manufacturer and its Barkhausen noise systems are utilized worldwide.
Self-Made Probes with Single Magnetizing Yoke
For certain applications, commercial probes may have coupling or accessibility restrictions as White [8] mentioned when justifying his self-made probe manufactured for feeder pipe inspection. Barkhausen measurement is qualitative and the set-up may vary considerably for each laboratory, as it is not yet standardized [17]. For example, the magnetizing and pickup coils may have a different number of turns, different diameters of wire and different widths. Many researchers [8,17,18,28,31,[39][40][41] have built their own Barkhausen noise systems for research use. Commonly the U-shape or rectangular ferrite yoke with a winding copper coil is used for magnetization. Stupakov [26] states that this Barkhausen noise probe design is similar to the industrial and commercial probes provided by Stresstech and IZFP, in addition to other types of pickups. Therefore, the use of self-made probes allows simultaneous use of other sensors with the BN pickup coil such as Hall sensors and induction coils [28]. In many studies e.g., [16,28,39,41], the Barkhausen noise search coil was equipped with Hall sensors. A vertical array of Hall sensors has been used to measure tangential magnetic surface fields at certain locations above the sample and field extrapolation technique used to determine the real sample magnetization [28], whereas an induction coil was utilized to control the sample magnetization [28].
Double Core or Tetrapole Probes
Besides the traditional single yoke and single pickup combinations for BN probes, various modifications have also been developed. A double-core measurement set-up called tetrapole was introduced by Vengrinovich and Tsukerman in 2004 [19]. The tetrapole probe includes two self-perpendicular electromagnets with pole pieces where the magnetic flux is created separately in each magnetizing coil [19]. The magnetic field is rotated in the surface plane of the sample and the linear superposition of two orthogonal magnetic fields is assumed. The flexible pickup is located in the middle of the coils. The use of tetrapole probes is justified in cases when there is anisotropy in the material as a tetrapole probe eliminates the need to rotate the probe between different measurement directions. Vengrinovich and Tsukerman have mainly studied stresses and created so-called directional diagrams (DD) of BN [19]. Augustyniak et al. [30] utilized a double-core magnetizing set-up, which used four coils that were wrapped around two perpendicular C-cores: called the X-core and Y-core referring to the measurement direction. The tetrapole BN probe has also been studied by Refs. [4,8,42]. White [42] introduced the tetrapole probe with either four cylindrical poles with magnetization and feedback coils on each pole or orthogonal U-cores with four legs with magnetization and feedback coils on each leg, depicted in Fig. 4 a and b, respectively. The measurement system was adapted for BN anisotropy measurements [8]. FEM modelling showed that cores made with ferrite showed too low a saturation flux density and therefore a core made from high saturation laminated alloys (Supermendur) was used. [8]
Surface Scanning Techniques and Moving
Coil Probes
Continuous Magnetic Barkhausen Noise (CMBN)
Conventional Barkhausen noise probes are usually used in the quality control of finished products or during their manufacture. However, traditional BN probes can be quite spaceand energy-consuming and thus, according to Hamfelt et al. [43], unsuitable for condition monitoring applications. New uses of Barkhausen noise in condition monitoring applications have generated a demand to change the conventional probe structure. The use of continuous magnetic Barkhausen noise (CMBN) measurements was first studied by Crouch [44], who utilized the concept of rotational permanent magnets for pipeline stress measurements. The magnet moving over the ferromagnetic sample produces a time-varying magnetic field that can excite BN if it is strong enough. The CMBN probe is an assembly of a magnet producing the magnetic field and a ferrite-cored coil for BN measurements [9]. Figure 5 shows a typical assembly of the CMBN device for measurements. CMBN has been studied by many researchers [9,11,45]. Continuous magnetic BN measurements can be made to detect anisotropy and the direction of the magnetic easy axis of ferromagnetic samples [45] as well as for studying the stress state [44]. Franco and Padovese [9] utilized the method to detect wall thickness loss and for the detection of plastic deformation of steel surfaces [11]. Franco Grijalba et al. [11] stated that the CMBN pickup coil has the same characteristics as a stationary BN pickup. The probe can be fixed in a stationary position as the samples are moved with an XYZ table [9]. The magnet and the pickup need to be positioned in a way to maximize the magnetic field. Franco Grijalba et al. [11] found that the pickup coil behind the magnet provided increased Barkhausen noise signal values and greater sensitivity. The speed also has an influence on the measurements. Caldas-Morgan et al. [45] noticed that the BN generation was less noticeable at lower speeds. The same effect was noticed by Franco Grijalba et al. [46]. However, during scanning, the probe needs to be kept still for at least half of a magnetization cycle [9]. The CMBN measurement method is a relative technique and requires a calibration procedure [46].
Permanent Magnet Probe
Barkhausen noise has also been utilized in condition-based maintenance inspection for condition monitoring [43,47]. Low power probe systems are required in the condition monitoring of fatigue damage for integrated bearings. Thus, a normal BN probe cannot be utilized as such and therefore, a BN probe was built with a permanent magnet that relies on the relative movement between the measured material and the probe. The BN probe consists of a solenoid coil, with or without a core, and a permanent magnet with magnetic flux parallel to the core of the coil. This kind of probe resembles a reluctance sensor with a non-varying reluctance circuit as stated by Hamfelt et al. [43].
Utilization of Finite Element Modelling in Barkhausen Noise Probe Studies
The finite element method (FEM) is a powerful engineering tool for all design, and includes research and development elements, such as building prototypes or optimizing parameters. This section is a summary of how different authors have utilized FEM in Barkhausen noise probe studies, and what kind of challenges have been experienced while doing so. Usually the general research question before the modelling is how the selection of the probe design parameters will affect the generation of magnetic fields used to excite BN in specimens, as stated by Prabhu Gaunkar et al. [14]. Without FEM, no information can be obtained about the Barkhausen noise probe, such as the strength of the generated magnetic field. Even though this is vital information, it cannot be measured from inside the material. By using FEM, several design problems can be solved, i.e., how changing the geometry and materials or how changing the input parameters affect the magnetic flux field. The modelling of the BN probe allows many different types of variations to be used in studying the magnetic flux densities and creation of the magnetic field. Typical variables in the probe assembly that can be modelled are the air gap or lift-off [40,48], the shape of the yoke [40,49], the number of turns of the magnetization coil [11], the shape of the pickup coil and the number of turns of the pickup coil [11].
In order to obtain usable results from simulations, a great deal of pre-existing information about the magnetizing [46] parameters of the BN probe is required. In the FEM software ANSYS, the coil is modelled as a uniform density load domain with a value of A/m 2 . In COMSOL software, this can be done by modelling the coil as a domain with the number of coil turns, coil diameter and input current/ voltage. The coil is then analysed with the Coil Geometry Analysis tool in COMSOL. In this step, the coil operation is studied in the simulations. The non-linear behaviour of the materials is required to ensure sufficient accuracy of the results in simulations that involve magnetic saturation. This can be taken into account with both ANSYS and COMSOL. In most publications, two commercial finite element programs: COMSOL Multiphysics and ANSYS were used, e.g., Franco et al. [9], Augustinyak et al. [29], Garstka and Stefanik [40], Laitinen et al. [50], Hao et al. [51], Persson et al. [52]. In addition, the Opera 3D program has been utilized in Barkhausen noise probe design [53].
COMSOL Multiphysics
Hao et al. [51,54] used COMSOL Multiphysics in 2D for calculating the effective permeability of an actual two-phase microstructure containing ferrite and austenite. Their experimental studies concentrated on detecting the ferrite ratio of dual-phase steels [51] and detecting decarburization [54] from the probe outputs. In [51], the effective permeability was calculated with different ferrite fractions and the results were compared to the probe response. The results were heavily affected by the shape and distribution of the ferrite domains. The first step was to model the effective relative permeability for the ferrite-austenite microstructure; the second step was to link the effective relative permeability changing with the ferrite fraction to the probe output, using a finite element probe output model with the particular probe geometry. In this way, different microstructures containing different amounts of ferrite and probe designs can be considered separately or in combination. Zhou et al. [55] continued a similar type of modelling work as Hao et al. [51] with COMSOL, extending the modelling to 3D and adding 2nd phase distribution relative permeability calculations to the model of ferrite-austenite phase mixtures. It was important to study the 3D aspect in the modelling because the probe interacts with the microstructure in three dimensions.
Prabhu Gaunkar et al. [14] studied the different magnetizing core tip curvature, core length and effects of core materials using the AC/DC module of COMSOL Multiphysics. Persson et al. [52] modelled a soft iron C-core with a magnetizing coil and a layered steel plate with COMSOL. The layers represented different steel phases, such as martensite, bainite and pearlite. The difference between the steel phases was made by assigning a relative permeability value resembling their real-life counterparts to each material. The current authors, Laitinen et al. [50], used the commercial COMSOL software with the AC/DC module for modelling the magnetization of the pickup coil with several assembly configurations, such as an E-core and a pot core.
ANSYS
Augustinyak et al. [30] studied the magnetization of a double core set-up using ANSYS, which included tetrapole magnetizing coils with time-dependent density loading. Pal'a et al. [49] studied the effect of an air gap from 50 to 200 µm between the probe and the sample, investigating the reproducibility of the measuring conditions. Pal'a et al. [49] also modelled different shapes of magnetizing yoke (plane or cylindrical) in 2D with ANSYS and in 3D with FEM. It was noticed that cylindrically shaped legs required higher magnetization currents compared with plane legs to gain the same tangential magnetic field generation. The cylindrical shape for yoke legs is only valid for stabilizing the magnetic field in a sample with large air gaps and large magnetizing currents. Garstka and Stefanik [40] studied the effect of changing the geometry of the magnetization yoke on the distribution intensity of the magnetic field using ANSYS. The electromagnetic calculations were based on the fundamental Maxwell field equations, and on Biot-Savart and Ampere's laws. They [40] studied square, rounded and concave profiled pole shoes. In addition, they also studied a case of how adding an air gap between the yoke and rounded sample changed the magnetic flux. The most favourable case, with the most uniform distribution of flux density and flux lines having the smallest relative gradient for magnetization, was obtained utilizing concave profiled pole shoes on rounded tubes (B = 1.5 T). The air gap gave the worst results; it decreased the value of B drastically at the surface (B = 50 mT). In this case, the most uniform contact with the sampled surface proved the most favourable case. Franco and Padovese [9] utilized ANSYS FEM in deciding the optimal placement for the pickup coil in a set-up featuring a permanent magnet. The simulation also took the non-linear magnetic behaviour of the materials into account.
Opera 3D
Laukkanen [53] performed comparisons between the probe model and measurements to verify the model. The model was created with a commercial program called Opera 3D from Vector Fields with an ELEKTRA steady-state solver.
Barkhausen Noise Pickup Modelling and Experimental Studies
Most modelling studies concentrate on the BN magnetizing coil alone [30,40,[51][52][53]. The modelling of magnetic fields is challenging as magnetizing coil modelling results do not reveal anything about the actual Barkhausen noise signal that comes from the sample to the pickup coil-it only provides information about the effect of the magnetization field on the pickup coil. However, it would be more important to gain information about how the pickup coil operates and its sensing capabilities. This is a hard task to achieve with modelling and thus a reverse approach might be needed: studying the pickup coil magnetization ability. The study [50] carried out by the current authors focused on pickup modelling. This approach was more valid for probe research and development design. It should be noted that pickup coil magnetization calculations only give the theoretical magnetic flux strength created by the pickup coil, from which suggestions about the strength and quality of the BN signal can be made. They are not directly proportional to each other and thus should always be verified by actual measurements. Franco Grijalba et al. [11] have concentrated on continuous magnetic Barkhausen noise (CMBN) pickup coils. In their experimental studies, the CMBN pickup coil parameters (number of turns, wire diameter and coil height) were changed. The coil having the highest number of turns and the thickest wire provided larger amplitude signals than the coil with a smaller number of turns with the thinnest wire, although the result for flaw capability separation was worse with the coil having the highest number of turns. Prabhu Gaunkar et al. [15] studied the effect of different sample material permeabilities on pickup coil magnetization and came to the conclusion that the pickup core permeability should be high to have the best sensitivity, but the effects of mutual inductance between the magnetizing coil and the sensing coil also need to be taken into consideration.
The main aim of the studies by McNairnay [4] and White [8] was to achieve the smallest sensing radius of the pickup coil to perform the most accurate sensing. The small sensing radius was performed by employing a ferrite core inside a coil which was inside a sheath and copper shield. This probe structure was called a pot core. Deveci [7] also studied this pot core probe design and its effect on experimental measurements based on White's [8] studies. The pot core probe in [7] had a new design, which consisted of a brass shield as the outermost layer, a ferrite core as the inner layer with a sheath cylinder made of ferrite between them. The sheath was believed to focus the sensing pickup area into a smaller area to gain better signal outcome. The result was that the pot core did not work significantly better than the traditional probe design without the shield and sheath.
Laitinen et al. [50] studied the effect of a similar kind of pot core pickup with a brass shield and compared the outcome of the experimental measurements with the standard pickup. Both pickups were used with external magnetization to study two hydrogen burn marks on a bearing. Based on the modelling results, the pot core pickup with brass shielding would have gained higher magnetization located in a wider area than the standard pickup. In the experimental measurements, the pot core with brass shielding gave a better spatial resolution of the burn marks than the standard pickup. The signal with the pot core pickup was much more concentrated and gave a smaller peak width for the shallow burn marks than with the standard pickup. Therefore, based on the studies of Deveci [7] and Laitinen et al. [50], it can be said that the pot core pickup works on a case-by-case basis. Deveci [7] used the magnetization located inside the same probe whereas Laitinen et al. [50] utilized external magnetization. Laitinen et al. [50] used COMSOL to study the magnetization plots of a standard pickup, pot core pickup and e-core pickup, which is a modification of the pot core where a ferrite core is connected to a ferrite sheath from the top with a connecting ferrite bridge. The modelling results showed that this ferrite bridge would increase the magnetizing effect on the surface of the studied steel. The difference between the pot core and e-core is the solid ferrite body directing the magnetic field flux into the component. The addition of copper shielding showed a decrease in magnetic flux densities in the surface for all pickup configurations.
Sensitivity of a BN Probe
Signal sensitivity can be said to be a result of the relationship between the applied input and the output signal. Magnetization of the studied sample creates BN events for the pickup coil to detect. If magnetization is not suitable or there are air gaps [18], naturally the possibility to record the BN events is also decreased. Thus, the pickup also plays an important role. The whole accuracy of the measurements depends on the accuracy of both the calibration and the measurement themselves, as stated by Vengrinovich et al. [56]. The repeatability of BN measurements might be challenging because the magnetic fields cannot be measured directly within the material [30]. The detected BN signal also depends on the stress-state, microstructural inhomogeneity, magnetizing field produced by the magnetizing coils, core geometry, probe-to-specimen coupling and spacing between the core tips of the sample, as Prabhu Gaunkar et al. [14] concluded. Therefore, the optimization of the probe configuration would improve the sensitivity, reproducibility and accuracy of the detected Barkhausen signals, as stated in [14]. In addition, the probe materials need to be carefully selected as pointed out by [15]. The most suitable configuration is case-dependent and should be selected based on the application and its requirements. For example, the use of tetrapole probes is justified in cases when there is anisotropy in the material as a tetrapole probe eliminates the need to rotate the probe between different measurement directions, whereas selfmade probes with variable coil turns allow more detailed sensitivity adjustment.
In modelling, the sensitivity of the probe was observed to be improved when a high-permeability core material with high saturation magnetization was utilized in the pickup coil with a large number of turns. As Tumanski [27] pointed out in his review, it is widely known and can be seen from the equation of Faraday's law that a high coil sensitivity can be obtained by using a large number of turns (n) and large active area (A). He [27] also stated that the optimization process for coil performance is no easy task. To make the coil probe with a ferromagnetic core more sensitive, the length of the core (or rather the ratio l D i where l is the length and D i is the inner diameter of the coil), should be made as large as possible, since the sensitivity is proportional to l 3 [27]. Thus, in the case of a high-permeability material, the sensitivity of the probe depends mostly on the geometry of the core [27]. Capo Sanchez et al. [17] carried out a comparison of different BN probe configurations with measurements and found the magnetizing yoke-pancake pickup combination to be the most sensitive, which is also the most commonly utilized assembly among commercial probes.
In the studies of Blaow and Shaw [57], it was established that the number of turns in the detection coil is an important parameter that controls the shape of the BN amplitude peaks. The sensitivity of the two-peak BN profile characteristics was also found to depend on the magnetization field strength. In a low magnetizing field, the BN peak height was observed to increase as the number of turns of the detection coil increased. Therefore, based on the BN profile when increasing the number of pickup coil turns, Blaow and Shaw concluded that the detection depth could be enhanced by the number of turns of the pickup coil [57]. In the case of continuous magnetic BN, the position of the probe and the magnet during motion was found to be an important factor for optimization of measurement sensitivity. The scanning speed and positioning of the probe need to be set in a way to maximize the magnetic field because BN generation is related to the magnetic field intensity [9].
Conclusion
In this paper, a review of the advances in the range of BN probes and modelling was presented. The BN probe has been studied but no thorough review or conclusions have been made so far. Thus, this review focused on combining the information regarding different types of BN probes and their modelling. The detailed comparison of different BN probes is challenging because they all seem to be built differently with different materials. Also, the materials, geometry and treatments in the studied samples vary and are not the same in the separate studies. BN measurements will only be consistent and comparable in cases where the magnetization distribution is achieved similarly in each measurement, and the circuit flux density control is an effective means for achieving this goal, as stated by White [18]. It only remains to be said that, besides basic studies, there is still a need to obtain better understanding of how various factors influence the measurement system. Optimizing the performance of probes for BN measurements can be carried out by means of probe modelling. Modelling with FEM will be utilized more in future to build probes more effectively. With FEM, changes in the probe design and materials can be easily modified to study their influence. The future trend of the utilization of FEM might be in modelling the actual microstructures of the studied samples to predict the magnetic field flux occurrence and to perhaps modify the probe materials based on the studied sample material. This assumption is based on the FEM microstructure model developed by Hao et al. [50] to predict relative permeability based on actual microstructures.
The challenges most authors face during magnetic field simulations are related to magnetic saturation, which requires a more refined mesh in the areas of saturation, and some optimization to converge on a solution. Finding the solution is often a simulation based on trial-and-error, which is highly time-consuming. It is also worth noting that there has not been any approach for modelling the Barkhausen noise phenomenon itself. This is due to the complex nature of the process, such as creating the magnetic domains for the material, which may not be possible with the current simulation tools. At present, if we wish to study the BN phenomenon itself, the approach of Hao et al. [51] and Zhou et al. [55] linking the effective permeability of the microstructure to the probe output appears to be the most feasible. | 2019-09-26T08:55:58.503Z | 2019-09-21T00:00:00.000 | {
"year": 2019,
"sha1": "f6bac4c214f667dd94a1080d350985aaaa82b09a",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s10921-019-0636-z.pdf",
"oa_status": "HYBRID",
"pdf_src": "SpringerNature",
"pdf_hash": "b2c82383bc039f321da57872707fc6921b793921",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
53287610 | pes2o/s2orc | v3-fos-license | Analyzing Image Segmentation for Connectomics
Automatic image segmentation is critical to scale up electron microscope (EM) connectome reconstruction. To this end, segmentation competitions, such as CREMI and SNEMI, exist to help researchers evaluate segmentation algorithms with the goal of improving them. Because generating ground truth is time-consuming, these competitions often fail to capture the challenges in segmenting larger datasets required in connectomics. More generally, the common metrics for EM image segmentation do not emphasize impact on downstream analysis and are often not very useful for isolating problem areas in the segmentation. For example, they do not capture connectivity information and often over-rate the quality of a segmentation as we demonstrate later. To address these issues, we introduce a novel strategy to enable evaluation of segmentation at large scales both in a supervised setting, where ground truth is available, or an unsupervised setting. To achieve this, we first introduce new metrics more closely aligned with the use of segmentation in downstream analysis and reconstruction. In particular, these include synapse connectivity and completeness metrics that provide both meaningful and intuitive interpretations of segmentation quality as it relates to the preservation of neuron connectivity. Also, we propose measures of segmentation correctness and completeness with respect to the percentage of “orphan” fragments and the concentrations of self-loops formed by segmentation failures, which are helpful in analysis and can be computed without ground truth. The introduction of new metrics intended to be used for practical applications involving large datasets necessitates a scalable software ecosystem, which is a critical contribution of this paper. To this end, we introduce a scalable, flexible software framework that enables integration of several different metrics and provides mechanisms to evaluate and debug differences between segmentations. We also introduce visualization software to help users to consume the various metrics collected. We evaluate our framework on two relatively large public groundtruth datasets providing novel insights on example segmentations.
INTRODUCTION
The emerging field of EM-level connectomics requires very large 3D datasets to even extract the smallest circuits in animal brains due to the high resolution required to resolve individual synapses. Consequently, at typical nanometer-level resolution single neurons in even a fruit-fly brain typically span over 10,000 voxels in a given orientation. An entire fly dataset which is less than 1mm 3 requires over 100TB of image data (Zheng et al., 2017). FIGURE 1 | Small segmentation errors locally can lead to large topological errors. The field-of-view for modern convolutional neuronal networks is a small fraction of the size of the neuron leading to potentially bad global mistakes. Segmentation evaluation is typically done on datasets only a few times bigger (in one dimension) than this field of view.
FIGURE 2 | Synapses are often located on thin neurites. The above shows a T5 neuron where many synapses are on the neuron tips. Each sphere represents a different synapse site.
These dataset sizes pose several challenges for automatic image segmentation, which aims to automatically extract the neurons based on electron-dense neuron membranes. First, image segmentation algorithms struggle with classifier generalizability. For a large dataset, there are greater opportunities for anomalies that are significantly outside of the manifold of training samples examined. Even with advances in deep learning (Funke et al., 2018;Januszewski et al., 2018), the size and high-dimensional complexity of neuron shapes allow even small segmentation errors to result in catastrophically bad results as shown in Figure 1. Independent of dataset size, image segmentation struggles in regions with image contrast ambiguity, inadequate image resolution, or other image artifacts. This is particularly prominent for small neurites where synapses often reside (Schneider-Mizell et al., 2016). In Figure 2, the synapses for the neuron reside on the small tips of the neurons.
It should follow that image segmentation should be evaluated on large datasets with additional consideration for the correctness of small neurites critical for connectivity. Unfortunately, this is not the case. The authors are aware of no publications for new segmentation algorithms that emphasize this. Recent work (Maitin-Shepard et al., 2016;Januszewski et al., 2018) have evaluated segmentation on large datasets, such as Takemura et al. (2015). But these works do not consider synaptic connectivity explicitly, which is the ultimate application of the image segmentation. Neither SNEMI (Arganda-Carreras et al., 2015) nor (CREMI, 2016) segmentation challenges use datasets that span large sections of neurons. While they have been instrumental to meaningful advances to the field, they are ultimately limited by their small size and can under-represent problems as shown in Figure 1. This occurs because the actual cause of the error is in only one small region, but the impact is observed in many more regions.
There are reasons large-scale, connectivity-based evaluations are uncommon. Importantly, evaluating large datasets requires considerable ground truth that is time-consuming to produce. The groundtruth dataset in Takemura et al. (2015) is an order-of-magnitude bigger than the other public challenges but took 5 years of human proofreading and is still over three orders of magnitude smaller than the whole fly brain. We believe connectivity-based metrics have not been readily adopted because (1) it requires the annotation of synapse objects which is an independent step of the typical segmentation workflows, (2) the segmentation optimization objectives used in classifier training focus on lower-level, local topology (Rand, 1971;Meilȃ, 2003), whereas connectivity is more global, (3) there are no proposed connectivity metrics that are widely adopted, and (4) there are no sufficiently large challenge datasets to meaningfully capture neuron connectivity. For algorithm designers, it is probably disconcerting to achieve poor evaluation scores based on connectivity that cannot be directly optimized in segmentation objectives without clever engineering and heuristics. While the training and local validation of segmentation is both practical and leading to significant improvements to the field, ignoring the higher-level objectives could lead to an over-estimation of segmentation quality and missed opportunities for more direct improvements for the target applications. We will show later that evaluating segmentation around synapses more directly results in less optimistic scoring compared to traditional metrics like (Meilȃ, 2003). Recent work in Reilly et al. (2017) also introduced a metric that more appropriately weighs the impact of synapses on segmentation, though it does not explicitly consider connectivity correctness between neurons.
To address these issues, we propose a segmentation evaluation framework, which allows one to examine arbitrarily large datasets using both traditional and newly devised applicationrelevant metrics. Our contributions consist of (1) new evaluation metrics, (2) novel mechanisms of using metrics to debug and a localize errors, (3) software to realize these evaluations at scale, and (4) visualization to explore these metrics and compare segmentations.
We advocate an "all-of-the-above" philosophy where multiple metrics are deployed. In addition, we provide an approach to decompose some of these metrics spatially and per neuron to provide insights for isolating errors. This overcomes a limitation in previous challenge datasets that mainly produce summary metrics over the entire dataset, which provides no insight to where the errors occur. By decomposing the results, our framework is useful as a debugging tool where differences between segmentations are highlighted. While ground truth is ideal for evaluating different segmentations to know which one is better in an absolute sense, these debugging features highlight differences even if directly comparing two test segmentations without ground truth. This is critical for practically deploying segmentation on large datasets. The best segmentation can often be discerned by quickly examining the areas of greatest difference. While this provides only a qualitative assessment, this information is useful for identifying areas where new training data could be provided. Also, if one samples some of these differences, potential impact on proofreading performance can be discerned. For instance, such analysis might reveal that the most significant differences are due to one segmentation having a lot of large false mergers, which tend to be time consuming to fix.
Beyond decomposing metrics in new ways, we introduce the following evaluations: • A novel, synapse-aware connectivity measure that better encapsulates the connectomics objective and provides intuitive insight on segmentation quality. • New strategies to assess segmentation quality with different definitions of connectome completeness, 95 providing a potentially more lenient and realistic optimization goal. This is motivated by research that suggests a 100% accurate connectome is unnecessary to recover biologically meaningful results (Takemura et al., 2015;Schneider-Mizell et al., 2016;Gerhard et al., 2017). • Ground-truth independent statistics to assess segmentation quality, such as counting "orphan" fragments and selfloops in the segmentation. These statistics provide additional mechanisms to compare two segmentations without ground truth.
The above is deployed within a scalable, clusterable software solution using Apache Spark that can evaluate large data on cloud-backed storage. We evaluate this ecosystem on two large, public datasets. Our parallel implementation scales reasonably well to larger volumes, where a 20 gigavoxel dataset can be pre-processed and evaluated on our 512-core compute cluster in under 10 min with minimal memory requirements. The comparison results emphasize the importance of considering the synapse connectivity in evaluation. We also show that groundtruth is not necessary to generate interesting observations from the dataset.
The paper begins with some background on different published metrics for segmentation evaluation. We then introduce the overall evaluation framework and describe in detail several specific new metrics. Finally, we present experimental results and conclusions.
BACKGROUND
Several metrics have been proposed for segmentation evaluation, where the goal is analyzing the similarity of a test segmentation S to a so-called ground truth G. We review four categories of metrics in this section: volume-filling or topological, connectivity, skeleton, and proofreading effort.
Volume-Filling or Topological
Topological metrics measure segmentation similarity at the voxel-level, so that the precision of the exact segmentation boundaries is less important than the topology of the segmentation. For instance, if the segmentation splits a neuron in half, the similarity score will be much lower than a segmentation that mostly preserves the topology but not the exact boundary. Example metrics of this class include the Rand Index (Rand, 1971;Hubert and Arabie, 1985), Warping Index (Jain et al., 2010), and Variation of Information (VI) (Meilȃ, 2003). Since VI will be discussed later in this work, we define it below as: where H is the entropy function. VI is decomposed into an oversegmentation component H(S|G) and an under-segmentation component H(G|S). A low score indicates high similarity.
Connectivity
Examining topological similarity using the above metrics can be misleading in some cases since small shifts in segment boundaries can greatly impact the scores as noted in Funke et al. (2017). Furthermore, as shown in Figure 2, the synaptic connections are often on the harder-to-segment parts of a neuron that only make a small percentage of overall neuron volume. One potential solution is to define S and G in Equation 1 over a set of exemplar points representing synapses, instead of all segmentation voxels as done in and Plaza and Berg (2016). A similar strategy of measuring groupings of synapses was introduced in Reilly et al. (2017), which additionally breaks down results per neuron making the results more interpretable. While these metrics better emphasize correctness near synapses, it is not obvious how to interpret error impact to connectivity pathways.
Skeleton
Similar to topological metrics, the works in Berning et al. (2015) and Januszewski et al. (2018) describe metrics based on the correct run-length of a skeleton representation of a neuron. This class of metric provides an intuitive means of interpreting data correctness, namely the distance between errors. In (Berning et al., 2015), the run length can be very sensitive to small topological errors if one tries to account for synapse connectivity since synapses can exist in small neuron tips or spine necks where segmentation errors are more prevalent due to the small size of the processes. While this can be useful to emphasize synaptic-level correctness, it can also under-value a neuron that is mostly topologically correct. Januszewski et al. (2018) proposes an expected run length metric (ERL) that proportionally weights contiguous skeleton segments. While ERL is the most topologically intuitive metric, it conversely suffers from under-weighting correctness for small process such as at dendritic neuron tips in Drosophila or spine necks seen in mammalian tissue.
Proofreading Effort
Tolerant-edit distance (Funke et al., 2017) and estimates of focused proofreading correctness time provide another mechanism to measure segmentation quality. Good segmentation should require few proofreading corrections (shorter edit distance) than bad segmentation. A segmentation that splits a neuron in half would be better than one with several smaller splits, since the former would only require one merge and the later several mergers. Designing interpretable edit distance formulations are challenging because different proofreading workflows could lead to very different proofreading reconstruction times. The usefulness of the above metrics often depend on the application. For practical reasons, mathematically well-formed metrics like VI and ERL that have few parameters are often favored. Metrics that better reflect connectivity are harder to define since they depend more on the target application or require the existence of synapse annotation which is currently predicted in a separate image processing step from segmentation.
Finally, there has been only limited exploration in using segmentation metrics as debugging tools. Presumably, this becomes a bigger concern when evaluating larger datasets. Notably, the authors in Reilly et al. (2017) recognized this challenge and describe a metric that allows intuitive insights at the neuron level. In Nunez-Iglesias et al. (2013), the authors decompose the VI calculation to provide scores per 3D segment. For instance, the over-segmentation VI score H(S|G) can be decomposed as a sum of oversegmentation per ground truth neuron g: Presumably, other metrics like ERL, can be used to provide neuron-level information for finding the worst segmentation outliers.
METRIC EVALUATION ECOSYSTEM
We introduce a metric evaluation ecosystem that is designed to assess the quality of large, practical-sized datasets. To this end, we propose evaluation paradigms that emphasize interpreting and debugging segmentation errors that make comparisons between two different segmentations. While having ground truth is mostly necessary to quantify whether one segmentation is better than another, meaningful comparisons are possible without laboriously generated ground truth since the metrics highlight differences and these differences can be readily inspected. In the following few paragraphs, we will discuss the overall philosophy of our efforts. Then we will explore in more detail novel metrics and the software architecture.
In this work, we do not advocate a specific metric, but instead recommend an "all-of-the-above" framework where for each dataset multiple metrics are used to provide different subtle insights on segmentation quality. While not every popular metric is implemented, our framework is extensible and can support customized plugins.
We provide feedback on segmentation quality at different levels of granularity: summary, body, and subvolume.
Summary
Each segmentation sample is evaluated with several scores applied to the whole dataset. These scores do not provide insight to where errors occur but provide a simple mechanism to compare two segmentation algorithms succinctly. VI and Rand index are two such examples. Section 3.4 introduces several new connectivity-based metrics.
Body
We provide per segment (or body) statistics with respect to segments from both datasets S and G (G need not be ground truth). For example, this includes the per-body VI score defined in Equation 2, which provides insights on where over and undersegmentation occur in the volume. We highlight a couple new body metrics in section 3.4.
Subvolume
When appropriate, metrics that are computed for the whole dataset are also applied to a regular grid of subvolumes that partition it. In this manner, the quality of segmentation can be assessed as a function of its location in the volume. This is useful for potentially detecting regions in the dataset where a classifier fails to generalize. For example, the framework runs VI on each subvolume. To partially disambiguate errors that originate in one region but propagate to another, distant region, we apply a local connected component algorithm to treat each subvolume as an isolated test segmentation 1 .
The evaluation framework can run over several distinct sets of comparison points. By default, segmentations are compared at the voxel level, i.e., the comparison points are all segmented voxels. If other sets of important points (such as synapses) are provided, analysis is similarly applied over these sets. The evaluation provides a mechanism to compare against oneself (no ground truth or alternative segmentation). We discuss metrics that enable self-evaluation in Figure 3. Comparisons to ground truth can be restricted to sparsely reconstructed volumes or dense labeling.
Metrics
In the following, we highlight a few novel metrics for evaluating segmentation, which is a subset of all metrics implemented in the framework. These new metrics are divided into the categories of summary, per-segment, and self-comparison.
Summary
We propose a metric to assess the connectivity correctness (CC) of the given segmentation S compared to ground truth G. At a high level, CC(S|G) defines the percentage of connections that match the ground truth connections. A connection is defined as an edge between two segments (neurons) that represents a synapse. There can be multiple connections between the same two segments. More formally: where x returns the set of synapse connections between two segments. A S (g i ) determines the optimal assignment of groundtruth segment g i to a segment in S (e.g., using the Hungarian matching algorithm). The matching is one-to-one and if there is no match x will be an empty set. In practice, an algorithm that greedily finds a set of matches by using greatest segment overlap with ground truth is likely sufficient since one would not expect the set of intersecting segments in S to a given segment in G to greatly overlap with intersection sets to other segments in G in a manner that would require joint optimization. This is true by construction in the scenario where every segment in S is either a subset of a given segment G or equal to a set of g. This metric is sensitive to both false merge and false split segmentation errors. If there is a false split, there will be fewer matching connections compared to ground truth. If there is a false merge between g 1 and g 2 , the one-to-one assignment A S ensures that A S (g 1 ) = A S (g 2 ) meaning that there will be no matching connections involving either g 1 or g 2 .
Additionally, we introduce a thresholded variant of the connectivity metric to emphasize the percentage of connection paths that are found with more than k connections. We modify Equation 3 to include this threshold and decompose into recall and precision components as defined below: The above metrics to measure the similarity between two connectomes have advantages over using a more general graph matching algorithm. First, by requiring an initial assignment of each segment to a groundtruth neuron (if a distinct match exists), the CC metric aims to better constrain the problem of measuring the similarity between two connectivity graphs, thereby avoiding the need for the computational complexity typical in general graph matching algorithms. Second, the CC metric allows one to express the matching in terms of individual neurons and number of connections preserved, which is more biologically intuitive compared to a general edit distance score.
In addition, to CC k , we define a class of statistics that analyzes the fragmentation of S compared to G based on the simple formula: where a high score indicates that S consists of many more segments than G. While very simple, this provides a lower-bound on the number of edits (or segments to "fix") to transform S into G. In practice, we find that S is typically an over-segmented subset of G and Frag provides a reasonable edit distance estimate. We can extend Frag by extracting a subset of S and G, S * and G * , that represent a less-than-100% correct segmentation. More specifically, we define a thresholded fragmentation score, where S * and G * are the smallest set of segments whose cumulative size reaches a specified size threshold, where size can be number of voxels or synapses. This trivially computed measure allows us to discern the number of segments required to produce a connectome that is X% complete.
Body
As described in Equation 7, VI can be decomposed to provide insight about the fragmentation of a given segment. If this score is applied with respect to segment g, it provides an over-segmentation score of g. If this applied with respect to segment s, it provides an under-segmentation score of s. We can alternatively decompose the VI calculation to report the over and under-mergers that intersect a given segment. We define the under and over segmentation score for g as: where P is the the probability of g (or percentage of g in G). This metric is useful to provide a simple score for the neuron that has the worst segmentation. This metric works most naturally over a densely labeled G since the impact of the false merging can be more accurately assessed.
Additionally, we modified the metric in Equation 3 to provide a score for each g the percentage of connections that are covered. We further note which bodies are the most correct by simple overlap, which is conceptually similar to examining the largest error-free run lengths often used in skeleton-based reconstructions.
Self-Compare
As mentioned, the ability to decompose the metrics at segment level allows one to compare two different segmentations. However, it is often useful to have some information on segmentation reliability when no comparison volume is available. One simple statistic that can be extracted is the number of segments that are needed to reach a certain volume threshold (as defined previously), which provide insights in regions that are relatively over-segmented compared to others. However, this metric can be misleading since neuropil regions vary in neuron packing density.
We introduce two metrics to better assess segmentation in the absence of ground truth: orphan segments and segmentation loops. Biologically, one does not expect a neuron to be a small fragment below a certain size K. A count of the number of segments below this threshold, provides a crude error measure. This will not uncover potential under segmentation errors. To find potential under segmentation errors, we note that neurons should have few connections to itself (self-loops). By counting the number of autapses or finding the segments that have a lot of autapses, we can detect potential false mergers. As segmentation gets better the effectiveness of using autapses as a proxy for false-merge errors is limited since such connections due exist in practice, such as in the Drosophila medulla connectome in Takemura et al. (2015). Therefore, the loop detector should be viewed as a mechanism to detect outliers due to either segmentation error or biological design and serve as a good entry point for analyzing a segmentation. Depending on the organism and the extent of the region being evaluated, additional metrics could be considered, such as ensuring that each segment has both inputs and outputs. We only formally consider orphans and self-loops in this work.
Architecture
We introduce an Apache Spark-based system for comparing two, large segmentations at scale. The implementation is built over the framework described in Plaza and Berg (2016) and is available at https://github.com/janelia-flyem/DVIDSparkServices as the EvaluateSeg workflow . The segmentation and synapse data is stored using DVID (Katz and Plaza, 2018). In general, segmentation compresses to a small fraction of the original EM data size and we do not observe fetching segmentation to be a bottleneck in the analysis workflow. However, evaluating on datasets that are significantly larger than the 1 gigavoxel datasets common in SNEMI and CREMI necessitates a framework that can compute metrics on a large-memory, multi-core, cluster environment.
An overview of the software workflow is shown in Figure 3. We partition the dataset into disjoint, equal-size subvolume for a region of interest (ROI). A local connected component algorithm is computed for each subvolume and other filters are applied, such as (1) dilating groundtruth segment boundaries to reduce the impact of small variations in the exact boundary between segmentation and (2) filtering out neurons that are not groundtruthed for sparse evaluation. If the ROI being analyzed is part of a larger segmentation, one can run a global connected component algorithm which ensures that segments that merge outside of the ROI are treated as separate objects within the ROI. The global connected component algorithm is computed by examining the boundaries between all subvolumes in parallel and determining which components have a connecting pathway through the ROI.
For each subvolume, we compute a contingency table between segments in S (when not doing a self-comparison) and G (where G is treated like ground truth unless otherwise specified). The overlaps computed between S and G allow many of the metrics to be computed per subvolume and then combined into global summary and body stats. This is done over the set of voxels and optionally any available synapse (or other point) data. In the current workflow, one of the largest, non-parallelized compute components is this final grouping of results. Future work to further reduce these non-parallel points is possible but not currently necessary for the experimented data sizes.
The framework allows additional plugins that conform to the API to be added without changing the surrounding framework. In circumstances where this partitioning and combination strategy will not solve a given metric algorithm, it is possible to define a completely custom workflow based on the input segmentation. The current framework does not implement ERL or other skeleton-based metrics, but our ecosystem should admit for its straightforward inclusion.
The statistics from this computation are collected into a file that can be easily parsed. However, the myriad of metrics can make interpreting results overwhelming, so we designed a single web page application in Javascript as shown in Figure 4 to improve accessibility. The web application groups similar stat types together displaying the list of summary stats and per-body breakdowns for provided metrics. A visualization tool shows a heat-map highlighting subvolume to subvolume variation in segmentation quality. The application also allows one to compare the summary results of two different segmentation evaluations. The web page application is available at https://github.com/ janelia-flyem/SegmentationEvaluationConsole.
EXPERIMENTS
We demonstrate our evaluation framework on two large, public datasets: a portion of the Drosophila medulla (Takemura et al., 2015) and mushroom body (Takemura et al., 2017). The medulla dataset segmentation and grayscale can be accessed at http://emdata.janelia.org/medulla7column, and the mushroom body dataset can be accessed at http://emdata.janelia.org/ mushroombody. Both datasets are around 20 Gigavoxels in size and contain over 100,000 synaptic connections. Since neither dataset is 100 percent accurate, we filter small orphan segments in the ground truth using options in the metric tool and we dilate ground truth neuron boundaries with a radius of two pixels. We compare these ground truths to initial segmentation generated using a variant of the algorithm developed in Parag et al. (2015). 2 A smaller portion of the optic lobe segmentation is also compared against a more recent segmentation algorithm (Funke et al., 2018). The purpose of the following experiments is to demonstrate the breadth of provided metrics, as well as, some insights that might impact how one analyzes segmentation results.
Summary Results
The evaluation service produced a series of summary stats. A subset of these are depicted in Figure 5. The stats are split into two broad categories: voxel-based and synapse-based. The voxelbased stats provide volume-relevant information. The synapsebased stats emphasize only the exemplar points that define each input and output for a synapse. 3 In both the mushroom body and medulla, we notice that there are very few false merge mistakes indicated by merge VI. Notably, the split VI is much higher when focusing near synaptic regions. The comparably higher values in the mushroom body highlight both the conservative segmentation used and the presence of very small, hard-to-segment processes. The thresholded segment count shows that to examine 50 percent of the synaptic points, a relatively small number of segments need to be examined compared to achieving 90 percent coverage. For FIGURE 4 | Evaluation web application. Web application that displays results and tools to visualize segmentation errors.
FIGURE 5 | Select metrics for the medulla and mushroom body dataset. The data shows voxel-based metrics like VI and less-common, but more useful synapse-based metrics. The histogram metric shows the many more segments are required to reach X percent of the total volume. both datasets, the connectivity correctness defined by Equation 3 is very low, in particular in the mushroom body where the neurites are very small. This indicates that the automatic segmentation is far from being useful for biological analysis without proofreading.
The summary results also report the worst body VI score and the segment ID number corresponding to this body. We show one example from the medulla in Figure 6. The evaluation service reports the biggest overlapping segments. Notice that the top 10 biggest fragments only cover a small portion of the complex neuron arbor.
We compare the baseline segmentation with a newer segmentation approach in Funke et al. (2018) for a subset of the medulla dataset in Figure 7. As expected, Funke et al. (2018) achieves a better score across all reported metrics. While the VI scores indicate significant improvement, the fragmentation thresholds and synapse connectivity clearly show the advantages for the newer segmentation. There are far fewer segments to consider to reach different levels of completeness as seen in Frag thres. Perhaps more significant is the much greater percentage of neuron connections found with the new segmentation. The CC metrics are sensitive to large neurons being correct in addition to the small synapse processes being correctly segmented. Metrics less sensitive to this level of correctness, like the VI numbers reported, might, in effect, over-rate the quality of inferior segmentation.
Unsupervised Evaluation
The previous results show comparisons between test segmentations and ground truth. As previously explained, the metric service is useful for comparing two segmentations directly even if one is not ground truth since there are many stats that highlight differences useful for debugging. For instance, while the VI between two test segmentations fails to suggest which one is better, it does indicate the magnitude of the differences, can indicate whether one segmentation is oversegmented compared to the other, and gives a list of bodies that differ the most, which can then be manually inspected to determine segmentation errors. But we also introduced stats that do not require a comparison volume. We evaluated both medulla and mushroom body in this way. In Figure 8, we see a heatmap highlighting the small orphan segmentation density over the subvolumes that partition both datasets. We define orphan as any segment with fewer than 10 synaptic endpoints. Visually, the diagram shows more errors in the alpha 3 lobe and proximal region of the mushroom body and medulla respectively. If we evaluate these regions separately against the ground truth, we observe that the supervised VI scores are consistent with the unsupervised visualization.
We were also able to find one neuron in the medulla dataset that had many autapses, which suggests a potential false merge. This worst neuron in the un-supervised analysis corresponds to the fourth worst body in the supervised analysis. This suggests that the autapse count can reveal false merge errors.
Performance and Scaling
These datasets are much larger than previous challenge datasets but are still much smaller than the tera to peta-scale datasets that are being produced. One obvious solution to handling larger datasets is to run the framework on a larger compute cluster.
We show the scalability of our framework by evaluating our two sample datasets with varying numbers of cores. The charts in Figure 9, shows a breakdown of runtime between the top parallelizable portion of the code and the bottom, sequential small overhead. As the number of cores increase we observe a speedup that is slightly less than linear to the number of added cores (indicated by the trendline). We observe that the sequential overhead indicated by the lowest two section of each bar is roughly constant and a small portion of this time (the lowest section) could potentially be partially parallelized with future optimizations.
The results in the table suggest that 512 cores can roughly process around 20 gigavoxels in around 5 min, or over 60 FIGURE 7 | Comparing two segmentations from a subset of the medulla dataset. Unsurprisingly, the more recent segmentation from Funke et al. (2018) performs better on all metrics (indicated by the highlighted boxes). In particular, Funke et al. (2018) achieves much higher CC scores finding 33 percent of all neuron connections with weight greater than or equal to ten synapses, compared to only 9 percent for the baseline. megavoxels of data per second, or 1 TB in a little over 4.5 h. Note that the comparison framework requires two datasets to be processed and this analysis includes the global connected FIGURE 8 | Orphan density map. For both the medulla and mushroom body sample, the orphan count density (an unsupervised statistic) appears greater (darker) in regions with worse synaptic VI compared to the other regions.
FIGURE 9 | Runtime of metric computation at different levels of parallelization. The line represents the optimal speedup for increasing the number of cores from the baseline 64 core implementation. The non-parallelized part of the framework represents the computation performed solely on the driver node and is indicated by the bottom two sections of each bar. This non-parallel time is around 129 and 187 s from the mushroom body and medulla respectively. components analysis, which is not necessary if segmentation is completely contained within the defined region. Also, note that medulla and mushroom body ROIs do not perfectly intersect the subvolumes, so more data is actually fetched to retrieve the entire 20 gigavoxel ROI.
In practice, we expect additional bottlenecks if there are a lot of small segment fragments which could lead to more computation in the sequential parts of the code and in shuffling data around on the network. Future work should aim to improve the performance when dealing with a large number of small fragments since its relevance to analysis is mostly in the aggregate and not at the individual fragment level. We do not observe slowness fetching the segmentation data, but the data could always be partitioned between multiple servers to allow for higher cumulative read bandwidth.
To further improve performance, we consider downsampling the segmentation. (A multi-resolution segmentation representation is available in DVID and does not need to be computed.) Figure 10 shows both datasets at original resolution and downsampled by a factor of 2, 4, and 8 along each axis. One might expect that downsampling the dataset considerably would greatly change the statistics particularly related to fragmentation due to presumably small synaptic processes. Perhaps surprisingly, a few key metrics have a consistent value when downsampling by 4x suggesting that significant computation reduction is possible since full resolution is unnecessary. For example, the fragmentation scores in these datasets, which provide a rough estimate of the number of merge edits required, is similar (within 20 percent) to full resolution. Once the resolution starts getting worse than 40x40x40nm, there is considerable impact on the synaptic VI and the number of thresholded segments. However, the significant differences reported between the two segmentations in Figure 7 are preserved even at the lowest resolution tested.
FIGURE 10 | Stability of various metrics when downsampling the dataset. When the voxel resolution is higher than 40 × 40 × 40 nm, the results are fairly consistent. When the voxel resolution is too low, several synapses on smaller neurites are missed. The 50, 75, and 90% connections number refers to the number of segments required to cover the specified percentage of connection endpoints.
CONCLUSIONS
In this work, we demonstrate a metric evaluation framework that allows one to analyze segmentation quality on large datasets. This work necessitated diverse contributions: new metrics that provide novel insights in large connectomes, a software framework to process large datasets, and visualization software to enable intuitive consumption of the results. All of these contributions, in synergy, were critical to enable segmentation evaluation in practical settings.
We implemented multiple metrics to provide different insights on segmentation. In particular, we introduced new connectivity-based metrics that clearly show that significant improvements are still needed to produce fully-automatic reconstructions, which seem to correctly reflect our observations in practice. Furthermore, we note that for purposes of comparison, it is possible to downsample the data significantly without significant impact on important metrics. Finally, we introduced the possibility of comparing two segmentations without ground truth, where evaluation can be done by manually inspecting the largest segmentation differences revealed by decomposing the metrics in different ways and providing useful visualizations, such as showing segmentation quality variation as a function of region location. We believe that this work should help accelerate advances in image segmentation algorithm development and therefore reduce bottlenecks in large connectomic reconstructions.
The diverse set of statistics produced by our workflow could make the task of comparing segmentations overwhelming, as one desires to know which is the best metric. This paper has taken an agnostic position to the best metric largely because it depends on the application. If one is concerned about optimizing proofreading performance, edit distance measures make the most sense. However, this is complicated because edit distance costs depend on the proofreading methodology. The fragmentation scores provide a very intuitive, parameterfree measure of segmentation quality if one has mostly tuned the algorithms to over segment, since the number of segments is a guide for the number of mergers required. To assess whether the segmentation can be used in a biologically meaningful way, our new connectivity metric will provide the best insight on the quality of the resulting connectome. For assessing general neuron shape correctness, ERL (which we do not currently implement) or VI can be used.
We expect additional improvement is needed to further parallelize sequential portions of the framework. Also, we believe that additional metrics should be invented that provide interesting insights for evaluating the connectivity produced from the segmentation. We have introduced a few metrics to this end in this paper. We advocate the inclusion of more metrics in evaluation to better understand the failure modes of segmentation, which will hopefully lead to the implementation of better algorithms.
AUTHOR CONTRIBUTIONS
SP devised and implemented the core methodology, JF contributed some image segmentation and metric discussions.
FUNDING
This study was funded and supported by Howard Hughes Medical Institute.
ACKNOWLEDGMENTS
Stuart Berg helped with the experimental setup and was instrumental in implementing the DVIDSparkServices ecosystem, where our metric service was implemented. Bill Katz provided support for the Big Data infrastructure through the software DVID. Alex Weston helped in developing the web application used for visualization. Lowell Umayam helped to collect data necessary for experimental analysis. We would especially like to thank the FlyEM project team at Janelia Research Campus for general discussions and support. | 2018-11-13T14:07:31.560Z | 2018-11-13T00:00:00.000 | {
"year": 2018,
"sha1": "d41a66fd86e99a8bbd8c3d5be9ab7904818e6bc8",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fncir.2018.00102/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d41a66fd86e99a8bbd8c3d5be9ab7904818e6bc8",
"s2fieldsofstudy": [
"Computer Science",
"Biology"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
214383628 | pes2o/s2orc | v3-fos-license | A comparative analysis between Mannheim peritonitis score and acute physiological and chronic health evaluation II score in predicting prognosis of patients of perforation peritonitis
Background: The early assessment and recognition of peritonitis patient is required in surgical emergency. Various scoring system have been designed successfully to assess the prognosis and outcome of peritonitis. The present study was carried out with an aim to evaluate the usefulness and severity of Mannheim peritonitis (MPI) score in comparison to acute physiological and chronic health evaluation II (APACHE II) scoring system for prediction of the outcome in patients with perforation peritonitis and thus decision making in perforation peritonitis. Methods: A prospective observational study was carried out at Department of Surgery, King George’s Medical University (KGMU), Lucknow for a period of one year from July 2018 to June 2019. A total of 100 patients were enrolled in the study. Results: Majority of patients were males compared to females. Maximum number of patients (40%) was aged 51-60 years. Maximum number of patients (42%) had duodenal perforation. A significant association between higher MPI scores and mortality was seen (p<0.001). Statistically, the association between APACHE II scores and mortality was significant (p<0.001). Conclusions: APACHE II had a slightly higher sensitivity as well as specificity as compared to MPI. MPI is easy to calculate but accuracy of APACHE II is more, compared to MPI.
INTRODUCTION
Peritonitis is inflammation of the peritoneum, the lining of the inner wall of the abdomen and which covers the abdominal organs. Peritonitis due to hollow viscus perforation continues to be one of the surgical emergencies which is considered to be the lifethreatening condition. Early evaluation by scoring system influences the management and prognosis. 1 Many scoring systems have been designed successfully to assess the prognosis and outcome of peritonitis. Those used were included as acute physiological and chronic health evaluation II (APACHE II) score, Mannheim peritonitis index (MPI) score, the peritonitis index Altona, the sepsis score, the Ranson score, Imrite score and the physiological and operative severity score for enumeration of morbidity and mortality. MPI score was developed by Wacha et al. 2,3 It was based on the retrospective analysis of the data from patients with peritonitis. The MPI is a specific score which has a very good accuracy and serves as an easy way to assess clinical parameters which allows the determination of the individual prognosis of patients with peritonitis. 4 Department of Surgery, King George's Medical University, Lucknow, Uttar Pradesh, India APACHE II score was developed by Knaus et al. 5 It was devised to stratify prognosis in a group of ill patients and for the determination of success of treatment.
Objectives of the present study was carried out with an aim to evaluate the usefulness and severity of MPI score in comparison to APACHE II scoring system for prediction of outcome in patients with perforation peritonitis and thus decision making in perforation peritonitis.
METHODS
A prospective observational study was carried out at Department of Surgery, King George's Medical University (KGMU), Lucknow for a period of one year starting July 2018 to June 2019. Patients were presented in the outpatient department or emergency wards of Department of Surgery with clinical features of perforation peritonitis, after clinical and radiological evaluation. Perforation was diagnosed either by chest Xray, X-ray abdomen (erect or lateral recumbant), ultrasonography of abdomen, computed tomography of abdomen or clinical evaluation.
The sampling frame of the study was bound by the following inclusion and exclusion criteria. Both sexes having 12-70 years of age group were admitted with diagnosis of perforation peritonitis with non-traumatic cause presenting within 72 hours of onset were included in the study. Patients aged less than 12 years and more than 70 years of age, having all traumatic cases and perforation more than 72 hours after onset; patients not undergoing surgery and colonic perforation cases were excluded from the study. n=2(Z α +Z β ) 2 ×S(1-S)/(S1-S2) 2 where; z α =1.96; z β =0.84; S=Pooled specificity=94.5%=0.945; S 1 =89%=0.89; S 2 =100%=1= 67.35273=68 Though the calculated sample size was 68, however, after adding for contingency and provision for loss to followup at 25%, author targeted a sample size of 85. Finally, 100 patients were enrolled in the study.
The MPI score was designed based on the retrospective analysis of data from patients with peritonitis, in which 20 possible and significant risk factors were considered. Among these 20 risk factors, only 8 approved to be of prognostic relevance which were entered into MPI and classified according to their predictive power (Table 1).
On the basis of clinical, laboratory or radiographic investigations, the APACHE II scores were calculated ( Table 2).
Ethical clearance was obtained from the Institutional Ethical Committee vide letter no.622/Ethics/2019. An informed consent was obtained from all the patients. The statistical analysis was done using statistical package for social sciences (SPSS) version 21.0 statistical analysis software. The values were represented in number (%) and mean±SD.
RESULTS
The present study was carried out to assess the usefulness of MPI and APACHE II scoring system in cases of perforation peritonitis. For this purpose, a total of 100 patients of perforation peritonitis were enrolled in the study. (Figure 1) presents the gender profile of patients.
Male preponderance was seen in the study. Glasgow coma score (GCS)=15 minus actual GCS A. Total acute physiology score (sum of 12 above points) B.
Chronic health points Total APACHE II score (add together the points from A+B+C) Table 4 presents the association of the two scoring systems with outcome. There was no mortality in patients with MPI 0-10 and 11-20. Out of 31 patients with score 21-30, a total of 7 (22.5%) died. On the other hand among 44 patients with MPI >30, a total of 22 (50.0%) died. On evaluating the data statistically, a significant association between higher MPI scores and mortality was seen (p<0.001). None of the patients with APACHE II score in 0-9 range died. A total of 4 out of 42 patients with APACHE II score in 10-19 range died and 25 out of 53 patients with APACHE II score >19 died. Thus, mortality rate was 0%, 9.5% and 47.2% respectively among patients with APACHE II score 0-9, 10-19 and >19 respectively. Statistically, the association between APACHE II scores and mortality was significant (p<0.001). Table 5 shows that mean MPI score of non-survivors was 32.90±4.56 which was significantly higher as compared to that of survivors who had mean MPI score of 23.62±7.74 (p<0.001). Mean APACHE II score of nonsurvivors (26.03±5.09) was significantly higher as compared to that of survivors (17.94±5.99) (p<0.001).
Sensitivity Specificity
indicated that majority of patients were males (73%) compared to females (27%). The present study was supported by Godara et al, which showed that majority of patients were from males. 8 The current study showed that majority of patients were of age group of 51-60 years which was similar to the study findings by Godara et al. 8 The origin of perforation peritonitis was from 6 different anatomical sites with most of the patients being observed under duodenal perforation and the study was supported by Godara et al, Arasu et al, and Malik et al. [8][9][10] The current study findings reported maximum number of patients, 45% had MPI >30 and 53% had APACHE II scores >19 which was supported by the study results of Kumar et al. 11 The present study carried out shows majority of mortality rate of MPI score >30 and APACHE II score >19 which was similar to the study carried out by Malik et al. 10 The MPI and APACHE II score of non-survivor was less in comparison to survivors with majority being 68 which was in contrast to the study findings by Kumar et al. 11
CONCLUSION
The present study evaluated and compared the prognostic efficacy of MPI and APACHE II scoring systems among patients of perforation peritonitis. The findings of the study showed that both MPI as well as APACHE II were good predictors of outcome among patients with perforation peritonitis, however, APACHE II had a slightly higher sensitivity as well as specificity as compared to MPI. MPI is easy to calculate but accuracy of APACHE II is more compared to MPI. In view of the dynamic changes in management strategies and emergence of newer techniques for management of perforation peritonitis patients, it is essential that continuous audit of the efficacy of existing and newer prognostic scoring systems should be carried out at regular intervals in order to update the management strategies in view of the changing mortality risk. | 2019-12-05T09:18:54.823Z | 2019-12-25T00:00:00.000 | {
"year": 2019,
"sha1": "7a0b7e6d933ad80f42f117cffc6568fd39680e67",
"oa_license": null,
"oa_url": "https://www.msjonline.org/index.php/ijrms/article/download/7554/5315",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "0f6b3741da3c6056b111d3df3c8375f5825d689e",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
247062744 | pes2o/s2orc | v3-fos-license | Social identity and risky leisure activities: implications for welfare and policy
In this paper, we build on theories in psychology and economics and link positional preferences to private agents’ identification with a social group, and the social norms present in that group. The purpose of our paper is to analyze behavioral, welfare, and policy implications of a link between private agents´ social identity and a risky leisure activity. Our results suggest that, when the outcome of the positional activity is uncertain, the over-consumption result that is associated with positional preferences in a deterministic framework need not apply to all agents in a social equilibrium. The reason is that agents have incentives to act with caution in order to avoid failure when the outcome of the socially valued activity is uncertain. We also show how policy can be used to improve the welfare within a social group where the risky leisure activity is positional.
Introduction
Social media platforms are filled with posts that display people who voluntarily engage in activities that could have dire consequences. Why do people voluntarily engage in activities that expose them to potentially fatal risk? One possible answer is that accomplishing risky tasks provide agents with social status, i.e., risky activities seem to be positional within some social groups.
The purpose of this paper is to analyze the implications of positional preferences associated with risky leisure activities for individual behavior, group behavior, welfare and policy. We focus on risky leisure activities for four reasons: (1) Risky 1 3 Social identity and risky leisure activities: implications… The last argument suggests that what individuals are positional about depends on their social identities. This is in line with the pioneering work by Akerlof and Kranton (2000), who showed that the desire to become an accepted member in a valued social group may drive people to overinvest (with respect to their ability) in status giving activities. It is also in line with theories in psychology, such as identity theory (Stryker 1968;Stryker and Serpe 1982;Burke and Reitzes 1981;Burke 1991;Stets and Burke 2000) and social identity theory (Tajfel et al. 1971;Tajfel 1982;Turner and Oaks 1986). These theories argue that an individual's sense of personal identity is a unique combination of different sub-identities linked to the macro, meso and micro social groups to which the individual belongs. Identity theories suggest that (i) individuals self-categorize into social groups, (ii) most individuals belong to a plethora of different social groups, and (iii) the types of behaviour and characteristics associated with social status will differ between groups due to the social norms prevalent in each group. Individuals gain self-esteem from their performance relative to other groups and relative to other group members for activities and behaviours that are valued within the social group (Festinger 1957;Rivis and Sheeran 2003;White et al. 2009). The extent to which an individual cares about her relative standing in the social hierarchy depends on her dependency on social affirmation and on how important membership to that specific group is (Hogg 2016). Several studies suggest that there is a link between social status, social comparisons and risk-taking behaviour, e.g., for drug use, and sun tanning (Leary et al. 1994;Aloise-Young et al. 1996;Miller-Johnson et al. 2003;Fischer et al. 2011). Jellison and Riskin (1970) find empirical support for that risk-taking is a signal of ability, and that most individuals want to be higher in abilities in comparison to others (i.e., a positional preference for ability).
The arguments presented above imply that agents can be positional in several dimensions; not only in the consumption/income dimension but also w.r.t. particular leisure activities. The latter observation constitutes the point of departure for this paper. We analyze how positional preferences w.r.t. a risky leisure activity affects the behavior and welfare in a social group where this activity is a central feature of the agents´ identity. The below analysis focuses on a leisure activity that is associated with risk of physical injury. Examples include, but are not are not limited to, activities such as mountaineering, backcountry skiing, downhill mountainbiking, scuba diving, parachuting, and long-distance running. In this context, we also analyze how policy can be used to improve the welfare within the social group. To address these issues, we set up a model where the agents have preferences over consumption, pure leisure and of their accomplishments in the leisure activity. Successfully performing a more difficult task provides the agent with a higher utility but there is a risk of failure, which increases with the difficulty level. Each agent´s identity is related to the leisure activity and depends on (i) how strongly the agent identifies herself as a, e.g., mountaineer or mountainbiker, and on (ii) how successful the agent perceives herself to be in comparison with other people in the social group.
In the below analysis, we focus on a situation where the social group consists of agents with high respectively low skill in the activity. In our base model, we assume that all agents compare their performance with the observed difficulty level performed by both high and low skilled agents. We thereafter evaluate the special case where all agents only compare themselves with the high-skilled group. This special case is consistent with arguments put forward by e.g. Eckerstorfer and Wendner (2013) who (in the context of consumption externalities) argue that it is more realistic to assume that some agents contribute more to the reference level than others. It is also supported by empirical research on positional preferences (Clark and Senik 2010), which find that most people engage in upward comparisons concerning income. Finally, Cowan et al (2004) argue that people, who do not belong to some elite group, aspire to mimic the consumption of the elite (aspiration effect) while people who belong to the elite can use consumption to outshine their peers (distinction effect).
Within this framework, we address the following topics: (i) Is there a qualitative difference in the behavioral response to an increase in the reference level associated with the positional activity when the outcome of the activity is certain compared with when the outcome is uncertain? (ii) Is there a qualitative difference in the behavioral response to an increase in the strength of the agent´s positional concern when the outcome of the positional activity is certain compared with when the outcome is uncertain? (iii) Under certainty, the keeping up with the Joneses effect implies that all agents over-consume the positional good in an unregulated equilibrium. Does the over-consumption result also apply when the positional activity is uncertain, or is it possible that some sub-group(s) may under-consume the positional good under uncertainty? (iv) What policy can be used to improve the welfare within a social group where the identity is linked to a risky leisure activity? To address the last question, we proceed along two routes. First, we pose the question of what policy instrument that would be needed to implement the first-best outcome and we show that an optimal design of fees to perform the leisure activity at different difficulty levels will be sufficient. However, a system of difficulty-adjusted fees is likely to be difficult to implement in practice because of the associated monitoring cost. Therefore, we also consider an alternative approach, namely that the risk of physical injuries means that the agent may require assistance (e.g., via rescue operations, or medical care). We show that if assistance is required, then the agent should pay an accompanying fee associated with this assistance. In the final part of the paper, we characterize the optimal design of such an assistance cost function. This paper makes two distinct contributions to the literature. The first is to relate positional preferences to a risky leisure activity with an uncertain outcome. We show that the standard keeping up with the Joneses effect (i.e. that an increase in the benchmark consumption associated with the comparison group has a positive effect on the individual agent´s consumption) may be reversed when the positionality activity is associated with an uncertain outcome. As a result, some people may under-engage in the leisure activity in the resulting equilibrium. This result can be related to the analysis conducted by Bakshi and Chen (1996) who considered the hypothesis that investors accumulate wealth to obtain wealth-induced social status. They found that when investors care about status and about "keeping up with the Joneses," the investors are less prone to take on risk. The second contribution is to show how a policy-maker can use participation fees or assistance fees to improve the welfare in a social group, which engages in a risky leisure activity. We show that a system of non-linear participation fees will be sufficient to implement the first-best 1 3 Social identity and risky leisure activities: implications… outcome whereas an optimal design of assistance fees can improve the welfare, albeit not all the way to the first-best level.
The outline of the paper is as follows. In Sect. 2, we characterize the individual agents' decision problem and in Sect. 3 we derive some comparative static results. In Sect. 4, we characterize the equilibrium within the social group and compare this outcome with the first-best outcome. Policy implications are addressed in Sect. 5 and the paper is concluded in Sect. 6.
2 The agent's decision problem Chang (2013) has pointed out that the concept of identity/self-esteem developed by Akerlof and Kranton (2000) can be used to motivate why relative income matters for subjective wellbeing (utility). Therefore, consider an agent who has preferences for consumption, c , pure leisure, z , and a risky activity. We define pure leisure as the time spent doing no activity at all. The preferences for consumption and pure leisure are captured by the sub-utility functions u(c) and v(z) , where u c , v z > 0 and u cc , v zz ≤ 0 . To keep the model as simple as possible, we assume that the agent is endowed with a fixed income w which means that consumption is given by c = w . To model the preference for the risky activity, we assume that the agent faces a continuum of activity related tasks which she can attempt to perform. Each activity/task is associated with a unique difficulty level which is captured by an index y which is monotonously increasing in the difficulty level and distributed over the interval y ∈ (0, y max ) . Activity y = 0 has zero probability of failure and is therefore risk-free. An activity with y > 0 will be labelled as risky because the probability of failure is positive. Activity y max is the most difficult task and therefore has the lowest probability of success. The subjective wellbeing/pride of successfully performing activity y is captured by the sub-utility function g(y) . We assume that this pride function has the following properties: If the agent fails in her attempt to perform a risky activity, the pride function takes the value zero.
The agent is part of a social group where all members have preferences for the risky activity. The social group consists of two ability types, 1 and 2, where ability type 2 is more skilled than ability type 1. An agent´s skill level affects the probability of successfully accomplishing a given activity y . Let p i (y) denote the probability that ability type i = 1, 2 successfully accomplishes activity y . We assume that p i (y) satisfies the following properties: The first assumption implies that there is no risk of failing the risk-free activity while the second assumption implies that the probability of success decreases with the difficulty level, but at a slower rate for the high-skilled type. The latter feature, in turn, ensures that the high-skilled type has a higher probability of successfully g(0) = 0, g y (y) > 0, g yy (y) < 0, lim y→0 g y (y) → ∞.
accomplishing a given risky activity than the low-skilled type, i.e. p 2 (y) > p 1 (y) holds for all y > 0 . Finally, by assuming that p i (y) is weakly concave, we incorporate the feature that the probability of success does not decrease at a slower rate as y becomes larger. An agent of type i derives utility, henceforth referred to as identity, from being a valued member of the social group. A person´s identity depends on how successful she is in performing the risky activity in comparison with the other members in the social group. When evaluating her performance, an individual of type i compares the difficulty level of the activity she attempts to perform with a reference level y i .
This reference level is a weighted average of the observed successful accomplishments made by the two ability types in the social group. If we let y i denote the activity that is accomplished by agent type i , then agent type i ´s reference level is determined by y i = x i y 1 + 1 − x i y 2 . Here x i ∈ [0, 1] is the weight that an agent of type i attaches to type 1 in the weighted average calculation of y i . The individual agent is myopic and treats the levels of y 1 and y 2 observed within the social group as exogenous in the weighted average calculation of y i . Agent type i ´s relative performance is captured by 2 Δ i = y i − y i and the agent´s identity, I i , is a function of this performance measure; I i = I Δ i . This identity function is increasing and concave in Δ i , and we use the normalization I(0) = 0 . For the analysis below, we note that I i y = −I i y < 0 , I i yy = I i yy < 0 and I i yy = −I i yy > 0. There is a time cost, h y i , associated with performing activity y i . We assume that this cost function satisfies h y > 0 and h yy ≥ 0 . The weak convexity reflects that it usually takes more time to prepare for, and to perform, a more difficult activity. Normalizing the time endowment to one, the time constraint can be written as z i = 1 − h y i . The overall utility associated with successfully performing activity y i is specified as follows 3 : where the parameters i and i reflect how important pride and identity are for the agent´s overall utility. Note that U i yy = i I i yy > 0 , i.e. the cross-derivative of the utility function w.r.t. the positional activity ( y ) and the reference level ( y ) is positive. By using an analogous definition as Dupor and Liu (2003), 4 we define U i yy > 0 to imply that the agent has "keeping up with the Joneses" preferences. For the analysis This measure is analogous to that used in much of the literature on positional consumption preferences. See e.g. Corneo and Jeanne (1997); Ljungqvist and Uhlig (2000); Aronsson and Johansson-Stenman (2008;. 3 Without loss of generality, we assume that w is the same for both agent types. 4 In a framework where agents have preferences over private consumption (the positional good), percapita consumption (the reference level) and leisure, Dupor and Liu (2003) define preferences to exhibit "keeping up with the Joneses" if the marginal rate of substitution between leisure and consumption increases with per-capita consumption. If the preferences are additively separable between, on one hand private consumption and per-capita consumption, and on the other hand leisure, Dupor and Liu point out that the definition above is equivalent with assuming that the cross-derivative of the utility function w.r.t. private consumption (the positional good) and per-capita consumption (the reference level) is positive.
3
Social identity and risky leisure activities: implications… below, let us define an agent´s degree of positionality w.r.t. successfully performing the risky activity as follows 5 : This measure shows the fraction of the overall utility increase from a marginal increase in y that is due to increased relative performance. The degree of positionality is zero if only absolute performance matters i > 0, i = 0 and one if only relative performance matters i = 0, i > 0 . When both absolute and relative performance matter i > 0, i > 0 , the degree of positionality takes a value between zero and one.
If the agent fails in her attempt to perform the activity of her choosing, the realized value of y is zero. In this situation, the pride function takes the value zero while the identity function takes the value I i,f = I 0 − y i < 0 where the super-index "f" stands for failure. Since the agent nevertheless has spent time preparing for, and attempting to perform the activity, the time cost h y i is still present. 6 The realized overall utility in the case of failure then becomes 7 : The agent chooses y i to maximize the expected utility Ũ i = p i U i + 1 − p i U i,f . Substituting the definitions of U i , U i,f and p i into Ũ i , and maximizing the resulting expression w.r.t. y i produces the first-order condition: Equation (4) shows the trade-off the agent faces when choosing y . At the optimum, the expected marginal benefit ( EMB ) of attempting to perform a more difficult activity is balanced against the marginal utility cost ( MUC ) associated with (3) 5 In models where an agent´s preferences are defined over private consumption, $$c$$, and relative consumption, $$\Delta =c-\stackrel{-}{c}$$, so that the utility function is given by $$U\left(c,\ Delta \right)$$, an agent´s degree of consumption positionality is defined as $$\alpha ={U}_{\Delta }/\ left({U}_{c}+{U}_{\Delta }\right)$$ (see e.g. Aronsson and Johansson-Stenman (2010)). The definition in Eq.
(2) is analogous with this definition. 6 One could also incorporate a direct utility cost associated with failure if failure is associated with injury. Including such a cost would not affect the qualitative results derived below. 7 It is also conceivable that attempting to climb a difficult mountain but failing may provide the agent with some satisfaction in the sense that "at least I tried". If this is the case then the sense of pride would still be there albeit not as strong as in the case of success. One way to model this feature would be to write the pride function in the case of failure as g(ky) , where k is a positive scaling factor which is situated in the interval 0 ≤ k < 1 . If k = 0 , then g(0) = 0 (as we have it in the text) but if k is nonzero, then 0 < g(ky) < g(y) . If the "at least I tried" argument would also apply to the identity function, then the identity associated with failure would be modified to read I ky − y . If we were to use these alternative specifications then the agent would, all else equal, have an incentive to choose a larger y than otherwise.
choosing a more difficult alternative plus the marginal utility loss associated with giving up pure leisure time ( −v i z h i y ). The MUC reflects that attempting to perform a more risky activity, on the margin, reduces the probability of success, where failure implies a utility loss which is given by the optimal choice which satisfies Eq. (4). This first-order condition implicitly defines y i,• as a function of i , i and y i , i.e. y i,• i , i , y i . In the Appendix, we show that y i,• > 0 as long as i > 0 . As for the second-order derivative, it is given by: Given the assumptions made above, it follows that the second-order condition Ũ i yy < 0 is satisfied which implies that the optimum is a global maximum. It can also be shown that if the two agent types are identical in all aspects except in the probability of success (i.e. 1 = 2 , 1 = 2 and y 1 = y 2 ) then y 2,• > y 1,• , i.e. the highskilled agent will choose to perform a more difficult activity than the low-skilled agent.
Behavioral effects
Let us now evaluate how an individual agent´s optimal activity choice is affected by changes in y i and i . A change in y i may reflect a changed behavior within the reference group while i may be interpreted as a taste shock whereby the agent attaches more importance to being a valued member of the social group. We begin with y i .
Differentiating Eq. (4) w.r.t. y i and y i,• produces: As a point of reference, consider first the outcome in the absence of uncertainty. In this situation p i = 1 and p i y = 0 for all y i ≥ 0 , in which case Eq. (6) reduces to This comparative static result reflects that an increase in y i , all else equal, has a negative impact on Δ i = y i − y i which, in turn, pushes up the marginal identity utility, I i y . When there is no risk of failure, the agent responds by choosing a higher y i . This result is a consequence of the fact that the agent has keeping up with the Joneses preferences along the lines defined by Dupor and Liu (2003), i.e. that U i yy = i I i yy > 0 . We therefore refer to this response as a conventional keeping up with the Joneses effect.
In the presence of uncertainty, the outcome is no longer clear-cut because the RHS of Eq. (6) is made up of two terms. The first −p i i I i yy ∕Ũ i yy > 0 captures the
3
Social identity and risky leisure activities: implications… Joneses effect discussed above while the second 8 −p i y i I i y − I i,f y ∕Ũ i yy < 0 reflects that an increase in the reference level magnifies the utility loss associated with failure. This potential utility loss provides the agent with an incentive to play it safe by choosing a lower y i than otherwise in order increase the probability of success. We will refer to this as the Cautionary effect of attaching more importance to an identity linked to a risky activity.
We illustrate the incentive underlying the Cautionary effect in Fig. 1 below where we, for notational convenience, omit the super-index i . Assume that the agent initially has made an optimal choice y • 0 conditional on the observed initial reference level y 0 (here the sub-index "0" refers to the initial value of a variable). The identity utility associated with success is therefore initially given by I y • 0 − y 0 which corresponds to point A in Fig. 1. If the reference level now increases to y 1 > y 0 (where the sub-index "1" refers to the subsequent value of a variable after a change has occurred), this affects the identity associated with success and moves the agent from point A to point B. At point B, the identity utility is given by I y • 0 − y 1 before reoptimization takes place (i.e. before the agent has revised her initial choice y • 0 ). The increase in the reference level also affects the identity associated with failure which is reduced from I −y 0 at point C, to I −y 1 at point D. Let (A − C) and (B − D) denote the vertical distance between points A and C, and points B and D, respectively, in Fig. 1. Since the identity function is concave, it follows that: This means that the identity loss associated with failure is amplified when the reference level increases which, ceteris paribus, increases the overall utility loss associated with failure; U − U f = g + I − I f . This provides the agent with an incentive to play it safe by choosing a lower y and is the rationale underlying the Cautionary effect.
Whether the Joneses effect dominates over the Cautionary effect depends on the properties of the identity and probability functions. Let us introduce the following short notations: (i) let i = −y i I i yy ∕I i y > 0 be a measure of how sensitive the slope of the identity function is to a change in the reference level, (ii) let i = − I i y − I i,f y ∕I i y > 0 measure the relative difference in the slope of the identity function between the point of success and the point of failure and (iii) let i,p = −y i p i y ∕p i > 0 be the elasticity of the probability function w.r.t. y i . By using these definitions, we can rewrite Eq. (6) to read where i = −p i ∕y iŨi yy > 0 . Since − i i I i y > 0 , we obtain the following results from (8): Proposition 1: Consider an increase in the reference level, y i . The effect on agent type i's optimal choice y i,• can be summarized as follows: In the Appendix, we use specific functional forms to exemplify when i − i i,p is positive/negative at the optimum.
Let us now turn to the effects of an increase in the importance attached to being a valued member of the social group, i.e. an increase in the parameter i . Differentiating Eq. (4) w.r.t. i and y i,• produces:
3
Social identity and risky leisure activities: implications… Also here we begin by noting that in the absence of uncertainty, Eq. (9) reduces to y i,• ∕ i = −I i y ∕U i yy > 0 . This response reflects that when the importance attached to the identity utility increases, then the marginal benefit of choosing a higher y i increases which induces the agent to choose a larger y i,• . Although this effect is not directly related to the sign of the cross-derivative U i yy = i I i yy , we will nevertheless refer to this response as a keeping up with the Joneses effect.
In the presence of uncertainty, the comparative static effect in (9) is made up of two components; −p i I i y ∕Ũ i yy > 0 which is the Joneses effect and −p i y I i − I i,f ∕Ũ i yy < 0 which reflects that if the agent attaches more importance than before to her identity, then the utility loss associated with failure will be magnified. 9 This potential utility loss provides the agent with an incentive to play it safe. As such, this is a Cautionary effect associated with attaching more importance to the personal identity.
Let us define i,I = I i y y i ∕ I i − I i,f > 0 to be the elasticity of the identity rent of success w.r.t. y i , where we use that I i y = I i − I i,f ∕ y i . By using the definition of i,I together with the definitions of i,p and i made above, we can rewrite Eq. (9) to read: We can now use this equation to evaluate when Joneses effect dominates over/is dominated by the Cautionary effect. The results can be summarized as follows: Proposition 2: Consider an increase in the importance, i , that an agent of type i attaches to her identity. The effect on the optimal choice y i,• can be summarized as follows: In the Appendix, we use specific functional forms to exemplify when i,I − i,p is positive/negative at the optimum.
3
An interesting special case arises if i = 0 , in which case the degree of positionality takes the value one 10 : Corollary: If only relative performance matters ( i = 1), then the first-order condition can be satisfied only if i,I > i,p holds at the optimum. Then y i, To show this result, note that when i = 0 , the first-order condition in (4) reduces to: where we have used the definitions of i,I and i,p to rewrite the first order condition on the form presented in the second row. From the latter expression, it follows that the first-order condition can only be satisfied if i,I > i,P holds at the optimum, in which case we know from part (i) in Proposition 2 that y i,• ∕ i > 0.
The myopic equilibrium vs the first-best outcome
A common result in the literature on positional preferences is that the incentive to keep up with the Joneses compels agents to consume more than they otherwise would have done which has a negative impact on welfare. The question we pose in this part is to what extent the over-consumption result holds when the positional activity has an uncertain outcome? To address this question, we compare the equilibrium with myopic agents with a first-best outcome.
Let us begin with the myopic equilibrium (ME). Recall that agent type i ´s demand function y i,• i , i , y i is implicitly determined by the first-order condition in Eq. (4). For convenience, we state the first-order conditions for the two agent types here: 10 We are thankful to an anonymous referee for making us aware of this possibility.
3
Social identity and risky leisure activities: implications… If we substitute y 1 = x 1 y 1,• + 1 − x 1 y 2 into (12), the resulting equation implicitly defines y 1,• as a function of y 2 , as well as a function of 1 , 1 and x 1 . Let us denote this reaction function RF 1 = y 1,• RF y 2 , 1 , 1 , x 1 . The other agent type´s reaction function is denoted RF 2 = y 2,• RF y 1 , 2 , 2 , x 2 and it is analogously defined by substituting y 2 = x 2 y 1 + 1 − x 2 y 2,• into Eq. (13). Differentiating RF 1 and RF 2 produces the following comparative static results (slopes) 11 : From these equations it follows that as long a x i y i,• ∕ y i < 1 , then the slope of agent type i ´s reaction function is determined by the sign of the partial derivative y i,• ∕ y i . Since we know from Proposition 1 that the sign of y i,• ∕ y i depends on whether or not the Joneses effect dominates over the Cautionary effect, it follows that also the slope of agent type i ´s reaction function depends on these effects. The myopic equilibrium is at the point where the two reaction functions intersect, and the equilibrium levels are denoted y 1,ME and y 2,ME . Figure 2 below illustrates the outcome when the Joneses effect dominates over the Cautionary effect for both agent types (i.e. y i,• ∕ y i > 0 holds for i = 1, 2 ), and the two reaction functions have positive slopes. In Fig. 2, we see that the two reaction functions will intersect, and the (14) Social identity and risky leisure activities: implications… myopic equilibrium will exist, if RF 2 has a flatter slope in y 1 , y 2 space than RF 1 . In the Appendix, we show that this happens if y i,• ∕ y i < 1 for i = 1, 2. 12 Figures 3a, b below illustrate two possible outcomes when instead it is the Cautionary effect which dominates over the Joneses effect for both agent types (i.e. y i,• ∕ y i < 0 for i = 1, 2 ), in which case the two reaction functions are negatively sloped.
Let us now turn to the first-best (FB) outcome. It is obtained by considering a social planner who maximizes the expected welfare within the social group. The expected utilitarian welfare function is given by W = N 1Ũ1 + N 2Ũ2 , where N i is the number of agents of type i within the social group. The social planner chooses y 1 and y 2 to maximize W , while taking into account how these choices affect the reference levels y 1 = x 1 y 1 + 1 − x 1 y 2 and y 2 = x 2 y 1 + 1 − x 2 y 2 . Substituting the latter two equations into the welfare function and maximizing w.r.t. y 1 and y 2 produces the following necessary conditions associated with the first-best optimum: To depict the first-best outcome y 1,FB , y 2,FB , we can use that Eq. (15) implicitly defines y 1 as a function of y 2 while Eq. (16) implicitly defines y 2 as a function of y 1 . Let us refer to these functional relationships as social reaction functions, and denote them SRF 1 and SRF 2 , respectively. Since the social planner accounts for the fact that y 1 and y 2 produce negative externalities on all agents in the social group via the reference levels y 1 and y 2 , while this is not taken into account by the myopic private agents, it follows that SRF 2 will be situated below RF 2 in y 1 , y 2 space while SRF 1 will be situated to the left of RF 1 in y 1 , y 2 space. The social reaction functions are depicted in Figs. 2, 3a, b below, and the first-best outcome is at the point where SRF 1 and SRF 2 intersect. Figure 2 illustrates the outcome when the Joneses effect dominates over the Cautionary effect. In this situation, there is over-consumption in the myopic equilibrium in the sense that y i,FB < y i,ME holds for i = 1, 2 . This over-consumption result is analogous to results derived in earlier literature on positional preferences, where it has been shown that inter-personal comparisons induce agents to engage in wasteful competition, thereby inducing agents to over-consume positional goods.
If, instead, the Cautionary effect dominates over the Joneses effect, the outcome is no longer clear-cut. Figure 3a illustrates an outcome where both agent types overconsume the acticity while Fig. 3b illustrates an outcome where one of the agent types, type 1 in this example, under-consumes the activity.
Let us take a closer look on the case where there is under-consumption among one of the agent types. To do this, let us consider the empirically interesting case where all members in the social group only compare themselves with the elite, which is the group of high-skilled agents of type 2 (see e.g., Cowan 2004;Clark and Senik 2010;Eckerstorfer and Wendner 2013). This implies x 1 = x 2 = 0 and y 1 = y 2 = y 2 , in which case the low-skilled agents do not contribute to the positional externality and the social first-order condition for y 1 reduces to W ∕ y 1 =Ũ 1 y = 0 . Since Ũ 1 y = 0 is the equation which implicitly defines the RF 1 , it follows that SRF 1 coincides with RF 1 when y 1 = y 2 = y 2 . Since RF 1 has a negative slope when the Cautionary effect dominates over the Joneses effect for agent type 1, it follows that also SRF 1 has a negative slope in this special case. As for the other social reaction function SRF 2 , we note that y 1 does not appear in the social first-order condition for y 2 when y 2 = y 2 . Therefore Eq. (16) defines the first-best level y 2,FB independently of y 1 . Hence, SRF 2 is a horizontal line which is situated below 13 RF 2 in y 1 , y 2 space. The outcome is illustrated in Fig. 4 where y 1,FB > y 1,ME and y 2,FB < y 2,ME . Hence, on a free market, positional preferences for risky activities may induce some agents to under-consume the leisure activity in the sense that the agent chooses a difficulty level which is too low from the perspective of the social planner. This result is consistent with research in social psychology, according to which individuals with a lower self-perceived status (e.g. due to a lower skill level) tend to avoid risk as a self-protective action (Baumeister et al. 1989;Wolfe et al. 1986).
Fig. 4 ME and FB under Cautionary dominance and elitist social preferences
13 Note that also RF 2 is a horizontal line in y 1 , y 2 space when y 2 = y 2 .
3
Social identity and risky leisure activities: implications… Since y 1,• RF ∕ y 2 = y 1,• ∕ y 1 when x 1 = x 2 = 0 , and since we know from Proposition 1 that the sign of y 1,• ∕ y 1 is determined by the sign of 1 − 1 1,p , we can combine these observations to obtain the following results: Proposition 3: Consider a social group made up of two agent types i = 1, 2, where all agents have positional preferences w.r.t. a risky leisure activity and compare themselves with the high-skilled agents of type 2. Compared with the first-best outcome, the outcome in the myopic equilibrium has the following properties: (i) In the myopic equilibrium, the high-skilled agent´s activity choice exceeds the first-best choice, i.e. y 2,ME < y 2,FB (over-consumption). (ii) If 1 < 1 1,p , the low-skilled agent´s activity choice in the myopic equilibrium falls short of the first-best choice, i.e. y 1,ME < y 1,FB (under-consumption). (iii) If 1 > 1 1,p , the low-skilled agent´s activity choice in the myopic equilibrium exceeds the first-best choice, i.e. y 1,ME > y 1,FB (over-consumption).
Policy implications
The positional externality has a detrimental effect on the welfare for all agents in the social group. From earlier studies, it is well known that public policy can be used (at least partially) to internalize the positional externality generated by interpersonal comparisons of income or consumption (e.g. Johansson-Stenman, 2008, 2010). In this part, we ask whether public policy also can play a role when the positional externality is related to a risky leisure activity? We focus on the case where all members in the social group compare themselves with high-skilled agents but it is straightforward to extend the results presented below into a more general framework. We proceed along two routes. First, we ask what type of policy instrument a policymaker needs to have at her disposal in order to implement the first-best policy. In the conventional literature on positional consumption goods, policy-makers can use taxes on income and/or positional commodities to improve the welfare. Unfortunately, such tax instruments may not offer viable alternatives when agents have positional concerns for leisure activities. Instead, we will use the fact that some risky leisure activities take place on publicly owned land. If the areas where agents perform their activities (e.g. trails or rivers) can be assigned unique index values based on their respective difficulty levels, and if the policy-maker can implement and monitor payment of fees, then the policy-maker has the possibility to impose an entrance fee to access an activity with a given index level y . These fees can e.g. be entrance fees to national parks, trailfees for mountain bikers/hikers or participation fees for long-distance running races. In Sect. 5.1, we show that an optimal design of this entrance cost function makes it possible to implement the first-best outcome defined in Section IV. However, a drawback of this approach is that it requires the policy-maker to monitor the activities on all public land. In addition, it does not include activities on private land. Therefore, we in Sect. 5.2, instead base the optimal policy on an alternative approach; namely that if public assistance or a rescue effort is required to bring the agent back to safe grounds, then he/she should pay for this effort.
First-best policy: an optimal entrance cost function
Assume that the policy-maker is able to implement and monitor a system where the agents are required to pay a fee to perform an activity with index/difficulty level y . We allow the fee facing the agents to be a nonlinear function of the activity´s difficulty index levels, (y) , and assume that the revenue generated by the entrance fees is used to cover some fixed expenditure E . These expenditures can, for example, be used on safety arrangements (e.g. shelters, emergency phones, maintainance of trails, etc.). The policy-maker´s budget constraint therefore becomes E = ∑ i N i � y i � and as mentioned above, we focus on the case where all agent compare themselves with the elite, i.e. y 1 = y 2 = y = y 2 .
The inclusion of an entrance cost implies that the private agent´s budget constraint is modified to read c i = w − y i . By using this modified budget constraint in the private agent´s decision problem, it is straightforward to show that the first-order condition for the optimal choice of y i in Eq. (4), now contains an additional term; − i y u i c . If the marginal entrance cost ( i y ) facing an agent of type i is positive, the additional term in the private first-order condition is negative and contributes to increase the marginal utility cost of attempting to perform a more difficult task. This modification of the model implies that in addition to being a function of i , i and y i , agent type i ´s optimal choice y i,• now also becomes a function of w and the parameters of the entrance cost function.
The policy-maker maximizes the social welfare function W = ∑ i N iŨi subject to the public budget constraint E = ∑ i N i � y i � . The decision variables are the parameters of the entrance cost function and the policy-maker recognizes the private time constraint in the maximization problem. It is also assumed that the policy-maker knows the proportion of each skill-type within the population (i.e. N i ∕ ∑ i N i is known) but the skill level of an individual agent may not be known by the policy-maker. This information asymmetry means that an agent of type i has the possibility to choose to perform the task with difficulty level y j,FB , where j = 1, 2 and i ≠ j , instead of difficulty level y i,FB where the latter is the first-best choice intended for agent type i . To prevent this behavior, we need to impose an incentive compatibility (IC) constraint for each agent type. However, it can be shown that these IC constraints will not be binding for any of the agent types as long as the first-best policy features y 1,FB < y 2,FB and this is the case that we will focus on. The Lagrange function associated with the policy-maker´s maximization problem can therefore be written as follows: where and are Lagrange multipliers associated with the policy-maker´s budget constraint and the reference level, respectively. By writing the reference level as an explicit Lagrange restriction, y will be treated as an additional (and artificial)
3
Social identity and risky leisure activities: implications… decision-variable in the policy-maker´s maximization problem. This approach makes it possible to define the quotient of Lagrange multipliers ∕ as the shadow price (in real terms) of y . This shadow price is interpretable as the marginal value that the policy-maker attaches to a reduction in the reference level, measured in terms of the revenue generated by the entrance fees. Before characterizing the optimal policy implied by the social first-order conditions, we recall that the impact of an increase in the reference level on agent type i ´s expected utility is determined by We can therefore use this equation to define agent type i ´s marginal willingness to pay for a reduction in the reference level as: where we in the second step have used that Ũ i w = u i c at the optimum. In the Appendix, we show that the policy outlined in the following Proposition replicates the first-best outcome defined in Sect. 4;
Proposition 4: The following policy reproduces the first-best choices:
The first equation in Proposition 4 shows that the shadow price is equal to the sum of the marginal willingness to pay for a reduction in the reference level. This shadow price constitutes the basis for constructing the optimal marginal entrance cost facing the high-skilled agent, which is determined by the third equation in Proposition 4. That equation shows that the positional externality is fully internalized when the marginal entrance cost facing the high-skilled agent is set equal to the sum of all agents´ marginal willingness to pay for a reduction in the reference level, scaled by N 2 . The scaling of the shadow price reflects that the individual highskilled agent only contributes a fraction 1∕N 2 to the positional externality. Finally, the second equation in Proposition 4 shows that there is no motive to distort the lowskilled agent´s choice as long as the marginal entrance cost facing the high-skilled agent fully internalizes the positional externality. Hence 1 y = 0.
Second-best policy: an optimal assistance cost function
The monitoring cost of implementing the optimal entrance cost function is likely to be overwhelmingly high. An alternative is to introduce a fee that is activated when things go wrong. Many types of risky leisure activities, including mountaineering, hiking, backcountry skiing, mountain-biking etc. take place in relatively remote areas. A failure in these activities can involve getting lost, injured, or energy depleted. The remote location, and rough terrain, means that rescue operations are costly. In many western economies, taxes fund these rescue operations. Now, suppose that an agent who needs public assistance to get to safety, will have to pay a fee for the effort needed to carry out the operation. In this part, we analyze the optimal design of such an assistance cost function. To address this issue, we modify the model outlined above as follows. First, we make the assumption that if an agent of type i fails to perform a task with difficulty level y i , she needs assistance with probability 1 − q i y i , and does not need assistance with probability q i y i . We assume that q i y i is type specific and decreasing in y i . If assistance is needed, the agent pays a fee where the size of the fee depends on the difficulty level of the task; i = y i . As in the previous section, we allow the assistance cost function to be a function of y i . If an agent succeeds with the activity or if she fails but does not need assistance, the private budget constraint is given by c i = w . However, if the agent fails with the task and needs assistance, then the private budget constraint is modified to read c i,f = w − y i . The utility associated with successfully preforming the activity y i , is defined by Eq. (1) while the outcome of failure is now uncertain and depends on whether assistance is needed or not. The expected utility associated with failure, denoted Ũ i,f , is given by: where q i = q i y i . The agent chooses y i to maximize Ũ i = p i U i + 1 − p i Ũ i,f and the private first-order condition becomes: where u = u(w) and u i,f = u w − y i . The terms in the first row are the same as those appearing in Eq. (4) while the terms in the second row are novel. The first term in the second row reflects that in the case of failure, there is now a risk of needing assistance which is associated with a cost. If the marginal assistance cost i y is positive, the first term in the second row contributes to increase the marginal utility cost of attempting to perform a more difficult activity. The second additional term, which is negative, reflects that attempting to perform a more difficult task is associated with a higher probability of needing assistance, where u − u i,f is the additional utility loss that is incurred if assistance is needed.
Turning to the policy-maker, we first note that the expected number of assistance missions provided to agents of type i is given by Ñ i y i = 1 − p i y i 1 − q i y i N i . The cost of assisting an individual agent who has tried, but failed to perform y i is captured by a cost function C y i which is increasing and convex in y i , i.e. C i y = C y y i > 0, C i yy = C yy y i > 0 . Since each agent who receives assistance after having attempted task y i pays the fee y i , the budget constraint that the policymaker (in expectation) faces when she determines the parameters of the assistance cost function can be written as The Lagrange function associated with the policy-maker´s maximization problem is specified as follows:
3
Social identity and risky leisure activities: implications… We solve this maximization problem in the Appendix. Before we present the results, define: where we have used that u i,f w = u f w w − y i > u w (w) to obtain the inequality on the RHS in (21). After these preliminaries, we are able to summarize the optimal second-best policy as follows: Proposition 5: The optimal design of the assistance cost function has the following characteristics: Since i < 1 , the first equation in Proposition 5 implies that the shadow price in the second-best optimum is not equal to the sum of the marginal willingness to pay for a reduction in the reference level. Instead ∕ < ∑ i N i MWP i y,w . To explain this result, we recall that the shadow price reflects the marginal value that the policymaker attaches to a reduction in the reference level, measured in terms of the revenue generated by the assistance fees. At the optimum, the optimal second-best policy will feature the equalization of the social marginal value of raising one extra dollar vis-à-vis the private marginal utility cost of giving up one extra dollar. Since the policy-maker is restricted to raise revenue from agents who need assistance, the private marginal utility cost of giving up one extra dollar will be given by u i,f w = u f w w − y i . Hence, the optimal second-best policy features = u i,f w for i = 1, 2 and since the agents who do not need assistance do not pay any fee, it follows that u w (w) < u i,f w for i = 1, 2 . The inability to distribute the cost of assistance equally between the two relevant states of nature (assistance needed or not needed) for an agent of type i means that the social cost of raising one extra dollar ( ) will be larger in the second-best than in the first-best. As a consequence, the shadow price will be smaller in the second-best than in the first-best, where we recall that the shadow price in the first-best equals the sum of the marginal willingness to pay for a reduction in the reference level. This explains why ∕ < ∑ i N i MWP i y,w in the second-best optimum.
Let us now turn to the optimal marginal assistance cost facing the low-skilled agent. The second equation in Proposition 5 shows that the marginal assistance cost facing the low-skilled agent, 1 y , is made up of two parts. The first part, C 1 y , reflects that the marginal assistance fee should be equal to the actual marginal assistance cost at the optimum. By setting 1 y equal to C 1 y , the low-skilled agent is provided with the correct incentive when it comes to internalizing the cost associated with assisting the agent. The second term in the formula for 1 y captures a revenue effect; if the low-skilled agent´s net revenue contribution is positive (negative), such that 1 > C 1 ( 1 < C 1 ), then the policy-maker has an incentive to increase (decrease) the tax base, i.e. increase (decrease) Ñ 1 y 1 , by implementing a smaller (larger) marginal assistance fee than otherwise. 14 Note that neither of these terms appeared in the corresponding formula defined in the previous section for the marginal entrance cost facing the low-skilled agent. 15 The first two terms in the formula for the marginal assistance cost facing the high-skilled agent, 2 y , that is presented in the third equation in Proposition 5 are similar to, and can be interpreted in a similar way, as the terms that appear in the formula for 1 y . The third term in the expression for 2 y serves the purpose of (partly) internalizing the positional externality. As in Sect. 5.1, the shadow price constitutes the basis for constructing the corrective part of the marginal assistance cost facing the high-skilled agent. Note, however, that the scaling differs compared with that which appears in the first-best. In the first-best, the shadow price is scaled by N 2 whereas the shadow price is scaled by Ñ 2 in the second-best. The explanation for why the shadow price is scaled by Ñ 2 in the second-best is that it is not certain that a high-skilled agent will need assistance. Therefore, the corrective part of the marginal assistance cost needs to be higher than otherwise in order to induce the highskilled agent to choose a level of y 2 closer to the first-best. However, since the private marginal utility cost of giving up one extra dollar is higher in the second-best than in the first-best (see the discussion following Proposition 5), the corrective part of the marginal assistance fee will nevertheless not be sufficient to implement the first-best outcome.
Concluding discussion
In this paper, we have developed a model where agents have positional preferences for a risky leisure activity, which holds a social value. Our theoretical analysis suggests that positional preferences can both increase and decrease risk-taking behavior. This result is in stark contrast to those derived under certainty where positional preferences always increases investments in the positional activity. We find that the elasticity of the probability of success w.r.t. the positional activity determines how the agent responds. If changes in the probability of success have a small impact on behavior, then positional preferences increases engagement in the risky activity. For sufficiently large values of the elasticity of the probability of success, however, the 1 3 Social identity and risky leisure activities: implications… agent instead responds by reducing her level of risk exposure (i.e. the agent responds by attempting to perform a less difficult task). The latter result has implications for the social equilibrium. To illustrate the effect, we focus on the empirically relevant case where all individuals compare their performance solely with the performance of high-skilled agents. Our policy analysis shows that, while positional high-skilled agents will attempt to perform too difficult tasks compared with the first-best outcome, the outcome for the low-skilled agents is ambiguous. If low-skilled agents choose to implement a self-protective strategy in the presence of interpersonal comparisons, they may end up attempting to perform less difficult tasks compared with the first-best outcome.
Our policy analysis suggests that a system of entrance fees, indexed to the difficulty level of the activity, can fully mitigate positional externalities. However, such a system is likely to be difficult to implement in practice. Another possibility is to link fees to rescue operations but since the assistance cost function is not able to target the agents´ decisions perfectly, the optimal design of this function would not be a sufficient policy tool to implement the first-best outcome.
Our analysis builds on a stylized model of a complex phenomenon. We would therefore like to discuss some limitations, and possible directions for future research. First, our model only includes two ability types. If the social group instead would consist of three or more types, then the interpersonal comparisons may become more intricate. For example, the lowest ability type may not aspire to mimic the behavior of the highest ability type but to the middle type, or an average of the middle and high ability type. In these situations, corrective policy is potentially much harder to implement because of the informational burden required to assess the interpersonal comparison patterns. If the corrective policy cannot be based on actual assessments of how individuals make comparisons between themselves, but instead is based on some (maybe imperfect) estimate of these patterns, an interesting question for future research is to what degree corrective policy in such situations could lead to improvements in welfare.
Second, we assume that agents compare the difficulty of their attempted task with a weighted average of the difficulty levels chosen by other agents. Since the activity is risky for all agents, some proportion of attempted tasks will fail. An interesting question is therefore if agents compare with the attempted difficulty level among others, or if the success rate matters? 16 Under incomplete information, positional agents have incentives to only display successful attempts (e.g., via social media). It is also possible that some agents send out false signals of success (e.g., a photo taken before giving up). This creates a bias in the informational flow, which forces agents to make guestimates of other agents' behavior. Since our recall is vulnerable to the availability heuristic (Tversky and Kahneman 1973), it is likely that agents will overestimate the frequency of successful attempts. This may affect agents' perception of both the success rate of others, and the objective risk associated with the activity. An interesting question for future researchers is therefore how policy interventions targeted at information about base frequencies and accident rates affect risk-taking behavior among positional agents with different skill levels. A complicating factor for this type of research is that agents may both engage in strategic ignorance (e.g., Thunström et al. 2016) and fall pray for optimism bias (e.g., DeJoy 1989).
Third, research in psychology on social identity threat (e.g., Scheepers and Ellemers 2005;Scheepers et al. 2009) suggest that members of high-status groups experience negative affect and heightened stress levels when their status is is threatened by the relative performance of low-status individuals, i.e., when low-status individuals catch up. In our setting, this implies that high skilled agents may compare with low-skilled agents, if low-skilled perform at a level that is sufficiently close to the performance level of the high-skilled. A number of studies (see Collins 2000 for a review) further suggest that upward comparisons can, in some cases, be associated with increased wellbeing. Positive utility effects arise when low-status agents see the performance of high-status agents as a reflection of a possible future for themselves. Such perceptions are only credible if the difference in performance is small enough. Taken together, this suggests that the utility function related to relative status concerns is more complex than the simple functions commonly used by economists. Our analysis thus represent a special case and a stark simplification. For future research, it would be very interesting to evaluate how e.g., a non-linear cost-function for downward comparisons among the high-skilled, affect the analytical results.
Appendix
Proof that y i,• > 0 when ˇi > 0 To show that y i,• = 0 cannot be an optimal solution to the agent´s maximization problem, let us evaluate the derivative Ũ i y at y i = 0 Here we have used p i (0) = 1 , g(0) = 0 and I 0 − y i = I f 0 − y i to simplify (21). Since the first term on the RHS approaches plus infinity (because lim y→0 g y (y) → ∞ ), while the remaining terms on the RHS are finite, it follows that Ũ i y | y i =0 > 0 . Hence, y i = 0 cannot satisfy the first-order condition Ũ i y = 0 . Instead, the concavity of the objective function implies that the optimal choice features y i,• > 0 and g i,• = g y i,• > 0 . If the agent does not have preferences for personal pride, i.e. i = 0 , then (21) reduces to In this special case, one cannot rule out that the two negative terms on the RHS in (22) are larger in absolute value, or equal to, the positive term on the RHS, in which case the agent would choose the corner y i,• = 0.
3
Social identity and risky leisure activities: implications… Proof that y 2,• > y 1,• when the agents are identical in all aspects except in the probability of success When the two agent types are identical in all aspects except in the probability of success, then the low-skilled agent type´s optimal choice y 1,• satisfies the first-order condition in (4) for 1 = 2 = , 1 = 2 = and y 1 = y 2 = y . Can the level y 1,• also satisfy the corresponding first-order condition for the high-skilled type? To address this question, let us substitute y 2 = y 1,• into the expression for Ũ 2 y . Then evaluate the difference Ũ 2 y | y 1,• −Ũ 1 y | y 1,• . We obtain where we on the LHS have used that Ũ 1 y | y 1,• = 0 . From the properties of the probability function, it follows that p 2 y 1,• − p 1 y 1,• > 0 and p 2 y y 1,• − p 1 y y 1,• > 0 . Hence, the expression on the RHS in (23) is positive. We therefore conclude that Ũ 2 y | y 1,• > 0 which, in turn implies that choosing y 2 = y 1,• cannot be an optimal solution for the high-skilled agent. Instead, the concavity of the objective function implies that the optimal choice of the high-skilled agent satisfies y 2,• > y 1,• .
Examples when − " p and " I − " p are positive/negative at the optimum To exemplify when − p and I − p are positive/negative at the private optimum, we use the following functional forms where we set y max = 1 . By using these functional forms, it follows that We set = 1 , = 0.01 , = 0.1 and y = 0.5 . The first-order condition in Eq. (4) can then be written as Equation (24) implicitly defines a negative relationship between the optimal choice y • and the marginal utility of leisure, k.
Let us first consider a value for y • such that > p and I > p hold at the private optimum. This can be achieved if we, for example, set k = 0.6186120962351367 in which case y • ≈ 0.20 satisfies (24). This implies Hence, > p and I > p . Next, let us consider a value for y • such that < p and I < p hold at the private optimum. This can be achieved if we, for example, set k = 0.00682772130532 in which case y • ≈ 0.47 satisfies (24). This implies Hence, < p and I < p .
The reaction functions and the myopic equilibrium
Substituting y 1 = x 1 y 1,• + 1 − x 1 y 2 into Eq. (12) and differentiating w.r.t. y 1,• and y 2 produces Note that we can rearrange (6) to read p 1 1 I 1 yy + p 1 Use this in (25) The expression after the second equality in (26) is the first equation in (14). The second equation in (14) is derived analogously.
A sufficient condition for the existence of a myopic equilibrium with positively sloped reaction functions is that y i,• ∕ y i < 1 holds for i = 1, 2 . To show this, recall first that as long as i > 0 , then an agent´s optimal choice features y i,• > 0 . This implies that the intercept of RF 1 with the horizontal axis in y 1 , y 2 space features y 1,• > 0 while the intercept of RF 2 with the vertical axis in y 1 , y 2 space features y 2,• > 0 . This is illustrated in Fig. 2 and implies that RF 2 is initially situated above RF 1 in y 1 , y 2 space. If RF 2 has a flatter slope than RF 1 in y 1 , y 2 space, the two functions will eventually intersect, as illustrated in Fig. 2, and this point is the myopic equilibrium. This implies that the myopic equilibrium will exist if the following two conditions are satisfied
3
Social identity and risky leisure activities: implications… Note that the inequality in (28) is satisfied if y 2,• ∕ y 2 < 1 , because then both the first term after the equality sign, 1 − x 2 ∕ 1 − x 2 y 2,• ∕ y 2 , and the second term, y 2,• ∕ y 2 , are less than one. Therefore, also the product of these terms is less than one so that y 2,• RF ∕ y 1 < 1 . A similar argument can be applied to show that the first inequality in (27) is satisfied if y 1,• ∕ y 1 < 1 . Hence, if y i,• ∕ y i < 1 holds for i = 1, 2 , then RF 2 has a flatter slope than RF 1 in y 1 , y 2 space and the myopic equilibrium exists.
Derivation of Proposition 4
We parameterize the entrance cost function as a fourth-order polynomial; y i = a 1 y i + a 2 y i 2 + a 3 y i 3 + a 4 y i 4 . The first-order conditions associated with the policy-maker´s maximization problem become (after we have used the private first-order condition for each agent type to simplify the resulting expressions) L a 2 = N 1 y 1 2 + N 1 1 y y 1 a 2 + N 2 y 2 2 + N 2 2 y y 2 a 2 − N 1 y 1 2 u 1 c − N 2 y 2 2 u 2 c − y 2 a 2 = 0, L a 3 = N 1 y 1 3 + N 1 1 y y 1 a 3 + N 2 y 2 3 + N 2 2 y y 2 a 3 − N 1 y 1 3 u 1 c − N 2 y 2 3 u 2 c − y 2 a 3 = 0, Applying Cramer´s rule to solve for 1 − u 1 c ∕ produces Since the first row in the determinant in the numerator is a linear function of the fourth row (multiplying each term in the fourth row by ∕ N 2 produces the first row), the determinant in the numerator is zero. Hence u 1 c = . Next, we apply Cramer´s rule on (34) to solve for 1 − u 2 c ∕ (33) Turning to the shadow price, we first substitute 1 y = 0 and 2 y = ∕ N 2 into (33). Then we use the definition of MWP i y,w in Eq. (18) together with u 1 w = u 1 c = . Finally, we divide by and rearrange. This produces ∕ = ∑ i N i MWP i y,w .
Derivation of Proposition 5
We parameterize the assistance cost function as a fourth-order polynomial; (y) = b 1 y + b 2 y 2 + b 3 y 3 + b 4 y 4 . The first-order conditions associated with the policy-maker´s maximization problem become (after we have used the private firstorder condition for each agent type to simplify the resulting expressions) where Ñ i y = − 1 − p i dq i ∕dy i + 1 − q i dp i ∕dy i N i > 0 . Dividing Eqs. (41)-(44) by and rearranging, produces the following equation system
3
Social identity and risky leisure activities: implications… Applying Cramer´s rule to solve for to solve for 1 − u 1,f c ∕ produces Since the first row in the determinant in the numerator is a linear function of the fourth row (multiplying each term in the fourth row by ∕ Ñ 2 produces the first row), the determinant in the numerator is zero. Hence u 1,f c = . Next, we apply Cramer´s rule on (46) to solve for 1 − u 2,f c ∕ Since the second row in the determinant in the numerator is a linear function of the fourth row, the determinant in the numerator is zero. Hence u 2,f c = . Furthermore tions defining 1 y and 2 y , produces = | 2022-02-24T16:17:26.855Z | 2022-02-22T00:00:00.000 | {
"year": 2022,
"sha1": "a51d060829bcd13a87c4e0d1b46f4f45f836ec73",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s00355-022-01394-7.pdf",
"oa_status": "HYBRID",
"pdf_src": "SpringerNature",
"pdf_hash": "e06d4431acf85cc1d256bc15da25e91bf1815fcd",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
214272771 | pes2o/s2orc | v3-fos-license | Hanafy Ahmed Youssef, DM, MRCS, FRCPsych
© The Authors 2020. This is an Open Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits noncommercial re-use, distribution, and reproduction in any medium, provided the original work is unaltered and is properly cited. The written permission of Cambridge University Press must be obtained for commercial re-use or in order to create a derivative work.
Hanafy Youssef, who died at the age of 80 on 21 January 2019, was a leading figure in Irish psychiatry. With his colleague John Owens, he set up the first community-based psychiatric service in Ireland. He was able to show that this pattern of service led to better outcomes for patients and their families. As a result, a number of other community-based services in Ireland and the UK were established.
Over a 20-year period in the 1980s and 1990s, he developed a highly productive research relationship with Professor John Waddington of the Royal College of Surgeons, Ireland. Together they published many papers on aspects of psychotic illness occurring in patients referred to the Cavan/Monaghan psychiatric services. Hanafy Youssef contributed to over 60 publications in the fields of psychotic illness, neurodevelopmental psychiatry and psychopharmacology. The relationship between the Cavan/Monaghan psychiatric services and the Department of Psychopharmacology at the Royal College of Surgeons, Ireland, continues to this day.
Hanafy had a strong interest in postgraduate psychiatric education. He was clinical tutor at St Davnet's Hospital and a member of the Irish Psychiatric Training Committee. He had a great capacity for explaining difficult concepts to medical and nursing staff. Indeed, as an educator, he was at the forefront of psychiatric education in Ireland. He was also a mentor to many overseas doctors who worked with him. He was an inspiration to them as they saw him become a leader in his specialty. He was a particularly strong supporter of women in medicine at a time when women doctors were still finding it difficult to climb the career ladder.
He had a great interest in the development of psychiatric services in less economically developed countries. In the 1970s he spent time in Zambia, developing clinics in rural areas and helping to confront the stigma often attached to psychiatric illnesses. From 1994, he spent 2 years in a professorial post in Trinidad and Tobago, where he helped improve the quality and prominence of undergraduate psychiatry training. He felt this was crucial if the finest doctors were to be recruited into the specialty. He then returned to take up a consultant post in liaison and general psychiatry at Addenbrooke's, a teaching hospital within Cambridge University Hospitals NHS Foundation Trust. In addition, he carried out voluntary work with charities in Yemen and Libya, providing both general medical and psychiatric care in areas where there was a shortage of doctors.
Hanafy was also a philanthropist, fiercely passionate about social justice, who took seriously the welfare of the less privileged. Together with his siblings he established Latifa's Orphanage House and a mosque in the suburbs of Alexandria. He worked tirelessly for the underprivileged and would offer free medical care to any in need.
Whether they spoke to him in person, heard him speak at international conferences or were his patients, Hanafy Youssef left a lasting impression on all he met. His knowledge, dedication to his field and commitment to his patients were widely appreciated.
He came from humble beginnings. He was born on 14 January 1939 in Alexandria, Egypt, to Ahmed Mahmoud and Latifa (née Taher). His father ran a hunting and tackle store there. Hanafy studied medicine at Alexandria University and undertook postgraduate training in the Department of Psychiatry at the University of Cairo. At university, he led an extremely active life, both in cultural pursuits and in sport. He authored several published novellas in Arabic, won a national poetry award and was the arts reviewer for the university paper. At the same time, he was a member of the university football and wrestling teams. He took an interest in science and politics, and became multilingual, adding French and Russian to the Arabic and English he spoke before entering university. After qualifying as a doctor, he spent 2 years as a medical officer in the Egyptian Army, including service during the Arab-Israeli Six-Day War.
He moved to Derry in 1971 to obtain further training in psychiatry and fell in love with Ireland and its people. This is where he met his wife Ann. He obtained a consultant psychiatrist's post at St Davnet's Hospital, Monaghan, in 1975 and spent most of his career there.
Hanafy retired in 2001 and enjoyed his retirement years in Armagh. He remained highly active, interested in medicine and writing about and closely following the struggle of his Egyptian compatriots for freedom and progress. He published several papers on the history of psychiatry, reviewed books on psychiatry and had several letters published in national broadsheets on a wide range of subjects. He was very proud of his family and their achievements.
He passed away peacefully surrounded by his family. He leaves Ann, his wife of 46 years, five children (Emma, Mahmoud, Latifa, Zahra and Omar) and three grandchildren.
Omar Youssef doi:10.1192/bjb.2020.1 © The Author 2020. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/ licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited. One of the most difficult things about writing a book for general readers on mental health (and, more particularly, psychiatry) is knowing how to structure itand where to begin. The authors of this very comprehensive book, a liaison psychiatrist (Ellen) and a writer-comedian (Deveny), have both experienced depression and they share their tales with considerable frankness and humility. However, instead of beginning with these engaging stories, they choose to start with a section on diagnosis and classificationwhich might unfortunately deter some from continuing.
Reviews
There are really useful sections on how to talk to friends who you think might need help and what happens when you go to see a mental health professional. However, my particular favourite has to be 'clues your shrink is a dud', which warns against those who claim excessive certainty, have a guru mentality and are excessively expensive. And therein lie clues that this book doesn't originate in the UK, but hails instead from Australia. The text has clearly been edited for the UK edition, with reference to clinical commissioning groups, mental health trusts and a list of UK organisations from which further help can be sought. However, there is, for example, no reference to the problems faced by those from UK ethnic minorities, the section on drugs mentions neither 'skunk' or 'legal highs' and the classification system is, of course, DSM.
Readers of this book would learn a great deal about mental health and illness from a biopsychosocial perspective. They might, however, be left with an idea that there is considerably more choice of professional and therapist in the National Health Service than in realityalthough this may of course be true if they can pay. Personally, I don't see any problem in asking your GP if they are good at mental health, and I wish it was easier to ask for second opinions. The authors tell us 'remember you are in charge!' but for many people seeking help it rarely feels that way. As interest in psychedelic research continues to increase, it is clear that a new conceptual framework is needed to investigate phenomena associated with altered states of consciousness. While research has focused primarily on their psychotherapeutic and entheogenic uses, few studies have dared to consider psychedelics as potent tools for enhancing cognition, conducting conceptual research and improving complex problem-solving. Sensing an opportunity to kick-start an intellectual revolution, Roberts introduces 'multistate theory' as a potential framework to guide new exploration of altered states of consciousness.
A core tenant of multistate theory is the rejection of the 'singlestate fallacy', which Roberts defines as the erroneous assumption that all worthwhile skills, abilities and knowledge reside in our default waking state. He argues that our default state is simply one of many possible states of consciousness (or 'mindbody states') that the mind can produce and | 2020-02-13T09:12:43.225Z | 2020-02-06T00:00:00.000 | {
"year": 2020,
"sha1": "1a615b365af6d4b9cb266215082816cbf03f2ef4",
"oa_license": "CCBYNCND",
"oa_url": "https://www.cambridge.org/core/services/aop-cambridge-core/content/view/173071B2C6D83D9470642C65835F029C/S2056469420000017a.pdf/div-class-title-hanafy-ahmed-youssef-dm-mrcs-frcpsych-div.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8344cdf8e2d7c95770a645fa17c165acd71dec5d",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
220833184 | pes2o/s2orc | v3-fos-license | Isolation and Identification of Some Probiotic Bacteria and Their Potential Role in Improving Immune Response and Resistance of Nile Tilapia (Oreochromis niloticus) in Comparison with a Commercial Product
This work aimed to retrieve a field isolate of probiotic from Nile tilapia (Oreochromis niloticus) and compare the obtained results with a commercial probiotic product through experimental studies. The study was conducted on 250 Nile tilapia. Ten fish were used to isolate the probiotic strain. Two isolates showed an in vitro inhibitory effect against pathogenic A. hydrophila. The isolate with the largest zone was identified by PCR. Sixty fish were used to test the safety of a potential probiotic. One hundred and eighty fish were used in a two-month feeding experiment. Fish were divided into 3 groups, group (1): the control, group (2): fed on potential probiotics, and group (3): fed on commercial probiotic (Organic Green™). The effects of tested products on the immune response were recorded in all groups. After one and two months of feeding experiment, blood and nonspecific immune parameters were evaluated. Disease resistance against Aeromonas hydrophila was evaluated through challenge experiment. The histopathology of the treated groups was fully recorded in comparison with the control group. The potential probiotic based on the in vitro antimicrobial activity test was identified as P. putida using routine and gel electrophoresis and 16S rRNA sequencing. During the first and the second month of experiment, there was a highly significant increase in the survival percent of the experimental fish in both treated groups with probiotics. In the first phase of the experiment, a significant increase in the haematocrit values and NBT, lysozyme activity, and phagocytic activity was seen in all treated groups in comparison with the control. The increase in the TLC was significant in the group fed with P. putida in comparison with the control group. In the second phase, a nonsignificant increase in the hematocrit values and significant increases in the NBT and phagocytic index were seen in P. putida and organic green groups in comparison with the control group. The TLC and DLC revealed nonsignificant changes in the treated groups in comparison with the control. The RLP in the groups treated with P. putida was higher than that in those treated with organic green. Although probiotics are an important management tool in aquaculture, it should be subjected to scientific laboratory tests and field measurements.
Introduction
Probiotics were firstly detected by Metchinkoff [1], who noticed that some acid-producing micro-organisms in fermented dairy products might prevent fouling in the intestine that led to a prolongation in the lifespan of humans. Today, probiotics are available in a variety of food products and have got wide applications in the control of cholesterol, cancers, and allergies [2] Lilley and Stillwell [3] mentioned that probiotics are substances secreted by a microorganism. Later on, probiotics were defined by several authors as microbial cell preparations that have a beneficial effect on the health and well-being of the host [4][5][6].
Probiotic was first recorded in fermented milk. After that, probiotics became popular with animal nutrition. Metchinkoff [1] suggested that people should consume fermented milk containing lactobacilli to prolong their lives. Accelerated aging is because of autointoxication (chronic toxemia), which is due to the toxins produced by gut microflora. e pathological reaction might be removed, and life expectancy could be enhanced by implanting lactic acid bacteria from yogurt [1]. Since then, researchers started investigations relating to the role of lactic acid bacteria in human and animal health.
Probiotics have been used in pigs as growth promoters [7], for lactose intolerance in rats, and antitumour and anticholestrolaemic effects in human [8,9]. e main fields of research with respect to probiotics are heart diseases, allergic reaction, cancer, and diarrhoea. e use of probiotics results in alleviation of lactose intolerance [10], relief from constipation [11], and antitumour activities [12]. Intestinal infections caused by Escherichia coli, Campylobacter fetus subsp. jejuni, Clostridium perfringens, and C. botulinum were reduced in man and animal with the presence of Lactobacillus supplements [4]. Bifidobacterium longum has been successfully used to reduce the later effects of antibiotic therapy [13].
However, probiotics of aquatic sources could be endogenous or exogenous microbiota, and the isolated probiotics from the endogenous microbiota may depend on genetic, nutritional, and environmental factors. As the ambient environment has a greater influence on the health status of the aquatic animals than for terrestrial animals, human probiotics obtained from the aquatic species have a much larger influence on the health status [14].
Disease outbreaks are increasingly being recognized as a major constraint in aquaculture production and the economic development in many countries. Conventional approaches, such as the use of disinfectants and antimicrobial drugs, have had limited success in the prevention or the cure of aquatic disease. Bacterial diseases, especially due to A. hydrophila, are responsible for heavy mortality in fish. Antibiotics are used to control these infections but may develop and spread antimicrobial-resistant bacteria and resistance genes [15]. e sensitivity of A. hydrophila isolates to some antibiotics revealed a high sensitivity reaction to cefquinome [16]. In the last decade, Roman [17] mentioned that A. hydrophila infection in fish is sensitive to some of the fourth-generation cephalosporins including the cefipime. Furthermore, there is a growing concern about the use and, particularly, the abuse of antimicrobial drugs not only in human medicine and agriculture but also in aquaculture where this could induce hazard through the development of cross-resistance to antimicrobials used in human medicine [18]. erefore, FDA's Center for Veterinary Medicine (CVM) regulates the manufacture, distribution, and use of animal's drugs. FDA also has established safe maximum residue limits (MRLs) for these drugs and other veterinary medications. Such strict regulations were taken to ensure that the treated animals are free from potentially harmful residues [19]. e concept of biological disease control has received widespread attention in the last decade; therefore, commercial probiotics are increasingly used in fish farming, but further investigations are required to identify the most suitable microbial preparations and doses for each fish species. Besides the high cost associated with purchasing a commercial product of probiotics, the variability in response to probiotics and the lack of reliable data hinder the use of such practices routinely in aquaculture [15]. An effective probiotic should be obtained from the same animal species. e underlying reason for this is that the intestines of individual species are sufficiently different from those of others, such that the isolates suited to those environments would not necessarily be suited to the intestine of others [15]. To ensure the required immunological response, alleviate the problem of introducing new microbial agents to our environment, and the high costs of commercial probiotics, isolation of native strains from native fish is important. For these reasons, the present study aimed to evaluate the effects of the isolated field strain of probiotics by determining their inhibitory effect against pathogenic A. hydrophila and evaluate their role in increasing the immune response, as well as the resistance of cultured tilapia fish to infection in comparison with other available commercial probiotics.
Fish.
A total number of 250 live and apparently healthy Nile tilapia (O. niloticus) of both sexes were collected from Fish Research Institute and used in this study. Ten O. niloticus (60 ± 5 g) were used to isolate the probiotic, 90 O. niloticus (60 ± 5 g) were used for the safety experiment, and 180 O. niloticus (30 ± 10 g) were used in the feeding experiment. Fish were kept in fiber glass tanks containing dechlorinated tap water and supplied with continuous air, and feces was siphoned daily. Fish were fed twice daily with a balanced diet at a rate of 3% body weight and kept for two weeks under observation for acclimation.
Bacterial Isolation and Identification.
Ten O. niloticus (5 apparently healthy and 5 with disease signs, each 60 ± 5 g) were randomly collected from earthen ponds. Bacteriological examination of the fish samples was carried out. Swab samples were taken from the internal organs (liver, kidney, gonads, stomach, and intestine) and gills; they cultured on tryptic soya broth (TSB) and incubated at 30°C for 1 to 2 days. Pure isolates were taken after subculture on tryptic soya agar (TSA). Identification of the strain was performed using biochemical tests according to [20] and the API 20 E strip system (Bio Merieux), as well as the molecular technique (the PCR product of the isolated strain was used in gel electrophoresis ( Figure 1) and 16S rRNA sequencing, and the strain was identified as P.putida). Pathogenic A. hydrophila strain was obtained as a reference strain.
Antimicrobial Activity Assay.
e bacteria were tested for their probiotic activity in vitro using an agar spot assay [21]. e probiotic strain was cultivated in trypticase soya broth (TSB) (BioLife Milano, Italy) and incubated at 30 1°C for 24 h. en, spots were made by pouring 10 mL of a wellgrown overnight culture of the probiotic strains in the centre of the trypticase soya agar plates. e plates were incubated overnight at 30°C, and the growth of the strains was checked the next day. After the spots were developed, a soft agar (composed of tryptone soya broth10.7% bacteriological agar containing 5% of an overnight culture of pathogenic strains of A. hydrophila in tryptone soya broth) was poured on the plates. Inhibition was recorded by measuring the absence of pathogen growth around the spots.
Basal
Diet. Pellets (0.5 cm) were prepared from locally available ingredients using a pellet machine (CPM California Pellet mill, San Francisco, CA, USA). e ingredients (Table 1) were mixed mechanically with a horizontal mixer (Hobartsmodel D300 T, Troy, OH, USA) at a low speed for 30 min after crushing the corn to a size of 0.5 mm using a omas-Willey Laboratory Mill Model 4. en, oil was added gradually to ensure an even distribution of the ingredients with an increase in the mixer speed for 5 min, during the time when 600 mL water was added. e pellets obtained were allowed to air dry at room temperature for 24 h. e required diet was prepared biweekly and stored in a refrigerator (4 1°C) for daily use.
Preparation of Feed with Probiotics.
Preparation of probiotic bacteria was carried out by inoculating the isolates (P. Putada) in TSB and incubating for 48 h at 30°C. ey were then centrifuged at 3000 ×g for 30 min. After centrifugation, the bacteria were washed twice with sterile saline, and the concentration of the final suspension was adjusted to 1 × 10 10 bacteria/ml in saline. e bacterial suspension containing the probiotic isolates was added to commercial food (containing 25% protein) to give 1 × 10 9 bacterial cells/g of diet for the viability experiment and 1 × 10 7 bacterial cells/g of diet for the feeding experiment, by mixing well with an automatic mixer. e pellets were dried in an oven at 45°C. To determine the viability of the probiotics, one half of the feed was stored in a refrigerator (4°C) while the other half was kept at 25 ± 1°C. For the feeding experiment, the feed was stored in a refrigerator at 4°C.
Organic Green ™ .
Organic Green ™ is a commercial product available in the market and manufactured by Hang Poong Industry, Inchon City, Korea. It is used to improve the growth and resistance of poultry and large animals. It is a mixture of probiotics, 1 kg of this product containing 1 × 10 11 bacterial cells each from Lactobacillus acidophilus, Bacillus subtilis, Saccharomyces, and Aspergillus oryzae according to the manufacturers. is commercial product was tested in the laboratory for isolation and identification of viable organisms. One dose of 1 g of Organic Green ™ /kg feed was mixed, and pellets were made. e pellets were prepared biweekly, air-dried at room temperature for 24 h, and stored in a refrigerator (4°C).
Safety of the Tested Probiotic Strains.
Ninety tilapia (65 ± 5 g) were divided into 2 equal groups, each with three replicates (each with 15 fish) and distributed randomly among 6 aquaria. e first group was intraperitoneally (I/P) injected with 0.5 ml L acidophilus fresh culture suspension containing 10 7 bacteria/ml, while the second group served as a control and was I/P injected 0.5 ml sterile saline (0.85% NaCl). Both the test and control group of fish were observed and fed on a basal diet containing 30% protein and water temperature was 26°C throughout the experiment. e mortality rate was recorded daily for 15 days.
Experimental Design.
One hundred and eighty Nile tilapia with an average body weight (30 ± 10 g) were divided into 3 equal groups, each with 30 fish. Each group was divided equally into 3 replicates (10 fish per each). e fish were acclimated in in-door fiberglass tanks for 14 days. Each tank was supplied with a well-oxygenated tap water. Group (1): the control was fed basal diet without bacteria, group (2): fed basal diet containing P. Putida, at a dose of 1 × 10 8 CFU/g, and group (3): fed basal diet with Organic Green ™ 1.0 g/kg diet. e prepared diet was transferred to plastic bags and stored in a refrigerator (4°C), and this preparation was repeated every two weeks. Fish were fed 6 days a week for 60 days. e dead fish were recorded and removed daily. International Journal of Microbiology anesthetized by immersion in water containing 0.1 ppm tricaine methane sulfonate (MS-222). Whole blood (0.5 ml) was collected from the caudal vein of each fish using syringes (1-ml) and 27-gauge needles that were rinsed in heparin (15 unit/ml), to determine the hematocrit values, NBT, and phagocytic activity tests. A further 0.5 ml blood sample was centrifuged at 1000 ×g for 5 min in order to separate the plasma. e latter was stored at −20°C to be used for the lysozyme activity test. For separation of serum, blood samples (0.5 ml) were withdrawn from the fish caudal vein, as before, and transferred to Eppendorf tubes without an anticoagulant. e blood samples were centrifuged at 3000 ×g for 15 min, and the supernatant serum was collected and stored at −20°C until used for the serum bactericidal test [22].
Hematological and Immunological Parameters
2.10.1. Hematology Parameters. Hematocrit capillary tubes were two-third filled with the whole blood and centrifuged in a hematocrit centrifuge for 5 min, and the percentage of the packed cell-volume was determined by using the hematocrit tube reader. e WBC count was determined by using a Neubauer hemocytometer [23,24]. Blood was diluted 1 : 20 with Turk's diluting fluid and placed in a hemocytometer. Four large (1 sq mm) corner squares of the hemocytometer were counted under the microscope. e cells touching the boundary lines were not counted. e total number of WBC was calculated in mm 3 × 103 [25]. e blood smears were prepared and stained with Giemsa stain for 30 min. One hundred leukocytes were identified, and the percentage values of different white cells were calculated according to [26].
Nitroblue Tetrazolium Activity (NBT)
. Blood (0.1 ml) was placed in microtiter plate wells, where an equal amount of 0.2% NBT solution was added and incubated at room temperature for 30 min. N dimethyl formamide 1 ml was added to a sample of NBT blood cell suspension (0.05 ml) in a glass tube and centrifuged at 3000 rpm for 5 min. e supernatant was measured using a spectrophotometer at 620 nm in 1 ml cuvettes [22,24].
Lysozyme Activity.
Chicken egg lysozyme (Sigma) was the standard, and 0.2 mg/ml Micrococcus lysodeikticus (lyophilized form) in 0.04 M sodium phosphate buffer (pH 5.75) was used as the substrate. Fifty µl of serum was added to 2 ml bacterial suspension, and the reduction in the absorbance at 540 nm was determined after 0.5 and 4.5 min incubation at 22°C [27].
Phagocytic Activity.
Phagocytosis assay was performed according to the method described by [28] with some modifications. One ml of the adjusted viable leukocytes suspension (leukocytes in RPMI1640 with 5% of pooled tilapia serum) was placed in sterile plastic tube, to which 1 ml of the prepared heat in activated C.Glabrata was added. e tubes were then incubated for 30 min at 27°C in a 5% CO 2 incubator. en, the tubes were centrifuged at 2500 rpm for 5 min, and the supernatant was removed. Slide smears were prepared from the deposit, air dried, and then stained with Leishman's stain.
2.11. Histopathological Examination. Tissue specimens including the liver, kidney, spleen, and intestine from each experimental group of treatments were collected by the end of feeding experiment (2 months). e collected specimens were immediately fixed in neutral buffered formalin 10%, dehydrated in ascending concentration of ethyl alcohol, cleared in two changes of xylene, blocked in paraffin, and sectioned at 5 µm using a rotary microtome. e microscopic tissue slides were stained with routine hematoxylin and eosin stain (H&E stain) and then covered with cover slips. e histopathological technique was performed according to Wallington (1980).
Response to Challenge Infections.
irty fish from each of the tested treatments (10 from each replicate) were collected and reared in a glass aquarium. ey were clinically examined, and blood samples were bacteriologically tested and proved to be free from bacterial infection. e treatment groups were subjected to challenge infections, after feeding with test diets for 1 month (15 fish for the first phase) and 2 months (15 fish for the second phase). e challenged bacteria were obtained as a reference pathogenic strain of A. hydrophila that were isolated previously from the liver of morbid O. niloticus and studied for pathogenicity.
A culture suspension of A. hydrophila was prepared by culturing in agar for 24 h, collected, washed and suspended in sterile saline 0.85%, and counted using Mc Firland standard tubes. en, fish were artificially infected by an intraperitoneal injection with 0.5 mL of culture suspension of pathogenic A. hydrophila containing 10 8 bacteria/L. e relative level of protection (RLP) among the challenged fish was determined [22,29] using the following equation: RLP(1/4)100 − percent of immunized mortality/ percent of control mortality × 100.
Statistical Analysis.
Analysis was performed to the measured growth and immunological parameters of the collected samples using the analysis of variance (ANOVA) and Duncan's multiple range test [30] (mean at a significance level of P < 0.05 ). Analysis was performed using Minitab (18) package.
Isolation and Identification of the Probiotic Isolates.
Five bacterial isolates were obtained from the intestinal tract of 10 fish (O. niloticus). e five isolates were investigated for their inhibitory activity against pathogenic A.hydrophila. Only two isolates showed an inhibitory effect against A.hydrophila. e inhibition zones were 15 and16 mm in diameter. e isolate that showed the largest inhibition zone (16 mm) in the in vitro antimicrobial activity test was preliminary identified as P.putida using (API20 E) strips with code 2204046. e PCR product of the isolated strain was used in gel electrophoresis ( Figure 1) and 16S rRNA sequencing, and the strain was identified as P.putida.
Safety and Survival Rate.
e intraperitoneal injection of O. niloticus with P. putida at a dose of 0.3 ml matching 3 × 10 7 CFU/ml was noticed to be safe, as well as causing no mortalities during a period of 15 days, indicating the safety of the isolated bacterial strain. During the first and the second month of experiment, there was a highly significant increase in the survival percent of the experimental fish in both treated groups with probiotics (the first group fed P.putida and the second group received Organic Green ™ ) when compared with the control group ( Table 2).
Hematological and Immunological Parameters.
e first phase of the experiment where the fish were given a basal diet mixed with probiotics for 1 month revealed a significant increase in the hematocrit values in all treated groups in comparison with the control. A significant increase in NBT, lysozyme activity, and phagocytic activity was seen in all treated groups in comparison with the control. e increase in the TLC was significant in the group fed with P. putida in comparison with the control group. Although the number of neutrophils had nonsignificantly increased in P. putidatreated groups in comparison with the control, the increase in TLC resulted mainly from the increase in lymphocytes and monocytes (Table 2).
In the second phase, the fish which were given a basal diet mixed with probiotics for 2 months revealed a nonsignificant increase in the hematocrit values. Significant increases in the NBT and phagocytic index were seen in P. putida and organic green groups in comparison with the control group. e TLC and DLC revealed nonsignificant changes in the treated groups in comparison with the control ( Table 2).
Relative Level of Protection of O. niloticus after Bacterial Challenge with A. hydrophila.
e RLP in the groups treated with P. putida was higher than that in those treated with Organic Green ™ ( Table 2).
Histopathological Findings.
e control group showed normal cellular details and tissue architecture with no marked degenerative changes or inflammatory reactions.
Tilapia Received Basal Diet Mixed with P. putida at a Dose of 1 × 10 8 CFU/ml for 2 Months.
e liver and kidney revealed mild vacuolar degeneration in the hepatocytes and renal epithelium. Focal proliferation of melanomacrophage cells was evident in the hepatopancreatic and renal tissues. e renal interstitial issue showed a mild edema and focal proliferation with leukocytes. e spleen showed congestion in blood vessels with focal proliferation of lymphocytes and mild proliferation of melanomacrophage centers (Figures 2(a) and 2(b)).
Tilapia Received Basal Diet Mixed with 1 gm/kg Diet
Organic Green ™ for 2 Months. e hepatopancreas and kidney exhibited minimal degenerative changes with aggregation of melanomacrophage cells around hepatopancreatic areas and in the renal parenchyma. e spleen revealed massive proliferation and activation of melanomacrophage centers all over the splenic parenchyma (Figures 3(a) and 3(b)).
Discussion
Currently, probiotics are available in a several food products, exclusively dairy products, due to the historical association of lactic acid bacteria with fermented milk. Probiotics are gaining importance for multiple benefits, e.g., treating lactose intolerance, hypercholesterol problem, cardiac diseases, and managing cardiac problems such as atherosclerosis and arteriosclerosis. Many probiotic products are present in the market place, supporting the evidence of health claims. New legislation governing the labelling of probiotics, such as indicating the species, strain, and number of bacteria present, is likely to come into force in the near future. Probiotics can be incorporated into a balanced diet to maximize good health. e main characteristics of microbes as candidate probiotics are to improve the health of their host, to antagonize pathogens, to have a colonization potential, and to be efficient in increasing the resistance to disease of their host. Gatesoupe [31] reported other beneficial effects of probiotics, e.g., competition with pathogens for nutrients or for adhesion sites and stimulation of the immune system. However, although probiotics may display multiple effects, possibly combining bacterial antagonism to some effects on the host, e.g., stimulating immunity or growth [32], competitive exclusion, enzyme activator [33], hormones inhibitor [34], immune response enhancement [35], and their modes of action, however, are not fully understood.
It is very important to characterize and identify the mode of action of the potential probiotics and their efficiency on the pathogen and safety. is can be achieved through in vitro and in vivo studies. e selection criteria of Gomez-Gil et al. [2] are based on the collection of background information; acquisition of potential probiotics; evaluation of the ability to outcompete pathogenic strains; and assessment of their pathogenicity and effect on the host; as well as economic analysis.
e United Nations has recommended some specifications to be considered when a probiotic is selected and approved [36] including viability of the probiotic to survive; colonization; competition against pathogenic bacteria; inhibiting pathogenic bacteria; resistance against other sanitary agents or disinfectants; and labelling according to the international nomenclature including dosing and the expiration date.
International Journal of Microbiology Bacterial diseases are responsible for heavy mortality in fish. Antibiotics are used to control these infections but may develop and spread antimicrobial-resistant bacteria and resistance genes [15]. As a consequence, prevention of fish disease by the application of live pathogen-antagonistic bacteria has received a widespread interest [37]. In the present study, samples from the intestine were cultured on TSB and incubated at 30°C for 24-48h. e isolated strains were purified through subculturing on TSA. Twelve bacterial isolates were obtained from the intestinal tract of tilapia fish (O.niloticus), and only two isolates showed an inhibitory effect against the pathogenic A. hydrophila. e isolate of the largest zone was identified as P.putida using API20 E and further molecular diagnostic tools. Similar findings were reported by Sebastião et al. [37] who isolated both P. putida (27%) from the spleen of tilapia fish and P. fulva (20%) from the skin and the kidney.
Abdel-Galil Ahmed et al. [38] also isolated many probiotic strains (Vibrio and Pseudomonas sp.) from endogenous and exogenous microbiota of a variety of species of marine fish. In our study, the antimicrobial activity assay was performed by applying agar disc diffusion method. e inhibition zones of isolated probiotics were 15 and 16 mm against pathogenic A. hydrophila. Related results were reported by Aly et al. [22,39] who recorded that B. subtilis and L. acidophilus inhibited the growth of A. hydrophila in vitro. Similar findings were also detected by Abdl El-Rhman et al. [40] who isolated and identified both M. luteus and Pseudomonas sp., after isolation from the intestine and gonads of Nile tilapia O.niloticus, and reported M. luteus and Ps. Mean ± SE having the same letter in the same row are not significantly different at P < 0.05. e intraperitoneal injection of O.niloticus with P. putida at a dose of 0.3 ml matching 3 × 10 7 CFU/ml was noticed to be safe, as well as causing no mortalities during a period of 15 days, indicating the safety of the isolated bacterial strain. is result of safety experiment revealed no mortalities after an intraperitoneal injection of O.niloticus with P. putida at a dose of 0.3 ml matching 3 × 10 7 CFU/ml during a period of 15 days, which was supported by Eissa and Abou [41] who tested the isolated P. fluroscens biovars I, II, and III and noticed that they were nonpathogenic and safe to O. niloticus. e first phase of the experiment revealed a significant increase in the hematocrit values in all treated groups together with an increase in the TLC that was significant in the group fed with P. putida in comparison with the control group. Aly et al. [22,42] proved that B. pumilus significantly increases the NBT values, hematocrit values, total leucocytic, and differential leucocytic count, with a significant increase in lymphocytes and monocytes in O. niloticus. Sakai et al. [35] mentioned that the nonspecific immune system of the fish can be stimulated by probiotics. Lysozyme has a bactericidal effect by destroying cellular walls of bacteria. Lysozyme also stimulates the phagocytic activity and participates in the regulation of immune cell differentiation and proliferation [43]. Fish Phagocytic activity represents an immediate response carried out by the phagocytes to kill the pathogenic bacteria as a part of their defense mechanism [44]. It is a key element in host defenses against bacterial infections (Kantari et al. 2008). Concerning the measured immunological parameters in the present study, the results of NBT, lysozyme activity, and phagocytic activity after one month of experiment showed a significant increase in the groups which received probiotics in relation to the control and could be attributed to the increased blood and immunological parameter. Aly et al. [22] reported that B. subtilis and L. acidophilus significantly increase the nitroblue-tetrazolium (NBT) assay, neutrophil adherence, and lysozyme activity and showed a significant increase in the serum bactericidal activity in O. niloticus. e relative level of protection (RLP) of Nile tilapiatreated groups with P.putida or organic green after challenging with A. hydrophila through the 1st and 2nd month of experiment was 62.5, 55.5, 52.5, and 44.4, respectively. ese results agreed with those of Aly et al. [22] who demonstrated that the relative level of protection (RLP) against P. fluorescens and A. hydrophila was significantly elevated in Nile tilapia fed on B. subtilis and Lactobacillus. Gram et al. [45] also recorded that P. fluorescens AH2 protects rainbow trout against challenge with V. anguillarum. Li and Gatlin [46] also found that yeast (GrostBioticR-A) protects hybrid striped bass against mycobacterial infection.
Conclusions
e concept of biological disease control using probiotics has received a widespread attention during the last decade for their cheaper value and more safety than antibiotics. In order to define the potential of probiotics in aquaculture, selection criteria are crucial. Data about the efficiency, mode of action, safety, durability, and economic cost benefit should be known. Although probiotics are an important management tool in aquaculture as an alternative to antimicrobials use and to the shortage in the vaccine availability, they should be subjected to scientific laboratory tests and field and economic measurements to be recommended in a large-scale aquaculture.
Data Availability
All required data are included in tables and figures.
Ethical Approval
All fish were maintained in accordance with the National and International Institutional Guidelines for the Care and Use of Animals for Scientific purposes.
Conflicts of Interest
e author declares no significant conflicts of interest. International Journal of Microbiology 7 | 2020-07-23T09:06:01.311Z | 2020-07-17T00:00:00.000 | {
"year": 2020,
"sha1": "0d429477bd82a6941b0f4cd503e4ed1a6b0cd509",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1155/2020/8865456",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "07a7834c9c900ab101e6264f3e388a23c81ea510",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
15949188 | pes2o/s2orc | v3-fos-license | Impulsive Behavior and Recurrent Major Depression Associated with Dandy-Walker Variant
Reported herein is a case of recurrent major depression with impulse control difficulty in a 33-year-old man with Dandy-Walker variant. He was diagnosed as having major depressive disorder a year before he presented himself to the authors' hospital, and had a history of three-time admission to a psychiatric unit in the previous 12 months. He was readmitted and treated with sodium valporate 1,500 mg/day, mirtazapine 45 mg/day, and quetiapine 800 mg/day during the three months that he was confined in the authors' hospital, and the symptoms were reduced within three months but remained thereafter. This is the only case so far reporting recurrent depression with impulse control difficulty associated with Dandy-Walker variant. This case implies that any cerebellar lesion may cause the appearance of recurrent depression with impulse control difficulty in major depressive disorder.
INTRODUCTION
The cerebellum is traditionally involved in the coordination and integration of the motor function. Its contribution to the modulation of higher-order functions is increasingly being recognized. 1 Patients with cerebellar diseases such as cerebellar tumor, stroke, and cerebellar atrophy have difficulty controlling their behavior and often show impulsive behavior, have difficulty concentrating, and have compulsive personality disorders. [2][3][4] Classic Dandy-Walker malformation represents cystic dilatation of the fourth ventricle and enlarged posterior fossa, complete or partial agenesis of the cerebellar vermis, elevated tentorium cerebelli, and hydrocephalus. Dandy-Walker variant has been introduced to describe variable hypoplasia of the cerebellar vermis with or without enlargement of the cisterna magna, communication between the fourth ventricle and the arachnoid space, and no hydro-Copyright © 2013 Korean Neuropsychiatric Association 303 cephalus. 5 Previous case studies reported that Dandy-Walker malformation was found to be associated with mental retardation, ADHD, schizophrenia, and bipolar disorder. [6][7][8][9][10][11] No previous case study, however, has reported findings of the association of major depression and impulse control difficulty with Dandy-Walker malformation, as shown in this case study. Thus, reported herein is a case of recurrent depression with impulse control difficulty in a 33-year-old man with Dandy-Walker variant.
CASE
A 33-year-old man presented depression, aggressiveness, and impulsivity in his visit to the authors' hospital and was thus compulsorily admitted. He had a two-year depressive episode, was diagnosed as having major depressive disorder a year earlier, and has been on medication irregularly. He had a history of three-time admission to a psychiatric unit in the previous 12 months. His initial admission was a year earlier, after attempting suicide by ingesting pesticide. One day before admission, he presented verbal abuse and aggression. He had no family history of psychiatric disorder, including depression. The patient had two-year, withdrawn university education and he had had various jobs but never managed to
Impulsive Behavior and Recurrent Major Depression Associated with Dandy-Walker Variant
Impulsive Major Depression with Dandy-Walker Variant keep any of them for more than two months. There was no evidence of substance use and head trauma in the history of the patient.
Brain MRI (Figure 1) disclosed an enlarged cisterna magna, hypoplasia of the cerebellar vermis, and dilated ventricles, indicating Dandy-Walker variant. Intelligence testing (K-WI-SC) confirmed a normal range of intelligence (IQ: 94; latent IQ: 105). The result of the psychological testing showed that the patient had aggressive perceptions of other people, and that he had poor impulse control ability. The patient was anticipated to be sensitive to trivial external stimuli, to easily become angry, and to be aggressive. Beck Depression Inventory (BDI) and Hamilton Depression Scale (HAM-D) tests were performed the day after his admission, and the patient obtained 23/63 and 28/50 scores, respectively.
Mirtazapine treatment was started at a dose of 7.5 mg/day and the dose was increased to 45 mg/day on the fifth day, for the patient's depression. Valproate treatment was started at a dose of 450 mg/day for the patient's aggressiveness and lack of impulse control, and the dose was increased to 1500 mg/day on the 12th day of the patient's confinement. About a week after the initiation of valproate treatment, the patient's aggressiveness and uncontrolled impulsive behavior were slightly alleviated. Quetiapine was added for further treatment of his aggression. Quetiapine was added at a dose of 100 mg/day on the 12th day of confinement, and the dose was increased to 800 mg/day on the 20th day. The BDI score decreased from 23 to 9, and the HAM-D score went down from 28 to 10 after three months of treatment. The patient still presented some uncontrolled impulse and aggression, although these had improved much in the three months.
DISCUSSION
Cerebellar lesions do not always cause ataxic motor syndrome. They also have manifestations as cerebellar cognitive affective syndrome (CCAS), including depression and various symptoms (e.g., aggression, psychotic disorder). The cognitive and psychiatric factors of CCAS have been conceptualized as "cognitive dysmetria concept". 12 This holds that a universal cerebellar transform facilitates the autonomic modulation of behavior around the homeostatic baseline, and the behavior being modulated is determined by the specificity of the anatomic subcircuits, or loops, within the cerebrocerebellar system. 13 The patient in this case study had major depressive disorder with impulsive behavior as a main problem, but his superficial relationships with other people, difficulty in building appropriate relationships, and frequent change of occupation despite his higher-than-average intelligence seemed to have been associated with his Dandy-Walker-variant-associated cerebellar lesion. The patient had had recurent depression and had shown impulsive behavior for a long time and was recently diagnosed with major depressive disorder. He had poorer effects compared to other patients with depression, which was a characteristic of the patient.
Leroi et al. 14 found that more than half of their patients with cerebellar degeneration had psychopathology including depression, personality change, psychotic disorders, and impaired cognition. In another study, the cerebellar vermis was smaller in the patients who experienced multiple episodes of depression. 15 This suggested that cerebellar vermal atrophy may be a late neurodegenerative event in those who have had multiple affective episodes. In this case study, the MRI result showed that the patient also had cerebellar vermal hypotrophy, which supports the aforementioned suggestion. This case study also showed the previous understanding of the role of the cerebellum and similar manifestations shown in case studies on Dandy-Walker malformation. In addition, the theoretical explanation that the cerebellum is associated with repeated depression and impulsive behavior has been confirmed via Dandy-Walker variant. The previously explained theories have been further developed via this case study, where superficial relationships with other people, repeated impulsive behavior, persistent exacerbation and repetition of depression, and resistance to treatment were unique features of the patient.
In conclusion, this case study showed that the cerebellum could be associated with depression with impulsive symptom, and expanded the understanding of Dandy-Walker-variantassociated cerebellar lesions. | 2018-04-03T00:49:55.435Z | 2013-09-01T00:00:00.000 | {
"year": 2013,
"sha1": "abe6ab293b616f36b7196db4a2d5ec91e558358b",
"oa_license": "CCBYNC",
"oa_url": "http://psychiatryinvestigation.org/upload/pdf/pi-10-303.pdf",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "abe6ab293b616f36b7196db4a2d5ec91e558358b",
"s2fieldsofstudy": [
"Medicine",
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
253310896 | pes2o/s2orc | v3-fos-license | Specific Graft Treatment Solution Enhances Vascular Endothelial Function
Background: Saline is still the most widely used storage and rinsing solution for vessel grafts during cardiac surgery despite knowing evidence of its negative influence on the human endothelial cell function. Aim of this study was to assess the effect of DuraGraft©, an intraoperative graft treatment solution, on human saphenous vein segments and further elaborate the vasoprotective effect on rat aortic segments in comparison to saline. Methods: Human Saphenous vein (HSV) graft segments from patients undergoing aortocoronary bypass surgery (n = 15), were randomized to DuraGraft© (n = 15) or saline (n = 15) solution before intraoperative storage. Each segment was divided into two subsegmental parts for evaluation. These segments as well as rat aortic segments stored in DuraGraft© underwent assessment of vascular function in a multichamber isometric myograph system in comparison to Krebs-Henseleit solution (KHS), a physiologic organ buffer solution. Results: Potassium-Chloride (KCL)-induced contraction depicted a tendency towards increase when treated with DuraGraft© compared to saline preservation of HSV segments (23.02 ± 14.77 vs 14.44 ± 9.13 mN, p = 0.0571). Vein segments preserved with DuraGraft© showed a significant improvement of endothelium-dependent vasorelaxation in response to cumulative concentrations of bradykinin compared to saline treated segments (p < 0.05). Rat aortic segments stored in saline showed significantly impaired vasoconstriction (3.59 ± 4.20, p < 0.0001) and vasorelaxation when compared to KHS and DuraGraft© (p < 0.0001). Conclusions: DuraGraft© demonstrated a favorable effect on graft relaxation and contraction indicating preservation of vascular endothelial function. Clinical Trial Registration Number: NCT04614077.
Introduction
Coronary artery bypass grafting (CABG) continues to be the "gold standard" for patients with complex multivessel coronary artery disease because of the superior longterm outcome.In the modern era of arterial revascularization, saphenous vein grafts (SVGs) still remain the most often used conduits for CABG worldwide [1,2].As widely known the graft patency is influenced by multiple factors.One of these is intimal hyperplasia progressing to vein-graft disease and graft failure [3].Various aspects and procedures have been studied and published with fundamental improvements like pedicled harvesting technique, no touch methods and of course external stenting of the vein graft [4,5].However, one topic was neglected for a long period but currently gains more and more attention [6].The type of graft storage, flushing and rinsing solution that is intraoperatively used to store the conduits can largely influence endothelial integrity and vessel function [7].Since the moment of grafting is the last time when the surgeon can influence the "auto-transplanted" vessel, every measure and care should be taken to ensure the best possible longterm outcome.This includes the intraoperative storage and flushing solution.The benefits of meticulous surgical handling, training, experience and various precautions during harvesting are redundant if the procedure is interfered by using the incorrect storage solution.In line with that, a subsection of the PREVENT IV trial demonstrated that using a buffered solution resulted in lower vein graft failure rates when studied by angiography at 12 to 18 months after surgery and showed generally a better clinical outcome compared to non-buffered acidic saline (pH value of 5.5) or autologous whole blood solutions (AWB) [8].Since AWB is only beneficial as long as inside the intact circulation and harmful once outside e.g. in the operative setting, due to an alkalytic pH (above 8) as a result of CO 2 loss we did not include AWB in the current experiments.The results for AWB in the literature are well reported and are undoubtedly not beneficial for the endothelium [9,10].Still heparinzed saline and even AWB are in use in the clinical setting, not only in coronary procedures but also for peripheral vascular surgery where the same principles apply.Therefore, novel approaches should limit the early endothelial dysfunction or preserve endothelial integrity to boost graft patency which has become a highly demanded clinical goal these days.
The current study compared the impact of intraoperative preservation of SVGs from patients undergoing CABG in a specific ionically balanced storage solution (Dura-Graft®, Maryzyme Inc, Jupiter, FL, USA) versus saline solution in the isolated organ bath testing system.Dura-Graft®, is also a pH-balanced physiological salt solution containing L-glutathione, L-ascorbic acid, L-arginine and other additives that protect the graft from the damaging effects of ischemia and handling during CABG.Dura-Graft® is a CE marked intra-operative graft storage solution and currently approved in Europe and several other global health care systems.
Since this is currently the only specific storage solution DuraGraft® was the target solution in this study to be compared to the still most widely used solution saline.In addition further characterization of the impact of saline on vascular endothelial function in rat aortic segments and the direct influence on human umbilical vein endothelial cells (HUVECs) was undertaken to specify initial findings.
Patients and Methods
The study was approved by the local Ethics Committee Nr.EK-20-219-1020 of the City of Vienna/Austria and registered as observational study by ClinicalTrials.gov.under the number NCT04614077.
Saphenous vein segment remnants were collected from 15 CABG patients after their informed consent.Patient's details are presented in Table 1.Within 15 patients undergoing aortocoronary bypass surgery, saphenous vein graft segments were randomized to DuraGraft© (n = 15) or heparinized saline (B Braun AG, Melsungen, Germany) (n = 15) solution before intraoperative storage.Each 2 cm long segment of human saphenous vein (HSV) was divided into two 1 cm long parts for twofold evaluation.In total n = 28 HSV segments were collected and n = 23 segments/conditions were used for endothelial-dependent and endothelial-independent vasorelaxation assessment.Special care was taken to exclude patients with concomitant diseases or medical treatment that could interfere with the outcome, testing methods with special focus on vessel wall reactivity and further pathophysiological vascular conditions.
The following inclusion criteria were applied: Age between 18-80 years.Planned CABG operation.Suitable vein grafts with the absence of blow outs, varicous veins or previous stripping.
Pregnant women were not included.Any disease of the lower veins.Any vasculitis.Hemoglobin A1c (HbA1c) levels >6.5% (mmol/mol).The segments of human saphenous vein were harvested in open technique.All patients underwent preoperative ultrasound scanning of the vein segments and varicose veins (outer diameter above 3.5 mm) were excluded.Special care was taken not to stretch by brisk handling or to touch the vessel frankly during the harvest procedure, therefore vein grafts were only harvested by experienced surgeons.
Assessing Vasoactivity in Human Saphenous Vein Segments Using Wire Myograph
Human Saphenous vein segments were cut out in 20 mm pieces, carefully pressure controlled flushed with 10 mL (mL, NaCl or DuraGraft®) at room temperature and placed in the solution they were assigned to (NaCl or Duragraft), any contact with any other substance was fully avoided.There was no mixture of the substances.Flushing, storage and testing was undertaken only with the given solution (NaCl or DuraGraft®).Each 20 mm long vein segment was divided into two study samples accounting for a total of 28 samples, as two sub-segments had to be excluded.The segments were then put immediately into preoxygenated (45 minutes oxygenation time) DuraGraft© or saline solution at normal room temperature and transferred to the laboratory in a sterile isolated box.The segments were kept under these conditions for a total of 60 minutes and then put into cold and oxygenated (5% CO 2 and 95% O 2 ) Krebs-Henseleit solution (KHS) containing (in mM/L) 119 NaCl, 4.7 KCl, 2.5 CaCl 2 •2 H 2 O, 1.17 MgSO 4 •7 H 2 O, 20 NaHCO 3 , 1.18 KH 2 PO 4 , 0.027 EDTA, 10.5 glucose) and the segments were gently cleaned from all connective tissue using Zeiss stereo preparation microscope (Carl Zeiss Meditec AG, Jena, Germany).After cleaning the parts of the saphenous vein, the subsegments were cut into 2 mm pieces and mounted onto a multi-chamber isometric myograph system (Model 620M, Danish Myo Technology, Aarhus, Denmark).The single organ chambers of the myograph were filled with heated (37 °C) and oxygenated KHS and each individual chamber was further heated and bubbled with oxygen during the whole procedure.To determine the resting tension we used the AD Instruments' LabChart® DMT Normalization Module (ADInstruments Inc., Colorado Springs, CO, USA) to mimic physiological conditions (target pressure: 20 mmHg).Segments were allowed to equilibrate for 30 minutes and resting tension was continuously adjusted during this period as described previously [11].Reference contractions were elicited by hyperkaliaemic (124 mM, KCl) solution.Precontraction of the human saphenous vein was achieved by norepinephrine (NE, 1 µM, Arterenol, Sanofi), respectively.Endothelial dependent and independent relaxation was tested by the cumulative dosage of bradykinin (Bra, 1 nM-10 µM, a nitric oxidedependent vasodilator, Sigma Aldrich) and sodium nitroprusside (SNP, 0.1 nM-1 µM, a nitric oxide-independent vasodilator Merck), respectively.The data were continuously recorded using the software program LabChart Pro (ADInstruments Inc., Colorado Springs, CO, USA).
Assessing Vasoactivity in Rat Aortic Segments Using Wire Myograph
To further test the potential vascular protective effects of DuraGraft©, segments of the abdominal aorta were used from Sprague Dawley rats.Male adult Sprague-Dawley rats (12-14 weeks old, body weight of 350-380 gram; Department for Laboratory Animal Science and Genetics, Himberg, Austria) were used.The experimental protocol was approved by the Ethics Committee for Laboratory Animal Experiments at the Medical University of Vienna and the Austrian Ministry of Science and Research (BMWF-66.009/0023-WF/V/3b/2016)and conforms with the Guide for the Care and Use of Laboratory Animals, published by the US National Institutes of Health (NIH Publi-cation No. 85-23, revised 1996) [12,13].Briefly, rats were anaesthetized by intraperitoneal injection of a mixture of Xylazine (4 mg/kg; Bayer, Leverksen, Germany) and Ketamine (100 mg/kg; Dr E. Gräub AG, Switzerland), heparin was injected (iv.femoral artery) and the abdominal aorta was collected as described previously [12].
After cleaning, the segment was cut into 2 mm sections and mounted onto a multi-chamber isometric myograph system (Model 620M, Danish Myo Technology, Aarhus, Denmark).The chambers were filled with one the following solution: (1) KHS (gold standard, positive control), (2) DuraGraft© and (3) physiological saline (0.9%).Each individual chamber was further heated and bubbled during the whole procedure.To determine the resting tension the AD Instruments' LabChart® DMT Normalization Module to mimic physiological conditions was used with a target pressure of 100 mmHg.Segments were allowed to equilibrate for 45 minutes and resting tension was continuously adjusted during this period as described previously [11,14].Reference contractions were elicited by hyperkaliaemic (124 mM, KCl) solution.Precontraction of the aorta segments was achieved by Phenylephrine (PE, 1 nM-1 µM, Sigma Aldrich).Endothelial dependent and independent relaxation was tested by the cumulative dosage of Acetylcholine (ACh, 1 nM-10 µM, a nitric oxidedependent vasodilator, Sigma Aldrich) and sodium nitroprusside (SNP, 0.1 nM-1 µM, a nitric oxide-independent vasodilator Merck), respectively.
Human Umbilical Vein Endothelial Cells (HUVEC) Cultivation and Cell Viability Measurement
Human umbilical vein endothelial cells (HUVECs, Lonza, Basel, Switzerland) were cultured in Medium 200 supplemented with Large Vessel Endothelial Supplement (LVES), 10% foetal bovine serum (FBS) and 1% penicillin and streptomycin solution and maintained at 37 °C and 5% CO 2 .The cells were cultured on 96 well plates at least for 24 hours.HUVECs were treated either with 100 µL of (1) Medium 200 served as control, (2) KHS, (3) DuraGraft© solution or (4) saline solution for 30 and 60 minutes followed by cell viability measurement.Briefly, XTT sodium salt was dissolved in warm PBS (1 mg/mL, Santa Cruz Biotechnology) and phenazine methosulfate was added in 25 µM final concentration.50 µL of XTT solution was used and the cells were incubated for 3-4 hours for the colour reaction.Absorbance was measured at 450/630 nm wavelength with a plate reader (Tecan SparkControl Magellan V2.2, Männedorf, Switzerland).
Materials and Reagents
All chemicals were purchased from Sigma Aldrich (Sigma Inc.Burlington, MA, USA) unless otherwise specified.Preservation solutions for all experiments contained 10 units/mL unfractionated Heparin.DuraGraft© was purchased from Somahlution Inc, Jupiter, Florida, United States.
Data Analysis
The contractile response was defined by the stress, which was calculated using the force generated by the vein rings.Vascular relaxation to Bradykinin and ACh was expressed as percentage of contraction to NE or PE, respectively.Differences in concentration-dependent relaxations induced by Bradykinine and ACh were analysed using twoway ANOVA followed by Bonferroni's test when appropriate.Differences between multiple groups were analysed using one-way ANOVA followed by Bonferroni's test.The number of experimental observations (n) refers to the number of vascular segments in respective experiments.
To demonstrate the good comparability of all cohorts, statistical testing for differences in baseline, procedural, and follow-up data has been performed.Depending on the variable's distribution continuous data are either expressed as means and standard deviation (+/-SD) or median and were analyzed with one-way analysis of variance (ANOVA).Categorical variables are expressed in absolute numbers and percentages.
All data were expressed as mean ± SD and were calculated using GraphPad Prism (Version 7.03; GraphPad Software Inc., San Diego, CA, USA).
Statistical significance was accepted when p < 0.05.
Results
We took maintenance of pH was vital to physiologic function and cellular viability of SVGs as given.In addition, it is important to state therefore that SVG segments after 30-60 minutes incubation with saline or DuraGraft© were transferred into Krebs solution and subsequently all measurements were performed in Krebs solution in this study.
Effect of Preservation Solution on Contractile Responses in SVGs
Maximal contraction in response to KCl (124 mmol) in Krebs solution, showed a tendency towards increase in contraction in DuraGraft© when compared to normal saline preservation HSV (23.02 ± 14.77 vs 14.44 ± 9.13 mN, p = 0.0571; segments/group, Fig. 1A).Next, when the resting tension reached a stable baseline, no attempt was made to adjust the HSV tension.After the equilibrium, the HSV segments were primed by the addition of norepinephrine (1 × 10 −6 M) to the organ bath in order to measure the maximal contraction in response to α 1 adrenoceptor activation.There was no difference in maximal contraction (% of KCl contraction) achieved by NE between the groups (Saline 85.31 ± 33.9% and DuraGraft© 72.66 ± 33.13%; p = 0.167, Fig. 1B).
Effect of Preservation Solution on Endothelial-Dependent and Independent Vasorelaxation in HSV
To investigate the potential protective efficacy of DuraGraft© on the vascular endothelium, we assessed the vascular reactivity of HSV segments in patients planned for elective CABG.The HSV segments were stored in DuraGraft© showed a significantly preserved endotheliumdependent vasorelaxation in response to cumulative dosage of Bradykinin in comparison to saline stored segments (Fig. 2A, p < 0.05).Endothelium-independent vasorelaxation was assessed by the response to cumulative dosage of SNP, and there was no difference between the two storage conditions (Fig. 2B).
Vasoconstriction and Endothelial-Dependent Relaxation in Rat Aortae
In the next step, rat aorta segments were used as a model of normal vascular tissue to further characterize and compare the vascular protective effects of DuraGraft© solution.The myograph chambers were filled with KHS, saline and DuraGraft© and the aorta segments from rat were mounted.Aorta segments that were kept in DuraGraft© showed comparable response to KCl (contraction; 20.97 ± 3.37 vs 22.87 ± 2.57 mN, Fig. 3A) and ACh (endothelium dependent relaxation; Fig. 3B) as KHS solution.In contrast, the aorta segments were kept in physiological saline showed significant impairment in response to both vasoconstriction (3.59 ± 4.20, p < 0.0001, Fig. 3A) and vasorelaxation when compared to Krebs Solution as well as Dura-Graft© preservation (Fig. 3B, p < 0.0001), respectively.
Cell Viability in HUVECs
Fig. 4C displays the morphology of HUVECs cultured in control medium (M200 Medium).The cells were aliquoted and kept in one of the following conditions: (1) control group (M200 Medium), (2) KHS, (3) physiological saline and (4) DuraGraft© solution for 30 minutes or 60 minutes.Then the cell viability was evaluated by XTT assay.Saline treatment for 30 or 60 minutes markedly inhibited cell viability (Fig. 4A,B compared to control, KHS and DuraGraft©, respectively (p < 0.0001).Of importance, cell viability was similar between the control and DuraGraft© group after 30 or 60 minutes incubation (Fig. 4C), suggesting DuraGraft© maintains endothelial cells viability and metabolism, indicating the fact that it is an optimal solution for preserving endothelial cells viability and function for at least 60 minutes.In addition, HUVECs were cultured under M200 Medium, KHS or DuraGraft© presented a confluent elongated shape, while those that were subjected to saline became spherical and non-confluence.
Discussion
Previous studies have already suggested that preservation in physiologic saline may harm vascular conduits and can accelerate the development of neointimal hyperplasia formation [3,[5][6][7][8].Still saline is one of the most or the most widely used solution for intraoperative graft storage or graft flushing besides AWB preparations or individual mixtures.Saline if non buffered has an acidic pH value of 5.5, the physiological pH of circulating blood is 7.32 pH.AWB becomes alkalotic once outside the humas circulation as described above.As Veres et al. [15] stated in 2015 storage with physiological saline and heparinized blood solutions is unable to protect the endothelium against cold ischaemia and warm reperfusion injury.Already in 2014, Harskamp et al. [8] examined the influence of the preservation solutions on vein graft failure using data from the PREVENT IV trial.Grafts were randomized to different groups of preservation solution consisting of saline, buffered saline and autologous whole blood.Grafts stored in buffered saline had significantly lower one-year vein graft failure rates compared to the other two groups, and were associated with a lower risk of five-year death, myocardial infarction and secondary revascularization, suggesting that intraoperative graft preservation is one of the key procedures in order to reduce graft failure risk.Despite these important clinical findings, heparinized normal saline is still widely used in coronary artery bypass grafting.Interestingly the first representative study was by O'Connell et al. [16] in 1974, conducted on the intima of arterial and not venous grafts.This study clearly demonstrated negative effects of normal saline (NS) on vascular endothelium and graft patency in a rabbit model.The topic of a specific graft storage or even treatment solution was neglected in cardiovascular research but gains in recent years more and more interest [6,7,17].The data from the current study showed a clear positive effect for DuraGraft© as a representative for a specific solution.As presented in the results the HSV segments that were preserved with DuraGraft© showed significantly preserved endothelium-dependent vasorelaxation in response to cumulative dosage of Bradykinin when compared to saline preservation.The solution itself is buffered and upholding the cell metabolism due to preserving glucose levels but also reducing oxidative stress and amino acid (L-Argine) related vasodilatation [17].The product is a relatively novel solution against endothelial-damage developed to efficiently protect the structural and functional integrity of the vascular endothelium.DuraGraft© is described as structural and functional endothelial stabilizer in aortocoronary bypass surgery, antioxidative, radical-scavenging, nitric oxide (NO)-synthetize-supporting, anticoagulant, isotonic structural and functional endothelial stabilizer for graft stabilization during venous and arterial aortocoronary bypass surgery.Saline does of course not provide any metabolism upholding elements and if not buffered has a direct damaging acidic effect on the endothelium.DuraGraft© alleviated in a recent study vascular function in vitro following ischemia-reperfusion injury [18].These results although representing data from in vitro animal studies were in line with the findings of this study conducted on human saphenous vein segements.
Currently a prospective observational registry with DuraGraft© targeting at 3000 patients undergoing an isolated CABG procedure or a combined procedure with at least one saphenous vein grafts or one free arterial graft is finished: EU Multicenter Registry to Assess Outcomes in CABG Patients: Treatment of Vascular Conduits With DuraGraft [VASC].Data on baseline, clinical, and angiographic characteristics as well as procedural and clinical events were and will be collected [3,17,19].Because preservation in the buffered solution represented by Dura-Graft© appeared to be superior to non-buffered saline in isolated rat aorta segments and relaxation in HSV, we concluded that maintenance of pH could be vital to physiologic function and cellular viability of HSV.Furthermore, a recent study by Tekin et al. [20] demonstrated that SVG stored in DuraGraft© had lower oxidative level and higher antioxidant capacity, both may contribute and partially explain the preservation of endothelial function as observed in another study by Szabó et al. [12].In addition, it is important to state that SVG segments after 40-60 min (comparative time as in the Operating room) incubation with saline or DuraGraft© were transferred and then kept in KHS solution with all measurements performed under this conditions, suggesting DuraGraft© storage solution effectively alleviates endothelial dysfunction [21].
As the next step further characterization to compare the vascular protective effects of DuraGraft© solution on rat aorta segments as a model system of healthy vascular tissue was conducted.The aorta segments that were kept in DuraGraft© showed comparable response to Potassium-Chloride and endothelium dependent relaxation solution.In contrast, the aorta segments that were stored in saline showed significantly impaired vasoconstriction and vasorelaxation when compared to KHS and DuraGraft© preservation.In line with the initial findings, DuraGraft® alleviates vascular dysfunction following ischemia and reperfusion injury by reducing nitro-oxidative stress and the expression of intercellular adhesion molecule-1 (ICAM-1), without leukocytes engagement in the rat model [12].Furthermore, the study by Pachuk et al. [7] in 2019 displayed in pig mammary arteries results exactly in line with the data above for an animal model set up for healthy vessel segments.Loss of HSV graft-cell viability was observed as early as 15 minutes post-exposure to saline whereas viability was maintained up to 5 hours' exposure to DuraGraft©.Histological analyses performed on pig mammarian artery (PMVs) demonstrated endothelial damage in PMVs stored in saline.Cytotoxicity assays demonstrated that saline-induced microscopically visible cell damage occurred within 60 minutes [22,23].In line with these findings, cell viability was evaluated in HUVECs by XTT assay in this study, saline (30 and 60 minutes) treatment markedly inhibited cell viability when compared to the control KHS and DuraGraft©, respectively (p < 0.0001).Of importance, cell viability was similar between the control and DuraGraft© group after 30 and 60 minutes incubation.These data suggested that DuraGraft© maintains endothelial cell viability and metabolism.DuraGraft© representing a specific storage solution the data of this study votes for the use of a specific solution for preserving endothelial cell viability and function in vascular procedures.Maintaining endothelail cell integrity may be also important to reduce the risk of graft occlusion, myocardial infarction and repeat-revascularisation due to early endothelial affection and later graft failure.In this context, recent clinical studies demonstrated that intraoperative graft treatment with DuraGraft showed a fa-vorable effect on saphenous vein graft wall thickness compared to saline treatment at 12 months follow up period [12,15,18,24,25].Current literature shows additionally that not only saphenous vein grafts but also arterial grafts benefit from a specific treatment solution.A recent study by Aschacher et al. [14] depicted lower levels of reactive oxygen species (ROS) after the treatment Interestingly in this study an increased expression of transforming growth factor β (TGFβ), platelet-derived growth factor α/β (PDGFα/β), and heme oxygenase-1 (HO-1) which are indicative for vascular protective function was reported after Duragraft exposure.Once more, as summarized in the literature and detected in this triple approach, saline is clearly not beneficial for the human endothelium whilst DuraGraft© being a representative for a specific storage and treatment solution demonstrated a significantly positive, at least superior to non-buffered saline solution effect.This study and the current results call for the stop of saline as vascular storage and graft flushing solution and for the use of a specific agent instead [7,12,14,15,[17][18][19][21][22][23][24][25][26][27].
Limitations
The study was limited by its single center design and patient's numbers in terms of the HSV samples.Although specific care was taken to avoid any influence on the vessel and endothelium in terms of preparation or transport some risk factors or patient's details might be present but not obvious at the time of hospitalization.In addition, animal study experiments confirm that DuraGraft© is as efficient as KHS in respect of vascular reactivity, and superior than saline.However, we have not stored the segments of aorta (rat) either in saline or that DuraGraft© prior to performing vascular reactivity assessment and performed on isolated aortic not on venous segments.However, we do not anticipate a difference between the artery and the venous segment in respect to vascular protection by Dura-graft©.This study represented a momentum snapshot of the influence and further longterm data is urgently needed to confirm the protective effects.The author WB is participating in the European Multicenter Trail VASC as national PI.Nevertheless, the protective effect of DuraGraft© was demonstrated on human specimens in this study.
Conclusions
Saline is still the most widely used storage and flushing solution for vessel grafts during cardiac surgery.Saline is clearly not beneficial for the human endothelium whilst DuraGraft© being a representative for a specific storage and treatment solution demonstrated a positive effect.This study and the current results call for the stop of saline as vascular storage and graft flushing solution and for the use of a specific agent instead.
Fig. 2 .
Fig. 2. Endothelial dependent and independent vasorelaxation-saphenous vein grafts.Effects of NaCl (black) and DuraGraft® (red) on vascular reactivity in the saphenous vein grafts.(A) Vein rings were precontracted with NE and relaxed with the cumulative dosages of Bradykinin.The Bradykinin response is expressed as percentage of the maximum NE response and baseline tension.(B) Sapneouse vein were precontracted with NE and relaxed with the cumulative dosages of sodium nitroprusside (SNP).The SNP response is expressed as percentage of the maximum NE response and baseline tension.Data are expressed as mean ± SD, n = 15 patients n = 23 segments/condition. | 2022-11-05T15:56:48.512Z | 2022-10-28T00:00:00.000 | {
"year": 2022,
"sha1": "e0488d3b4bdcc9c38d44876d337209d176b2f416",
"oa_license": "CCBY",
"oa_url": "https://www.imrpress.com/journal/RCM/23/11/10.31083/j.rcm2311368/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7d5dc100ea914bb89ddd05de6fd9764c2eb613a6",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
267114646 | pes2o/s2orc | v3-fos-license | ArchEthno - a new tool for sharing research materials and a new method for archiving your own research
The archiving of ethnographic material is generally considered a blind spot in ethnographic working methods which place more importance on actual investigations and analysis than on how archives are constructed. A team of computer scientists and ethnographers has built an initial tool for sharing ethnographic materials, based on an SQL relational data model that suited the first survey processed but proved difficult to transpose to other surveys. The team developed a new tool based on dynamic vocabularies of concepts which breaks down archiving into three stages. Firstly ethnographers can select and contextualise their survey materials; secondly they structure them in a database according to the research question discovered during their survey; finally, they share this data with other researchers subject to the opinion of an ethics committee whose members are competent in ethnography.
INTRODUCTION
From the 2000s onwards, it became extremely difficult for sociologists and anthropologists to defend their publications which were increasingly being questioned by respondents to their surveys or other third parties.Cases with varyingly high profiles destabilised certain researchers and even the discipline as a whole.Therefore why should researchers bother with the delicate matter of sharing unpublished material?This then is the context which has led to the open data policy promoted in European-level research policies being perceived by the ethnographers' community as a pointless and even harmful injunction imposed on them from above in the name of transparency in science [Genèses, 2022].Ethnographers are artisans who provide a specific type of material namely their field notes based on ethnographic encounters [Pina Cabral, 2011].These are distinct from the data constructed by large public or private organisations and also from archives the preservation of which depends on political decisions.The validity of ethnographers' findings is now subject to peer review which is not always competent to judge the ethnographic approach and also to multiple forms of review by the people concerned by the research, whether as interviewees or because they feel authorised to speak on behalf of those who were.These discouraging observations risked calling into question the actual exercise of the ethnographer's profession.This situation led to a team of ethnographers, sociologists, anthropologists and photographers working with data base and computer engineers to develop a method to protect respondents and ethnographers against the circulation of responses and practices which ethnographers have access to because a relationship of trust has been established [Béliard & Eideliman, 2008].The intention was also to protect them from suspicions and tactlessness within the collective ethnography teams themselves [Laferté, 2016] and in the public space [Avanza, 2016].This work took as its starting point a multidisciplinary collective survey whose ethnographic materials were archived in print format.The method for achieving this form of protection is based on an initial sideways step.It does not address individual ethnographers and instead relies on partial pooling of ethnographic materials selfregulated by professional ethnographic researchers and academics in compliance with ethnographic deontology and scientific ethics.Producing these materials is expensive in terms of time and emotional cost as is their analysis.It is similarly costly to format them for re-use.The fundamental idea underpinning our team's work is that such materials are a useful resource for other research, particularly multidisciplinary research.We work on the complex task of making these materials available to researchers from other disciplines and in this way are breaking with the dichotomy established over the last forty years in the worlds of science and art in the name of the materials' heritage value.This dichotomy posits the idea that the value (and even the price) of certain materials is immense if their producer is famous and zero if the producer is unknown.It was imposed on the scientific community from outside and has led to the destruction of a great deal of material.Within the scientific community itself, it has created a certain form of impunity as regards the theft of research material when creators do not have the means to publish their results quickly.Our method proposes a profound modification in the way materials are disseminated with their value being linked to their quality rather than to the status of their producer.This quality is gauged by the work invested in making them available and by the scientific interest of their re-use.This form of transformation enables us to re-open the question of the cumulative nature of knowledge produced by the sciences studying societies.
I IS ARCHIVING ETHNOGRAPHIC MATERIALS IN 2020 IN A DEADLOCKED SITUATION?
Who are ethnographic materials archived for and why and how is this done?Over the last twenty years or so, these questions have been the subject of much discussion in the international scientific community working in the two disciplines that have forged the ethnographic tradition since the nineteenth century -social and cultural anthropology (which is commonly referred to as ethnology in France) and the so-called qualitative sociology.
The 2000s -a turning point
Many post-publication conflicts between researchers and survey respondents marked the beginning of the 21st century, revealing the structural weakness of ethnographers as academics and social scientists studying the contemporary world in different socio-historical contexts and situations.To give just a few examples, in 2006 the Quebec anthropologist Natacha Gagné was challenged on the international scientific scene by other anthropologists claiming that a non-Maori had no right to speak about Maori cultures [Gagné, 2008].A few years later, her university did not defend her when a former survey respondent made an accusation against her.In France, the Thérame affair [Weber, 2008] highlighted the different issues facing an interview respondent who was also a writer and a sociologist of literature working on an analysis of the field of 1970s French literature.The former's self-image and reputation as an author were at stake while the latter was more concerned by her contribution to scientific knowledge of the historical risks linked to the construction of value in the literary field.There was a mix-up involving two judgement criteria -firstly did the sociologist damage the writer's self-image and reputation?And secondly does the writer's complaint invalidate the increase in knowledge her work provided?This situation could undoubtedly have been avoided if the sociologist had asked the writer to read her article before publication or if a judge had been able to help them come to a negotiated compromise.An initial summary report of the situation in France [Laurens and Neyrat, 2010] showed that controversies tended to be much more violent when they had a political dimension or highlighted the structural weakness of ethnographers in unambiguous power relationships, for example relations between a boss and employees carrying out an incognito investigation.The Alice Goffman affair in 2015 combined all these ingredients -a woman, a brilliant anthropologist and heir to Erving Goffman's reputation, was dragged through the mud.She was accused of being complicit in abuses committed by her interviewees while she was observing them and then actually accused of lying [Avanza, 2016].The mix of morality, politics, judicialisation and criminality became a white-hot issue.
Two anthropological traditions that know nothing of each other
Do these conflicts testify to a shift in the balance of power between observers and the observed or even peer judgement being worryingly replaced by public judgement?To judge this, we need to take into account the existence of two distinct anthropological traditions.Anthropology's editorial, encyclopaedic and museographic traditions have long promoted a genuine policy of publishing and conserving heterogeneous materials, texts, sound and visual recordings that are often linked to objects.This tradition dates from the earliest ethnographic expeditions and has successfully alimented some splendid museographic experiments around the world1 as well as grandiose encyclopaedic hopes 2 .In France, several national institutions such as the Muséum d'Histoire Naturelle, the Musée de l'Homme, the Musée du Quai Branly and the Musée des Arts et Traditions Populaires which has been succeeded by the MUCEM 3 are at the forefront in showcasing these ancient collections.Their collections tend to be of more interest to social science historians than to contemporary ethnographers.An exception would be cases in which they engage in dialogue with movements that aim to revive indigenous arts and customs [Le Gonidec, 2020].Another tradition stemming from Malinowski's first survey published in 1922 merged the two roles of investigator and researcher in a single ethnographer-author role [Stocking, 1983].This tradition spread in France in a form of opposition to the museum tradition [Segalen, 2019].For a long time it was spared the radical criticism of the scientific pretensions of social and cultural anthropology that occurred in the United States and Great Britain.
Incertitude about the status of field diaries
An ethnographer's most important prime material is his or her field diary but this tool has never had a unified status.The publication of Malinowski's 'A Diary in the Strict Sense of the Term' [Malinowski, 1967] written between 1914 and 1920 created shockwaves.The ethnographer revealed the hidden sides of a revered and radiant work, Argonauts of the Western Pacific, the ethnographers' bible to this day [Malinowski, 1922].His diary was not initially intended for publication and the author revealed therein his doubly difficult relationship with the members of the colonial society he was forced to mix with and with the natives of the colonised societies he was observing and studying by vocation.Anthropologists were often careful to destroy their diaries -at least partially -to protect their allies in the field and themselves in conflict situations.This was notably the case in colonial contexts like for example Evans-Pritchard [Grootaers, 2001] or criminal contexts for example Tarrius [2017] or Alice Goffman [Yanow, 2021].The distinction between a personal diary and an investigative research diary [Weber, 1991] coincided with the ethnographic method spreading beyond anthropology [Beaud & Weber 1993, Parent & Sabourin, 2016] to reach the field of design [Léchot Hirt et al., 2015].
The lexicon of qualitative issues
From 1994 to 2020, ethnographers were encouraged to deposit their materials to facilitate their re-use through several major initiatives aimed at depositing and sharing ethnographic materials.The most important initiatives of this kind in Europe were Qualidata in Great Britain4 and Bequali in France5 .Taking stock of these initiatives shows they have all come up against the following difficulties: • the reluctance of researchers to deposit what they consider to be their own professional property has led university authorities to attempt to make this legally compulsory (difficult to implement except in exceptional cases) and to offer financial incentives which is now effective but still viewed as yet another administrative constraint [Duchesne & Noûs, 2019]; • the target audience's low level of interest in re-using such materials has led the promoters of platforms of this kind to target captive audiences like students despite the risk of discouraging future ethnography researchers from using the tool; • there have been continuing communication difficulties between researchers and archivists [Wasamba, 2013] despite some successes, for example in the field of the political history of social science research [Wolikow, 2003]; • a terminological disagreement has arisen between ethnographers who refer to materials and academic authorities who refer to data with the latter influenced by the powerful models of statistical surveys and administrative data.
In the social sciences, ethnography corresponds to the inductive phase of scientific research: researchers do not know precisely what they want to prove before the beginning of their fieldwork.Sometimes they have questions and choose places, events and people they think relevant to find answers.Sometimes the scientific question emerges from the analysis of the first ethnographic encounters.In no case the ethnographer is able to follow biomedical research's protocolised methods or hypothetico-deductive methods from economics [Olivier de Sardan, 2008].Additionally, ethnography is based on a doubly private form of participant observation in which ethnographers and their interviewees commit to a greater extent than to purely professional obligations.This means ethnography now needs to redefine the circle in which its observations are shared.This is the idea foreseen by Geertz in 1988 when he asked the somewhat anxious question: "Who do we need to convince now?Africanists or Africans?
And of what?" [Geertz, 1988].Today, this question is posed in a more precise fashion.Who can we open up our intermediate results and confidential materials to and why?
The lexicon of data. Do open data and the General Data Protection Regulation (GDPR) constitute a double bind?
Additional difficulty has arisen because of the European-level General Data Protection Regulation (which prohibits the retention and dissemination of personal data) combined with regulations stipulating that the head of a research project is responsible for applying the law.
Universities and research organisations are beginning to measure the likely impact of all this on research based on the ethnographic method.Some universities have appointed a GDPR referent to work on protecting students, particularly those working on PhDs, from potential misuses of these laws by certain employers or respondents who may aim to retain a monopoly on legitimate discourse about the population they represent.With a few exceptions, university data platforms mainly disseminate tools and methods for using quantitative data which have only recently become available.As is well known, the two models for the relationship of researchers with empirical material are statistics (reinforced by the Big Data boom) and therapeutic trials.Teachers and students in the social sciences still sometimes find themselves referred to ethics and scientific integrity committees that do not have the competence to pronounce on issues involving the ethnographic method.These committees follow the model of Institutional Review Boards in the English-speaking world which have long been badly adapted to ethnographic research [Desclaux, 2008].This situation has led to researchers developing avoidance tactics or taking radical stances against injunctions to open up research data [Genèses, 2022].
Science archives -heritage or accumulation?
The French science archives sphere [Charmasson, 2006] has acquired high-quality instruments as the history of science and technology has developed, driven by major heritage institutions like the Collège de France, the Bibliothèque Nationale, Ecole des Chartes, prestigious libraries, national anthropology museums and the National Archives.However, the study of some of the most important archives in the history of fieldwork [Müller and Weber, 2003] has shown that it is difficult to create collections that bring together archives from among various institutions over the course of turbulent political and social histories.How should the value of research materials be assessed?According to the value of their authors?Of their subjects?Of the new questions such materials can help to explore?In a world overwhelmed by existing digital data that is however eminently fragile due to the rapid obsolescence of their writing formats, how can we create a new science of data that takes into account the skills of producers, historians, digital specialists and future users in all their unpredictable variety?
Despite misunderstandings between the different professions working on this new data science, ethnography could in fact prove an excellent example of how to bring together robust multidisciplinary teams before materials and data are made available and also before enabling other new multidisciplinary teams to appropriate these materials and data.
The development of new practices for surveys and for analysing materials
In the meantime, today's young ethnographers are working with no metaphorical safety net on developing investigation and analysis methods aimed at saving time (sometimes illusorily) and accumulating material with the potential to give them a head start in the race to publish their work.Certain of these researchers seem caught up in a specific form of quantophrenia which involves obtaining as many documents 'of interest' (recorded interviews, photographs, videos, thousands of pages of field diaries, etc.) as possible despite the risk of losing all control of the thought process about the issue at hand.Others find themselves blocked by anxiety deriving from their own uncertainty about the rules they are supposed to follow in the field and in the construction of their research object.Some use their smartphones to record 'on the spot' interviews and are then faced with the current tendency to discredit 'undercover' survey practices or with technical problems linked to exporting files designed to trap their users within a commercial system.Novice ethnographers can also find themselves faced with data security and privacy issues of varying levels of justification and thus entangled in arcane GDPR bureaucracy.
Other practices are developing for editing rough field notes and classifying materials.Ethnographers are as subject as anyone to the temptations of mainstream consumer cognitive tools that promise to turn amateurs and enthusiasts into professionals for the right price.They can however find something positive for their work in this.To manage and process their proliferating research materials, they can use professional tools that are more or less adapted to their requirements such as the classification of photographs on tropy 6 .Ethnographers are at the crossroads of several disciplinary traditions in the cultural professions -linguistics, with its tools for the analysis of language material; the history of literature with its tools for the analysis of texts and the history of art with its tools for the analysis of images.They can also use office tools for automatic transcription like Sonal7 , translation and report writing or alternatively ask themselves questions about all such tools.The method ArchEthno proposes is located between technophilic euphoria and technophobic anxiety and aims to collectively take back control of the profession of ethnographer.Like the profession of archaeologist, this profession is historically attached to writing, images and schematisation rather than computing, as [Fabry, 2021] has demonstrated.
Towards ethnographic lucidity?
The effects of these new practices on the methodology of field surveys and their results have yet to be examined or debated.The practice of open questioning enabled ethnographers to explain their academic status, their projects and their purpose to respondents and guaranteed the confidentiality of the information respondents being asked to provide.This had established itself as one of the norms of ethnography in a context of inter-acquaintance [Beaud, Weber, 2010].It is clear that this practice was overtaken by the widespread practice of covert questioning in the 2010s combined with the sometimes overt aim of 'deceiving' to be able to 'denounce' a given issue.Proponents of this aggressive approach seem unaware of its effects and it has also been countered by forms of anxiety about confidentiality issues.We may reasonably hope a new era is dawning with an ethnography that is aware of its effects.This could be based both on ethnographies in extreme conditions [Shukan, 2016] and on a renewed alliance between photography and the social sciences [Dantou et al., 2020].
II THE GENESIS AND CHARACTERISTICS OF THE ARCHETHNO METHOD
With this hope for a new era of ethnography in mind, since 2013 we have been working on the development of an initial tool for sharing structured materials within a multidisciplinary team and then on enhancing it so that it can be adapted to several different studies [Dantou, 2014, Blum, 2017, Blum, Goudet and Weber, 2022].We have found a solution that can be adapted to materials shared in multidisciplinary teams and those produced and deposited by individual researchers.The production, transformation and uses of this tool have directly resulted in the ArchEthno method.The tool consists of a software suite with three elements that enable researchers to construct a dictionary, enter their materials and metadata locally and allow authorised persons to consult their work on the internet.These are accompanied by instructions for use and a guide to archiving ethnographic materials.The method requires the involvement of an ethics committee for researchers working in institutions or, failing that, professional legal advice.
Researchers alone should decide whether to publish or share scientific material
A dual conviction influenced the method's development.It became clear to us at a very early stage of the project that researchers had to control the entire research cycle and also that they needed to be helped to identify the issues linked to sharing and publishing the hitherto concealed aspects of their research -from the production and collection of materials to their transformation into reusable data.This conviction can be summed up by the following four points: • researchers alone are able to document their own materials (each material has an identity document: who produced it, when, where and in what circumstances).We believe it is counter-productive to entrust the 'investigation of the investigation' to anyone other than researchers except in the case of historical research based on archival collections; • researchers alone are entitled to structure their materials according to the research question being explored which changes as the survey and initial analyses progress; • researchers alone are capable of implementing a multi-disciplinary approach based on common goals; • researchers need to discuss the elements they feel are worth sharing with other researchers and need to be supported by an ethics committee or a legal advisor who will consider and discuss the consequences of sharing with them.
In short, the question of publishing research materials is a fully scientific issue rather than of a technical or legal nature.The technical and legal obstacles need to be overcome so researchers can then rely on the evaluation system of thesis and recruitment juries and the scientific publishing system with its reviewers and others in publishing professions so they can propose the 'material' component that is complementary to the published article or book. .The original survey has also been the subject of several sociology theses based on its ethnographic complements.
ArchEthno2020: a paradigm shift in IT
The ArchEthno2017 prototype constructed around a relational database enabled researchers to think in depth about the conditions for standardising and rationalising ethnographic field data.It also introduced the issues of systematising the documentation of survey conditions and the long-term archiving of digital or digitised material.The database's structuring was based on a set of metadata that was hard to transpose to other surveys.Any changes would have required help from IT experts to modify the data model, adapt the old data to new formats (with a risk of corruption during this work) and review the data entry and retrieval interfaces.This all meant the model was not viable in the long term.
Extending the usages and simplification of the model
A paradigm shift was therefore required and a new tool (ArchEthno2020) was developed between 2017 and 2020 to overhaul the initial prototype.This involved a shift from a rigid SQL structure to approaches based on the semantic web where metadata is structured in modular and scalable concept dictionaries which also provide finely tuned confidentiality management for all the data and metadata.The structuring of metadata and the conditions for its reusability are These concept dictionaries are stored in ad hoc files and act as configuration files for a set of generic software components.They are only written once and definitively.Any extensions or modifications that may be required will be achieved by modifying the concept dictionary but not the software stack.The generic software stack is made up of 4 components: • a dictionary editor used to create or modify a dictionary; • a data entry client -an application for entering and storing data on a personal computer; • a centralised database which enables users to merge data entered via the data entry client; • a consultation website that enables users to view the data in the centralised database using a web browser.
These four components make up a coherent whole in which: • the software's fine level of granularity means confidentiality can be managed and enables researchers to think about their decisions as and when they need to rather than all at once in advance while following a protocol; • a data construction framework enables compliance with scientific research presentation and documentation standards while respecting the inventiveness of researchers carrying out their analyses (in a Structuring module); • a guide to confidentiality decisions helps raise researchers' awareness of the practical implications of their confidentiality decisions before these are submitted to an ethics committee in combination with the user guide which has been simplified as much as possible; • a data entry guide linked to each dictionary available in free access provides guidance for researchers in structuring their material and for those who want to learn about the conditions data are produced under before requesting access to confidential materials (a search engine can be used to check the existence and interest of confidential material).
Even researchers lacking IT skills or the support of IT specialists can use ArchEthno2020 to develop the tool to adapt it to the requirements of their research.The complete software solution was developed as part of an industrial support project funded by the Université Paris Sciences & Lettres (PSL).It has the following strengths and original features: • it can manage the polymorphous nature of data and metadata.This requires ongoing development through concept dictionaries that enable data and their production contexts to be characterised in a unique and persistent way; • it integrates the crucial need for confidentiality in the humanities and social sciences whereby certain data and metadata can only be made available to authorised users.The levels of visibility and authorisations are managed by dictionaries and can be modified when the data is entered.The tool can manage an infinite number of levels of confidentiality; • it offers innovative solutions combining the ability to openly cite data with the need for confidentiality and respect for privacy.
The concept dictionary approach facilitates interdisciplinarity and enables the migration of data between disciplines.Also, from a technical standpoint, the data entry and retrieval interfaces are automatically generated from the dictionaries.
Using mutualisation as a guide to moving back and forth between material and hypotheses
The tool possesses flexibility and the capacity to be used as an instrument for a structuring approach during the dictionary creation phase.These factors open up prospects for wider use than the dissemination of research material.These could include helping researchers to structure field notes and formulate sociological reasoning or to pool materials, questions and hypotheses during a survey's collective phases.In both cases, the tools offer users the certainty of being able to return to the ethical issues involved in disseminating the material with the help of trusted third parties once the analysis is complete.Thus researchers have the peace of mind required for the sociological imagination.
We have proposed basic structuring in three modules that we consider suited to the ethnographic approach.The 'Matériaux/Materials' module can be used to describe each brick of the field diary as the survey progresses.The 'Recherche/Research' module recalls the approach's institutional framework.The 'Structuration/Structuring' module can be used to construct an analysis of the material with reference to the research question as it is being stabilised and to formalise hypotheses that have been tested, invalidated or retained and possibly expanded or made clearer.
The prospect of sharing materials, in the framework of a teaching relationship or a research team, made us aware of the tool's richness.Before sharing, description is required.Mutualising materials helps build sharing communities, truly brings a team into existence and enables collective work to be brought to a close through deciding together the subsequent re-opening opportunities that will be possible.Sharing materials enables researchers to progress in structuring them.This approach is particularly well-suited to multi-disciplinary research which it provides with a practical framework focused on the most important points.What is the research question?How is work divided between team members according to their specialisation?How can any misunderstandings about each discipline's specific approach be avoided?Finally, when it comes to scientific publication, editing materials opens up new ways of sharing so who will be entitled to question decisions taken as regards structuring the research and analysing the materials?The ArchEthno tool offers the ethnographic practice of translation between several social worlds (those of survey respondents, scientific publications and possibly of decision-makers or journalists) a life-size test of what needs to be translated, simplified and finally highlighted or stressed.
Solitary and collective usages of ArchEthno
When a researcher is carrying out a survey alone, ArchEthno can help him/her to clean up field notes without undue 'fetishising' of their textual, narrative or rhetorical dimension.It also helps researchers as the survey progresses to integrate materials that are crucial to the analysis of how the survey was carried out like exchanges of emails, text messages or visual notes.These all make it possible to reconstruct the institutional and material conditions for survey appointments for example.Typed research diaries retain their temporal structure and their character as nonmodifiable material.They are also used to help formulate a research question as the analysis progresses along with theoretical hypotheses based on constant 'back-and-forth' references to scientific concepts and analysis of the materials.The tool then helps ethnographers to put into practice the sometimes mysterious suggested rules on how to develop a "description in concepts" [Passeron, 1991] of an observed event, interaction or situation.
During the collective survey training courses that introduce students to the ethnographic approach, the tool can help make researchers more aware of the two stages of work in this research field.The first is actually carrying out a survey which involves the need for patience, disappointments and meeting people which requires a researcher to reformulate a question that 'speaks' to the respondents as much as to sociologists.The second is an analysis that uses the materiality and content of the collected textual, auditory and visual notes to reflect on the nature of the investigative relationship and the interactions' dynamics before formulating hypotheses that will need to be tested by using sociological materials (description of infrastructures and institutions, even those that are initially invisible to the investigator).ArchEthno can thus become a tool for centralising the various techniques, for making a break with the interviewer's ethnocentrism and which helps sharpen an ethnographer's eye and ear.These include replaying recorded interviews, focusing on the spatio-temporal dimensions of the way interaction occurs and progresses and finally questioning the meaning attributed by others to their own behaviour.
ArchEthno 2020 and the FAIR principles
The ArchEthno2020 tool was constructed based on the principles of FAIR data (Findable, Accessible, Interoperable, Reusable).The data, metadata and the structures used to model the data are all FAIR-ised.All the FAIR principles have been followed and implemented.
• The data is 'findable' because the tool offers a search portal for users to consult existing archives and their metadata.We also intend for this service to be indexed, offered and hosted by the Huma-Num TGIR (very large research infrastructure) as a generic tool for use by different disciplinary communities.• The tool's confidentiality management function ensures a balance between accessibility and reusability.All data have a unique persistent identifier that is public and resolvable even if they are confidential.Once the PID has been resolved the information available depends on the user's level of accreditation.A PID is a public reference to data present in the information system but only authorised users will be able to view the private content of the public reference.In this way, we are complying with the European principle of "as open as possible, as closed as necessary"9 by guaranteeing the traceability and reproducibility of scientific work and making it public even if 'closed' data are involved.• The technological choices our solution is based on mean that interoperability is dependent on the definition of concept dictionaries which set the epistemological limits for each piece of data (production context, potential field of validity, underlying assumptions, modelling) beyond which the piece of datum could become a source of error.As stated above, the process of definition of the dictionary complies with the FAIR Semantic recommendation [http7, http13].
A community of developers and users (data producers and site visitors) has been constructed to make this a long-term development effort.The classification into different levels of confidential information for each survey is submitted to an ethics committee at least 75% of whose members actually practice the ethnographic method.When a team of ethnographic producers linked to a laboratory receiving public funding -including students, PhD students, post-doctoral fellows, independent service providers, contract or tenured engineers or teachers -opens part of its metadata and data it needs to consult the ethics committee for its laboratory.Ethnographic producers who are not in a publically-funded laboratory will be informed that they can consult a legal specialist before opening their research data.The ArchEthno method is not just the software solution involved and obviously still needs to find its place in the socio-technical system being constructed around the challenges linked to opening up research material.It will need to fit in with all the platforms dedicated to anthropology currently being set up on most continents.Indeed [Murillo, 2018] has proposed the idea of an Open Anthro Source that could link these together in ways that are still to be invented.The method will also have to become a link in the chain running from the 'primary' production of material and data by past and present researchers right up to the various 'postproduction' stages whether these involve real or virtual dissemination or editorial development.
Publishing the dictionary and the data. Depositing and reusing material. Examples and future prospects
First and foremost ArchEthno is a tool for research.As such, its success depends on academic recognition of the work required to construct the database in a form similar to software copyright, film credits, printed credits or book colophons.This work includes software development, dictionary design (each dictionary put online is credited to its authors who are duly cited by the producers who deposit their materials and by database visitors) and the formatting and editing of material and data by the ethnographer and his/her editors (each collection put online is credited to its producers and duly cited as a scientific publication).The division of editorial work will become routine as people gradually publish the first editions of materials and data from ethnographic surveys.As with digital publishing, a great deal of attention will be paid to dating the various phases of the work and to the risks of free labour being exploited and added value being captured by the front office [Godechot, 2006].
The first databases made available on ArchEthno2020 will be as follows: • the first Medips Alzheimer database constructed with ArchEthno2017 will be put online on ArchEthno2020 in 2024 along with its dictionary10 .Its confidential material will be deposited to be accessible to authorized readers; • the dictionary [Goudet, Vieujean and Weber 2022] used to design [Blum, Goudet, Weber 2022] will be put online in 2024 along with non-confidential material and data.Confidential data will be gradually deposited as they are prepared.• Special mention needs to be made of databases designed to promote and disseminate students' work carried out during field surveys that combine the ethnographic method with the tradition of documentaries.The ethnographic and documentary traditions do not work with the same responses to crucial anonymity and confidentiality issues [Béliard andEideliman, 2008, Dantou andWeber, 2015].Also, until currently they do not share the same financial stakes linked to the trade in works [Weber, 2011] with a few exceptions.Furthermore, they do not have the same ways of involving respondents in the process of publishing the results of the survey.
Thanks to these first experiments, we will be able to propose criteria and processes to assess the value of research material: number of scientific publications based on this material; precisions on the conditions of its production, including analysis of the fieldwork relation and restitutions workshops; pedagogical issues; interest of the material for multidisciplinary teams.
In order to avoid arbitrary selections we will participate to collegial discussions with dean's offices, archivists, scholars and students.
Other forms of usage are currently being examined.
• In art history and fields involving artistic creation, the ArchEthno tools could be used to document works in anthropology museums' (Musée du Quai Branly -Jacques Chirac, regional ethnology museums, MUCEM, BNF) collections of objects and photographs more effectively.This is particularly the case when the dissemination of certain information could be potentially dangerous for the respondents or their families.This is seen as more of an issue than the 'restitution' of such works and opening them to these people.From this point of view, contemporary ethnographers are the most aware of issues linked to the unwelcome circulation of confidential information [Tarrius, 2017].• In history and archaeology, the ArchEthno tools could be used to construct multidisciplinary databases covering survey contexts, events, stratigraphic units and graphic and photographic documents including confidential geolocalised metadata via a dictionary constructed by several people.The tool can of course be used to store and revisit primary documentation from archaeological digs and also to publish heritage archives produced by an archaeology laboratory over several decades.
More broadly, ArchEthno could have three other forms of usage: • methodological usage through the dissemination of the concept dictionary design method to professional ethnographers.The team is considering the possibility of adapting the tool to metadata in other languages and alphabets; • multi-disciplinary usage through sharing materials equipped with metadata within teams of researchers in which ethnographers work alongside statisticians (sociologists, economists, geographers, demographers, epidemiologists) or environmental science researchers (ecologists, geologists, hydrologists, etc.); • a technological usage by transferring the technical solution to other disciplines like archaeology or astrophysics.
Conclusion
The Archethno method responds to several types of difficulties encountered by researchers in different situations and in different disciplinary traditions and not just in the field of ethnography.Such difficulties are linked to epistemological, ethical or scientific integrity issues.
• From an epistemological standpoint, the method respects researchers' inventiveness as they can publish their own dictionaries when they decide to do so.It also makes it possible to combine 'objective' (documentation of a context) and 'subjective' material (where the investigative relationship is made explicitly clear) without opposing these or reducing their specific features.• From an ethical standpoint, the method allows confidentiality decisions to be made explicitly clear as all materials submitted are initially confidential by default.Also confidentiality restrictions are lifted as late as possible when any confidentiality issues have been fully analysed and the procedures for lifting confidentiality restrictions have been designed.Anonymisation techniques cannot be standardised as each research project has its own individual priority issues to manage.• From a scientific integrity standpoint, the method authorises the citation of data whose content remains confidential and also provides a publication platform that de facto protects ethnographers' work from plagiarism.
Finally, the ArchEthno method makes it possible to clarify and deal with the tensions raised by the delimitation of 'sharing collectives' in contemporary societies that would do anything to defend what they see as their individual borders.We live in a society of cultural hyperconsumption that authorises or values the transparency of practices but also develops sophisticated secrecy techniques.Putting data sharing issues forward for analysis as issues linked to power [Butnaru, 2023] could serve as an effective dispositif or device for 'cooling' conflicts.In a world that is currently suffering from no longer knowing how to think about borders [Debray, 2010], the method could possess an inherent strategy for 'slowing down' cognition to maintain our capacity for analysis when we are faced with the endless deluge of unwanted information that characterises the contemporary public space [Fenoglio and Fleury, 2022].
Figure 1 .
Figure 1.ArchEthno2017's SQL relational data model (1a) compared with the ArchEthno2020 model based on dynamic concept vocabularies (1b).The simplification carried out between 2017 and 2020 means the structure of the database now only contains references to concepts whose description and hierarchisation are the subject of the dictionary.
ArchEthno 2017 for the Medips-Alzh-2005)2003 surveyDisseminating research material requires multidisciplinary skills ranging from IT techniques (creating and structuring relational databases or websites, managing security or encrypting confidential information) to regulatory (for example the new National Plan for Open Science) and legal aspects.Isolated researchers or those working in small teams can find it very difficult to bring together the skills required to ensure materials are effectively disseminated.A team of researchers carried out the 2004 Medips-Alzheimer exploratory survey[Joël and Gramain, 2005, Medips 2006] which involved a large-scale collaboration between economists from LEGOS (Paris Dauphine) and anthropologists and sociologists from the Centre Maurice Halbwachs (ENS).The project received several public (the Ministry of Research's ACI Blanche Young Researchers initiative, 2000) and private(Fondation Médéric Alzheimer, 2003-2005)grants.When the print files corresponding to the ethnographic analysis of the 91 'cases' that made up the survey were deposited in the ENS library in 2011 new funding from the National Solidarity Fund for Autonomy (CNSA) helped extend the survey to other populations and this combined with the TransferS Labex's support made it possible to archive the files digitally.Agnès Tricoche, an archaeologist and data engineer joined the team and worked with us between 2014 and 2017 on the creation of our first shared ethnographic data entry tool (ArchEthno2017) based on a relational database.The underlying data is from research carried out by LEGOS at Paris-Dauphine and the 'Surveys, Fieldwork and Theory' team at the Centre Maurice Halbwachs into the vital family and professional care provided for fully dependent people.This 'Medips-Family' research was carried out in several phases.The most standardised phase involved a survey carried out in 2004 by a dozen or so interviewers of 91 people with Alzheimer's-type disorders contacted through 5 institutions in the Paris region.
For each case involving a person receiving care, the survey material was made up of a set of questionnaires completed by the people contacted, the interviewer's field diary and a summary sheet.A number of additional documents on the institutions were involved in the survey.The survey was based on an original combination of the ethnographer's field diary and structural econometrics.It resulted from long-term research carried out by a health economist and a kinship anthropologist.A simplified version was reproduced by the National Institute for Statistics and Economic Studies (INSEE) in the context of its Handicap Santé surveys 8 The concept dictionaries and their construction procedures comply with the criteria defined in the FAIR Semantic recommendation[http11].The tool's design was discussed at the 'Digital practices in History and Ethnography' session at the 11th plenary meeting of the Research Data Alliance in spring 2017 [http7].Its innovative solutions for openly citing data while fully respecting confidentiality and privacy requirements were considered of interest for applying the GDPR to the HSS with no negative consequences for researchers' working conditions.The tool was developed in dialogue with the main French open data stakeholders -the INSEE, the Secure Data Access Centre (CASD), the National Institute for Demographic Studies (INED) and the Bulletin de Méthodologie Statistique -and also in relation to the social and cultural anthropology of contemporary worlds in Europe (Portugal, Spain, Netherlands, the European Anthropological Association) and the United States (Princeton, Harvard for medical anthropology, the American Anthropological Association). | 2024-01-24T17:57:09.354Z | 2024-01-16T00:00:00.000 | {
"year": 2024,
"sha1": "6cb1716453501170a67768eb8454289f666c0070",
"oa_license": "CCBY",
"oa_url": "https://jdmdh.episciences.org/12873/pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "f0f20e331d65c09cf5382030e7a7ffb2f58ac1ef",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
11554441 | pes2o/s2orc | v3-fos-license | Demonstration of Binding of Neuronal Calcium Sensor-1 to the Cav2.1 P/Q-Type Calcium Channel
In neurons, entry of extracellular calcium (Ca2+) into synaptic terminals through Cav2.1 (P/Q-type) Ca2+ channels is the driving force for exocytosis of neurotransmitter-containing synaptic vesicles. This class of Ca2+ channel is, therefore, pivotal during normal neurotransmission in higher organisms. In response to channel opening and Ca2+ influx, specific Ca2+-binding proteins associate with cytoplasmic regulatory domains of the P/Q channel to modulate subsequent channel opening. Channel modulation in this way influences synaptic plasticity with consequences for higher-level processes such as learning and memory acquisition. The ubiquitous Ca2+-sensing protein calmodulin (CaM) regulates the activity of all types of mammalian voltage-gated Ca2+ channels, including the P/Q class, by direct binding to specific regulatory motifs. More recently, experimental evidence has highlighted a role for additional Ca2+-binding proteins, particularly of the CaBP and NCS families in the regulation of P/Q channels. NCS-1 is a protein found from yeast to humans and that regulates a diverse number of cellular functions. Physiological and genetic evidence indicates that NCS-1 regulates P/Q channel activity, including calcium-dependent facilitation, although a direct physical association between the proteins has yet to be demonstrated. In this study, we aimed to determine if there is a direct interaction between NCS-1 and the C-terminal cytoplasmic tail of the Cav2.1 α-subunit. Using distinct but complementary approaches, including in vitro binding of bacterially expressed recombinant proteins, fluorescence spectrophotometry, isothermal titration calorimetry, nuclear magnetic resonance, and expression of fluorescently tagged proteins in mammalian cells, we show direct binding and demonstrate that CaM can compete for it. We speculate about how NCS-1/Cav2.1 association might add to the complexity of calcium channel regulation mediated by other known calcium-sensing proteins and how this might help to fine-tune neurotransmission in the mammalian central nervous system.
Ca v 2.1 P/Q-type channels are responsible for the entry of Ca 2+ into synaptic terminals in many brain regions. 1 P/Q-type function is therefore pivotal in the regulation of neurotransmitter release and communication throughout the central nervous system (CNS). Indeed, mutations in the gene encoding the human P/Q α 1 subunit are responsible for a number of clinically relevant neurological disorders. 2 An interesting feature of many Ca v proteins (Ca v 1.x and Ca v 2.x), including the P/Q channel, is that channel activity is itself regulated by Ca 2+ . 3 Ca 2+ -dependent channel modulation manifests as two distinct phenomena termed Ca 2+ -dependent inactivation (CDI) 4 and Ca 2+ -dependent facilitation (CDF). 5 CDI represents a mechanism by which channels become refractory to further opening in the presence of a sustained stimuli, whereas CDF manifests as the opposite process whereby Ca 2+ augments channel currents. 6 Specific Ca 2+binding proteins can respond to the entry of Ca 2+ through Ca v channels by interacting directly with regulatory motifs present in the intracellular domains of the α-subunit. 7−10 These interactions, by currently poorly understood mechanisms, alter channel behavior to elicit either CDI or CDF. CDI and CDF both have a direct impact on synaptic neurotransmitter release, the strength of synaptic signaling, and ultimately synaptic plasticity. Understanding the basic modes of Ca v regulation at the molecular level therefore generates insights into more abstract, higher-level, processes, including learning, memory acquisition, and reasoning.
Of the characterized Ca 2+ -binding proteins that control Ca v activity, the ubiquitous small EF-hand-containing protein calmodulin (CaM) has been most intensively studied. CaM has been shown to interact with a consensus isoleucineglutamine "IQ" motif present in the carboxy terminus of the channel, although the precise details of the interaction differ for Ca v 1.x and Ca v 2.x subtypes. 11,12 A second Ca 2+ -dependent CaM-binding motif immediately C-terminal to the IQ domain has also been characterized in Ca v 2.1 P/Q-type channels. 13 This CaM-binding domain (CBD) has been found to influence P/Q channel activity in some studies 13 but not others, 7 and its true significance remains somewhat controversial.
A number of CaM-related Ca 2+ -binding proteins are enriched in the mammalian CNS and have been found to modulate the properties of Ca v channels. In all instances thus far characterized, channel regulation is distinct from that observed with CaM, suggesting nonredundant functions that have been thought to permit complex modes of channel regulation and neuronal signaling. 12 At present, two CaM-related Ca 2+ -binding proteins, Ca 2+ -binding protein-1 (CaBP1) and visinin-like protein-2 (VILIP-2), have been shown to directly interact with P/Q-type α-subunits to exert unique regulatory outcomes. 10,14 Large numbers of additional CaM-related small EFhand-containing proteins are also expressed in the mammalian CNS. Neuronal Ca 2+ sensor-1 (NCS-1) is a CaM-related protein that is evolutionarily conserved from yeast to humans 15 and has been implicated in specific human neuronal disorders. 16−19 Evidence of a regulatory role of NCS-1 on P/ Q-type channels has been reported on the basis of physiological experiments in mammalian cells, including an effect on CDF, 20−22 and a genetic study in Drosophila, 23 although no direct interaction between the two proteins has thus far been demonstrated.
In this paper, we aimed to use exclusive yet complementary analytical methods to rigorously determine if there is a direct interaction between NCS-1 and the α-subunit of the P/Q-type channel and to provide the first evidence of an interaction between a defined region of the C-terminal domain of the P/Q α-subunit and NCS-1. We show that a segment of the P/Q αsubunit encoding the IQ motif and CBD interacts with NCS-1 in a Ca 2+ -dependent manner. Using mutagenesis and NMR experiments, we further refine the NCS-1-binding site to the IQ motif and show that the interaction is competitive with CaM. We provide further evidence using an ex vivo HeLa cell model that NCS-1 and the C-terminal tail of Ca v 2.1 interact. These novel observations expand the repertoire of P/Q-type interacting proteins and add to the potential variety of modes by which such channels can be controlled.
Cell Culture and Plasmid Transfections. HeLa cells were cultured in DMEM supplemented with 10% (v/v) fetal bovine serum, 1% (v/v) penicillin/streptomycin, and 1% (v/v) nonessential amino acids. All cells were maintained in a humidified 95% air/5% CO 2 atmosphere at 37°C. Cells were plated onto sterile 13 mm round coverslips at a density of 0.25 × 10 6 cells/well. After 24 h, cells were transiently transfected with the indicated expression vectors using GeneJuice transfection reagent (Novagen) according to the manufacturer's protocol. For single and double transfections, 1 μg of each plasmid was used.
Cell Fixation and Confocal Imaging Analysis. Twentyfour hours post-transfection, cells on coverslips were washed with phosphate-buffered saline [137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 PO 4 , and 2 mM NaH 2 PO 4 (pH 7.4)] and then fixed with 4% (v/v) formaldehyde in PBS for 6 min at room temperature. Coverslips were subsequently air-dried and mounted onto microscope slides using Prolong antifade glycerol (Life Technologies). Fixed cells were imaged using a Leica TCS-SP2 confocal system (Leica Microsystems, Heidelberg, Germany) with a pinhole set to 1 Airy unit and a 63× oil immersion objective with a numerical aperture of 1.3. Images were exported as TIFF files and compiled, processed, and analyzed with ImageJ and CorelDraw X6 applications.
In Vitro Binding Assays. For the In vitro protein binding assays, 5 μM GST fusion protein (GST-P/Q, GST-P/Q IM→EE , or GST-CBD) was immobilized onto 30 μL of glutathioneagarose resin (Thermo Scientific) that had been prewashed in binding buffer (BB) [150 mM KCl, 20 mM HEPES (pH 7.4), 10% (v/v) glycerol, 5 mM NTA, 5 mM EGTA, 1 mM DTT, and 0.1% (v/v) NP-40] or BB supplemented with 1 μM free Ca 2+ (BB+Ca 2+ ) by incubation for 30 min with constant agitation at 4°C. GST-free NCS-1 and CaM were added to samples at concentrations of 5 μM (total final binding assay volumes of 100 μL) and incubations continued for 1 h with constant agitation at 4°C. For competition binding assays, 5 μM GST-P/Q was prebound to glutathione-agarose resin in BB +Ca 2+ by incubation for 30 min at 4°C with constant agitation; 5 μM NCS-1 was then added to samples in the presence of increasing concentrations of CaM ranging from 0 to 10 μM. Binding reactions were continued for 1 h at 4°C while the mixtures were constantly agitated. For the Ca 2+ dose dependency of the NCS-1/GST-P/Q interaction, 1 μM GST-P/Q was incubated with 1 μM NCS-1 in the presence of increasing concentrations of free Ca 2+ . For all binding assays, glutathioneagarose pellets were collected by centrifugation (3000 rpm for 1 min at 4°C) and washed three times with 1 mL of BB or BB +Ca 2+ and bound proteins were extracted by boiling of final bead pellets for 5 min in 50 μL of SDS dissociation buffer [125 mM HEPES (pH 6.8), 10% (w/v) sucrose, 10% (v/v) glycerol, 4% (w/v) SDS, 1% (v/v) β-mercaptoethanol, and 2 mM EDTA]. Proteins were resolved on 4 to 12% Tris-glycine gradient gels (Novex, Life Technologies) and transferred to nitrocellulose membranes for Western blotting by transverse electrophoresis.
Western Blots. Nitrocellulose filters were blocked by incubation in a blocking solution [3% (w/v) skim milk powder in PBS] for 1 h at room temperature. Filters were subsequently incubated with a primary antibody [rabbit anti-NCS-1 (1:1000 26 ) or rabbit anti-calmodulin (1:500) (AbCam)] diluted in a blocking solution overnight at 4°C with constant agitation. Filters were washed three times in PBS supplemented with 0.05% (v/v) Tween 20 (PBST) and twice with PBS before being incubated with a HRP-conjugated species specific secondary antibody (1:400, anti-rabbit HRP, Sigma) in a blocking solution for 1 h at room temperature. Filters were washed three times with PBST and twice with PBS prior to application of ECL reagents and visualization of immunoreactivity using a Chemidoc automated gel/blot documentation system (Bio-Rad). Densitometry analysis of developed Western blots was performed using Quantity-1 (Bio-Rad). For the Ca 2+ dose dependency of the NCS-1/GST-P/Q interaction, densitometry data were analyzed by applying nonlinear curve fitting with OriginPro8 (OriginLab).
Synthetic Peptide. The peptide used corresponds to residues 1903−1929 of the human P/Q receptor. The synthetic P/Q peptide, TVGKIYAAMMIMEYYRQSKAKKLQAMR (hereafter termed the PQIQ peptide), was purchased from GenicBio. The peptide was delivered >95% pure.
Protein Expression and Purification. NCS-1 was expressed in Escherichia coli strain BL21(DE3) (Novagen) and purified as previously described. 27 Expression was induced overnight at 18°C; cells were harvested, resuspended into lysis buffer [50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 5 mM CaCl 2 , and Complete EDTA Free Protease Inhibitor (Roche Applied Science)], lysed, centrifuged, filtered, and loaded onto the Hiprep 16/10 Phenyl FF High Sub (GE Healthcare) column that was pre-equilibrated with buffer A [50 mM Tris-HCl (pH 7.5), 200 mM NaCl, and 5 mM CaCl 2 ]. The column was extensively washed with buffer A and NCS-1 eluted using Milli-Q water. The eluted protein was buffer exchanged into 50 mM Tris (pH 7.4) and 500 mM NaCl, and the N-terminal His tag was removed by incubating the protein overnight at 4°C using TEV protease (1:20 TEV protease:NCS-1 molar ratio). His-tagged uncleaved NCS-1 and TEV protease were separated for cleaved NCS-1 using a HisTrap FF 5 mL affinity column (GE Healthcare). The cleaved NCS-1 was further purified using a Superdex 75 Hiload 26/60 (Amersham Biosciences) sizeexclusion column [50 mM Tris-HCl (pH 7.5) and 150 mM NaCl]. The purity of the sample was assessed by sodium dodecyl sulfate−polyacrylamide gel electrophoresis (SDS− PAGE) and deemed to be >95% pure. The eluted peak was concentrated and stored at −80°C.
Calmodulin was cloned into a pET-15b (Novagen) vector (gift from A. Kitmitto, University of Manchester, Manchester, U.K.) and expressed in E. coli strain BL21(DE3) (Novagen) at 37°C. Harvested cells were resuspended in lysis buffer [50 mM Tris (pH 7.5) and 2 mM DTT], disrupted using a French press (Sim Aminco) at 1000 psi, and centrifuged, and the supernatant was recovered, CaCl 2 added to a final concentration of 2 mM, and the protein solution loaded onto a 20 mL HiPrep phenyl sepharose hydrophobic column (GE Healthcare), pre-equilibrated with CaM buffer A [50 mM Tris (pH 7.5), 200 mM NaCl, 2 mM CaCl 2 , and 0.5 mM DTT]. The column was washed with CaM buffer A, followed by buffer B [50 mM Tris (pH 7.5), 0.5 mM CaCl 2 , and 0.5 mM DTT], and pure calmodulin eluted with CaM buffer C [50 mM Tris (pH 7.5), 1 mM EGTA, and 0.5 mM DTT]. The protein was further purified using a 26/60 Superdex 75 (GE Healthcare) sizeexclusion column pre-equilibrated with CaM gel filtration (GF) buffer [50 mM Tris (pH 7.5), 200 mM NaCl, 10 mM CaCl 2 , and 0.5 mM DTT]. The sample was loaded before isocratic elution. The purity of the sample was assessed by SDS−PAGE and deemed to be >95% pure. CaM was dialyzed against water and lyophilized before being stored at −20°C. 15 Healthcare) size-exclusion column. The eluted protein was dialyzed against Milli-Q water and stored as lyophilized samples. Confirmation of the identity of purified PQIQ 1909−2035 was achieved via matrix-assisted laser desorption ionization time-of-flight mass spectrometry performed on unlabeled PQIQ 1909−2035 prepared using a procedure identical to that described above.
Isothermal Titration Calorimetry. ITC experiments were performed using a MicroCal ITC 200 instrument. Because of limitations in PQIQ peptide solubility, NCS-1 was titrated into the peptide. Ca 2+ /NCS-1 at 1 mM was prepared by buffer exchange using a PD10 column equilibrated in 50 mM Tris (pH 7.5), 50 mM NaCl, and 5 mM CaCl 2 , and the PQIQ peptide solution at 100 μM was prepared using the same buffer. If necessary, minor adjustments were made to the pH of the peptide solution. Experiments were conducted using 200 μL of 100 μM PQIQ peptide in the cell and 60 μL of 1 mM NCS-1 in the syringe at 25°C. One injection of 0.2 μL, followed by 20 injections of 2 μL, was made with a 180 s spacing to allow the baseline to return after each injection. All experiments were performed in triplicate. The data were analyzed with a one-site (three-parameter) curve fitting conducted using the MicroCalsupported ITC module within Origin version 7.
Spectrofluorimetry. To monitor the intrinsic tryptophan fluorescence of NCS proteins, 28,29 purified recombinant NCS-1 at a concentration of 1 μM in a buffer [50 mM Tris (pH 7.5), 50 mM NaCl, and 5 mM CaCl 2 ] was excited at room temperature with 280 nm wavelength light and the emission measured between 290 and 410 nm with a slit width of 20 nm using a CARY Eclipse spectrofluorimeter. The PQIQ peptide (stocks ranging from 50 μM to 1 mM) was then added to give an incremental increase in peptide concentration, and emission spectra were acquired after each addition. Experiments were performed in triplicate. The data for the measured tryptophan fluorescence change at each peptide concentration were fit to a logistic equation using nonlinear curve fitting in OriginPro version 9.0. NMR Spectroscopy. NCS-1 was prepared in Tris buffer (pH 6.8) in the presence of 5 mM CaCl 2 . NMR spectra were recorded at 298 K on Bruker Avance II 800 and 600 MHz spectrometers equipped with cryoprobes. Data were processed using Bruker Software TopSpin and analyzed using CCPN. 30 Sequence specific assignment of the PQIQ peptide was achieved using homonuclear two-dimensional TOCSY, COSY, and NOESY data. 13
■ RESULTS
The aim of this study was to test, using a range of complementary methods, the possibility that the small calcium-sensing protein NCS-1 directly interacts with the Ca v 2.1 P/Q-type Ca 2+ channel as has been indirectly suggested in prior functional studies. 20−23 With the knowledge that other known small EF-hand Ca 2+ -binding proteins interact with the α-subunit of the P/Q channel predominantly through motifs located in the C-terminal tail, we reasoned that the same domains represent the most likely sites for an NCS-1 interaction. In initial experiments, we generated a recombinant, GST-tagged, form of the P/Q α 1 subunit C-terminal tail domain corresponding to residues 1898−2035 (Figure 1, GST-PQ). This fragment encompasses both the IQ motif (IM in the P/Q-type channel) and the CaM-binding domain (CBD). 10 Two further constructs derived from this precursor were the CBD alone [residues 1950−2035 (Figure 1, GST-CBD)] and an IQ motif mutant in which the P/Q channel IM residues at positions 1913 and 1914 were mutated to glutamic acid to (Figure 1, GST IM→EE ). This mutation has been previously shown to abolish CaMdependent CDF and CDI of the human P/Q-type channel. 7 We used these constructs to investigate whether recombinant NCS-1 could associate with P/Q-type C-terminal regulatory domains in vitro. NCS-1 exhibited Ca 2+ -dependent binding to GST-PQ but was not observed to bind to either GST-CBD or GST IM→EE under any tested conditions (Figure 1A), suggesting a specific interaction with residues in the IQ region of the channel. Averaged densitometry data from three independent binding experiments are listed in Table 1.
A Ca 2+ titration of binding of NCS-1 to GST-PQ confirmed the Ca 2+ dependency of the interaction and allowed us to derive a K d of 0.7 μM based on densitometry quantification of Western blot data ( Figure 1B). For comparison, we performed the same binding analysis with CaM ( Figure 1C). Similar to our results with NCS-1, CaM displayed Ca 2+ -dependent binding to GST-PQ but no detectable binding under any conditions to either GST-CBD or GST IM→EE .
Our initial data suggested that both NCS-1 and CaM could interact with the same IQ motif region of the rat P/Q-type channel, and therefore, we extended our binding studies to evaluate the potential existence of a common, overlapping, binding site for both proteins. Competitive association of both proteins with GST-PQ was assessed by binding of NCS-1 to GST-PQ in the presence of increasing concentrations of recombinant CaM (Figure 2). Western blot results from these experiments highlighted a clear displacement of NCS-1 from GST-PQ that was first apparent at a CaM concentration between 0.3 and 1 μM (Figure 2A). Densitometry analysis of the Western blot data confirmed a competitive association between NCS-1 and CaM for GST-PQ ( Figure 2B). Our initial experiments indicated that the IQ-like motif of the P/Q-type channel is important for interaction with both NCS-1 and CaM, and therefore, in subsequent NMR and ITC analyses, we focused on only this region of the channel.
Further testing of the ability of NCS-1 to bind the P/Q-type channel and to delineate the NCS-1-and CaM-binding regions of the P/Q-type channel was performed using NMR experiments. A sample of 15 N-labeled P/Q spanning residues 1909− 2035 (encompassing much of the IQ motif and the CBD) was prepared. Addition of CaM shows chemical shift perturbations of glycine and tryptophan residues, whereas NCS-1 has no effect on either group of residues (Figure 3). Glycine and tryptophan residues are located in the CBD region but not in or around the IQ segment. Hence, it is clear that NCS-1 does not bind to the CBD of the P/Q channel, although there is an indication that there is weak binding of CaM to this site.
Knowing from the experiments described above that the major site of binding of NCS-1 on the P/Q C-terminal domain was likely to be the IQ motif, we used a synthetic PQIQ peptide in subsequent experiments. The intrinsic tryptophan fluorescence of NCS-1 was used to measure interactions with the PQIQ peptide (Figure 4), using an approach similar to that described for the interactions with the D2 dopamine receptor peptide. 27 NCS-1 has two tryptophan residues; addition of the PQIQ peptide resulted in a decrease in fluorescence and allowed titration of the peptide over a range of concentrations ( Figure 2A). The data for the change in fluorescence versus concentration were analyzed using a logistic fit. The dose response upon titration with the PQIQ indicated half-maximal binding at 1.036 ± 0.07 μM with a Hill coefficient of 1.7. Hill coefficients greater than 1 would be consistent with a 2:1 stoichiometry of peptide binding to NCS-1 because NCS-1 itself was monomeric. 27 ITC shows that the interaction is entropically driven. Because of the tendency of the protein complex to aggregate at the high peptide (100 μM) and protein concentrations (1 mM) required to observe binding by this method, it was not possible to obtain reliable equilibrium binding constants using ITC. The affinity obtained for the interaction of NCS-1 with the PQIQ peptide is almost 1 order of magnitude lower than the reported affinities involving the calmodulin lobes. From ITC experiments, K d values of 51 ± 20 and 4.32 ± 0.39 nM were obtained for binding of the Ca 2+ /N and Ca 2+ /C lobes, respectively, to the IQ domain of the PQ channel. 31 The two lobes in NCS-1 are intimately linked, and attempts to express the two NCS-1 lobes as independently folded domains soluble at micomolar concentrations have met with little success so far. Therefore, experiments that aimed to disentangle the relative affinities of the two NCS-1 domains for the PQIQ peptide have not been possible.
To determine which region of the PQIQ peptide forms the NCS-1-binding site, we again used NMR. In these experiments, [ 13 C, 15 N]NCS-1 was titrated into a sample of the unlabeled PQIQ peptide and 13 C-and 15 N-filtered TOCSY and NOESY data were acquired. Analysis of the side chain resonances, in particular the methyl and tyrosine resonances, reveals selective line broadening for 14 residues, YAAMMIMEYYRQSK ( Figure 5A). Hence, a subset of the 27 amino acids in the polypeptide sequence forms the main interaction site for NCS-1 binding, with the residues beyond this core peptide region further enhancing the interaction. The length of this motif correlates well with the recently reported structure of NCS-1 in complex with the C-terminal peptide from the human D2 dopamine receptor (Protein Data Bank entry 2YOU). The 14-amino acid motif is short enough for it to bind NCS-1 with a stoichiometry of 2:1 (peptide:NCS-1) and supports the fluorescence binding data.
NMR 1 H− 15 N HSQC of NCS-1 in the presence of a 2-fold excess of PQIQ shows resonances that are universally broadened although still discernible to confirm that the protein Figure 5B). Addition of equimolar amounts of unlabeled CaM/Ca 2+ to the preformed 15 N-labeled NCS-1/ PQIQ complex led to a complete reversion of the 1 H− 15 N HSQC C spectrum to that of free NCS-1. This is further confirmation that NCS-1 and CaM bind to similar sites on the PQ channel, with, as expected, CaM binding the peptide sufficiently strongly to displace it from NCS-1.
The experiments described above show that a direct interaction of NCS-1 with the P/Q C-terminus could be detected using three distinct in vitro biochemical approaches. To provide evidence of the interaction of NCS-1 with the P/Q channel regulatory domain in a more physiological setting, we studied the localization of these proteins ex vivo in cultured HeLa cells. For these experiments, we returned to the use of a longer recombinant construct encompassing both the IQ and CBD domains of the P/Q C-terminus. We first examined the cellular localization of both NCS-1-mCherry 18 and a Cterminally YFP-tagged variant of P/Q 1898−2035 to which an N-terminal myristoylation consensus sequence had been attached ( myr PQ-YFP). In previous studies, this approach has been used to successfully direct isolated soluble domains of plasma membrane proteins to the plasma membrane for functional analyses. 25 NCS-1-mCherry exhibited a predominantly perinuclear localization consistent with its association with the TGN in addition to some plasma membrane localization ( Figure 6A). 32 myr PQ-YFP was observed to target the plasma membrane and/or large cytosolic puncta ( Figure 6A). Plasma membrane association of myr PQ-YFP suggested that myristoylation of the peptide was accurately targeting a fraction of PQ to the expected location. Interestingly, in cells exhibiting myr PQ-YFP puncta, co-expressed NCS-1-mCherry failed to localize to the TGN or plasma membrane and instead colocalized with myr PQ-YFP on the enlarged punctate structures ( Figure 6B). We quantified these phenomena and were able to demonstrate that in 39.5 ± 9.2% of cells cotransfected with NCS-1-mCherry and myr PQ-YFP puncta were observed and that in 100% of this subpopulation NCS-1 protein was associated with large myr PQ-YFP positive punctate structures ( Figure 6C, NCS-1-mCh + myr PQ-YFP). This was similar to the number of cells exhibiting puncta in myr PQ-YFP/mCherry controls [43.6 ± 3% ( Figure 6C, myr PQ-YFP + mCherry)]. In contrast, control cells expressing NCS-1-mCherry along with EYFP protein exhibited no observable formation of puncta ( Figure 6C, NCS-1-mCh + EYFP). It should be noted that
■ DISCUSSION
In this paper, we set out to examine whether NCS-1 is able to interact directly with the C-terminus of the α 1 subunit of the P/ Q-type channel. To avoid a potentially misleading indication of a positive interaction, we tested this using three separate in vitro biochemical/biophysical assays using either a long construct from the P/Q C-terminus or a synthetic IQ domain peptide as well as a test through cellular colocalization. The use of multiple techniques gives a more rigorous approach to addressing the question of whether a significant direct interaction does exist. In all cases, we were able to detect a direct interaction between NCS-1 and P/Q-type Ca 2+ channels. P/Q-type channel regulation by small Ca 2+ -binding proteins is essential in the mammalian CNS for the processes of CDI and CDF that drive short-term presynaptic plasticity. 6 Alterations in synaptic plasticity of this form in turn influence information processing and neural network behavior that underpin the normal functioning of the CNS. 33 A large body of experimental evidence that describes at the molecular level how CaM is able to regulate all Ca v -type channels, including those of the P/Q subtype, has accumulated. CaM is constitutively associated with the P/Q α subunit and is considered an auxiliary subunit of the channel. Apo-CaM is believed to bind to the P/Q α subunit at a site outside of, but in the proximity of, the Ca 2+ -dependent IQ-binding site, as determined using a Forster resonance energy transfer approach in cultured mammalian cells. 34 It has been suggested that this arrangement locates CaM within striking distance of the regulatory IQ motif to ensure fast responses to Ca 2+ influx on channel opening. 34 Ca 2+ -dependent binding of CaM to the IQ motif is then able to modulate both CDI and CDF depending on the precise nature of the Ca 2+ signal. Mechanistically, this system has been resolved in elegant experiments showing that Ca 2+ entry through individual P/Qα channels selectively activates the CaM C-lobe to mediate CDF. Larger Ca 2+ signals emanating from multiple local channel opening events conversely activate the CaM N-lobe to drive CDI. 3,7,35 Our data confirming a Ca 2+ -dependent interaction of CaM with the P/Q containing the IQ motif are consistent with previous biochemical 7 and structural analyses. 31 A second CaM-binding site, the CaM-binding domain (CBD), has been reported in related studies of P/Q−channel interactions. 5,13 In this study, although we have been unable to replicate binding of CaM to this region of the rat α 1A CBD using pull-down experiments, the NMR data suggest that CaM, but not NCS-1, binds to the proposed CBD, albeit rather weakly. The weak nature of this interaction might explain why there are discrepancies in the literature regarding the significance of the CBD. Pull-down-type experiments select for high-affinity interactions, whereas NMR approaches are able to detect lower-affinity transient binding events consistent with these results. Our data help reconcile apparently conflicting experimental data as some studies do observe a CaM−CBD interaction 5,13 whereas others do not. 7 NCS-1, a multifunctional Ca 2+ sensor conserved from yeast to humans, has documented roles in the regulation of membrane trafficking, 36,37 ion channel activity, 38,39 dopamine receptor signaling, 27,40 long-term depression, 41 autism, 18,42 and memory acquisition. 43,44 NCS-1 exhibits enriched expression in mammalian brain tissue, 45 and many of the aforementioned functions are related to neuronal activity. NCS-1 interacts with a number of targets in common with CaM, 24 and a previous functional study identified links between NCS-1 and CDF of P/Q-type channel activity in rat primary neurons. 20 Other studies have additionally indicated a cellular connection between NCS-1 and P/Q-type channel activity in bovine adrenal chromaffin cells 21,22 and Drosophila. 23 With this information in mind, we have tested the possibility that NCS-1 can directly bind to P/Q regulatory elements in common with CaM and other EF-hand-containing Ca 2+ sensors. 10,14 Our data show that NCS-1 can associate with the IQ motif of rat P/ Qα 1A in a Ca 2+ -dependent manner and that this binding site must overlap to some extent with that of CaM because CaM can compete for binding to the same region leading to NCS-1 displacement. This is the first example of a CaM-related Ca 2+sensing protein behaving in a manner almost identical to that of CaM with respect to IQ binding. Both CaBP1 10 and VILIP-2 14 interact with the CBD or require this domain for functional activity. CaM binds predominantly to the IQ domain with a possible weak interaction with the CBD, and NCS-1 binds to only the IQ domain in the presence of Ca 2+ . The weak CaM− CBD interaction that we have observed in this study might also help to explain why CaBP1 and VILIP-2 are able to effectively compete with CaM at this site. 10,14 The precise details of the NCS-1 interaction are still to be elucidated. In rat primary neurons, NCS-1 appears to play a role during CDF, 20 and in Drosophila, loss of the NCS-1 orthologue Frequenin leads to a reduced level of entry of Ca 2+ into neurons and defective synaptic transmission. 23 These data suggest that at least one function of NCS-1 is in positively regulating P/Q channel opening. Our data now provide an attractive biochemical explanation for these cellular studies and act as a platform for further cellular and structural investigations studying the interplay between CaM and NCS-1 during regulation of P/Q channel activity. CaM is expressed in all neurons, and one possibility is that specific neuronal populations that also express NCS-1 are able to utilize this Ca 2+ sensor as an additional modulator of P/Q activity. NCS-1 has a Ca 2+ affinity higher than that of CaM 15,46 and therefore could potentially interact with the IQ motif over a different range of local Ca 2+ concentrations to modulate activity in a manner independent of CaM. As Ca 2+ concentrations increase, CaM could then displace NCS-1 to exert its documented roles in CDF and CDI. The N-and C-lobes of CaM, as described, exert distinct modes of P/Q channel regulation. NCS-1, unlike CaM, does not have distinct lobes but rather a compact globular structure. 47 The orientations of the N-and C-lobes with respect to one another differ in the two proteins, giving rise to different conformations of the ligand-binding site. Residues from both lobes in NCS-1 together form a large, contiguous solvent-exposed hydrophobic ligand-binding crevice, resembling the palm of a partially open hand. In the Ca 2+ / CaM−PQIQ complex, the hydrophobic binding pocket is more enclosed, resembling a closed hand that envelops the PQIQ peptide. It is, therefore, not possible to model a structure of NCS-1 in complex with the PQIQ peptide based on the Ca 2+ / CaM complex structure. It is likely that binding the PQIQ peptide would require a significant conformational change to the overall structure of NCS-1 to accommodate the peptide.
Using a combination of in vitro biochemical, NMR and ITC biophysical, and cellular approaches, we provide the first evidence of a direct interaction of NCS-1 with P/Q-type voltage-gated Ca 2+ channels. Voltage-gated Ca 2+ channel modulation has direct implications for synaptic plasticity and complex neural processing in higher organisms. Our work expands the number of small Ca 2+ -sensing proteins that can interact with the essential P/Q-type neuronal channel and provides further insights into the complexity of the regulation of voltage-gated Ca 2+ channels in the mammalian CNS. Extensive further work will be required to determine the full structural basis for the interaction of NCS-1 with the P/Q IQ domain. | 2016-05-04T20:20:58.661Z | 2014-09-04T00:00:00.000 | {
"year": 2014,
"sha1": "ba1f6629868aa6669406bb4bbd8242b8b5eed4ad",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1021/bi500568v",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "ba1f6629868aa6669406bb4bbd8242b8b5eed4ad",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
5090079 | pes2o/s2orc | v3-fos-license | Ankle-Foot Continuous Passive Motion Device for Mobilization of Acute Stroke Patients
Purpose: To develop a continuous passive motion (CPM) device for the passive motion of the paretic ankle-foot and investigate the effect of continuous passive motion of bedridden, hemiparetic acute stroke patients. Methods: 49 patients with stroke were investigated. Results in stroke patients (device group) were compared with those of 15 control subjects (manual group) also with stroke but not treated by device. The period of the treatment was 7 days; the duration was 30 minutes per day by CPM device in the device group. The efficacy of the device was evaluated by scales used in the clinical routine (6th item of National Institutes of Health Stroke Scale (NIHSS), Modified Ashworth Scale (MAS), modified Rankin Scale (mRS)). Ankle’s passive range of motion (PROM) and flexible equinovalgus deformitiy were measured every day with a goniometer. Results: 6th item of NIHSS score improved by −0.76 (SD = 0.56) points in the device group (p < 0.001) compared to the baseline values; the mean change in the manual group was −0.33 (SD = 0.62) points (p = 0.055). The mean of MAS decreased significantly by −0.53 (SD = 1.12) point in the device group (p < 0.001). The ankle’s mean plantar flexion PROM increased by 3.41 (SD = 5.19) degrees in the device group (p < 0.001). Significant improvement of the mean dorsiflexion in the PROM of the ankle was also detected (p = 0.019). The equinovalgus improved significantly by −5.12 (SD = 8.02) degrees (p < 0.001) in the device group. The scores of the mRS also improved significantly in the device group (p < 0.001). Conclusion: In the early phase of rehabilitation, ankle-foot continuous passive motion device treatment combined with manual therapy improved the ankle’s PROM better than manual therapy alone; in addition, device treatment decreased the foot’s equinovalgus, improved the 6th item NIHSS score, and decreased the severity of spasticity.
Introduction
Stroke is an important health care issue resulting in chronic disability and its early rehabilitation is still a challenge.
The most common motor impairment is the hemiparesis of the upper and lower limb.The impaired lower limb is a major challenge in the rehabilitation of post stroke patients.
Caused by circulatory disturbance, the paresis endangers patients in terms of survival and improvement outlooks.Spasticity, deep venous thrombosis, and decubitus ulcers can all be contributed to by the decreased ability of lower leg movements [1].
In stroke patients, spasticity disrupts the functional use of weakened muscles mainly in the distal joints of the paretic lower limb.Structural changes to muscle fibers and connective tissue may contribute to alterations in intrinsic mechanical properties [2].
Several studies reported spasticity rates of 4% to 27% in the early time course (1 to 4 weeks); moreover, Lundstörm et al. and Sommerfeld et al. detected spasticity in the first week after first ever-stroke [3]- [5].
Following stroke, paresis in the dorsiflexor muscles that raise the foot and move the ankle during walking results in the plumping down of the foot after heel strike, and lugging toes during swing [2].
Moreover, there is a strong relationship between ankle plantar flexor weakness and knee hyperextension during the midstance phase [6].
All these complications may cause unsafe gait, accidental falls, and fractures such as hip fractures; also, these dysfunctions may provoke additional complications such as contractures and/or deformities [2].
Common deformities following stroke are equinovarus, equinus, and equinovalgus deformities.Equinovarus is caused by spasticity in the plantar flexors and invertors, and paresis of the dorsiflexors and evertors of the foot.Equinus is caused by the spasticity of the three-headed calf muscle (triceps surae) [7].
Equinovalgus is also a frequent finding; the base and the heel are rotated away from the midline.It can be a flexible or a fixed deformity.
Equinovalgus is caused by spasticity in the gastrocnemius, soleus, and peroneus brevis muscles [8], and weakness of the tibialis anterior and tibialis posterior muscles.These dysfunctions result in external rotation of the foot during the gait cycle, and create instability at push-off.Contractures and loss of passive ankle range of motion are common problems after stroke [9]- [11].
Gracies investigated the time course of muscle contracture.In mouse soleus muscles, after only 24 hours of unloading, there was a 60% shortening of muscle fiber length, and evidence of sarcomere disorganization [12].
Muscle shortening caused by reduced joint mobility increase the resistance to passive joint movement, and can provoke deformities and pain [13].
Active and passive movement (by a physiotherapist) of the paretic leg is essential for rehabilitation and for prevention or improvement of spasticity and contractures.
Due to a world-wide shortage of physiotherapists (especially on weekends and holidays), the demand for simple devices that can replace or at least assist physiotherapists is increasing.
To meet this challenge was our main motivation for developing the device.
The second aim of our study was to measure whether our device therapy with manual therapy are as effective as manual therapy (or even better) in acute stroke patients.
The ankle-foot continuous passive motion device passively moves the paretic foot (plantar and dorsiflexion) and is equipped with a special alarm system to prevent extreme extension or flexion.
The frequency and amplitude of the movement can be adapted to the patient's personal needs; mobilization can be automatically stopped if the patient requires an increase or decrease of mobilizing strength or frequency.
Participants
Forty-nine acute ischaemic and haemorrhagic stroke patients (mean age: 62.2 years (SD = 12.38); female/male distribution: 18/31) were treated by manual therapy and by ankle-foot CPM device every day for a week (device group).
Our control group (manual group) consisted of 15 acute stroke patients (mean age: 71.3 years (SD = 10.39);female/male distribution: 7/8) who did not receive treatment with the ankle-foot CPM device, only manual therapy by a physiotherapist.
Demographic and clinical data are summarized in Table 1.
Manual and ankle-foot CPM device therapy were started between 24 and 48 hours after stroke.Inclusion criteria: acute ischemic or hemorrhagic stroke confirmed by clinical investigations and computed tomography in the device and manual groups.
All patients were bedridden and suffering from hemiparesis with modified Rankin Scale (mRS) 3 to 5 [14].None of our patients were able to get out of bed without assistance.Definition of mild and moderate paresis (6 th item of National Institutes of Health Stroke Scale (NIHSS)): 1) Point: drift; leg falls by the end of the 5-second period but does not hit the bed 2) Points: some effort against gravity; leg falls to bed by 5 seconds but has some effort against gravity Definition of severe paresis (6 th item of NIHSS): 3) Points: no effort against gravity; leg falls to bed immediately 4) Points: no movement [15].Patients were excluded from investigation for the following reasons: primary brain tumor, brain metastasis, Parkinson's disease, multiple sclerosis, rheumatoid arthritis, ankylosis of the ankle or foot joints, trauma or surgery of the foot or ankle; patients with chronic stroke symptoms or psychiatric symptoms were also excluded.
No patients received oral or parenteral antispastic therapy (e.g.botulinum toxin).
The study was approved by the Regional and Institutional Ethics Committee (DE OEC RKEB/IKEB 3772-2012) and each patient gave written informed consent.
The Ankle-Foot Continuous Passive Motion Device
The ankle-foot continuous passive motion device (prototype device) mobilizes the paretic foot repeatedly across the ankle's entire range of motion between plantar and dorsiflexion (up to 60 degrees).
Angle, power and speed parameters were individually adapted to the patients in the device group in order to reach maximum motion intensity below the pain threshold.
If the device detects active resistance, movement is stopped automatically.During the investigation, neither the extension nor the flexor motion was observed to reach the threshold of pain.
The ankle-foot continuous passive motion device is driven by a 24-volt direct current (DC) motor.The operating frequency range extremes are 7 times per minute from plantar to dorsiflexion (altogether 60 degrees) at maximum speed and 4 times per minute from plantar to dorsiflexion at minimum speed.
Resistance parameters were typically set between 1 to 7 Nm maximum resistance torque in dorsiflexion, and 1 to 10 Nm maximum resistance torque in plantar flexion.
Randomization and Treatment Procedures
Our study consisted of two periods.In the first period, patients received standard physiotherapy (manual therapy in the manual group).Subsequently, treatment was supplemented by the ankle-foot continuous passive motion device in the second period (device group).
Patients were assigned to groups randomly, consecutively according to selection criteria.In both groups (before and after any therapy) the Modified Ashworth Scale, modified Rankin Scale, 6 th item of NIHSS and goniometer measurements were performed.We used the goniometer for measurements of the range (in degrees) of motion of passive ankle plantar or dorsiflexion, and flexible equinovalgus deformity before and after treatment.
During the assessment of passive range of motion, the ankle was in a neutral position with the knee fully extended and the hip in a neutral position (i.e. the hip was not in rotation).The neutral position of the ankle is at 90 degrees between the leg and the foot (Figure 4 and Figure 5).Active range of motion and muscle strength was not measured.
The daily duration of ankle-foot continuous passive motion device therapy was 30 minutes on the paretic limb; the average total treatment period was 5.5 days (range 4 to 7).
Consisting of 11 items, the National Institutes of Health Stroke Scale quantifies clinical status impaired by stroke.The 6 th item of NIHSS only evaluates lower leg movements [15].
A separate evaluation of the 6 th item was necessary because other items (e.g.facial nerve, gaze etc) are not necessarily related to the improvement or worsening of lower leg status, but might obscure changes specific to our anatomical region of interest.
We investigated the spasticity of the soleus muscle because the soleus muscle plays an important role in plantar flexion of the foot and maintaining a standing position.The Modified Ashworth Scale is a quick and easy way to measure spasticity [16] (Figure 6).
The modified Rankin Scale evaluates the degree of disability or dependence in the daily activities of post stroke patients or patients with other neurological disabilities.
The scale runs 0 to 6, from perfect health without symptoms to death [14] (Figure 7).
Statistical Analysis
Within-subject differences were calculated by subtracting baseline values from pre-session measurements of the last available session, and tested for significance using Student's paired t tests or Wilcoxon's matched-pairs signedranks tests, subject to normality assumptions being satisfied or not.The statistical package Stata was used for data analysis [17].
Results
Because of the wide individual variability of the passive range of motion and equinovalgus position, first we compared patient by patient changes before and after therapy, then we evaluated the differences and compared them with those observed in the manual group.
The improvement of the 6 th item (NIHSS) did not reach the level of significance (p = 0.055).
The passive range of motion of the ankle plantar flexion increased massively by an average of 3.41 (SD = 5.19) degrees in the device group (p < 0.001) and significant improvement of the mean dorsiflexion in the PROM of the ankle was also detected (p = 0.019).
The changes in modified Rankin Scale before and after treatment are shown in Figure 11 and Figure 12.
To our knowledge, this is the first study to evaluate the efficacy of a passive plantar and dorsiflexion device for the reduction of flexible equinovalgus deformity and ankle spasticity in acute stroke patients.Spasticity with or without other complications (e.g.equinovalgus, Achilles contracture, etc.) has a negative impact on global outcome after stroke [20].
Therefore, prevention and therapy of contracture or abnormal foot position is of high priority.Besides pharmacotherapy, passive mobilization and stretching of the paretic lower extremity is one of the options.Regular mobilization helps prevent contractures and reduce spasticity [21].
A number of publications have reported about the positive impact of stretching on connective tissue prolifera- tion, muscle fiber atrophy, and serial sarcomere loss in immobilized muscles of mice [22]- [24].Gomes et al. investigated the effect of passive stretch, applied for 30 minutes to the rat soleus muscle, on myogenic differentiation (myoD) and they found that a single session of passive stretching increased myoD gene expression, a factor related to muscle growth [25].
Kamikawa et al. have proposed passive stretching to preserve skeletal muscle tone in patients who are unconscious or paralyzed [26].
However, manual motion or stretching requires permanent physiotherapist presence and participation, and treatment efficacy depends on her/his skill to measure range of motion limits or "end feel" [27].
Passive motion devices can be effective alternatives to manual passive motion because they also improve the ankle's range of motion [28] and deformities of the foot.
However, contrary to our work, none of the cited articles tested the new device on acute, bedridden stroke patients (the earliest treatment onset in the cited articles was 7 months after stroke), and none of the publications report a test battery as complex as the one we used for evaluation (passive range of motion before and after therapy, modified Ashworth Scale, equinovalgus, modified Rankin Scale, 6th item of NIHSS, NIHSS).Our device is designed to be used on bedridden patients (as the majority of stroke patients is bedridden in the acute phase), while the majority of the devices published about so far could be used only on standing or sitting pa-tients [27] [29] [30], which is extremely difficult in the acute phase of stroke.
Contrary to Wu et al., Zhang et al., and Selles et al., we studied not only the improvement of gait, reflexes, or Ashworth scale, but also the impact of treatment on global outcome scales such as NIHSS and modified Scale to verify the usefulness of our device on acute stroke patients in a more complex manner [27] [29] [30].Our device is user friendly because it is easy to carry to the patient's bed and adjustable to individual treatment needs (angles, power, and speed); and when the device detects a resistance, it stops automatically.The therapy did not result in any pain or injury.
As far as we know, there is no similar device in available literature that was tested on acute stroke patients and proved to be useful for daily bedside application through an evaluation scheme as complex as described above.
A noteworthy observation is that not only was therapy by the device as effective as the manual approach, but better in terms of some investigated parameters: 1) In the manual group, the passive range of motion of plantar and dorsiflexion, equinovalgus deformity of the foot, and modified Ashworth Scale score did not change significantly, while the same parameters improved significantly in the device group.
2) In the device group, where we performed manual therapy as well, our ankle-foot continuous passive motion device also significantly improved disability scores such as NIHSS score, 6 th item of NIHSS score, and modified Rankin Scale score, while in the manual group, only NIHSS score improved.
The explanation could lie in the fact that, as known from literature, passive movement of a paretic limb can elicit flow and/or metabolic changes in the corresponding cortical and/or subcortical area of the contralateral hemisphere.Motor recovery (induced by arm training) has been observed to be associated with distinct changes of brain activation (regional cerebral blood flow increase as ascertained via positron emission tomography) in bilaterally distributed sensory and motor systems [31].Activation during passive movement could also be detected in the contralateral primary motor, sensory cortices, and in premotor cortical and subcortical regions by functional MRI [32].
Carel et al. found that prolonged passive training of the wrist can change the cortical representation of passive movements and induce over-activation of the primary sensorimotor cortex and supplementary motor areas in healthy subjects as assessed by functional MRI.
Carel et al. [33] and Weiller et al. [34] detected much more widespread cortical and subcortical activation by passive movement of the wrist or elbow than Mima et al. [35] observed; in the latter study, the middle finger was moved passively, and activation could only be detected in the contralateral cortical somatosensory area.
We hypothesize that the reason why cortical and subcortical activation is more intense by mobilization with the device than by only manual mobilization is because the device is characterized by a greater stability in strength, frequency, and form of repetition parameters.Moreover, the stimulation surface is larger (the device stimulates the whole surface of the foot, as opposed to contact area limitations inherent in manual therapy).
In summary, the advantages of our ankle-foot continuous passive motion device are: -Lightweight and easy to use; -Personalized therapy (adjustable angles, power, and speed); -Bedside application on bedridden patients; -No risk for patients (automatically stops in case of overextension or flexion); -Potential applications for quantitative evaluation of antispastic therapy.
The limitations of the device and of this study include: -The plane of movement of the ankle-foot device is confined to the sagittal plane; -We do not know how long the effect of the treatment lasts; -In this study, we applied semi-quantitative clinical scales; -We investigated a heterogeneous sample of stroke patients, similarly to most of the available literature.
Figure 1 .
Figure 1.The ankle-foot continuous passive motion device.
Figure 2 .
Figure 2. Control unit of the ankle-foot continuous passive motion device.
Figure 3 .
Figure 3. Plan view of the ankle-foot continuous passive motion device.
Figure 4 .
Figure 4.The goniometer.Measurement of the passive range of plantar and dorsiflexion.The term "passive range of motion" refers to the radius of motion achieved by external force without using the patient's own muscles.
Figure 5 .
Figure 5. Measurement of the flexible equinovalgus deformity of the foot.
Figure 6 .
Figure 6.Score definitions of the Modified Ashworth Scale.
Figure 7 .
Figure 7. Score definitions of the modified Rankin Scale.
Figure 8 .
Figure 8. Degree of equinovalgus before first and before last session in the manual and device groups.
Figure 9 .
Figure 9. Degree of dorsiflexion passive range of motion before first and before last session in the manual and device groups.
Figure 10 .
Figure 10.Degree of plantar flexion passive range of motion before first and before last session in the manual and device groups.
Figure 11 .
Figure 11.Functional outcome on the modified Rankin Scale in the device group.
Figure 12 .
Figure 12.Functional outcome on the modified Rankin Scale in the manual group.
Table 1 .
Characteristics of patients in the manual and device groups.
Table 2 .
Changes (absolute difference and relative improvement) after therapy compared with baseline values in NIHSS, NIHSS 6 th item, passive range of motion (plantar and dorsiflexion), degree of equinovalgus deformity and Modified Ashworth Scale score.
None of our patients (in either group) developed any complications such as pneumonia or deep venous thrombosis/pulmonary embolism.The changes in equinovalgus, dorsiflexion and plantar flexion before first session and before last session in the manual and device groups are shown in Figures 8-10. | 2017-10-24T02:13:12.246Z | 2015-04-03T00:00:00.000 | {
"year": 2015,
"sha1": "662b0e35750541722e0b4c68c492e835a86c22e6",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=55361",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "662b0e35750541722e0b4c68c492e835a86c22e6",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
3341147 | pes2o/s2orc | v3-fos-license | A rare case of nonresterilized reinforced ETT obstruction caused by a structural defect
Abstract Rationale: Various factors can cause ventilatory failure after endotracheal tube (ETT) intubation, which is associated with increased patient morbidity and mortality. Patient concerns: A 76-year-old woman who was diagnosed with a hemopericardium and suspicion of a major-vessel injury due to dislocation of the clavicular fracture fixation screw. Diagnosis: Non-resterilized reinforced ETT obstruction caused by a structural defect. Intervention: Endotracheal tube was exchanged. Outcomes: The ventilator profile showed rapid improvement. Lessons: Anesthesiologists should consider that a non-resterilized reinforced ETT may be defective. An ETT defect can cause high PIP and ETT obstruction without kinking or foreign materials.
Introduction
Ventilatory failure during general anesthesia is a principal cause of morbidity and mortality; therefore, there is a need to confirm airway patency. Among the many airway maintenance methods, endotracheal tube (ETT) intubation is the most general and safe method for general anesthesia. However, ETT intubation does not guarantee airway patency, and the ETT itself can become a source of airway obstruction. [1] Furthermore, it is difficult to detect an ETT defect quickly, and the results of the defect can be misjudged as being other clinical conditions.
Here, we present a rare case of nonresterilized reinforced ETT obstruction caused by a structural defect; the hemodynamically unstable patient was undergoing emergency explorative sternotomy for hemopericardium.
Case report
A 76-year-old woman (weight, 52 kg and height, 158 cm) classified as American Society of Anesthesiologists (ASA) physical status IV E, was admitted due to sudden chest discomfort and dyspnea. She had hypertension, hyperlipidemia, and a history of surgical fixation of the clavicle with a screw for a clavicular fracture 5 days previously. She underwent chest radiography and chest computed tomography (CT) in the emergency room (ER) and was diagnosed with a hemopericardium and suspicion of a major-vessel injury due to dislocation of the clavicular fracture fixation screw. Her hemoglobin level was 5.9 mg/dL, suggestive of ongoing bleeding. An emergency operation was planned, and 2 pints of packed red blood cells were immediately transfused. Her mental status was alert with spontaneous ventilation. Vital signs were blood pressure of 174/89 mm Hg, heart rate of 72/min, body temperature of 36.2°C, and respiratory rate (RR) of 36. Oxygen saturation was 90% under 3 L/min oxygen through a nasal cannula. Breathing sounds in both lung fields were clear without crackling or rhonchi.
On arrival in the operating theater without premedication, standard monitoring devices including electrocardiography, pulse oximetry, and an oscillometric noninvasive blood-pressure cuff were applied. After a few minutes of preoxygenation, general anesthesia was induced using 10 mg intravenous etomidate (0.2 mg/kg), 50 mg remifentanil (1 mg/kg), and 50 mg rocuronium (1 mg/kg) for neuromuscular blockade. Edinburgh, UK). The insertion depth was 22 cm from the upper incisors. Anesthesia was maintained using oxygen, medical air, 1 minimum alveolar concentration (MAC) of sevoflurane, and continuous infusion of remifentanil.
After intubation, both lung fields were auscultated using a stethoscope; there were no signs of bronchospasm, with clear lung sounds, no wheezing, and normal capnography. The ventilator was set in volume-guaranteed, pressure-controlled mode at a tidal volume of 250 mL. However, the ventilator failed to reach the intended tidal volume of 250 mL and achieved a tidal volume of only 170 mL, due to limitation in peak inspiratory pressure (PIP) of 35 cmH 2 O. We considered the high PIP to be caused by the abundant pericardial hematoma burden increasing intrathoracic pressure. We used the recruitment maneuver and applied positive end expiratory pressure (PEEP) with low tidal volume and high RR, instead of increasing inspiration pressure to maintain hemodynamic stability. An indwelling right radial artery cannula and a right internal jugular central venous catheter were placed for monitoring, blood sampling, and fluid resuscitation.
At that time, the patient's vital signs deteriorated, with ongoing bleeding; therefore, an immediate surgical procedure became necessary. We maintained the tidal volume setting at 160 mL, RR of 16/min, and PEEP of 5 cmH 2 O and the operation began. The patient began to become desaturated after massive bleeding from the sternotomy, despite delivery of a 1.0 fraction of inspired oxygen. The oxygen saturation measurement was not reliable due to low perfusion, and the arterial line did not work appropriately for blood pressure monitoring and blood sampling. An arterial blood gas analysis (ABGA) was required to assess the adequacy of ventilation and blood transfusion; however, access to the sampling site was limited due to positioning and the ongoing procedure. Therefore, we had to wait until the main surgical procedure had been completed. After the innominate artery laceration had been found and repaired, the right femoral artery was cannulated and blood sampling was performed. The ABGA showed severe respiratory acidosis (pH 7.22, partial pressure of carbon dioxide 72.3 Torr, and partial pressure of oxygen 61 Torr). Lung compression was released once the hematoma was removed, but PIP remained higher and tidal volume lower than expected. The possibility of an ETT obstruction arose and was carefully investigated. After the recruitment maneuver, we could not advance the tracheal suction tip through the ETT, and luminal narrowing of the ETT was detected (Fig. 1).
The tube was immediately exchanged, and the ventilator profile showed rapid improvement. The lungs ventilated with a tidal volume of 420 mL and peak airway pressure was 19 cmH 2 O with the new ETT. In addition, the ABGA results showed that the respiratory acidosis had been corrected. Almost the full length of the ETT that was removed was dissected and the metal frame was found to be distorted (Fig. 2). We assume that the faulty tube caused the increased PIP.
Discussion
ETT intubation is an essential element of general anesthesia and is needed for accurate testing to establish a secure airway. High PIP after ETT intubation can be caused by various factors such as respiratory disease and position, operation type, and problems with the ventilator or airway equipment. [1] Among possible problems with airway equipment, the ETT can become obstructed by mucus, blood, or kinking. Rarely reported cases such as dissection of the internal wall, herniation of the cuff, or detachment, can also obstruct the ETT. [2] Thus, accurate ETT intubation is not enough to guarantee a patent airway, and the ETT itself may become a source of airway obstruction. [1] Ventilatory failure from an airway obstruction can lead to a lifethreatening event; therefore, it is important to check the possible causes of ventilator failure in advance. However, such confirmation can easily be overlooked, particularly in emergency cases. Several previous cases have reported a reinforced ETT obstruction. In most of these cases, a tube defect occurred after reusing a sterilized tube. [1,3,4] Other cases of introduced nonresterilized ETT obstructions can be explained by a bleb or air formation during manufacturing, which expanded after nitrous oxide exposure. [5,6] Another case report detailed an entire dissection of the inner wall of a reinforced ETT because of an aggressive stylet. [7] The presumed etiology of our case was a tube-manufacturing malformation or stylet aggression of the inner wall.
During the procedure, we did not consider the possibility of a tube defect. First, we considered that the massive hematoma could increase intrathoracic pressure and PIP. [8] Moreover, PIP did not increase immediately after intubation as usually occurs with an ETT defect. Therefore, the patient's high PIP was considered due to the hemopericardium. Moreover, there was no unusual resistance when the stylet was inserted and withdrawn at the moment of intubation. ETT dissection by an aggressive stylet was therefore not considered. Finally, there was neither kinking of the ETT nor change in the appearance of the ETT.
Exchanging the ETT was the solution in this case and it was not difficult or complicated. However, this situation could have worsened if ventilatory failure had been misjudged as a lung problem. Standardized checklists are available for similar conditions. The checklist confirms that critical steps are not missed and can be used in multiple situations, including on the ward, in the intensive care unit, in the operating room, and particularly during emergencies. [9] In our case, airway equipment was checked by anesthetic nurses who usually check only for ETT cuff leaks and the external shape of the tube. If the anesthetic staff had checked the airway equipment more carefully, the distorted metallic frame might have been detected before intubation.
In conclusion, anesthesiologists should consider the possibility that a nonresterilized reinforced ETT may be defective. ETT defects can cause high PIP and an ETT obstruction without kinking or foreign materials. Additional careful confirmation of the checklist is an important and useful step in emergency situations. | 2018-04-03T01:07:13.478Z | 2017-12-01T00:00:00.000 | {
"year": 2017,
"sha1": "575ec94408503e8848ca74630457d4c942f12da4",
"oa_license": "CCBYSA",
"oa_url": "https://doi.org/10.1097/md.0000000000008886",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "492d93ab417fffa6dce54ebde29a2fe474498bf0",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
110502993 | pes2o/s2orc | v3-fos-license | Evaluation of Thailand’s Drug Abuse Resistance Education Project (DARE) Provincial Police Region 1
Aim of this research is to study the factors affecting the problems and obstructions of the Drug Abuse Resistance Education (DARE) project. This research evaluates project administration, the processes of cooperation among key persons, success of gaining knowledge, positive attitudes and behaviour of students after rehabilitation. The research also suggests methods to improve and develop the operation of the project in the area of the Provincial Police Region 1. Research methodologies were qualitative and quantitative and the participants were interviewed in detail about their field observations as well as their participation in discussions. The key success factors included the support of the heads of relevant organisations. Positive aspect of the DARE project is that it can prevent youths from taking drugs for some time. The negative aspects were the lack of understanding of the police towards project objectives. The suggestions for improvements mostly included providing mobile standard equipment for DARE police and the operation of DARE youth camps in schools, establishment of DARE police clubs and the operation of DARE police groups, which included the commissioned officers and the warrant officers, and curriculum improvement to suit Thai culture and drug situations.
Introduction
Thailand still faces the problem of illicit drug trafficking and smuggling from the Golden Triangle and neighbouring countries.'Methamphetamine' still remains a major seized drug.Royal Thai Government's approach is to solve drug problems in a comprehensive and systematic manner, from prevention and suppression to rehabilitation of drug addicts.Thailand's Drugs Control Strategy is prevention.For the 'General group', it will launch anti-drugs campaign and provide drug information in various forms to give knowledge about drugs.For the 'Sensitive group', it will perform surveys and collect information, carry out anti-drugs activities and monitor sensitive locations.For the 'Educational group' it will establish a core group (teachers, parents, students) in educational institutions and communities to watch drugs abusers and develop teaching and learning curriculum to prevent taking drugs.
The Drug Abuse Resistance Education (DARE) project or DARE police is one of the major drug prevention projects which focus on the youth target groups organized by police officers.DARE project gives kids life skills they need to avoid getting addicted to drugs, avoid gangs and also violence.DARE was found in 1983 by the Police Academy in Los Angeles and has proved successful in preventing drug abuse and other problems related to it.DARE is now implemented in schools and in more than 43 countries around the world. 1 In the United States, DARE is one of the most widely used substance abuse prevention programmes, targeted at school-aged youth.In recent years, DARE has been the country's largest single school-based prevention programme in terms of federal expenditure, with an average of three quarters of a billion dollars spent on its provision annually. 2DARE operates in about 80 percent of all school districts across the United States and in many foreign countries. 3RE's primary mission is to provide children with information and skills they need to lead drug and violence free lives, by educating them about drug menace through programmes at elementary, middle and high school levels. 4Lesson plans for DARE focus on four major areas.
• to provide accurate data about using drugs, alcohol as well as tobacco.
• to teach students decision-making skills.
• to show students how to recognize and resist peer pressure, and, • to inform students about positive alternatives to drug use.DARE officers are working with children to raise their self-esteem, teach them how to make decisions on their own, and help them identify positive alternatives to drugs.By role-playing, the DARE curriculum emphasises the negative consequences of drug use and reinforces skills to resist peer pressure.Moreover, DARE is a cooperative effort by the police, schools, parents, and communities -all four working together to help children make right choices to abstain from drug use. 1 One of the unique features of DARE is the use of uniformed police officers as instructors assigned to classrooms.
The DARE curriculum is designed to be taught by police officers whose training and experience give them the background necessary to answer sophisticated questions, often posed by young students about drugs and crimes. 1 Prior to entering the DARE program, officers underwent 80 hours of special training in areas such as child development, classroom management, teaching techniques and communication skills.Forty hours of additional training were provided to DARE instructors to prepare them to teach the high school curriculum on prevention of drug abuse.
For the follow-up and evaluation of DARE, it was found that there was a significant relationship between earlier DARE participation and less use of illegal, more deviant drugs (e.g., inhalants, cocaine, LSD ) in a development sample, and not in a validation sample. 5RE project was founded in Thailand in 1999.The Office of Narcotics Control Board of Thailand (ONCB) is considered as the office of the DARE project, which supported the national strategy of solving drug problem in high risk youth groups.The objective of the project is to educate youth about the resistance to drugs, the negative action of gangs and anti-violence in youth groups.The ONCB has given financial support to the group of researchers from Suan Dusit Rajabhat University of Thailand to organise the research project to evaluate the DARE project in the area of Provincial Police Region 1, in 2010.
Objectives
1. To study the models of the DARE project in Thailand and foreign countries 2. To study factors affecting problems and obstructions for successful implementation of the project 3. To evaluate project administration, cooperation of DARE police with the schools, provincial organisations as well as the parents, the success of raising knowledge and positive attitudes, behaviour of the students and to prevent them from drug abuse 4. To suggest ways to improve and to develop the operation of the DARE project in the area of Provincial Police Region 1
Methods
Research methodologies were qualitative and quantitative.To gather data for qualitative methodologies, in-depth interviews and field observations were conducted.The sample groups were the head administrator, the administrators of the project, resources persons of the DARE project, teachers, parents, local administrators and DARE police in the area of Provincial Police Region 1 (the central part of Thailand), who were involved in the project, out of a total of 64 persons.The researchers also used the focus group discussions of concerned persons to comment and to confirm the result of the research from 12 persons.For the quantitative approach, the researchers used questionnaires to gather data from 360 students involved in the project, in the area of Provincial Police Region 1.
Key factors of success
In the operation, the project was supported by the head administrators and the head of the police stations.Continuous training for knowledge and skill development of DARE police also helped the programme.It included voluntary participation of DARE police, the support and cooperation from administrators and teachers in schools, the continuous support from the governors and also local administrators.
Problems and obstructions
In the operation of the project, there were many obstacles such as instability of the status of police of the area, lack of financial support for the operation, administration and special activities of the project, as well as lack of public support for the project in the province.Seeking the support from relevant people might be seen as a benefit.The length of the DARE curriculum had to be covered in a short period of time.Lack of parentchildren relationships, lack of DARE police officers in some areas and the length of the DARE curriculum which was not suitable for the current drug situation in Thailand were some problems.
Positive aspects
The DARE project can prevent youth from taking drugs for some time.DARE police officers in uniform can be satisfied that they have educated students about risks of taking dangerous drugs.School administrators as well as teachers can participate in the project to inform students such risks and to prevent students taking drugs.Students can also inform their friends and parents the necessity to reject such drugs.Students have rational thinking and were interested in studying in DARE classes.The project was worthwhile compared to the cost of the drug treatments and rehabilitation.Many people are interested to see that the DARE project continue and expand, in spite of its ineffectiveness and even harm to students.About two million baht is spent for the programme each year, and thousands of police officers are assigned to educate students about the risks of taking dangerous drugs.
Negative aspects
Police officers do not understand the objectives of the project.Some details of the curriculum were not suitable for Thai culture and society.Lack of financial support for learning equipment and teaching and the short duration of teaching will make the students become the risk group in taking drugs.Lack of continuous support and cooperation from local administration is also a negative aspect.There are many reasons for the remarkable ineffectiveness of the DARE project.For example, it is based on out-dated theories of learning and human behaviour and it failed to distinguish between legal substances and illegal drugs.It views all use as abuse and it presents a view of substance use inconsistent with what most students see in their environment.
Suggestions for urgent implementation
Suggestions for urgent operation are organising mobile standard equipment for DARE police, curriculum improvement to suit Thai culture and the drug situation in Thailand.When increasing financial support is given for the project operation, the budget should be sent directly to the key officers who worked in the area.Increase in the cadres of DARE policemen and police women in every area, integration of DARE curriculum in the official curriculum of primary and secondary schools and the development of continuous follow-up and evaluation operation of the project are some other suggestions.
Suggestions for project operation and curriculum
Suggestions
Conclusion
The DARE project or DARE police is one of the major drug prevention projects, which stress on the youth target groups and was organized by the police officers.The ONCB considers the DARE project as one of its drug prevention projects, which supports the nation's strategy of solving the drug problem among youth.The objectives of the research were to study the models of the DARE project in Thailand and foreign countries, to study the factors affecting the problems and obstructions of the project, to evaluate the project administration and the role of DARE police officers and their cooperation with school administrations, provincial organisations and parents, success of organising awareness campaigns, positive attitudes to prevent students taking drugs, and to suggest ways to improve and to develop the operation of the DARE project in the area of the Provincial Police Region 1. Research methodologies are both qualitative and quantitative; the qualitative methodology used in-depth interviews and field observation methods.Sample groups were the head administrator, the administrators of the project, and resources persons of the DARE project, teachers, parent, local administrators and DARE police in the area of Provincial Police Region 1.The present study showed that it is not a 100% successful program.
for project operation and curriculum are operation of DARE youth camps in schools, establishment of a DARE police clubs, and operation of DARE police groups which included Commissioned Officers and Warrant Officers.Other suggestions for the success of the project are separation of DARE police training in the operation and skills in teaching practical courses, finding financial resources to support the project, preparing an official directory of DARE students and DARE, expansion of DARE operation in government and private schools, having a website of DARE project for the members concerned, promotion of a strong cooperation among DARE police, teachers in schools, local administration and with relevant organizations. | 2019-04-13T13:07:21.149Z | 2013-08-07T00:00:00.000 | {
"year": 2013,
"sha1": "ab6bdc9c947c5f27a797480a019a9fb820f48d86",
"oa_license": "CCBY",
"oa_url": "http://ijptsud.sljol.info/articles/10.4038/ijptsud.v1i1.5914/galley/4674/download/",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "2d00b8abd39de5f23b6fe70680049c3791021a65",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Engineering"
]
} |
271079826 | pes2o/s2orc | v3-fos-license | Small and Large Extracellular Vesicles of Porcine Seminal Plasma Differ in Lipid Profile
Seminal plasma contains a heterogeneous population of extracellular vesicles (sEVs) that remains poorly characterized. This study aimed to characterize the lipidomic profile of two subsets of differently sized sEVs, small (S-) and large (L-), isolated from porcine seminal plasma by size-exclusion chromatography and characterized by an orthogonal approach. High-performance liquid chromatography–high-resolution mass spectrometry was used for lipidomic analysis. A total of 157 lipid species from 14 lipid classes of 4 major categories (sphingolipids, glycerophospholipids, glycerolipids, and sterols) were identified. Qualitative differences were limited to two cholesteryl ester species present only in S-sEVs. L-sEVs had higher levels of all quantified lipid classes due to their larger membrane surface area. The distribution pattern was different, especially for sphingomyelins (more in S-sEVs) and ceramides (more in L-sEVs). In conclusion, this study reveals differences in the lipidomic profile of two subsets of porcine sEVs, suggesting that they differ in biogenesis and functionality.
Introduction
Seminal plasma is the complex biofluid that surrounds spermatozoa during and after ejaculation.It is composed of a mixture of the secretions from the functional organs of the male reproductive tract, mainly from the epididymis and the accessory sex glands [1].Seminal plasma contains a wide variety of molecules that play key roles in crucial reproductive processes, such as the modulation of sperm functions [2,3] and the induction of maternal immune tolerance to spermatozoa and embryos [4,5].Some of these biomolecules, including proteins [6], lipids [7], small non-coding and regulatory RNAs [8], are not only circulate freely in seminal plasma, but are also loaded into cell-derived membrane-surrounded nanovesicles, called extracellular vesicles (EVs), where biomolecules are protected from the degradative activity of seminal plasma enzymes [9,10].
Extracellular vesicles play a key role in cell-to-cell communication and circulate freely in all body fluids [11].Seminal EVs (sEVs), like those of any other body fluid, are a heterogeneous population with different phenotypic characteristics [6,12,13].This heterogeneity suggests the existence of different sEV subsets, which may have different cell origin, biogenesis pathways, cargoes, target cells and functional performance [14].The isolation and separate characterization of EV subtypes circulating in body fluid remains a scientific challenge [15].
Two subsets of sEVs of different sizes, termed small sEVs (S-sEVs) and large sEVs (L-sEVs), were successfully isolated from porcine seminal plasma using a size-exclusion chromatography (SEC)-based isolation procedure [16].Such size-associated EV subsets are in line with the recommendation of the International Society for Extracellular Vesicles (ISEV) to use terms referring to phenotypic characteristics, such as size, to define EV subsets [17].The proteomic profile of these two sEV subsets revealed quantitative differences between S-sEVs and L-sEVs in several proteins [6].Such differences in protein cargo support the concept of the coexistence of different sEV subsets in porcine seminal plasma, which may have differences in their functionality, as EV functions are closely related to their cargo [14,18].
In addition to proteins, EVs also contain lipids, a diverse family of macromolecules that perform essential biological functions, including structural ones, as they are essential components of membranes, energy storage and signaling [19].The lipid composition of EVs is determined by the biogenesis pathway in the cell of origin and contributes to their intracellular trafficking, as well as their uptake and functional response in recipient cells [18,20].The lipid content of sEVs has received much less attention than that of proteins or RNAs [10], and quantitative lipidomic studies of EVs are needed [21].Quantitative studies of the lipidomic profile of sEVs are essential for a better understanding of the underlying mechanisms of EV biogenesis and protein packaging, as well as for the exploration of sources of molecular biomarkers of fertility in sEVs [22].
The aim of the present study was to provide a detailed lipidomic profile of two differently sized porcine sEV subsets using high-performance liquid chromatography (HPLC) coupled to high-resolution mass spectrometry (HRMS), a high-throughput and highly sensitive technology that allows comprehensive and quantitative analysis of a broad repertoire of lipid species in any biological sample, including EVs [23].To the best our knowledge, there are only four studies that have reported the relative amounts of some lipid classes in sEVs, two in humans [24,25], one in horses [26], and one in pigs [7].Only the study by Brouwers et al. was performed using MS-based technology for relative lipid quantification [25], the other three studies used methods based on thin-layer chromatography (TLC), an older technology that is less sensitive than MS-based technology [27].
Lipids Identified in sEVs by LC-MS Lipidomic Analysis
The targeted lipidomic analysis identified and quantified a total of 157 lipid specie belonging to four major lipid categories, namely sphingolipids (SP), glycerophospholip ids (GP), glycerolipids (GL), and sterol lipids (ST).The identified and quantified lipid spe
Comparison of Lipidomic Profile between S-sEVs and L-sEVs
Qualitative differences between S-sEVs and L-sEVs were found only in two CE cies.Specifically, CE (16:0) and CE (20:2) were identified and quantified only in th sEVs.They accounted for 1.3% of the total lipid species that were identified and quanti in the S-sEV samples.In contrast, quantitative differences were found in virtually all id tified and quantified lipid classes.The lipid to protein ratio of the two subsets of sEVs different, being 5:1 for S-sEVs and 10:1 for L-sEVs.L-sEVs had higher relative amoun < 0.01) of the four major lipid categories, namely SP, GP, GL, and ST, compared to S-s (Figure 4A-D).All the 14 identified and quantified lipid classes were more abundan L-sEVs than in S-sEVs (p < 0.05), regardless of their lipid category (Figure 5A-D).The 36 SP species identified and quantified were six ceramides (Cer), four dihydroceramides (DHCer), 12 sphingomyelins (SM), six dihydrosphingomyelins (DHSM), four hexosylceramides (HexCer), and four ceramide dihexosides (CDH).The 53 GP species were 25 phosphatidylcholines (PC), 15 phosphatidylethanolamines (PE), four lysophosphatidylcholines (LPC), and nine ether-linked PE (PE O-).The 59 GL species were eight diacylglycerols (DG) and 51 of triacylglycerols (TG).Finally, the nine ST species were free cholesterol (FC) and eight cholesteryl esters (CE) (Figure 3 and Table S1).Lipid species identified and quantified in both sEV subsets were uploaded to the Vesiclepedia database (Study ID: 3594) [28].
Comparison of Lipidomic Profile between S-sEVs and L-sEVs
Qualitative differences between S-sEVs and L-sEVs were found only in two CE species.Specifically, CE (16:0) and CE (20:2) were identified and quantified only in the S-sEVs.They accounted for 1.3% of the total lipid species that were identified and quantified in the S-sEV samples.In contrast, quantitative differences were found in virtually all identified and quantified lipid classes.The lipid to protein ratio of the two subsets of sEVs was different, being 5:1 for S-sEVs and 10:1 for L-sEVs.L-sEVs had higher relative amounts (p < 0.01) of the four major lipid categories, namely SP, GP, GL, and ST, compared to S-sEVs (Figure 4A-D).All the 14 identified and quantified lipid classes were more abundant in L-sEVs than in S-sEVs (p < 0.05), regardless of their lipid category (Figure 5A-D).The four lipid categories were similarly distributed in S-sEVs and L-sEVs.GL was the dominant category with more than 50%, while ST was the minority with less than 1% (Figure 6A).Regarding the lipid classes within each lipid category, in the SP category, SM, Cer and HexCer were the most predominant lipid classes in both sEVs subsets.However, SM and Cer were present in different proportions in S-sEVs and L-sEVs (p < 0.05).SM were proportionally more abundant in S-sEVs than in L-sEVs, whereas Cer were proportionally more abundant in L-sEVs than in S-sEVs (Figure 6B).In the GP category, PC and PE were the predominant lipid classes in both sEV subsets, with no differences between sEV subsets (Figure 6C).In the GL category, TG was the more predominant lipid class in both sEV subsets, and the distribution of TG and DG was similar in L-sEVs and in S-sEVs (Figure 6D).In the ST category, FC was the predominant lipid in both sEV subsets, but FC and CE were present in different proportions in S-sEVs and L-sEVs (p < 0.01).FC was proportionally more abundant in L-sEVs than in S-sEVs, whereas CE were proportionally more abundant in S-sEVs than in L-sEVs (Figure 6E).The four lipid categories were similarly distributed in S-sEVs and L-sEVs.GL was the dominant category with more than 50%, while ST was the minority with less than 1% (Figure 6A).Regarding the lipid classes within each lipid category, in the SP category, SM, Cer and HexCer were the most predominant lipid classes in both sEVs subsets.However, SM and Cer were present in different proportions in S-sEVs and L-sEVs (p < 0.05).SM were proportionally more abundant in S-sEVs than in L-sEVs, whereas Cer were proportionally more abundant in L-sEVs than in S-sEVs (Figure 6B).In the GP category, PC and PE were the predominant lipid classes in both sEV subsets, with no differences between sEV subsets (Figure 6C).In the GL category, TG was the more predominant lipid class in both sEV subsets, and the distribution of TG and DG was similar in L-sEVs and in S-sEVs (Figure 6D).In the ST category, FC was the predominant lipid in both sEV subsets, but FC and CE were present in different proportions in S-sEVs and L-sEVs (p < 0.01).FC was proportionally more abundant in L-sEVs than in S-sEVs, whereas CE were proportionally more abundant in S-sEVs than in L-sEVs (Figure 6E).
Discussion
In the present study, the lipidomic profiles of two subsets of EVs of different sizes isolated by SEC from porcine seminal plasma (sEVs), termed small sEVs (S-) and large sEVs (L-), were characterized.Lipidomic studies of EVs are still limited, and those characterizing the lipidomic profiles of EV subpopulations are even more limited.To the best of our knowledge, only a few studies have compared the lipid profile of EV populations of different sizes, all of which are secreted in vitro from body cells or single cell lines [29][30][31][32].Therefore, the present study is the first to compare the lipidomic profile of EV subsets of different sizes circulating in body fluids.The results of the above-mentioned studies showing differences between small and large EVs in the relative amounts of several lipid species are consistent with those reported here in that the relative amounts of many lipid classes differ between S-sEVs and L-sEVs.
Separation of the two sEV subsets was achieved by centrifugation at 20,000× g, a g-force that allows the sedimentation of the large particles, including the L-sEVs, while leaving the smaller particles, including S-sEVs, in the supernatant [33].Both subsets of sEVs were then separately purified by SEC, a procedure that allows the recovery of a large number of intact and minimally contaminated EVs [15].The samples of sEVs were characterized using an orthogonal approach and the results confirmed that S-sEVs and L-sEVs differed in size, as expected, but also in concentration and morphology, in agreement with previous studies [6,34].Flow cytometry analysis showed that the isolated sEVs in both S-sEV and L-sEV samples exhibited a high degree of purity, a necessary requirement to obtain reliable results in EV composition studies such as the present one [21].It should be noted that the major non-EV particles co-isolated in EV samples are lipoproteins, which are characterized by a hydrophobic lipid core [35].Lipoproteins are usually co-isolated with EVs due to their similar size to smaller EVs (≈25-80 nm for lipoproteins).This can lead to the overestimation of some lipid species [21,36,37].
Extracellular vesicles are membranous structures derived from cell membranes.Therefore, the lipid repertoire of EVs is expected to consist mainly of cell membrane lipids.However, they may contain small amounts of other cytosol-derived lipids acquired during biogenesis [11,21].Indeed, the most abundant lipids found in both sEV subtypes were, in descending order, TG, PC, SM, DG, and Cer.These results would be consistent with those reported by Peterka et al., who found that TG, PC and SM were some of the most abundant lipids in EVs isolated from human plasma [38].Most of the lipids found in the sEVs are typical structural lipids of cell membranes.In particular, PC, SM and Cer are abundant structural lipids in plasma and endosomal membranes [39].In contrast, TG and DG are lipid classes present in the inner space of the bilayer but are not thought to play a structural role [38].This is the first study to report the presence of TG and DG in sEVs, lipid classes that have also been found in EVs isolated from human blood plasma [38], mouse urine [40], and human brain tissue [41].The enrichment of TG in both sEV subsets should be viewed with caution, as TG are important components of lipoproteins and lipid droplets [42], which are common contaminants of EV samples, as discussed above.However, it is unlikely that the high levels of TG found in porcine sEV samples are solely due to lipoprotein contamination.This contention would be supported by flow cytometry results showing a very high percentage of CFSE+ events in both sEV samples, which would be membrane-intact sEVs [43]; and by the relatively low amount of CE, another lipid class abundant in lipoproteins [21,44].Furthermore, TG have been found at high levels in EVs produced in vitro by cell lines [45], which are less contaminated by lipoproteins than those isolated from body fluids [21].The high levels of TG found in sEVs may be due to the binding of free lipoproteins circulating in seminal plasma to sEV membrane, which has been shown to occur in EVs cultured in vitro [37,46].It should be noted here that the comparison of the lipidomic profile of EVs isolated from different body fluids or different cell cultures may not be scientifically accurate.The lipid composition of EV samples is strongly influenced by the isolation procedure used [47], in addition to the expected natural influencing factors such as the biogenesis pathway, the cell of origin and its functional state, i.e., physiological or pathological [11].Among other reasons, the variety of methods used for EV isolation leads to the unpredictable presence of non-EV particles in isolated EV samples, including lipoproteins as mentioned above.This can lead to inaccurate results regarding the lipid composition of EVs, which makes comparisons between studies of little scientific value.
In addition to the structural role of lipids in EVs, their role in cell-to-cell communication must also be considered.Lipids are one of the active molecules that facilitate the transport EVs from source cells to target cells [11,47].With respect to sEVs, it is well known that they bind to and deliver their cargo to spermatozoa, thereby regulating important sperm functions such as capacitation [10].Leahy et al. showed enrichment of the sperm proteome after sEV binding, which would indicate transfer of proteins from sEVs to spermatozoa [48].Lipid transfer from sEVs to sperm remains to be demonstrated [49], although the easy transfer of lipophilic dye between epididymosomes and spermatozoa [50] suggests that lipid exchange occurs between sEVs and spermatozoa.
Lipids are known to be involved in key sperm functions such as capacitation.Certainly, key steps in sperm capacitation are the removal of cholesterol from the sperm surface and the subsequent remodeling of membrane lipid to increase membrane lipid fluidity [51,52].Sperm subjected to stressful conditions, such as cryopreservation, undergo so-called premature or false capacitation, rendering them incapable of fertilization [53].Binding of sEVs to spermatozoa prevents premature or false capacitation and stabilizes sperm membranes, including acrosomal membranes [54][55][56].This role of sEVs may be mediated, at least in part, by the transfer of certain lipid species from sEVs to spermatozoa.Indeed, some of the lipids found in the sEVs are associated with processes related to sperm physiology.For instance, some specific PC and PE species that were detected in both sEV subsets, namely PC (36:1), PC (38:4), and PE (34:4), have been proposed as sperm motility markers [57].In addition, PC in particular, but also PE, is known to play a key role in the prevention of phospholipid movement by binding to seminal plasma proteins, thereby maintaining sperm membrane stability [58].It has also been shown that the stabilizing effect of PC on the sperm membrane prevents cold shock-induced ultrastructural damage in porcine spermatozoa [59].The degradation of PC by phospholipase A2 results in LPC, which has been measured in S-sEV and L-sEV samples and could be another sEV lipids that modulates sperm functionality, specifically the acrosomal reaction, as moderate amounts of LPC have been shown to induce the acrosomal reaction in bovine and human spermatozoa [60][61][62].PC can also be degraded to DG, which are abundant in the sperm membrane and whose reduction is associated with infertility in bulls [63].Concentrations of TG and FC in seminal plasma have been shown to be significantly higher in cats with good than poor semen quality [64].Loss of sperm FC is a key early step in capacitation, as noted above.SM plays an important role in slowing down FC loss, thereby influencing the timing of sperm capacitation [65].Therefore, it is plausible that one of the ways by which sEVs avoid premature capacitation is through the transfer of SM to spermatozoa.L-sEVs would be more effective in modulating sperm function via these lipid-dependent pathways because they contain higher levels of all these lipid classes than S-sEVs.Recent evidence has shown that sEVs decrease lyso-PC in porcine spermatozoa [66].Thus, sEVs may also be involved in the removal of "unneeded" molecules from spermatozoa, as suggested by Leahy et al. [48].
Lipidomic analysis revealed qualitative differences in two CE species, specifically CE (16:0) and CE (20:2) between the two sEV subsets, which were identified in S-sEVs but not in L-sEVs.CEs are energy storage lipids formed by the esterification of cholesterol to fatty acids [67] and, as mentioned above, are abundant in the hydrophobic core of lipoproteins.Therefore, the interaction of S-sEVs with lipoproteins circulating in the seminal plasma may be responsible for the presence of these two CEs in S-sEVs.Small EVs are more likely than large EVs to be surrounded by a peripheral coronal layer composed of active molecules, including lipoproteins [68][69][70].These fatty acid-bearing cholesterol lipids may indicate the pool of fatty acids available to synthetize complex lipids [71].One of the two CEs present only in S-sEVs was CE (16:0), termed cholesteryl palmitate, which is formed by the esterification of cholesterol and palmitic acid (C16:0).Recently, Barreca et al. developed an innovative approach for the metabolic labeling of a homogeneous population of small EVs based on the tracking of a fluorescent palmitic acid analogue, called Bodipy FL C16 [72].They successfully demonstrated the localization of BODIPY FL C16 in specific intracellular compartments, including the endoplasmic reticulum and the endolysosomal compartment, while it was absent from the cell membrane of EV-producing cells.Therefore, they hypothesized an association between palmitic acid and the intracellular origin of a specific subpopulation of small EVs.The exclusive presence of CE (16:0) in S-sEV samples would support that these samples are enriched with exosomes (i.e., endosome-derived EVs) whose size matches that of S-sEVs.The other uniquely identified CE in S-sEVs, CE (20:2), has not been previously reported.Whether these two lipids play a relevant role in S-sEVs is therefore unknown.In contrast to the few qualitative there were clear quantitative differences in the lipid profile of S-and L-sEVs.L-sEVs had higher relative amounts of the four major lipid categories and of each of the 14 identified lipid classes.Since the number of sEVs/µg protein was similar in both sEV subsets, size differences between S-and L-sEVs could be the reason for these quantitative differences.In this and previous studies [21], it has been shown that the lipidomic profile of EVs is largely composed of membrane lipids.Therefore, it is reasonable to assume that larger EVs have a greater relative amount of lipids than smaller EVs due to their larger membrane surface area.
In addition to the relative amount, the proportion of each lipid category and class was also compared between S-sEVs and L-sEVs.The four lipid categories were similarly distributed in S-sEVs and L-sEVs, which would be consistent with the results reported by Durcin et al. and Kim et al. in subtypes of EVs of different sizes from adipocytes and a specific cell line (DU145 HRPC), respectively [29,31].In terms of lipid classes, SM and Cer were observed to have the most notable proportional differences between the two subsets of sEVs.Hydrolysis of SM to Cer is a known pathway in the biogenesis of both exosomes and ectosomes [11].It is plausible that this pathway is more active in L-sEVs than in S-sEVs, because L-sEVs had a higher Cer:SM ratio than S-sEVs.It is interesting to note that L-sEVs contain more ectosomes than exosomes [73].
The quantitative lipidomic differences between S-and L-sEVs suggest differences in functional roles between the two sEV-subsets.Indeed, a recent functional study in pigs by Barranco et al. showed that L-sEVs were able to modulate sperm metabolism, whereas S-sEVs had no detectable effect [74].The different molecular composition of Sand L-sEVs may explain the observed functional differences between the two subsets of sEVs.This would be supported by the results of Barranco et al. who reported quantitative differences between the two subsets of sEVs in proteins related to sperm functionality [6].Unfortunately, lipidomics does not have the same background of functional data as proteomics.Therefore, investigating the role of sEV lipids in modulating sperm function remains a challenge.
Animals and Seminal Plasma Samples
Animal procedures were performed in accordance with international guidelines for the protection of animals used for research purposes (Directive 2010-63-EU) and the experiment carried out was approved by the Bioethics Committee of the University of Murcia at its meeting on 25 March 2021 (research code: CBE: 367/2020).The reagents used were provided by Merck (Darmstadt, Germany) unless otherwise stated.
Nine ejaculates were collected from nine mature, healthy, and fertile Pietrain boars (one ejaculate per boar) housed in an artificial insemination (AI) center of AIM Iberica (Topigs Norsvin España SLU).The boars were routinely used to produce AI semen doses for commercial AI programs, and the ejaculates used in this study met the quality criteria established for the commercial production of AI semen doses, namely greater than 200 × 10 6 sperm/mL with more than 70% motile spermatozoa and more than 75% morphologically normal spermatozoa.Samples of 15 mL of each ejaculate were centrifuged twice at 1500× g for 10 min at room temperature (RT) (Rotofix 32A, Hettich Centrifuge UK, Newport Pagnell Buckinghamshire, UK) to collect seminal plasma.The resulting nine seminal plasma samples were randomly pooled three to three to generate three separate pools, each containing three seminal plasma samples.One tablet of a protease inhibitor cocktail (Roche protease inhibitor cocktail complete™, Basel, Switzerland) was added to each seminal plasma sample and the samples were then stored at 5 • C until the isolation of sEVs.
SEC was performed using home-made columns consisting of 10 mL of Sepharose CL-2B ® (Sigma Aldrich ® , Burlington, MA, USA, Merck KGaA) packed into 12 mL filtration tubes (Supelco ® Filtration Tubes 12 mL, Bellefonte, PA, USA).The supernatant and pellet samples from each 20,000× g centrifugation were separately subjected to SEC.Twenty eluted fractions of 500 µL each were collected sequentially in each SEC.The EV-enriched fractions (7 to 9) from each SEC were pooled to create a single sEV sample.Thus, two sEV samples were obtained from each of the three seminal plasma pools, one enriched in smaller sEVs (S-sEVs) from the SEC of the 20,000× g centrifugation supernatant and one enriched in larger sEVs (L-sEVs) from the SEC of the 20,000× g centrifugation pellet.In total, six samples of sEVs, three S-sEVs and three L-sEVs, were generated and stored at −80 • C (Ultra Low Freezer; Haier Inc., Qingdao, China) until used for either EV characterization or lipidomic analysis.
Characterization of sEVs
The sEVs from the three samples of S-sEVs and the three samples of L-sEVs were characterized according to the minimal information for studies of extracellular vesicles 2018 (MISEV2018) guidelines [17].For this, orthogonal characterization including total protein concentration, particle concentration, particle size distribution, sEV morphology, sEV protein-specific markers, and sEV purity was performed as described below.
Total protein concentration was measured using the bicinchoninic acid assay (Thermo Scientific™ Pierce Micro BCA, Waltham, MA, USA) according to the manufacturer's instructions.Each sEV sample was analyzed under two different conditions, lysed and unlysed.Lysis was performed by mixing (1:1, v:v) sEV samples with lysis solution (0.1% of Triton plus 0.1% of sodium dodecyl sulfate in fPBS) and the mixture was incubated at 37 • C for 30 min under gentle agitation (≈50 rpm).Absorbance was measured at a wavelength of 570 nm (PowerWave XS; BioTek Instruments, Winooski, VT, USA).Two technical replicates were analyzed for each of the six biological samples.Results are expressed as µg per mL.
Particle concentration was measured by NTA using a NanoSight LM10 (Malvern Instrument Ltd., Malvern, UK) equipped with a 405 nm laser and a complementary scientific semiconductor metal oxide camera.Measurements were analyzed using NTA software (version 3.3.;dev build 3.3.104)with min track length, max jump distance, defocus set to auto, and the detection threshold set to five.The camera level was set to 15 and five 30 second videos were recorded at a 30 fps.Two technical replicates were analyzed for each of the six biological samples.
Particle size distribution was measured by DLS analysis using a Zetasizer Nano ZS system (Malvern Panalytical, Malvern, UK) operating at 633 nm at 25 • C and recording the backscattered light at an angle of 173 • .Samples of 50 µL were loaded into 10 mm cuvettes.Light scattering was recorded for 150 s and 3 measurements were taken per sample.Dispersion Technology v.5.10 software (Malvern Panalytical) was used to convert the DLS signal intensity to size distribution.Particle diameter was determined from the maximum peak of the normal function.The intensity-based distribution was recalculated to volume and the results recorded as volume and intensity size distributions.Three technical replicates were analyzed for each of the six biological samples.
The morphology of S-sEVs and L-sEVs was evaluated by TEM using a JEOL JEM 1011 microscope (JEOL Ltd., Tokyo, Japan).Sample preparation and visualization were performed according to the protocol described by Théry et al. with slight modifications [75].Briefly, 10 µL of sample was fixed in 2% paraformaldehyde for 30 min and applied to carbon-coated copper grids for 15 min.Samples were washed with fPBS, fixed in 1% glutaraldehyde, and washed with distilled water.The samples were then counterstained with 1% uranyl acetate and immersed in 0.5% methylcellulose.After drying, the samples were imaged at 80 kV.
The expression of the specific protein markers CD63 and HSP90β and the purity of the sEV samples were analyzed by flow cytometry using a high-sensitive flow cytometer (CytoFLEX S, Beckman Coulter, Life Sciences Division Headquarters, Indianapolis, IN, USA) equipped with violet (405 nm), blue (488 nm), yellow (561 nm), and red (638 nm) lasers according to the technical procedure described by Barranco et al. [34].The accuracy of the flow cytometer for EV input and counting was confirmed using recombinant EVs expressing green fluorescent protein on their membrane surface (SAE0193, Merck).The optical setup was adjusted to use the side scatter (SSC) information from the 405 nm laser (violet SSC-A).The forward scatter (FSC) and violet SSC-A were set to a logarithmic scale and the fluorescence channels were set to a logarithmic gain.Samples were analyzed in low flow mode (10 µL/min) with a minimum of 10 4 events per sample.Distilled water was used as the sheath fluid, while 0.1 µm-fPBS was used to ensure removal of background noise.Analysis was adjusted for intact sEVs.For this, samples were incubated with CellTrace™ CFSE (C34554, Thermo Fisher Scientific) and only CFSE-positive particles were considered sEVs.
All relevant data from the experiments were submitted to the EV-TRACK database (EV-TRACK ID: EV240049) [77].
Lipidomic Analysis
Lipidomic analyses were performed by the Research Unit on Bioactive Molecules (RUBAM) of the Institute of Advanced Chemistry of Catalonia (IQAC, Barcelona, Spain), member of the Spanish National Research Council.Lipid extraction and lipid identification and quantification were performed as described by Simbari et al., with minor modifications [78].
Determination of Total Protein Concentration
Total protein concentration of S-sEVs and L-sEVs was determined by bicinchoninic acid assay (Thermo Scientific™ Pierce Micro BCA) using BSA as a standard according to the manufacturer's instructions.Samples were then frozen, lyophilized, and resuspended in 0.2 mL of MilliQ water.
Samples were vortexed and sonicated until they appeared dispersed and extracted at 48 • C overnight and cooled.Samples were then evaporated to dryness and stored at −80 • C until analysis.Prior to analysis, 150 µL of methanol was added to the samples, centrifuged at 13,000× g for 5 min and 130 µL of the supernatant was transferred to a new vial and injected.
Liquid Chromatography-High-Resolution Mass Spectrometry
Lipids were analyzed by liquid chromatography high-resolution mass spectrometry (LC-HRMS).LC-HRMS analysis was performed using an Acquity ultra-high-performance liquid chromatography (UHPLC) system (Waters Co, Wexford, Ireland) coupled to a timeof-flight detector (LCT Premier XE).Full scan spectra were acquired from 50 to 1800 Da, and individual spectra were summed to produce data points every 0.2 s.Mass accuracy at a resolution of 10,000 and reproducibility were maintained using an independent reference spray (leucine enkephalin) via the LockSpray interference.Lipid extracts were injected onto an Acquity UHPLC BEH C8 column (1.7 µm particle size, 100 mm × 2.1 mm, Waters Co) at a flow rate of 0.3 mL/min and a column temperature of 30 • C. The mobile phases were methanol with 2 mM ammonium formate and 0.2% formic acid (A)/water with 2 mM ammonium formate and 0.2% formic acid (v:v) (B).A linear gradient was programmed as follows: 0.0 min: 20% B; 3 min: 10% B; 6 min: 10% B; 15 min: 1% B; 18 min: 1% B; 20 min: 20% B; 22 min: 20% B. For HRMS analysis, the capillary voltage was set to 3.0 kV, the desolvation temperature was set to 350 • C, and the desolvation gas flow was set to 600 L/h.
Positive identification of compounds was based on the accurate mass measurement with an error <5 ppm and their LC retention time, compared to that of a standard (92%).Quantification was performed on the extracted ion chromatogram of each compound, using 50 mDa windows.Linear dynamic range was determined by injecting mixtures of internal and natural standards as described above.In the absence of authentic standards, identification was made by accurate mass measurement, elemental composition, calculated error, double bond equivalents, and retention time.The relative amount of each lipid species was measured as pmol equivalents to available standards, normalized to protein content and expressed as pmol per mg protein.The LIPID-MAPS consortium lipid nomenclature is used in this paper [79,80].
Statistical Analysis
Statistical analysis was performed using GraphPad Prism version 10.1.0for Windows (GraphPad Software, San Diego, CA, USA).The normal distribution of the data was assessed using the Shapiro-Wilks test.Data were analyzed by paired t-test or one-way ANOVA followed by Tukey's multiple comparison test for comparisons between two or more groups, respectively.The results of the sEV characterization are presented as the mean ± SD, unless otherwise noted.Statistical differences were considered when p values were <0.05.
Conclusions
In conclusion, porcine seminal EVs have a complex lipid composition, including 157 lipid species belonging to 14 different lipid classes.Most of the identified lipids were structural to cell membranes.However, some diacylglycerol and triacylglycerol species were also identified, which may originate from circulating free lipoproteins in seminal plasma bound to the sEV membrane.The two subsets of sEVs of different sizes showed qualitative and quantitative differences in lipid profile.The most notable quantitative differences were in sphingomyelins and ceramides, with the former in higher proportions in small sEVs and the latter in large sEVs, differences that may reveal different biogenepathways.
Figure 1 .
Figure 1.Phenotypic characterization of large (L) and small (S) porcine seminal extracellular vesicles (sEVs).Box plots showing (A) total protein concentration and (B) particle concentration measured by nanoparticle tracking analysis.(C) Particle size distribution measured by dynamic light scattering analysis (blue line indicates S-sEVs and red line indicates L-sEVs).(D) Representative transmission electron microscopy images showing the morphology of sEVs.**** p < 0.0001.Box plots: Boxes enclose the 25th and 75th percentiles, whiskers extend to the 5th and 95th percentiles, and line represents median.
Figure 1 .
Figure 1.Phenotypic characterization of large (L) and small (S) porcine seminal extracellular vesicles (sEVs).Box plots showing (A) total protein concentration and (B) particle concentration measured by nanoparticle tracking analysis.(C) Particle size distribution measured by dynamic light scattering analysis (blue line indicates S-sEVs and red line indicates L-sEVs).(D) Representative transmission electron microscopy images showing the morphology of sEVs.**** p < 0.0001.Box plots: Boxes enclose the 25th and 75th percentiles, whiskers extend to the 5th and 95th percentiles, and line represents median. | 2024-07-10T15:20:34.195Z | 2024-07-01T00:00:00.000 | {
"year": 2024,
"sha1": "8583845a362221115f0e8adac2a35509a8d03142",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "5ad1b50b02b896eadda06fc0169e5ea6b8fa4133",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
263939270 | pes2o/s2orc | v3-fos-license | A lab-on-chip platform for simultaneous culture and electrochemical detection of bacteria
Summary A simple, cost-effective and miniaturized lab-on-a-chip platform has been developed amenable to perform simultaneous cultivation and detection of bacteria. A microfluidic chamber was integrated to screen-printed electrodes for electrochemical detection of bacteria. The temperature required for the bacterial culture was provided through the optimized laser-induced graphene heaters. The concentration of bacteria was quantified accurately with the three-electrode system in the range of 2 × 104 to 1.1 × 109 CFU/mL without any need of biological modifications to the electrodes. The viability of cultured bacteria in the microfluidic device was also confirmed through fluorescent imaging. Furthermore, the metabolic activity of the cultured bacteria was validated through a miniaturized microbial fuel cell. Furthermore, the specificity of electrodes was also performed through electrochemical technique. Finally, a handheld and portable lab-on-a-chip platform was realized by 3D packaging, integrated with a portable potentiostat for real-time and on-field applications.
INTRODUCTION
sensing in the range of 4.2 3 10 2 CFU/mL to 4.2 3 10 5 CFU/mL (Li et al., 2012). In another work, electrochemical technique was implemented to detect E. coli using oxidation reaction of nickel on a rotating disc type electrode. It was reported that the sensor was capable of rapid detection of bacteria in the order of 10 4 CFU/mL (Ramanujam et al., 2021). Silver nanoparticles (AgNP) were also utilized for direct detection of E. coli. The established phenomenon that adherence of AgNP to bacterial cells results in increase of current when maintained at a suitable potential was utilized to detect the concentration of bacteria as a proof of concept (Sepunaru et al., 2015). Alternately, gold electroplated interdigitated electrodes were modified with anti-E. coli antibodies and were utilized for sensing E. coli through electrochemical impedance spectroscopy (EIS) (Ghosh Dastider et al., 2018). A similar work, where gold electrodes were immobilized with E. coli using antibody-binding method was reported to detect bacteria of concentration up to 10 5 CFU/mL (Xu et al., 2020).
It is evident from the available literature and recent attempts that in most of the work attempted at microfluidic platform, electrochemical sensing approach was utilized for bacterial samples. However, for detection in the above reported works, the samples were already cultured in a conventional platform to the required concentration level and then electrochemical analysis was performed on a bulk platform. This leads to a lack of the understanding of the growth of the bacterial concentration with time. Hence, there is a need for a simple, low-cost, reliable, effective, and a handy device that is capable of performing incubation and simultaneous detection of the growth of bacteria instantaneously. In order to achieve this, present work is focused in the direction to realize how to miniaturize the conventional incubation and detection platforms into a single handheld device. Earlier reported works (Zhou et al., 2019) suggest that culture of bacteria can be effectively done in (1) microchannels (Balagaddé et al., 2005;Chen et al., 2010), (2) microchambers (Held et al., 2010;Wu et al., 2015), and (3) microdroplets (Ho et al., 2020;Kaushik et al., 2017;Watterson et al., 2020). In addition, it was also reported that the reduced sample volumes used in microfluidic devices minimize the time taken for bacteria culture compared with the conventional incubator (Kaushik et al., 2017). Although bacterial culture was implemented in microfluidic channels, simultaneous detection of the bacterial growth was not monitored through electrochemical detection techniques, to the best of our knowledge. iScience Article Motivated with such gaps and requirements, in this work, screen-printed carbon electrodes were fabricated and integrated to a microfluidic chamber. The integrated device (shown in Figure 1) is provided with the required temperature for bacterial growth and the timely growth of the bacteria was recorded electrochemically. In addition, the viability of the cultured bacteria was validated through fluorescence imaging. Furthermore, the metabolic activity of the cultured bacteria was corroborated with an enzymatic biofuel cell. Finally, a handheld device was realized and integrated with a portable potentiostat, enabling the usage of the device for on-field applications.
Electrochemical detection of bacteria
In order to comprehend the growth of bacteria in the microfluidic device, two types of electrochemical detection techniques were performed: (1) cyclic voltammetry and (2) chrono amperometry. The peak currents in both the techniques were observed, and calibration plots were plotted through which the growth of bacteria is related in terms of current obtained in the detection technique due to the enzymatic activity of E. coli. In order to achieve this, 7 samples of known concentrations were chosen (2 3 10 4 -1.1 3 10 9 CFU/ mL). In brief, one sample of bacteria with 2 3 10 4 CFU/mL was infused into the microfluidic device and another sample of same concentration was kept in a conventional incubator with orbital shaker at 36 C. Simultaneously, the temperature was continuously provided for the microfluidic device through the LIG heater. Later, at concentration of 2 3 10 4 CFU/mL, electrochemical analysis was performed for both the samples individually in two different devices. The time for the first reading, i.e., for concentration of 2 3 10 4 CFU/mL, was set to be taken at zero hours and the readings were continued up to 12 h with an interval of 2 h. After electrochemical analysis, the sample in the device was allowed to culture in the device itself, whereas another sample was kept in the incubator for culturing in conventional way for further timebased analysis of bacterial growth in the two cases. After 2 h, electrochemical measurements were recorded for both the samples, i.e., sample incubated in conventional incubator and sample incubated in the device. In a similar fashion, experiments were repeated for a period of 12 h. After 12 h, no significant changes in the current in either of the electrochemical techniques were identified and hence within 12 h, 7 different concentrations were recorded.
Cyclic voltammetry for detection of bacteria
In general, among all electrochemical techniques cyclic voltammetry gives a clear comprehension of transfer rate of electrons with the electrodes for a given analyte (Elgrishi et al., 2018). Few earlier works reported iScience Article the implementation of CV for detection of bacterial concentration (Sun et al., 2019). In this context, initially cyclic voltammetry was performed to investigate the effect of growth in bacteria with time on the voltammograms. Cyclic voltammetry was performed for a period of 12 h in an interval of 2 h. Because the bacterial culture in a conventional incubator is a well-established protocol, initially, CV was carried out for the samples incubated conventionally, and these voltammograms were considered for comparison with that of the graphs obtained for sample in the microfluidic device. Figure 2 shows voltammograms performed for different samples at different time intervals. Figure 2A represents the cyclic voltammetric curve for bacteria sample cultured in the conventional way, and Figure 2B represents the curves obtained for culture performed in the microfluidic device. As can be seen from the Figure 2A, an oxidation peak current was observed at a potential of 0.8 V, and no reduction peak was observed. This potential of 0.8 V specifically corresponds to the peak of LB media as shown in Figure S2. With time and continuous provision of optimal temperature, it was very obvious that the concentration of the bacteria would be multiplying both in the microfluidic device or in the incubator. It can also be observed from the figure that the peak potential is gradually reducing with time similar to the works reported in literature (Sun et al., 2019).
This implies that the peak of the LB is gradually decreasing, which must be possible only with the increase in the bacterial concentration in the media. The increase in concentration of bacteria reduces the amount of LB media, which corresponds to reduction of peak in the cyclic voltammogram. The similar trend can be seen in Figure 2B, for sample cultured in microfluidic device, which goes in line with the voltammograms obtained for conventionally cultured bacteria. In addition, the peaks obtained in Figure 2B are more distinguishable compared with that of Figure 2A. This can be related to the gradual increase of bacterial concentration in the microfluidic device. However, in order to comprehend the growth rate of bacterial concentration, a calibration plot has been considered for concentration against the observed peak currents from the CV graphs. It can be seen from Figure In addition, a comparison plot ( Figure 3B) was considered to compare the conventionally incubated sample versus the microfluidic device wherein a linear fit was observed with a regression coefficient of 0.99. This suggests that the culture of bacteria on a microfluidic device is equally competent to that of the culture in a conventional incubator.
Chronoamperometry for identification of bacterial growth
In addition to cyclic voltammetry, chronoamperometry technique was also implemented to understand the nature of curves with increase in concentration of bacteria in conventional as well is in microfluidic device. In general, chronoamperometry provides the measure of current with respect to the time at a fixed potential. As per the CV graphs, because the peak was distinguished at around 0.8 V, the same potential was fixed for CA analysis. Each experiment was run at a fixed potential of 0.8 V for a span of 120 s. As can be seen from Figures 4A and 4B, there is a consistent decrease in the current with the increase of concentration similar to iScience Article that was observed in the case of CV graphs. It can also be noted that after an initial decrement, the current stabilizes, and no further change in the current was observed. This suggests that a constant current is showcased for individual concentrations. Furthermore, the decrease in current values is observed in both the cases, i.e., conventionally cultures and that in the microfluidic device. This nature of curve suggests that the current gradually reduces with the increase in concentration of bacteria and this goes in line with the works reported earlier (Ramanujam et al., 2021).
In addition, a calibration plot ( Figure 5A) for concentration against current was plotted similar to the calibration plot of CV, whereby very minute changes in the current values were observed for both the cases. This calibration plot also helps in identifying the concentration of an unknown sample based on the current value within the limits of linear range 2 3 10 4 to 1.1 3 10 9 CFU/mL. The linear relationship followed the Equation 3 Besides the calibration plot, a comparative graph ( Figure 5B) for currents obtained with conventionally cultured bacteria against the current obtained with the microfluidic device was plotted, which exhibited a linear trend with a regression coefficient of 0.97. This imply that the culture performed in our microfluidic device is comparable with the one performed in a conventional apparatus.
Fluorescent imaging for confirmation of live/dead bacterial cells
With the curves obtained from CV and CA, it can be analyzed that due to enzymatic activity, transfer of electrons is taking place and proper oxidation peak is being identified. However, to confirm that the decrease in peaks is owing to the growth of the bacteria, fluorescent imaging was performed to distinguish if the bacteria is alive or dead after the culture period. In order to achieve this, live/dead viability kit was utilized to identify dead and live bacteria. In brief, the viability kit includes two stains: (1) SYTO 9 and (2) propidium iodide, which gets binded to the live and dead bacterial cells, respectively. After 12 h of incubation in the conventional incubator and microfluidic device, the samples were taken out and were mixed with the bacterial stains and are left out in a dark place for 15 min. After the short incubation period, the samples were examined under fluorescence microscopy for three different conditions: (1) conventionally cultured bacteria, (2) bacteria cultured in the microfluidic device, and (3) sample in microfluidic device at 50 C. Figures 6A and 6B correspond to the condition of sample cultured conventionally, i.e., for the sample that is cultured in the incubator at 36 C in the presence of orbital shaker where it can be observed that very few bacterial cells were dead after incubation.
After 12 h, there was actively populated bacteria with very less number of dead bacterial cells. Figures 6C and 6D represent the sample cultured in microfluidic device wherein the growth of bacteria is equivalent to that of the conventionally cultured. However, the number of dead bacteria is a bit higher when compared with that of sample cultured in incubator. In another condition, bacteria were left in the microfluidic device, iScience Article and a temperature of 50 C was constantly maintained. The fluorescent image of that sample showed that all the bacterial cells were dead ( Figure 6E). Thus, the fluorescent images clearly showcase that there is definite growth in the microfluidic channel and the detection of their growth through the screen-printed electrodes is truly reliable.
Corroboration for microbial fuel cell for metabolic activity
It is clearly evident from the fluorescent images that there was growth of bacteria with time in the microfluidic device and was simultaneously detected using a three-electrode system. However, besides the growth of bacteria, its metabolic activity is relatively important for its utilization for real-time applications. In this regard, to comprehend the metabolic activity of the cultured bacteria, a microbial fuel cell (MFC) was implemented. Earlier, few works reported the usage of bacteria as a fuel for running a fuel cell (Rahimnejad et al., 2015). So, as reported in one of our previous studies, a 3D-printed-paper-based microbial fuel cell was utilized to understand the effect of cultured bacteria over obtained power density . In brief, an anode and a cathode were fabricated using silver nano ink and carbon paste, respectively. A further enhancement of the cathode was done by modifying it with MnO 2 nanoparticles. After fabrication of the electrodes, silver ink was used to provide electrical contacts. The performance of the microbial fuel cell, with three different samples (1) plain LB media (2) bacteria cultured in conventional setup, and (3) bacteria cultured in microfluidic device were considered to comprehend the efficiency of the bacterial fuel. Initially, 100 mL of LB media was drop-casted on the anode, and the readings were recorded once a stable open circuit voltage (OCV) was attained. The same was repeated for the other two cases. After the stabilization, the polarization curves were plotted to obtain maximum power and current density in all the three cases ( Figure 7). In case of plain LB media, a maximum power density of 4 mW/cm 2 was recorded, whereas the sample with bacteria delivered higher power output. The cultured bacteria using conventional setup generated a power density of 61 mW/cm 2 and that of sample incubated in microfluidic device was found to be 43 mW/cm 2 . The reduced power density obtained from the sample cultured in microfluidic device can be attributed to the fact that more amount of dead bacterial cells was existing compared with the conventionally cultured bacteria as can be seen from Figure 6. However, these values of power densities confirm the successful growth of bacteria in the microfluidic device whose metabolism iScience Article has produced quite a good amount of power density in comparison to other MFC available (Rewatkar et al., 2021). This suggests that the metabolic activity of bacteria cultured in the microfluidic device was relatively close to that of the sample cultured in conventional incubator.
Realization of a handheld incubator cum detector for growth indication of bacteria
In order to realize a miniaturized and an all-in-one platform, a 3D printed packaging was designed and realized to integrate the microfluidic device with the heater and the detection equipment ( Figure 8). The voltage to the heater is supplied by a buck booster, and the electrochemical readings were taken using a handheld potentiostat (Sensit smart [PalmSens, Netherlands]) connected to a smartphone. Once the setup was ready, sample was infused into the microfluidic device, and the required voltage is supplied to the heater to provide sufficient temperature for the bacteria for incubation. Initial readings for plain LB media and the bacterial growth were shown in the inset of Figure 8.
Specificity
In order to perform the specificity test on the fabricated device, two different bacteria namely Shewanella and Streptococcus were utilized. Initially, a blank sample of the media (Luria Bertani) was run and the current values were noted. Later on, cultured bacteria (E. coli, Shewanella, and Streptococcus) with almost a iScience Article similar concentration of 2.1 3 10 9 CFU/mL was considered for determining the current values using chronoamperometry technique (Figure 9). An initial run for blank resulted in higher values of current. As bacteria were introduced, there was significant changes in the values of current. It was identified that three different values of currents were recorded for the three bacteria but however, a mixture of all the three resulted in a current values similar to that of the E. coli, which is in line with the reported work (Ramanujam et al., 2021). This suggests that the current values in the mixed solution was nearly identical to that in the E. coli-only solution, confirming the sensor's ability to detect signals from E. coli in particular. This work delves upon developing a miniaturized lab-on-a-chip platform, which was capable of simultaneously performing bacterial culture and detection of the bacterial concentration. Electrochemical techniques, such as cyclic voltammetry and chronoamperometry, were employed to investigate the growth of bacteria using screen printed electrodes. The device was capable of quantifying the bacterial growth within the range of 2 3 10 4 to 1.1 3 10 9 CFU/mL without any biological modifications to the electrodes. The viability and metabolic activity of the cultured bacteria were confirmed through fluorescence imaging and microbial fuel cell, respectively. 3D packaging was done to realize a handheld device integrated with a portable potentiostat expanding its usage in on field applications. In addition, specificity studies show that the device is capable to detect signals from E. coli in particular. These results pave a way to realize a miniaturized incubation cum detection platform that can be implemented for bacterial culture with greater advantages such as simple, reliable, quantifiable device with minimal sample volume, low cost and minimal power consumption. The future developments of this work would be to achieve lower detection limits and antibiotic susceptibility in microfluidic droplets in the integrated device. In addition, the device can also be modernized by integrating it with mobile-based fluorescence module for on chip viability indication. Furthermore, the integrated device can also be utilized for on-chip cell culture applications.
Limitations of the study
The viability of the culture from the device can be further improved by modifying the device to have a shaking module that replicates the condition of the conventional incubator.
STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
Lead contact
Any information and requests for resources should be addressed to and will be responded by the lead contact, Sanket Goel (sgoel@hyderbad.bits-pilani.ac.in).
Materials availability
Our study did not generate any new unique reagents.
Data and code availability
Data reported in this paper will be shared by the lead contact upon request.
Our study did not report any unpublished custom code, software, or algorithm.
Any additional information required to reanalyse the data reported in this paper is available from the lead contact upon request.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
This study does not use experimental methods typical in the life sciences.
Materials
Conductive carbon ink was purchased from Engineered Materials system, Inc. Ag/AgCl was procured from ALS Co. Ltd., Japan. Multi walled carbon nanotubes (MWCNT) in powder form was procured from Sigma Aldrich, India. Polyimide sheet was purchased from Dali Electronics, India. A CO 2 laser (VLS 3.60) was purchased from Universal Laser Systems, AZ, USA. Thermal camera was procured from FLUKE TECHNOLO-GIES PVT. LTD., India. Potassium ferricyanide was procured from AVRA chemicals. PDMS was purchased from Dow corning, USA. Luria Bertani was purchased from SRL chemicals.
Fabrication of screen printed electrodes
Screen printing technology has been chosen fabricate three electrode system over glass substrates. Initially, a mask of required dimensions has been prepared on a poly vinyl chloride (PVC) sheet using a commercial CO 2 laser. The mask was adhered to the glass substrate. Later, a conductive carbon ink was laid over the mask and squeegee was used to spread the carbon ink over the required design uniformly. The substrate was then placed in an oven at 65 C for 1 h for complete drying. After drying, the PVC sheet was removed leaving the three electrode system on the glass substrate. In order to perform the electrochemical analysis, one of the three electrodes, to work as a reference electrode, was modified with Ag/ AgCl ink and was kept in hot air oven at 65 C for a period of 15 min for drying. | 2022-10-21T15:07:42.188Z | 2022-10-01T00:00:00.000 | {
"year": 2022,
"sha1": "c55e4dae221f15507ad029c20bd5bd1d8abbd1de",
"oa_license": "CCBY",
"oa_url": "http://www.cell.com/article/S2589004222016601/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a0e7450dd615dfc5419175dc57c914c61cd85853",
"s2fieldsofstudy": [
"Materials Science",
"Engineering"
],
"extfieldsofstudy": [
"Medicine"
]
} |
56025942 | pes2o/s2orc | v3-fos-license | Can family risk-factors moderate the link between psychopathy and Life-History Strategy?
Life History Theory is an explanatory evolutionary framework which explains differences in fitness-relevant outcomes using the characteristics of the environment and individual organisms. Basically, individuals can be positioned somewhere on the r/K continuum of the Life History Strategy (LHS): a K or slow strategy represents later maturity and reproduction, a smaller number of offspring with higher investment in them, while the r (or fast) strategy follows the opposite pattern. Previous research offered evidence that psychopathy can represent a trait associated with fast LHS. In the present research we examined the relations between the family risk-factors, a four-factor model of psychopathy and the LHS in a sample of male convicts (N=181). The results have shown that a manipulative and deceitful interpersonal style is associated with slow LHS while shallow affect and antisocial tendencies are related to fast LHS. The interactions between psychopathy and family risk-factors revealed that parental criminal behaviour enhances the relation between fast LHS and psychopathic traits, including the manipulative interpersonal style. The findings are in accordance with the Life History Theory and provide a deeper understanding of the preservation of psychopathy in contemporary populations.
Key concepts of Life History Theory
In order to successfully adapt to the environment, organisms must make decisions on when and how to allocate energy and resources which are limited.Biological adaptation is a complex process, mostly associated with life-time reproduction success as the best indicator of fitness (Zietsch, Kuja-Halkola, Walum, & Verweij, 2014).However, in order to successfully reproduce and bear offspring, organisms must invest in somatic growth and maintenance, survival, gaining resources and other outcomes related to fitness.Since organisms have limited resources, energy and time, the decisions related to their investment are accompanied by trade-offs: every decision that has been made has some costs and benefits (e.g.investing in individual survival means that energy cannot be allocated to reproduction).The sum of choices regarding energy and resource allocation that individuals make in their life time depicts their Life History Strategy (LHS).The evolutionary conceptual framework which analyzes Life History Strategies, their causes, dynamics and outcomes is Life History Theory (Roff, 2002).
Most authors emphasize three crucial trade-offs in LHS and all of them are related to reproduction (Del Giudice, Gangestad, & Kaplan, 2016).The first comprises current and future reproduction: organisms may reproduce at an earlier age or delay reproduction in order to invest in growth and resources.The Second trade-off distinguishes between quality and quantity of offspring: the former one is related to lower number of offspring with higher investment in them while the latter is based on the opposite pattern.Finally, there is a trade-off between mating and parenting: after the reproduction, individual can invest energy in offspring or in the search for another mate.Based on these trade-offs, we can distinguish two basic strategies (Figueredo, Vasquez, Brumbach, Sefcek, Kirsner, & Jacobs, 2005).Slow or K strategy is related to a later age of reproduction and smaller number of offspring with high parental effort.Fast or r strategy is associated with early reproduction, followed by a large number of offspring and low parental investment.K and r are not two distinct strategies; they are opposite poles of a single dimension of the LHS.It should be noted that neither one of them is superior: both are equally successful depending on organisms' characteristics and environmental conditions.The human species is dominantly under the slow LHS; however, there are differences between individuals regarding their LHS (Figueredo, de Baca, & Woodley, 2013).
Two main factors influence what strategy would be developed by an individual.The first one is associated with the characteristics of the organism.LHS is partially heritable (Figueredo, Vásquez, Brumbach, & Schneider, 2004), which implies the existence of genes that influence the phenotypic development of LHS.The second factor is the environment.Harsh, unpredictable environments with scarce resources will facilitate fast LHS; in contrast, supportive, predictable environments with sufficient resources will enable the development of the slow strategy (Kogan, Cho, Simons, Allen, Beach, Simons, & Gibbons, 2015).The studies have shown that family characteristics are an important predictor of human LHS.Individuals who grew up in families with frequent conflicts, poor parent-child relationships or BETWEEN PSYCHOPATHY AND LIFEHISTORY STRATEGY?who experienced some form of maltreatment tended to develop the fast LHS (Rickard, Frankenhuis, & Nettle, 2014).
Psychopathy and Life History Theory
Psychopathy is a syndrome of behavioural traits consisting of the following characteristics: 1) manipulative and deceitful interpersonal style, 2) shallow and callous affect characterized by a lack of empathy, fear and guilt, 3) disinhibition and lack of impulse control and 4) antisocial and criminal behaviour (Hare, 2003).One of the most prominent models of psychopathy was proposed by Robert Hare and collaborators and in this model these four traits are labelled as Interpersonal, Affective, Lifestyle and Antisocial psychopathy features (Hare & Neumann, 2009).Psychopathy represents a construct that has been predominantly explored in the context of criminality and forensics because it is a reliable predictor of criminal behaviour and recidivism (Leistico, Salekin, DeCoster, & Rogers, 2008).In fact, psychopathy is related to various types of socially undesirable phenomena such as aggressiveness, violence and immorality (Međedović, 2015).It is considered to be the core of human "dark" personality traits (Međedović & Petrović, 2015).
Recently, psychopathy has been explored in an evolutionary context, too.One of the topics regarding its evolutionary status concerns its position on the LHS continuum.Since psychopathy is characterized by disinhibition (absence of planning and establishing long-lasting goals) and affective callousness, which indicates low potential for emotional investment, researchers assumed that it represents a trait constellation associated with fast LHS (Jonason, Webster, Schmitt, Li, & Crysel, 2012).Indeed, empirical findings confirmed this hypothesis, revealing that the key link between psychopathy and fast LHS is the association between psychopathy and short-term mating (Jonason, Koenig, & Tost, 2010).However, there is empirical data that did not confirm this relationship.When indicators of LHS, psychopathy, risk taking and mating effort were subjected to factor analysis, the LHS markers constituted a latent factor which was distinct from psychopathy (Gladden, Figueredo, & Jacobs, 2009).The explanation for these unequivocal results was offered by analyzing the scores on narrow psychopathy traits.When subordinate traits were analyzed, it was shown that impulsive and antisocial psychopathy traits were indicators of fast LHS while manipulative and affective characteristics represented slow LHS (McDonald, Donnellan, & Navarrete, 2012).
Goals of the present research
The last described research highlighted the importance of taking into account the narrow traits of psychopathy.In the present research, the four-factor model of psychopathy (Hare, 2003) is explored and the scores of all four traits are used in the analysis.LHS is operationalized by the age of the first sexual conduct and the indicators of unrestricted sexuality.The onset of sexual activity is considered as one of the crucial indicators of LHS (Carlson, Mendle, & Harden, 2014), while unrestricted sexuality was used as measure of fast LHS in previous research as well (Eisenberg, Campbell, MacKillop, Modi, Dang, Koji Lum, & Wilson, 2007).Finally, it is important to notice that all of the earlier studies on the link between psychopathy and LHS failed to include environmental factors in the design.Since the environment is very important for Life History Theory, and among environmental factors family characteristics are considered as central, we also included detrimental family indicators in the analysis.As far as we are aware, there is no study that has analysed these constructs in such a design so far.
Several hypotheses can be derived from the previous data and theory.When psychopathy is considered, two competing hypotheses can be postulated: 1) all of the psychopathic traits are related to fast LHS and 2) Interpersonal and Affective characteristics are associated with slow LHS, while Lifestyle and Antisocial features are linked to fast LHS.The third hypothesis involves the environmental factors: 3) detrimental and depriving family characteristics are enhancing positive associations between psychopathy and fast LHS.
Sample
The sample consisted of 181 male participants.All subjects participated on a voluntary basis.Verbal informed consent was collected from all participants.Subjects were prisoners serving their sentence in two penal institutions in Serbia.Mean age of participants was 35.7 years (SD=10.3).On the average, research participants finished secondary school (43%), a smaller number of subjects had only primary education or lower (31%) while the rest had higher education (26%).Participants were convicted for various crimes, including the violent and non-violent offences.All participants had elementary reading skills.
Measures
The PCL-R (Hare, 2003) is used to explore psychopathy.It is based on a rating method which involves a structured interview and extracting data from the participants' prisoner files.The raters were psychologists from the Institute of Criminological and Sociological Research (four researchers, including the author of this report).The raters were mostly balanced in BETWEEN PSYCHOPATHY AND LIFEHISTORY STRATEGY?their evaluation of psychopathy (for details see: Međedović, 2015).Four psychopathy traits were measured via 18 items (Hare & Neumann, 2009): Interpersonal (α=.70),Affective (α=.64),Lifestyle (α=.73) and Antisocial (α=.74) psychopathic features.The rating scale on PCL-R items had 3 responding scores: 0 -the participant does not exhibit rated traits/behaviours or has the opposite characteristics, 1 -the participant does exhibit rated traits or behaviours in some degree, and 2 -the participant exhibits rated traits or behaviours in a high degree.
Family risk-factors were operationalized as negative characteristics of family functioning: substance abuse, maltreatment and presence of criminal behaviour in participants' families.These data were collected from the participants' prison files.Every file contained various pieces of information about the convict, collected from the Court, social work services or prison staff.The variables were coded as binary measures, where 0 represented the absence of the examined factor in the family, while 1 represented the presence of the examined factor.
The Life History Strategy was measured by several indicators.Participants were asked to provide information about the age when they had had their first sexual conduct, to estimate the total number of sexual partners that they had had and the number of occasions in which they had cheated on their partner.The last indicator concerned the situations in which they had two or more parallel partner relationships.The specific numbers that participants provided were taken.
The interviews were held with each participant individually, with the average lasting time of an hour and fifteen minutes.After the interview, the researchers obtained the relevant information from prisoners' files.
Prediction of the fast LHS by psychopathy and family-risk factors
Before we explored the relations between psychopathy, family riskfactors and LHS, we had conducted a Principal Component Analysis (PCA) on the indicators of sexual behaviour.If all of the indicators measure single construct (in this case LHS), then only one latent component must be extracted.Indeed, the results of the PCA revealed one latent dimension with an Eigenvalue of 1.62, which explained 40.37% of original indicators variance.Multiple partner relations had the highest loading on the extracted component (.80), followed by infidelity (.79), total number of sexual partners (.44) and the age of first sexual encounter (-.40).It is clear that the extracted factor depicts the fast LHS.The score of the respondents on this component was saved in the database and set as a criterion measure in the regression.Psychopathy traits and family risk-factors were set as predictors.The age of participants was also included in the models because its variance needed to be controlled in the analysis.A significant regression function was obtained (R2=.15;F=3.01; df=8; p<.01).The results of the regression analysis are presented in Table 1.Three psychopathy traits have had independent contribution to the prediction of the fast LHS.Antisocial and Affective traits are positively associated with the criterion while the contribution of Interpersonal style is negative.The effect sizes for all three predictors are similar.Lifestyle characteristics are positively related to the criterion measure.However, when all other variables are controlled, this relation drops to zero.Neither of family-risk factors has a significant contribution to the prediction.
Family risk-factors as moderators of the relation between psychopathy and LHS
In order to evaluate the possible moderating role of the environment on the link between psychopathy and fast LHS we analyzed the interactions between psychopathy and family characteristics.Psychopathy measures were centred in zero and interactions were calculated as the products of the centred measures and binary scores of family risk-factors.Afterwards, interaction variables were entered in a new block in the regression analysis.Three statistically significant interactions were found.In the next step, psychopathy traits were dichotomized using the median score, in order to obtain graphical representations of interactions.Finally, the two-factor ANOVA SPSS procedure was used to produce graphs.The interactions obtained in the prediction of the fast LHS are presented in the following graphs.
BETWEEN PSYCHOPATHY AND LIFEHISTORY STRATEGY? Graph 1. Interactions between parental criminal behaviour and psychopathy traits in the prediction of fast LHS
The first detected interaction (left chart) reveals the moderation which parents' criminal behaviour has on the link between psychopathic Interpersonal style and fast LHS (β=.28, p<.01; ΔR2=.06).In the families where criminal behaviour of the parents was absent, the highest LHS scores were found in the persons with low Interpersonal style.However, in the families where this risk-factor was present, the individuals with more pronounced Interpersonal features had the fastest LHS.The second interaction (the middle chart) shows that the highest LHS scores were detected in the individuals who grew up in families where criminal behaviour was present and had highly pronounced Affective psychopathic traits (β=.18, p<.05; ΔR2=.02).Criminal behaviour of the parents was present in the third interaction, too (the right chart).Similar to the former, the participants who grew up in the families where criminal behaviour was present and had high scores on Antisocial tendencies had the fastest LHS (β=.25, p<.01; ΔR2=.05).
Psychopathy traits as predictors of fast LHS
Three psychopathic traits have a positive relation with fast LHS: Lifestyle, Affective and Antisocial features.The independent contribution to the prediction of the latter two is significant.The present data is in accordance with earlier findings.Previous research showed consistent relations between psychopathic lifestyle, antisocial behaviour and fast LHS (McDonald et al., 2012).We can plausibly assume that proximate mechanisms of this relation can be reflected in the facilitated excitation and sensation seeking towards sexual stimuli (Kastner, & Sellbom, 2012).The relation between these psychopathy traits and fast LHS is probably based on impulsivity, because this trait is a consistent and reliable predictor of unrestricted sexuality (Kahn, Kaplowitz, Goodman, & Emans, 2002).The link between Affective psychopathic characteristics and fast LHS was also expected because shallow and callous affect does not enable emotional bonding which characterizes the long-lasting partner relations.Furthermore, this trait indicates dysfunctional partner relations characterized by violence, which is not impulsive but premeditated in its nature (Stanford, Houston, & Baldridge, 2008).When all of these findings are taken into account it is not surprising that psychopathy not only facilitates fast LHS but also represents a disposition towards sexuallyrelated criminal behaviour (Hawes, Boccaccini, & Murrie, 2013).
The only trait that deviates from the described pattern is Interpersonal style.It is associated with slow LHS.This finding partially corroborates the previous findings of McDonald and colleagues (2012).The authors explained the link between the Interpersonal features and slow LHS in the following manner: this trait is characterized by high confidence, dominance and self-esteem.These features are more prominent in slow LHS because they enable acquirement of resources and social status, which in turn delays reproduction.The findings that these traits correlate positively with achievement motives and dispositions (Ross & Rausch, 2001) further corroborate this line of reasoning.
Detrimental family factors enhance the link between psychopathy and fast LHS
Although we can agree with the argumentation that Interpersonal features are indicators of slow LHS, the obtained interactions revealed that psychopathy traits are associated with fast LHS if detrimental factors are present in participants' families.This applies to Interpersonal psychopathy features, too.When risk-factors are absent, Interpersonal style is negatively associated with fast LHS; however, in participants with parental criminal behaviour this association changes direction.All of the detected interactions have confirmed the basic assumption of Life History Theory: scarce, unpredictable and stressful environments enable the development of fast LHS.A manipulative and deceitful interpersonal style can be related both to the fast and slow LHS, depending on the presence of parental criminal behaviour.Shallow and callous affect is basically related to the fast LHS, but this relation is enhanced in criminogenic families.The same can be said for Antisocial and criminal behaviour.The findings from the present study join the large amount of data which show that various detrimental family factors serve as environmental triggers for the development of the fast LHS (Carlson, Mendle, & Harden, 2014;Rickard et al., 2014).
It is important to underline that of all family risk-factors examined in the present study, criminal behaviour of the parents proved to be the most BETWEEN PSYCHOPATHY AND LIFEHISTORY STRATEGY?powerful moderator of the association between psychopathy and the fast LHS.Previous data also revealed the links between criminal behaviour and the fast LHS (Yao, Långström, Temrin, & Walum, 2014), especially in lifecourse persistent offending (Boutwell, Barnes, Deaton, & Beaver, 2013).Since LHS is in part heritable (Figueredo et al., 2004) it does not surprise that individuals with parental criminal behaviour exhibit fast LHS.However, the data obtained in the present research suggest that criminal behaviour within the family does not predict fast LHS directly but indirectly, by interacting with the psychopathic characteristics of individuals.
Concluding remarks
Present data corroborate and deepen the existing knowledge about the connection between psychopathy and LHS.Evidence suggests that psychopathy is a fast Life History trait.It seems that a manipulative interpersonal style is related to slow LHS, thus confirming the possibility that narrow psychopathy traits can represent different Life History Strategies (McDonald et al., 2012), which is in accordance with the first proposed hypothesis.However, when detrimental family conditions are present, even this trait facilitates the development of fast LHS.This result shows that all psychopathy traits could be associated with fast LHS, at least under certain conditions, which corroborates the second hypothesis.These findings suggest that the first and second hypotheses are in fact not mutually exclusive.Both of them can describe empirical data adequately, depending on the environmental factors.Detected interactions confirmed the third hypothesis: the presence of risk-factors in the family enhances the positive connection between psychopathy and fast LHS.
The findings from the present research have contributed to our understanding of how psychopathy is maintained in contemporary human populations.Individuals with pronounced psychopathy have a distinctive sexual strategy: the one which is based on earlier onset of sexual activity, frequent changing of partners and low investment in the relationship.This strategy could lead to earlier reproduction -a hypothesis which has recently been empirically confirmed (Međedović, Petrović, Želeskov-Đorić, & Savić, 2015).These findings are not only of theoretical significance, but also deepen our understanding of psychopathy in an evolutionary context and contribute to the knowledge of why psychopathic individuals engage in sexually-related crimes and partner violence: it may be the result of the dysfunctional aspects of fast LHS.
The present research has several limitations.The age of the first sexual conduct and unrestricted sexual behaviour are only some of many operationalizations of the fast LHS.Obtained findings should be replicated using other measures of LHS.Family risk-factors in the present research were measured by binary indicators, which did not allow for a more precise distinction between them (e.g.physical or verbal maltreatment; alcohol or drug abuse, etc).Measuring family-factors in a more precise manner could provide more detailed data of the link between family dysfunction, psychopathy and LHS.Future research should pay attention not only to family factors but to other indicators of a depriving, uncertain environment which could interact with psychopathy in the prediction of LHS (e.g.poverty, social exclusion etc).Finally, the present data were obtained on the convicts' sample, which limits their generalizability.The examination of the relations between the environment, psychopathy and LHS in the general population would be fruitful.
Table 1 .
Prediction of fitness measures by psychopathy and family risk-factors | 2018-12-07T08:02:16.279Z | 2016-01-01T00:00:00.000 | {
"year": 2016,
"sha1": "0ff2fd89cf0e18bfb948c3d17cb6a98f1c0d969d",
"oa_license": "CCBYSA",
"oa_url": "https://scindeks-clanci.ceon.rs/data/pdf/0352-7379/2016/0352-73791601023M.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "0ff2fd89cf0e18bfb948c3d17cb6a98f1c0d969d",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Psychology"
]
} |
237977852 | pes2o/s2orc | v3-fos-license | A REVIEW ON IMPACT OF CLIMATE CHANGE ON FOOD INSECURITY ON AGRICULTURE
Food Insecurity is one of the major problems in the present situation and also it is continuing from the past several generations. Impact of change in climate is the main cause for this condition. Insecurity is due to several factors such as drought, Heavy Rainfall etc. Global warming increases the temperature, mainly during night hours. It is harmful/damaging to grain filling stage which eventually leads to loss in weight of the grain of many crops. Climate change is the continual adaptation of temperature and exemplary weather patterns in a place, and also it may cause weather patterns to be less certain. Change in the climate with raised/hoisted atmospheric Carbon-di-oxide concentration and temperature sometimes known for impact of plant photosynthesis. studies have thought-about impacts on cropping systems for situations wherever world mean temperatures increase by 4°celsius or additional. For five periods within the close to term and future, information (n=1090) ar planned within the 20-year amount on the horizontal axis that features the point of every future projection amount. Changes in crop yields are relative to late-20th-century levels. Information for every timeframe adds to 100%. © IPCC, 2014, fifth
INTRODUCTION
Agriculture sector in India is vulnerable to climate change. Green house gases are a major source of Climate change which contribute to the greenhouse effect (Sutherst, R.W,1996). However, the changes in climate have far reaching the impact on agricultural production, which are feasible to challenge food security in the forthcoming (IPCC Chapter 5, 2010). Change in the Climate may have harmful effects on irrigated crop yields over agro-ecological regions both because of the rise in temperature and change in water availability (Gurdeep Singh Malhi et. Al, 2020). Climate change is a transformation in analytical properties of the climate system that exists over a period of time or indefinitely-conventionally at least 30 years, it is the continual adaptation of temperature and exemplary weather patterns in a place, and also it may cause weather patterns to be less certain. These unpredicted weather patterns can make it challenging to sustain and grow crops in regions that depend on farming because the expected temperature and rainfall level may no longer be depended on and also it has also been associated with other destructing weather (5) Declining Arctic ocean ice: each the extent and thickness of Arctic ocean ice has lessened quickly over several decades.
(6) Glacial retreat: Glaciers are withdrawing nearly all over round the worldtogether with within the Alps, Himalayas, Andes, Rockies, Last Frontier and continent.
(7) Ocean acidification: From the start of the economic Revolution, acidity of the surface ocean waters has been magnified by nearly half-hour. The number of carbonic acid gas absorbed by the higher layer of the oceans is increasing by regarding 2 billion tons per annum. Food security is one of the main distress associated with the climate change. Changes in the climate have an effect on the food security in complicated ways. It shows an impact on the crops, livestock, forestry, fisheries and aquaculture, and can cause serious social and economic consequences in the form of reduced incomes, eroded livelihoods, trade disruption and adverse health impacts (Moors E. et al., 2013). However, it is important to note that the net impact of climate change depends not only on the extent of the climatic shock but also on the substantive vulnerabilities (FAO, 2016). According to the FAO (2016), both the biophysical and the social vulnerabilities decide the net impact of the climate change on food security.
Much of the literature on the impact of change in the climate on food security, although, has focused on just one aspect of the food security, i.e., food production. Climate change presents an additional stress on India's long-term food security challenges as it impacts the food production in several ways. For one, it may cause compelling increases in inter-annual and intra-seasonal variability of the monsoon rainfall. According to World Bank estimates, based on the International Energy Agency's current policy scenario and other energy sector economic models, for a global mean warming of 4°Celsius, there will be a 10% increase in annual mean monsoon intensity and a 15% increase in year-to-year variability in the monsoon precipitation. Indian agriculture, and thereby India's food production, is highly liable to climate change largely because the sector continues to be highly sensitive to the variability in monsoon. Ultimately, about 65% of the India's cropped area is under rain-fed. Lobell et.al (2012) found that wheat growth in Northern India is highly sensitive to temperatures greater than 34°C. The IPCC report of 2007 sounded similar concerns on wheat yield: a 0.5°celsius rise in the winter temperature is prone to reduce the wheat yield by 0.45 tons per hectare in India.
Impact of the worldwide temperature change on Food Security:
"Food security exists once all individuals, within the slightest degree times, have physical and economic access to decent, safe and alimentary food that meets their dietary wants and food preferences for an energetic and healthy life". (World Food Summit, 1996) Food security is one in every of the foremost distress associated with the worldwide global climate change.
Amendment within the climate has an effect on the food security in difficult ways during which. It shows an impact on the crops, livestock, forestry, fisheries and cultivation, and should cause serious social and economic consequences within the kind of reduced incomes, worn livelihoods, trade disruption and adverse health impacts. However, it is necessary to note that net impact of worldwide temperature change depends not solely on the extent of the climatically shock however conjointly on the substantive vulnerabilities. In keeping with the global organization agency (2016), each the biophysical and thus the social vulnerabilities decide net impact of the worldwide global climate change on food security. Much of the literature on the impact of amendment within the climate on food security, although, has centered on only 1 side of the food security, i.e., food production. Global temperature change presents an additional stress on India's semi permanent food security challenges because it impacts the food production in some ways within which.
For one, it should cause compelling will increase in inter-annual and intra-seasonal variability of the monsoon downfall. in step with International Bank for Reconstruction and Development estimates, supported the International Energy Agency's current policy situation and different energy sector economic models, for a worldwide mean warming of 4°Celsius, there will be a tenth increase in annual mean monsoon intensity and a V-day increase in year-to-year variability within the monsoon precipitation. Indian agriculture, and thereby India's food production, is extremely liable to global global climate change mostly as a results of the planet continues to be sensitive to the variability in monsoon. Ultimately, concerning sixty fifth of the India's cropped space is beneath rain-fed. Lobell et.al (2012) found that wheat growth in Northern Bharat is extremely sensitive to temperatures bigger than 34°C.
The IPCC report of 2007 measured similar considerations on wheat yield: a zero. 5° celsius rise within the winter temperature is susceptible to reduce the wheat yield by zero.45 tons per area unit in Bharat.
STRATEGIES TO MITIGATE IMPACT OF CLIMATE CHANGE IN AGRICULTURE AND FOOD SECURITY:
The different adaptation strategies being formulated and developed at individual, organizational, and institutional level to avert the negative impacts of the climate change are as under- | 2021-08-27T17:09:42.863Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "0d18eb8600c81ba7e11e9fbafce104f41fdef2e4",
"oa_license": null,
"oa_url": "https://doi.org/10.51193/ijaer.2021.7311",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "8b302df8f0852b8603464316309114258305f9d1",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Geography"
]
} |
209468323 | pes2o/s2orc | v3-fos-license | Hydroxyurea-induced Tongue Hypermelanosis and Transverse Melanonychia
Hydroxyurea (HU) is a commonly used medication for myeloproliferative neoplasm (MPN) and is usually well tolerated. Cutaneous toxicity of HU is well known and can be seen in several manifestations. We report a case of a man with MPL gene mutation associated with essential thrombocytosis, who had a rare mucocutaneous toxicity with diffuse tongue, skin and nail discoloration. Mucocutaneous toxicity is usually a benign condition and self-resolves after discontinuation of the medication. It can lead to patient anxiety and medication discontinuation. The mechanism for development of HU-induced mucocutaneous hyperpigmentation is poorly understood.
Introduction
Hydroxyurea (HU) is a well-tolerated oral chemotherapeutic drug frequently used in the treatment of myeloproliferative neoplasms (MPNs), including essential thrombocytosis (ET). HU exerts its antitumor activity by inhibiting the enzyme ribonucleotide reductase and thereby inhibiting DNA synthesis [1]. While HU is a well-tolerated agent, it is known to have skin toxicity such as skin ulceration, oral aphthosis and mucocutaneous hyperpigmentation. There have been case reports of nail hyperpigmentation and tongue hyperpigmentation, but having both is rare [2,3]. We report a unique case with development of diffuse tongue hyperpigmentation, nail hyperpigmentation in all nails, knuckle hyperpigmentation, and palmar and plantar hyperpigmentation soon after the initiation of HU for the treatment of ET.
Case Presentation
A 61-year-old African American man with past medical history of hypertension, diabetes mellitus and osteoarthritis was noted to have persistent isolated thrombocytosis on routine laboratory testing. He had no lymphadenopathy or hepatosplenomegaly. Laboratory testing showed a white blood cell count of 6.2 K/µL, hemoglobin 12 gm/dL, mean corpuscular volume 80, red cell distribution width 15 and platelet count of 1,197 K/µL. He underwent a bone marrow aspiration and biopsy which showed a hypercellularity for his age. Trilinear hematopoiesis was present with myeloid and erythroid maturation and increase in megakaryocytes noted, with some clustering. Molecular testing showed MPL exon 10 mutation was present; no mutation noted in BCR/ABL1, JAK2, and CALR; and a diagnosis of ET was made. The patient was started with HU 1,500 mg once daily and aspirin 81 mg. He tolerated the medications well and in three months his platelet count normalized. At that visit, the patient complained that he had developed diffuse tongue bluish/back discoloration ( Figure 1). He had also noted darkening of the skin over the knuckle and the palms and soles as well ( Figure 3).
FIGURE 3: Hyperpigmentation of the palmar and plantar skin.
The nails were not thickened and brittle, and the surrounding skin was normal. There was no swelling, warmth, tenderness or any other abnormalities. The patient was concerned about the cosmetic appearance and chose to discontinue HU. The platelet count went back up to 850 K/µL. He was started on anagrelide with normalization of the platelet count.
Discussion
ET is a BCR-ABL1 negative MPN associated with high platelet production leading to persistent thrombocytosis (platelet count ≥ 450 × 10 9 /L) and predisposition for vascular events [4]. Patients with ET have JAK2 mutation at 60%-65%, CALR mutation 20%-25%, MPL mutation 3%-5% or triple-negative 10% [5]. A study of 509 MPN patients with MPL mutation showed that 7% of them had a somatic mutation of MPL exon 10 [6]. A combination of HU and aspirin is the standard first-line treatment for high-risk and intermediate-risk ET [4]. The previous reports of HU-induced cutaneous toxicity are from older publications, and the driver mutations of ET are not well known. This is the first report of HU-induced cutaneous toxicity in an African American patient with ET with MPL mutation.
Several chemotherapeutic agents are known to cause mucocutaneous hyperpigmentation, most commonly cyclophosphamide, platinum agents and doxorubicin [7]. HU is a well-tolerated oral chemotherapeutic drug frequently used in the treatment of MPNs, including ET. It is an antimetabolite and exerts its antitumor activity by inhibiting the enzyme ribonucleotide reductase and thereby inhibiting DNA synthesis [1]. While HU is a well-tolerated agent, it is known to have skin toxicity such as skin ulceration, oral aphthosis and mucocutaneous dyschromia. A large review of HU-induced skin toxicity found non-ulcerative skin toxicity to occur in 1.6% of all patients [8]. It led to permanent discontinuation in 52% patients [8]. There have been two case series of patients with mucocutaneous hyperpigmentation [3,9]. This case is a unique example of one patient having palmoplantar, skin, nail and tongue (mucosal) discoloration.
The HU-induced nail discoloration may be transverse, longitudinal or diffuse, with longitudinal being the most common. Skin and palmoplantar discoloration is frequently seen with nail changes. The tongue discoloration is rare. The usual onset of discoloration is several months, and the temporal relation between start of HU and time of onset helps with diagnosis. While there is no known relationship between HU dose and hyperpigmentation, reports of tongue ulceration related to HU have been reported to occur with higher doses i.e., HU 3,000 mg per day [10]. A review of more than 3,000 patients with MPNs showed only less than 1% patients to have cutaneous dyschromia, and the median daily dose of HU was 1 gm at time of toxicity. The mean time on HU to toxicity was 60 months [11]. Interestingly, there was no difference in median dose in patients having skin ulceration, skin dyschromia or other mucosal adverse effects [11].
In a patient presenting with HU-related hyperpigmentation, it is important to rule out mimicking conditions such as subungal melanoma, especially in one nail involvement. Finger nails may have more hyperpigmentation then toenails due to slower growth of the nail, as was seen in our case as well. Discontinuation of HU frequently leads to slow improvement in the hyperpigmentation.
The mechanism of HU-induced hyperpigmentation is not clearly understood, but it is hypothesized to be due to activation of melanocytes by HU which leads to increased melanin | 2019-12-12T10:37:21.924Z | 2019-12-01T00:00:00.000 | {
"year": 2019,
"sha1": "32819b8c997a7dbc351160a8406bca86351e361a",
"oa_license": "CCBY",
"oa_url": "https://www.cureus.com/articles/24969-hydroxyurea-induced-tongue-hypermelanosis-and-transverse-melanonychia.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8a5e1f1f5240f2efe660b6a69c46e546962bf5c8",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
226455038 | pes2o/s2orc | v3-fos-license | Interaction of Central Bank Independence and Transparency: Bibliometric Analysis
This paper summarizes the arguments and counterarguments within the scientific discussion on the central bank independence and central bank transparency interaction. he main purpose of the research is to define the substantial relationships between central banks’ independence and transparency based on scientific research results. Systematization literary sources and approaches for the definition of the central bank's independence and transparency indicate that there is no generalized approach to the hierarchy of these concepts. Existing empirical studies have shown that the independence and transparency of the central bank can be interconnected components to increase the effectiveness and flexibility to solve tasks regarding the provision of currency stability and financial stability in general. Methodological tools of the research methods are Google Trends and VOSviewer instruments. The presented results of an empirical analysis of trends regarding the dynamics of the search query number by the keywords “central bank independence” and “central bank transparency” show a decrease from 2004 to the present, with a significant predominance of central bank independence over transparency throughout the study period. The results of bibliometric analysis of keyword coincidence while writing scientific articles show that researchers of the central bank independence very rarely link its solution with the transparency aspects study, while researchers of central bank transparency often study this problem in direct connection with their independence analysis. Along with this, the issues of transparency of central banks are often studied simultaneously with the analysis of their independence. The research empirically confirms and theoretically proves that the central bank’s independence is the primary category that creates the conditions for the study of its transparency. The study results provide a scientific ground to study the strength and direction of the relationship between the levels of transparency and independence of the central bank, identifying quantitative indicators to describe these relationships and can be useful for further research to ensure the central bank efficiency.
Keywords: Bibliometric Analysis, Central Bank, Coincidence, Google Trends, Independence, Interaction, Transparency, VoSviewer.
Introduction
Trends in democratization and informatization of society necessitate the growth of publicity of processes in all spheres of the state governance. Given the significant level of interconnectedness of economic processes in all fields of the national economy and the decisive role of the banking system in ensuring economic development, disclosure of economic information on banking regulation is a prerequisite for improving the financial planning of real sector entities, financial institutions and households long-term financial stability and a prerequisite for progressive economic development. Besides, increasing the level of publicity and transparency minimizes the possibility of political influence on the decisions of the central bank, which allows it to maintain the appropriate level of its de facto independence and form a positive perception in society. Thus, an essential area in the study of the effectiveness of banking regulation is to determine the role of independence and transparency of the central bank. Thus, the study of the essential and substantive relationships between the categories of "central bank independence" and "central bank transparency" becomes relevant. It enables to form a basis to substantiate the causal relationships between them.
Literature Review
Central bank development trends show a close relationship between their independence and transparency. Thus, some authors consider transparency as a component of central bank independence. Therefore, Fouad et al. (2019) proposed the composition of the elements of the de jure and de facto indices of independence of the central bank, which includes indicators of transparency and accountability. On the other hand, transparency is a condition for increasing the independence of the central bank and relates to an increase in its accountability level, which allows to increase the effectiveness of monetary policy (Dincer and Eichengreen, 2014).
On the other hand, scholar Šmídková (2013) considers three components of ensuring an effective institutional environment for central banks: independence, transparency and accountability. Each of these concepts has its measures, functional purpose and components. At the same time, there are closer links between transparency and accountability, the causal nature of which also remains unclear (Dumiter, 2014).
The need for comprehensive development of independence and transparency of central banks is also evidenced by the main types of these categories, identified by researchers. Thus, the central bank transparency index developed by Eijffinger and Geraats (2006) includes five main components: political transparency (characterizes the level of openness of central bank policy objectives); economic transparency (related to access to economic information used by the central bank for decision-making); procedural transparency (concerning the disclosure of the basic approaches and rules of monetary policy implementation); transparency of monetary policy (characterizes the disclosure of policy decisions, as well as forecast parameters); operational transparency (reflection of the level of results achieved during the implementation of the monetary policy of the central bank). This context indicates the need to identify a set of causal relationships that arise between the independence and transparency of central banks. Crowe and Meade (2007) argue that the independence and transparency of the central bank are important for effective governance and are complementary concepts. Thus, the growth of central bank independence is associated with lower inflation, while increasing transparency contributes to improving the quality of inflation forecasts. Other scholars analyze the independence and transparency of central banks as separate categories. Thus, independence is associated with the central bank's freedom to choose its own tactics and monetary policy tools. At the same time, transparency characterizes the central bank's willingness to take public responsibility for its decisions (Bernanke, 2010). However, there is a certain specificity regarding the connection of these categories with other aspects of the development of central banks. For example, the independence and transparency of the central bank are determinants of inflation in the country. At the same time, transparency is characterized by a close connection with the level of depth of the country's financial markets and the strength of political institutions, while for the independence of central banks, such a connection is not found.
In terms of determining the relationship between the categories of independence and transparency, one should pay attention to the recommended values for their level. Thus, most studies confirm that the independence of the central bank should reach the highest possible level. On the other hand, it is worth paying attention to the position of the authors who justify the need to determine the optimal level of transparency of the central bank, taking into account the fact that too high its level leads to information overload and may lead to growing uncertainty in the private sector policies that affect their own forecasts. An empirical study in the context of quality assurance of inflation forecasts at different levels of central bank transparency confirmed the authors' hypothesis and justified the need for an optimal level of transparency at which society has sufficient access to general information, while certain specific data that can be interpreted differently, and, accordingly, to determine the growth of uncertainty, should be kept in closed access (van der Cruijsen et al., 2010).
Lustenberger and Rossi (2017) confirm that the central bank's independence and transparency have a similar effect on the quality of macroeconomic and financial forecasts. Besides, empirically, the transparency of the central bank, along with its independence, has a stimulating effect in terms of achieving the central bank
Methodology and Results
At the first stage of the study, we will analyze the scientific interest in the categories of "central bank independence" and "central bank transparency", which has formed in recent years in the world community. We use the Google Trends toolkit, which estimates the number of search queries in the Google system as a whole in the world and by region. The results of this analysis are presented by Fig. 1. The total number of queries in the world during the maximum available period was selected to determine the generalized trends for analysis. From the data of the figure, we can conclude that there is a gradual reduction of scientific interest in the study of these two categories during the research period. The next step is to conduct a deeper analysis of the relationships between these categories that have emerged in preparing research in the world. We first analyze the scientific articles that were devoted to the study of central bank independence. The sample is made up of more than 700 articles indexed by the Scopus database, in which the words "central bank independence" are contained in the title, abstract or keywords of the paper. The tool of the analysis is the VOSviewer software, the application of which allows to determine the coincidence of keywords that occur simultaneously when writing articles related to the research of different aspects of central bank transparency. The results of the study are shown in Fig. 2. The research tools enable us to distinguish different categories, the closeness of which occurs due to the frequency of coincidence of each of the two concepts in the list of keywords in a particular study. Besides, the frequency of occurrence and the combination of individual keywords allows you to identify clusters that include concepts that are most often jointly studied by scientists. The size of the circle for each concept illustrates the frequency of its occurrence in the total sample. Thus, most often the work devoted to the study of central bank independence also concerns the investigation of monetary policy and inflation. At the same time, the analysis of the composition of clusters formed by groups of keywords is a subject of considerable scientific interest. At this stage of the study, eight clusters were formed, each of them includes from 8 to 3 words ( Table 1). The category "central bank independence" is included in cluster 3, which contains 7 keywords. It should be noted that the content of the concepts included in this cluster illustrate the goals that can be achieved directly or indirectly as a result of increasing central bank independence (governance, economic growth) or the conditions enhancing the role of central bank independence einsuring (globalization, global financial crisis). On the other hand, the category of "central bank transparency" was included in cluster 6, along with the concept of accountability, as well as categories that indicate the importance of ensuring the transparency of the central bank in the context of transformations (transitional economy, institutional reform). At the same time, such a division does not suggest that central bank transparency is a necessary aspect that must be explored when studying central bank independence.
On the other hand, there is a fairly close relationship between cluster 3 and cluster 2. Given the fact that cluster 2 includes categories which the central bank should take into account (price stability, inflation targeting, financial stability etc.) or confront (financial crisis), it is quite natural and shows that the independence of the central bank is not a target category, but a basic one, which creates conditions for increasing its efficiency in achieving the final goals of central bank functioning.
This stage of the study did not allow us to definitively substantiate the substantive relationship between the categories of independence and transparency of central banks. That is why we will conduct a similar study of the composition of keywords in articles focused on the study of transparency of central banks. The sample of scientific articles on this topic included in the Scopus database is about 150 publications. The results of the analysis of the composition of keywords using the VOSviewer toolkit are presented by Fig. 3. The results obtained at this stage show that the most common in research next to the category of "central bank transparency" are monetary policy and inflation, which are similar to previous results and are quite natural, given the basic conceptual framework of the central bank. At the same time, attention should be paid to the cluster structure of keywords identified in central bank transparency studies ( Table 2).
Unlike the previous stage, within this block, the transparency and independence of the central bank belong to the same cluster. Besides, this cluster includes the concepts of financial stability, information asymmetry and uncertainty, which indicates the maximum effectiveness of the interaction of independence and transparency of the central bank in combating financial and information risks. Cluster 2, which includes market indicators of the banking system and methods used in monetary policy, is close to this cluster. It should be noted that in the study of central bank transparency, the categories of independence and transparency of the central bank fall into one cluster, while in a similar analysis, the categories of "central bank independence" belong to different clusters. This allows us to conclude that the independence of the central bank is the primary category, which, in turn, creates the basis for research and development of transparency of the central bank.
Conclusions, Discussion and Recommendations
The study showed that there are close substantive and practical links between the categories of "central bank independence" and "central bank transparency". Existing empirical studies have shown that the independence and transparency of the central bank can serve as tools to increase the effectiveness of its main goals, such as price or financial stability, or become complementary components in improving the quality of the central bank's core tasks. A preliminary analysis of trends in the study of central bank independence and central bank transparency showed a declining trend in search queries from 2004 to the present, with a significant predominance of central bank independence over transparency throughout the study period. The results of bibliometric analysis of the coincidence of keywords in scientific articles devoted to the study of independence and transparency of the central bank, showed that the investigations of central bank independence concern the study of aspects of its transparency in a limited number of cases. Along with this, the central bank transparency is often studied simultaneously with the analysis of their independence. It enables to determine that the independence of the central bank is the primary category that creates the conditions for the study of its transparency. Obtained results provide a scientific basis for studying the strength and direction of the relationship between the levels of transparency and independence of the central bank, measured as quantitative indicators. | 2020-08-06T09:09:19.886Z | 2020-01-01T00:00:00.000 | {
"year": 2020,
"sha1": "1bcd5f5260526ec84a25a10fbe2e0bf03ed84ec2",
"oa_license": "CCBY",
"oa_url": "https://armgpublishing.sumdu.edu.ua/wp-content/uploads/2020/07/10.pdf",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "aa36fb33a611855eecbc15f67f5fcffae3858327",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Business"
]
} |
250975193 | pes2o/s2orc | v3-fos-license | Intestinal type adenocarcinoma of the endometrium with signet ring cells, a rare aggressive variant
Highlights • Intestinal type mucinous adenocarcinoma is a rare variant of mucinous carcinoma of the endometrium.• Intestinal differentiation in the endometrium can manifest with a wide spectrum of morphologies including signet-ring cells.• E-cadherin expression is downregulation in the signet-ring cell component in all three cases.• Cases described provides further evidence of the aggressive nature of intestinal type mucinous adenocarcinoma.
Introduction
Endometrial cancer is the most common gynecological malignancy in the United States. Over 63,000 new cases are reported annually and 1-9% are reported to be mucinous adenocarcinoma of the endometrium (MACE). According to the 2014 World Health Organization (WHO), mucinous adenocarcinoma is defined as an endometrial carcinoma with >50% cells showing mucinous differentiation. In addition, the WHO suggests that mucinous adenocarcinoma is an independent category of endometrial cancers and not a subcategory of endometrioid endometrial carcinoma (Ardighieri et al., 2020).
Intestinal type mucinous adenocarcinoma (iMACE) is a rare and unusual variant of mucinous carcinoma that was first reported by Berger in 1984(Berger, 1984. Recently, an exhaustive systematic review of the literature performed by Ardighieri reported a total of 9 cases (Ardighieri et al., 2020;Berger, 1984;Trippel et al., 2017;Rubio et al., 2016;Nieuwenhuizen, 2007;Buell-Gutbrod, 2013;Mogor et al., 2019;Zheng et al., 1995). Of these cases, only Berger reported the presence of signetring cells as a morphological feature of intestinal type mucous adenocarcinoma. Berger reported no evidence of disease on follow up of their cases (Berger, 1984). To our knowledge, these are the only cases of intestinal type mucinous adenocarcinoma with signet ring morphology and thus prognosis is difficult to determine. In other parts of the body signet-ring cell carcinoma (SRCC) is usually associated with a highgrade tumor and has a poor prognosis (Boyd et al., 2010;Wang et al., 2016). Interestingly, loss of E-cadherin expression in colorectal SRCC has been shown to be a independent prognostic factor of patients' overall survival. We describe in this case report the loss of E-cadherin expression and the presence of signet-ring cells in three cases of intestinal type mucinous adenocarcinoma.
Case presentation
Case A. An 80-year-old para 9, with no past gynecologic care, presented to the emergency department with the complaint of one week of post-menopausal bleeding. A pelvic ultrasound was performed which showed a markedly distended and abnormal endometrium measuring 3.3 cm, and an enlarged right ovary measuring 4 cm without discrete mass. She was referred to a gynecologist who performed an endometrial biopsy which showed FIGO grade 2 endometrioid adenocarcinoma with focal mucinous features. She was brought to the operating room by a gynecologic oncologist who performed total abdominal hysterectomy, bilateral salpingo-oophorectomy, right pelvic lymph node excision, omentectomy and appendectomy. Findings at the time of surgery included a ten-week gestational size uterus with irregular contour, enlarged right ovary grossly suspicious for tumor involvement, enlarged right common iliac lymph node and grossly normal appendix and omentum. Frozen section of the right ovary revealed adenocarcinoma with possible signet-ring cells. Due to this histologic finding, an appendectomy was performed. Excision of the enlarged right common iliac lymph node was performed and node was found to be positive for metastatic adenocarcinoma on frozen section. There were no other enlarged lymph nodes and systematic lymphadenectomy was deferred. The patient recovered well post-operatively. Final pathology demonstrated gland forming architecture with focal intestinal type mucinous glandular differentiation and signet-ring cells. The tumor was deeply invasive through the myometrium, involved cervical stroma, bilateral ovaries, right pelvic lymph node, and omentum. Peritoneal cytology was negative. The signet ring cell component of the carcinoma was positive for both cytokeratin 7 and cytokeratin 20, positive for CDX2, negative for PAX8 and estrogen receptors, and focally, very weakly positive for Ecadherin (Fig. 1). The staining for the DNA mismatch repair (MMR) proteins was intact. A upper and lower gastrointestinal workup was completed which included endoscopy, colonoscopy, and relevant imaging with results negative for a gastrointestinal primary malignancy. The patient declined chemotherapy treatment and she died eleven months post-surgery from diffuse metastatic disease to the lung, liver, bones, and brain.
Case B.
A 65-year-old para 0, who was compliant with routine wellwoman gynecologic care, presented to her gynecologist with the complaint of several months of post-menopausal bleeding. She had a pertinent gynecologic history of cervical dysplasia that was treated with cryotherapy in the distant past, with normal Pap smears subsequent. A pelvic ultrasound showed an endometrial thickness of 12 mm, and endometrial biopsy was interpreted as adenocarcinoma, mucinous type with focal areas appearing endometrioid, but the primary site of origin was indeterminable. Tumor markers CA-125 and CEA were checked and within normal limits, 6 and 0.9 respectively. The patient underwent CT imaging of the abdomen and pelvis, which revealed a 1.9 cm right lateral uterine wall mass. Pre-operative endoscopy and colonoscopy was negative for a gastrointestinal primary malignancy. She was taken to the operating room by a gynecologic oncologist who performed robotic assisted total laparoscopic hysterectomy, bilateral salpingooophorectomy and sentinel pelvic lymph node dissection. Findings at the time of surgery included omental adhesions to the left fallopian tubes, grossly normal uterus, bilateral adnexa, appendix, and omentum. Final pathology showed a mucin producing invasive adenocarcinoma with focal squamous differentiation which involved the entire endometrial cavity and serosa but spared the cervix. The bilateral adnexa and all lymph nodes were negative for metastases. The pelvic washings were negative. Immunohistochemical studies were compatible with iMACE ( Table 2). The signet-ring cell component of the carcinoma was focally weakly to moderately positive for cytokeratins 7 and 20, positive for CDX2, negative for PAX8 and estrogen receptors and focally, very weakly positive for E-cadherin ( Fig. 1). In situ hybridization analysis for high-risk HPV subtypes showed rare punctate signals in some tumor cells, the interpretation of which was uncertain. The staining for the DNA MMR proteins was intact. She was referred to radiation oncology for adjuvant therapy, but the patient declined any further treatment. Patient has been seen for regular follow-up with her gynecologic oncologist and is clinically without evidence of disease at 36 months since surgery. Case C. A 73-year-old para 3, compliant with routine well-woman gynecologic care, presented to the emergency department with two months of post-menopausal bleeding. Pelvic ultrasound revealed an endometrial thickness of 6 mm. She had a pertinent gynecologic history of right salpingo-oophorectomy for a tubo-ovarian abscess and cervical dysplasia status post loop electrode excisional procedure in the distant past with subsequent normal Pap smears. A Pap smear was performed which showed atypical cells of undetermined significance, HPV positive (high risk positive, 16/18/45 negative). Patient was evaluated by a gynecologic oncologist where an abnormally hardened cervix was appreciated, and cervical biopsy, endocervical curettage and endometrial biopsy were performed. On rectovaginal exam there was suspicion for left parametrial shortening. Patient was referred to medical oncology for neoadjuvant chemotherapy for presumed advanced cervical or endometrial primary malignancy. Patient declined chemotherapy and was referred to gastroenterology for repeat screening, with upper and lower endoscopy noncontributory. She underwent a radical abdominal hysterectomy, left salpingo-oophorectomy, and pelvic lymphadenectomy. Intra-operative findings included a flush, hardened cervix with a small amount of necrotic tumor prolapsing through the endocervix. The parametria were grossly not involved, the rectosigmoid colon was densely adherent to posterior uterine wall and the appendix and omentum were grossly normal. Immunohistochemistry confirmed intestinal differentiation ( Table 2). The signet-ring cell component of the carcinoma was negative for cytokeratin 7 and positive for cytokeratin 20 (suggesting a colorectal and appendiceal-type immunohistochemical profile), positive for CDX2, negative for PAX8 and estrogen receptors, and completely negative for E-cadherin (Fig. 1). MMR staining was not performed, but next generation sequencing showed the tumor to be MSI stable. The patient expired six months after the initial surgery.
Intestinal differentiation has been postulated by Ardighieri et al. to denote a more aggressive variant of mucinous adenocarcinoma of the endometrium. The stage of disease at presentation was IA for 50% (4/8) of the cases described. However, of the patients with low-stage disease and a follow-up available, two (2/3) had disease recurrence in the vagina, vulva, and peritoneum (Ardighieri et al., 2020). This demonstrated that, despite early stage of presentation, the disease often recurs. All three cases presented with late stage disease and two patients died of disease, with the third disease free at three years post therapy. (Table 2). This not only further confirms Ardighieri's findings, but also indicates that those individuals with signet-ring cell morphology may have a potentially worse prognosis.
E-cadherin is a calcium-dependent cell-to-cell adhesion molecule found in epithelial tissue and implicated in cellular migration. Alterations in E-cadherin expression have been linked to decreased cell-cell adhesion, increased metastatic potential, tumor dedifferentiation, and Focally weakly positive Focally weakly positive deep myometrial invasion in endometrial carcinoma. Wang et al. showed that the signet ring cell component of colorectal adenocarcinomas was often negative for E-cadherin, and the loss of E-cadherin expression conferred a worse clinical prognosis (Wang et al., 2016). In their turn, Humar et al. showed that the loss of E-cadherin initiated carcinogenesis of gastric signet ring cell carcinoma (Humar et al., 2009). Consistent with the previous studies, our series showed that the expression of E-cadherin was lost or markedly decreased in the signet ring cell component of endometrial mucinous adenocarcinoma, although the low number of extant cases does not allow to determine its influence on the clinical behavior of these tumors. The presence of signet ring cells and other morphologic features of intestinal differentiation in uterine tumors increases the probability of theses tumors being metastases from extragenital organs. Thus, ruling out metastasis from the gastrointestinal system or the breast is imperative. Immunohistochemistry has been shown to have limited value, since both gastrointestinal metastases and intestinal-like mucinous endometrial carcinomas express intestinal-like markers and are negative for PAX8 and estrogen and progesterone receptors. Thus, to rule out metastasis, the time-consuming process of imaging and endoscopy must be performed (Mogor et al., 2019). The differential diagnosis of the cases presented also includes intestinal differentiation in mucinous adenocarcinoma of other regions of the female genital tract, including the cervix and ovary. This can be particularly challenging in advanced-stage disease, when involvement of multiple regions of the female genital tract can be seen at time of diagnosis. It has been proposed that negative HPV molecular testing, negative immunostaining for p16, or both, can aid in excluding an HPV-related primary cervical carcinoma (Trippel et al., 2017). However, the relationship between HPV and the intestinal phenotype is not well understood. An argument can be made for designating case C of our series as a primary cervical carcinoma, given the extensive involvement of the cervix. However, p16 immunostain was negative, which favors either an endometrial origin, or a rare gastrictype adenocarcinoma of the cervix, which is not related to HPV. The latter two diseases would be difficult to differentiate from each other, especially if the tumor centers on the LUS and involves both the cervix and the corpus uteri.
Traditionally, the treatment and prognosis of endometrial carcinoma have relied on histotype and grade of the tumor. This system has been challenged by the development of the ProMise (Proactive Molecular Risk Classifier from for Endometrial Cancer) algorithm which classifies endometrioid adenocarcinoma into four distinct prognostic subtypes and is based upon genomic abnormalities. MSI-H is one such subtype and is classified as a tumor with MLH-1-promoter hypermethylation or mutations of the DNA MMR genes. Recently, Ardighieri and Trippel postulated that this tumor classification system can potentially be extended to iMACE due to the loss of MSH2/MSH6 and the presence of MLH-1-promoter hypermethylation, respectively (Ardighieri et al., 2020;Trippel et al., 2017). These genomic abnormalities were not seen in our cases.
In summary, we describe three rare cases of intestinal type mucinous adenocarcinoma with signet-rings cells. These cases demonstrate the aggressive nature of intestinal differentiation in the endometrium and suggests that the presence of signet-rings cells and loss of E-cadherin expression may render a poorer prognosis as seen in other parts of the body. However, more cases are needed to further study the biologic behavior of this morphological variant. This case report also adds to the body of literature that supports the concept of divergent molecular profiles, including cell adhesion molecule expression, in different types of endometrial carcinoma. | 2022-07-23T15:19:29.516Z | 2022-07-21T00:00:00.000 | {
"year": 2022,
"sha1": "791625295438eb4499ff8af776f7c4cf55701fc4",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "ed13118d618a8933b72e672510dea8bec54babfe",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
247853468 | pes2o/s2orc | v3-fos-license | PIK3CA Mutations Drive Therapeutic Resistance in Human Epidermal Growth Factor Receptor 2–Positive Breast Cancer
The phosphatidylinositol 3-kinase (PI3K) pathway is an intracellular pathway activated in response to progrowth signaling, such as human epidermal growth factor receptor 2 (HER2) and other kinases. Abnormal activation of PI3K has long been recognized as one of the main oncogenic drivers in breast cancer, including HER2-positive (HER2+) subtype. Somatic activating mutations in the gene encoding PI3K alpha catalytic subunit (PIK3CA) are present in approximately 30% of early-stage HER2+ tumors and drive therapeutic resistance to multiple HER2-targeted agents. Here, we review currently available agents targeting PI3K, discuss their potential role in HER2+ breast cancer, and provide an overview of ongoing trials of PI3K inhibitors in HER2+ disease. Additionally, we review the landscape of PIK3CA mutational testing and highlight the gaps in knowledge that could present potential barriers in the effective application of PI3K inhibitors for treatment of HER2+ breast cancer.
abstract
The phosphatidylinositol 3-kinase (PI3K) pathway is an intracellular pathway activated in response to progrowth signaling, such as human epidermal growth factor receptor 2 (HER2) and other kinases. Abnormal activation of PI3K has long been recognized as one of the main oncogenic drivers in breast cancer, including HER2-positive (HER2+) subtype. Somatic activating mutations in the gene encoding PI3K alpha catalytic subunit (PIK3CA) are present in approximately 30% of early-stage HER2+ tumors and drive therapeutic resistance to multiple HER2targeted agents. Here, we review currently available agents targeting PI3K, discuss their potential role in HER2+ breast cancer, and provide an overview of ongoing trials of PI3K inhibitors in HER2+ disease. Additionally, we review the landscape of PIK3CA mutational testing and highlight the gaps in knowledge that could present potential barriers in the effective application of PI3K inhibitors for treatment of HER2+ breast cancer.
PHOSPHOINOSITIDE 3-KINASE PATHWAY AND ITS ROLE IN HUMAN EPIDERMAL GROWTH FACTOR RECEPTOR 2-POSITIVE BREAST CANCER
The phosphoinositide 3-kinase (PI3K) pathway plays a key role in growth, proliferation, and survival of cancer cells. 1 PI3K transmits signals from oncogenic receptor tyrosine kinases (RTKs), 2 including human epidermal growth factor receptor 2 (HER2), platelet-derived growth factor receptor, insulin growth factor 1 receptor, and others. 3,4 Activated PI3K catalyzes phosphorylation of phosphatidylinositol diphosphate (PIP2) to phosphatidylinositol 3-phosphate (PIP3), which recruits serine-threonine kinase AKT to the membrane. Activation of AKT and mammalian target of rapamycin (mTOR) complex is a key consequence of RTK-based signaling. 5,6 AKT reduces apoptosis and promotes proliferation, epithelial-mesenchymal transition, invasion, metastases, and angiogenesis. 5 mTOR, a catalytic subunit of two protein complexes, mTORC1 and mTORC2, is the major regulator of cell growth. 7 Phosphatase and tensin homolog (PTEN) acts as a negative regulator of PI3K by dephosphorylating PIP3 to PIP2. 5,8 However, PTEN is frequently disabled in cancer cells by loss of heterozygosity, inactivating mutations, or epigenetic silencing, 5 which can augment the effects of PI3K activation (Fig 1).
Class I PI3K contains four isoforms of the catalytic subunit: p110α, p110β, p110δ, and p110γ. 9 These isoforms are subdivided according to their connections to upstream signaling. Class IA (p110α, p110β, and p110γ) can be activated by RTKs; and class IB (p110γ) by small G-proteins. 8 p110α and p110β are expressed ubiquitously, whereas p110δ and p110γ are mostly found in hematopoietic cells. 10 The p110γ isoform of PI3K regulates migration of CD4+ T lymphocytes 11 and may play a role in anticancer immunity. 12 Although p110α is the most studied in cancer, activity of any class IA PI3K isoform can sustain cell proliferation and survival. 9,13 PI3K class IA consists of a p110 catalytic subunit and a p85 regulatory subunit. Genes PIK3CA, PIK3B, and PIK3D encode PI3K catalytic subunit isoforms p110α, p110β, and p110δ, respectively. 1 Notably, p110α or p110β inactivation differentially modulate RTK signaling: knockout of p110α allows all phosphosites to be occupied by p110β decreasing RTK signaling, whereas knockout of p110β leads to binding of p110α, which has a greater kinase activity and increases RTK signaling. 14,15 Out of all isoforms, p110α is particularly important in breast cancer because of its high RTK-dependent kinase activity and frequent activating mutations: mutations in p110α are seen in approximately 35% of breast tumors, compared with p110β (, 5%) and p110δ (not reported). 14 PI3K p110α (encoded by PIK3CA) is critically important for the development of breast tumors overexpressing HER2. HER2 signaling is mediated almost exclusively through p110α. 15,16 PIK3CA knockout mice are completely resistant to HER2 transgenemediated tumor formation, 15 whereas mice bearing transgenic overexpression of HER2 and activating mutations in PIK3CA develop mammary tumors faster than those bearing only an HER2 transgene. 17 HER2-positive (HER2+) breast cancer is an aggressive subtype characterized by rapid growth and visceral and brain metastases. Nearly 54,000 US women are diagnosed with HER2+ breast cancer every year. Despite advances in treatment, approximately 25% of patients experience recurrence within 5 years. 18 Most patients with HER2+ metastatic breast cancer (MBC) develop resistance to HER2-targeted inhibitors (HER2i) leading to progressive disease and death. 19,20 PIK3CA mutations are present in approximately 30% of HER2+ breast tumors (Table 1), contributing to aggressive tumor behavior and poor treatment outcomes. 21 A largescale shRNA screen identified PI3K pathway as a major modulator of sensitivity to HER2 monoclonal antibody trastuzumab. 21 In vitro, PIK3CA-mutant HER2+ human breast tumor cell lines are resistant to trastuzumab and maintain AKT phosphorylation despite treatment. 15,21 In vivo, PIK3CA-mutant HER2+ breast tumors grow despite treatment with trastuzumab, another HER2-mAb pertuzumab, or HER2 tyrosine kinase inhibitor (TKI) lapatinib; this drug resistance is reversed by addition of PI3K inhibitors (PI3Ki). 17 In contrast to other HER2i, the antibody-drug conjugate ado-trastuzumab emtansine (T-DM1) is active in PIK3CA-mutant breast tumor cell lines and xenograft models, 22 perhaps because of its unique mechanism of action via HER2-receptor mediated delivery of chemotherapeutic payload, which is independent of signaling downstream of HER2. 23 PIK3CA mutations are associated with poor outcomes in patients with HER2+ early breast cancer. [21][22][23][24][25][26] In NeoALTTO, 27 GeparSixto, 28 and NeoSphere 23 studies, PIK3CA mutations were linked to low rates of pathologic complete response after neoadjuvant chemotherapy and HER2i, with 13%-20% absolute decrease in complete response rates in patients with PIK3CA-mutant versus wild-type tumors. In the adjuvant setting, PI3K/PTEN/AKT alterations were associated with poorer prognosis (hazard ratio, 1.35; 95% CI, 1.01 to 1.79) in the APHINITY study of adjuvant trastuzumab and pertuzumab. 29 In contrast to APHINITY, the ExteNET trial of adjuvant neratinib (an irreversible TKI of HER1, HER2, and HER4) did not show the difference in outcomes between patients with PIK3CA-mutant and wildtype tumors. 30 Similarly, KATHERINE trial showed that benefits of adjuvant T-DM1 on invasive disease-free survival are independent of PIK3CA mutational status. 31 Among patients with HER2+ MBC enrolled in the docetaxel, trastuzumab, and pertuzumab arm of the CLEOPA-TRA study, those whose tumors had PIK3CA mutations experienced shorter progression-free survival (PFS) when compared with patients with PIK3CA wild-type tumors (13 v 22 months, respectively; hazard ratio, 0.67; 95% CI, 0.50 to 0.89). 24,26 In the EMILIA trial, patients with PIK3CAmutant tumors had shorter PFS and overall survival on capecitabine plus lapatinib treatment, but not on T-DM1 treatment, 22,24 underscoring potential activity of T-DM1 against PIK3CA-mutant disease. Activity of T-DM1 against PI3K-mutant HER2+ MBC was confirmed in the TH3RESA trial, where T-DM1 was compared with treatment of the physician's choice. 32 Although published clinical trials suggest that T-DM1 and neratinib may be active against PIK3CA-mutant HER2+ breast cancer, the majority of the HER2i have diminished efficacy in this setting. 21,23,27,28,33 Considering the presence of activating PIK3CA mutations in approximately 30% of HER2+ tumors and the strong association of these mutations with resistance to HER2i, there is an unmet need to develop combination regimens blocking HER2 and PI3K. In this review, we summarize the targeted agents that block the PI3K pathway and ongoing clinical trials of these agents in HER2+ disease. Additionally, we highlight the landscape and challenges in PIK3CA mutational testing. Our goal is to provide a reference for translational and clinical investigators to accelerate research on PIK3CA inhibitors in HER2+ breast cancer.
Throughout this paper, we used the American College of Clinical Pathology's definition for HER2+ 34 and hormone receptor-positive (HR+) breast cancer. 35 To ascertain the current landscape of clinical trials testing of PI3Ki in HER2+ disease, we performed a PubMed search of the published, English-language scientific literature through January 2021 using the keywords "breast cancer," "HER2," "human epidermal growth factor receptor 2," "PI3K," "PIK3CA," and the generic names of specific inhibitors. The same keywords were used to search the NIH Clinical Trial Database ClinicalTrials.gov.
CHALLENGES IN DEVELOPING EFFECTIVE AND TOLERABLE PI3K INHIBITORS
Despite the key role of PI3K in numerous tumor types, development of effective and safe PI3Ki has been a major challenge. Reasons include off-target effects of nonselective PI3Ki leading to intolerable toxicities, drug resistance because of activation of alternative oncogenic pathways, mutations in PI3K downstream effectors, or PTEN inactivation. 36 Pan-PI3Ki suppress activity of all PI3K class I isoforms and include buparlisib, 37-39 pictilisib, 40 47 and SF1126. 48 With the exception of copanlisib, no pan-PI3Ki have been approved for clinical use because of a lack of clinical activity and/or significant safety concerns. 36 One of the most extensively tested pan-PI3Ki is buparlisib, which has results available from three randomized trials in HR+/ HER2-MBC. [37][38][39] Although it showed antitumor efficacy, including activity in brain metastases, its clinical development was abandoned because of toxicity. More than 20% of patients on buparlisib had grade ≥ 3 elevation of liver enzymes, and there were cases of severe depression leading to suicide attempts. 38 Another pan-PI3Ki pictilisib failed to show improvement in PFS in HR+ breast cancer, perhaps because 24% of patients in the pictilisib arm discontinued treatment and an additional 24% required dose reduction because of adverse events (AEs). 40 Development of pilaralisib, PX-866, and CH5132799 has been stopped because of lack of activity, whereas SF1126 is still in clinical trials. 36 Unlike other pan-PI3Ki, copanlisib is administered intravenously and has a better therapeutic index, 36 with hyperglycemia and nausea being the most frequent yet manageable AEs. 43 The drug demonstrated low incidence of high-grade GI and liver toxicity. 49 Copanlisib showed activity in patients with hematologic malignancies, breast cancer, and endometrial cancer. 43 Copanlisib is US Food and Drug Administration (FDA)-approved in patients with relapsed refractory follicular lymphoma irrespective of PI3K mutational status on the basis of the results of the CHRONOS-1 trial that showed durable responses in 59% of patients. 49 Copanlisib is now being studied in solid tumors and hematologic malignancies in combination with chemotherapy, antihormonal therapy, targeted agents, or immunotherapy. 36 Further efforts to suppress the PI3K pathway progressed primarily in two directions. To improve efficacy and overcome drug resistance, dual PI3K/mTOR inhibitors were formulated, such as apitolisib, dactolisib, and gedatolisib. Alternatively, to improve the toxicity profile, isoform selectivity was prioritized and selective inhibitors were developed, such as taselisib (inhibiting p110α, p110γ, and p110δ, but sparing the p110β isoform) and alpelisib (selective p110α inhibitor).
Dual PI3K/mTOR inhibitors, dactolisib and apitolisib, have not achieved the hoped-for clinical efficacy, mainly because of frequent dose-limiting toxicities, such as diarrhea, hyperglycemia, mucositis, and liver toxicity. [50][51][52] Broadening the spectrum of inhibition seemed to increase the toxicity of these inhibitors disproportionately to antitumor activity, likely because of the fundamental role of PI3K/ mTOR pathway in normal tissues. Gedatolisib differs from other PI3K/mTOR inhibitors because of its intravenous administration and is better tolerated. Although studies of gedatolisib in endometrial cancer (NCT01420081) 53 and hematologic malignancies (NCT02438761) were terminated because of low activity, studies evaluating combinations of gedatolisib with antihormonal drugs and targeted inhibitors of HER2, poly(ADP-ribose) polymerase, and CDK4/6 are ongoing in breast cancer (NCT03911973, NCT02626507, NCT03698383) as well as lung, pancreatic, and head and neck cancers (NCT03065062). The first selective inhibitor of PI3K (taselisib) spared the p110β subunit, while inhibiting p110α, γ, and δ. From a signaling standpoint, this approach could alleviate the severity of some AEs without sacrificing antitumor efficacy. However, a phase III clinical trial of taselisib (SANDPIPER) showed only modest activity and a challenging safety profile (diarrhea, nausea, vomiting, abdominal pain, stomatitis, fatigue, hyperglycemia, and rash), 54 leading to cessation of its clinical development.
On the basis of the strong association with RTK signaling and frequent mutations in human cancers, PI3K p110α has become the clear target for inhibition. The selective PI3K p110α inhibitor alpelisib demonstrated a manageable safety profile and prolonged PFS among patients with the PIK3CA-mutant, HR+ MBC in the phase III clinical trial SOLAR-1. 55 Main side effects were hyperglycemia, diarrhea, and rash. At a median follow-up of 20 months, PFS was 11 months (95% CI, 7.5 to 14.5) in the alpelisibfulvestrant group compared with 5.7 months (95% CI, 3.7 to 7.4) in the placebo-fulvestrant group. 55 Alpelisib received FDA approval for the HR+ MBC in May 2019, becoming a milestone of success on the difficult path toward PI3K inhibition in cancer. Although patients with brain metastases were not included in SOLAR-1, case reports indicate potential activity of alpelisib in brain metastatic disease. 56 Given the challenges of developing effective and tolerable PI3Ki, researchers have explored inhibition of PI3K downstream effectors AKT and mTOR. A first-generation mTOR inhibitor everolimus is FDA-approved in HR+ MBC on the basis of the BOLERO-2 study, 57 whereas the secondgeneration mTOR inhibitor, sapanisertib, and AKT inhibitors, capivasertib and ipatasertib, are in early clinical trials. Compared with everolimus, which predominantly inhibits mTORC1, sapanisertib may have an advantage because of combined inhibition of mTORC1 and mTORC2. 58 mTORC2 directly phosphorylates AKT, and this escape pathway is suppressed by sapanisertib. 58
CURRENT CLINICAL TRIALS OF INHIBITORS OF THE PI3K PATHWAY IN HER2+ DISEASE
Agents suppressing PI3K pathway were studied in several clinical trials in patients with HER2+ MBC ( Table 2). The first-generation mTOR inhibitor, everolimus, has been tested in HER2+ disease in phase III randomized placebo controlled clinical trials BOLERO-1 59 and BOLERO-3. 60 BOLERO-1 evaluated everolimus versus placebo in combination with trastuzumab and paclitaxel as a first-line therapy for HER2+ MBC. PFS did not differ between the everolimus and placebo groups. Although in the subgroup of patients with HR-/HER2+ disease PFS was 7.2 months longer on everolimus compared with placebo, it did not meet the prespecified criteria for significance. 59 In BOLERO-3, women with HER2+ transtuzumab-resistant MBC previously treated with taxanes were randomly assigned to everolimus or placebo in combination with vinorelbine and trastuzumab. Median PFS was 7.0 months in the everolimus and 5.8 months in the placebo group (P = .0067). However, this small improvement in PFS came at the cost of increased toxicity, such as cytopenias, stomatitis, and fatigue, in the everolimus group. Serious AEs were reported in 42% of patients on everolimus and 20% of patients on placebo. 60 Both BOLERO-1 and BOLERO-3 trials were conducted in a biomarker-unselected population of patients. These trials were not practice-changing because of two possible reasons: (1) the relatively weak activity of everolimus in suppressing the PI3K pathway, with mTORC2 mediating sustained AKT activation, and (2) the absence of a biomarker selection strategy. Subsequent exploratory analysis of the combined BOLERO-1 and BOLERO-3 trials suggested that patients with HER2+ MBC with aberrant PI3K pathway activation could derive significant PFS benefits from everolimus, whereas patients whose tumors lacked such activation do not benefit from mTOR inhibition. 61 Two clinical trials of buparlisib in HER2+ disease have been completed. In the phase IB PIKHER2 study, patients with trastuzumab-resistant, PIK3CA mutation-unselected, HER2+ MBC were treated with a combination of buparlisib and lapatinib. The duration of treatment was 4-60 weeks, with a median duration of 40 weeks at maximum tolerated dose. The observed disease control rate was 79%, the clinical benefit rate (CBR) was 29%, and one patient obtained a complete response. AEs included diarrhea, nausea, rash, depression, anxiety, an increase in transaminases, and asthenia. 62 The phase II trial NeoPHOEBE randomly assigned HER2+ early breast cancer patients regardless of PIK3CA mutation status to receive either buparlisib or placebo plus trastuzumab in the first 6 weeks and then buparlisib or placebo with trastuzumab and paclitaxel. 63 Although no significant differences were noted in the pCR rate between the buparlisib and placebo arms (32% v 40%), a near-significant trend was observed in the overall response rate (68.6% v 33%; P = .053) and a significant decrease was noted in Ki67 levels (75% v 26.7%; P = .021) favoring buparlisib in the subgroup of patients with HR+/HER2+ tumors. 63 Only eight of 50 enrolled subjects had PIK3CA mutations. 63 The study planned to recruit 256 patients but suspended recruitment early because of hepatotoxicity. Both PIKHER2 and Neo-PHOEBE did not include PIK3CA mutations as a biomarker for selection, potentially affecting efficacy outcomes.
The results are available from a phase I study of alpelisib and T-DM1 in patients with HER2+ MBC who had progressive disease on trastuzumab and taxanes. In evaluable patients, overall response rate was 43% and CBR was 71%, with a median time on study of 7.6 months. 64 Notably, even in patients with prior T-DM1 exposure, CBR of this combination reached 60%. 65 This result is intriguing because T-DM1 is more active in patients with PIK3CA-mutant HER2+ breast tumors compared with other HER2-targeted agents. 22,32 However, activation of the PI3K pathway may be partially responsible for acquired resistance to T-DM1, 66 and inhibition of this pathway could induce resensitization to this agent. This early trial did not select patients on the basis of PIK3CA mutation status, although approximately 50% of patients had tumors with PI3K pathway aberrations. 1,65 Grade≥ 3 AEs were observed in 59% of patients, and the most common events were rash, hyperglycemia, anorexia, and hypertension, but all were noted as manageable. 64,65 Investigators concluded that the combination of alpelisib and T-DM1 is tolerable and has activity in patients with trastuzumab-resistant HER2+ MBC. 64,65 The results from these early trials indicate that PI3Ki will likely be essential in the treatment of HER2+ disease, and underscore the importance of a biomarker to identify patients who are most likely to benefit from addition of PI3Ki.
CLINICAL TESTING FOR PIK3CA MUTATIONS: THE COMPANION DIAGNOSTIC VERSUS NONCOMPANION DIAGNOSTIC APPROACH
One challenge in clinical studies of targeted agents is the development of a biomarker to identify potential responders. This biomarker should be fine-tuned to be specific, but also sensitive enough not to miss those who may derive clinical benefits. Challenges in the biomarker identification and development of a companion diagnostic (CDx) test can influence clinical fate of targeted agents.
Shortly after the 2019 FDA approval of alpelisib, the Therascreen PIK3CA RGQ PCR Kit (QIAGEN Manchester Ltd, Manchester, UK) CDx test was approved for the identification of PIK3CA mutations in tumor tissue and/or plasma circulating tumor DNA. This assay (hereafter referred to as the CDx) is a real-time quantitative polymerase chain reaction (qPCR) assay that detects 11 mutations within exons 7, 9, and 20 (Table 3). These mutations can also be detected by using a variety of other platforms, including next-generation sequencing (NGS). A wide spectrum of NGS-based assays is used in clinical care, and most, if not all, cover the hotspots included in the CDx. Although the range of coverage of PIK3CA coding regions varies, nearly all NGS-based assays will detect alterations beyond those identified by the CDx. The ability of NGSbased tests to detect additional alterations raises important questions about the clinical implications of such findings.
Different groups have performed analyses to quantify the PIK3CA mutations in breast tumors that are missed by CDx testing alone. 67,68 In an assessment of 5,813 breast tumors from publicly available data sets, the overall mutation rate of PIK3CA in the 763 HER2+ breast cancer specimens evaluated was 31%. 69 This is similar to the prevalence of PIK3CA mutations previously observed in HER2+ tumors assessed for hotspot mutations within exons 7, 9, and 20. 23 The mutation rate in HR+/HER2-(4,055 specimens) and triple-negative (995 specimens) tumors was 42% and 16%, respectively. 69 The distribution of specific mutations across the three subtypes was similar, with five missense mutations (H1047R, H1047L, E542K, E545K, and N345K) constituting approximately 70% of all PIK3CA mutations. 69 Specimens were obtained from patients across all clinical stages of disease and included some metastatic lesions. [69][70][71] A comparison of the mutations identified within these data sets with the mutations tested by CDx showed that 20% of tumors harboring PIK3CA mutations would be missed by the CDx, including approximately 25% of mutations identified in the HER2+ subtype. 69 Notably, mutations with at least in vitro evidence of oncogenic activity were observed to occur at a higher rate than some of the variants included as part of the CDx. 69 Corroborating these findings is an NGS-based assessment of 5,549 tumors from patients with MBC from a commercial laboratory database. 67 Authors found more than 70 activating mutations not covered by the CDx, with 26% (626/2,435) of all activating mutations undetectable by the CDx. 67 Moreover, in a retrospective reassessment of PIK3CA alteration status (originally determined by CDx) of tumors from SOLAR-1 participants, NGS revealed that 16% (28/175) of tumors originally designated as nonmutant did in fact harbor PIK3CA alterations. 68 Retrospective analysis of SOLAR-1 reported 60 additional PIK3CA mutations and five copy-number alterations. 68 The identification of tumors with non-CDx-detectable PIK3CA mutations is of clinical importance, as non-CDx alterations may be oncogenic and may respond to alpelisib or another PI3Ki. 72 To ascertain the spectrum of non-CDx PIK3CA mutations interrogated with alpelisib treatment to date, we performed a PubMed search of the published, Englishlanguage scientific literature through January 2021 using the terms "PIK3CA," "alpelisib," and "BYL719." Preclinical and clinical studies were included regardless of tumor type. For in vitro studies using cell lines, PIK3CA mutation status was determined using the publicly available DepMap Portal data set. 73 We initially identified 53 publications; however, studies evaluating only CDx mutations, or mutations of unknown significance were excluded. Twenty publications evaluating 37 alterations not detected by the CDx with a known or likely activating effect on the basis of OncoKB curated information annotated in the cBioPortal database were identified (Table 4, Fig 2). [94][95][96] Alterations included additional point mutations within the C2, helical, and kinase domains, mutations within the p85 binding domain, amplifications, oncogenic deletions, and a nonstop frameshift mutation at the transcript terminus (Table 4, Fig 2). In addition, two cell lines, two transduced cell lines, and one patient were reported as harboring a likely or known activating non-CDx mutation in combination with a known CDx-activating mutation ( Table 4, Fig 2). Notably, several non-CDx mutations occur within the helical and kinase domains at or very near residues of CDx mutations (ie, E542*, E545*, Q546*, and G1049*). It is possible these mutations are detected by the CDx and misattributed. Such cross-reactivity can occur in PCR-based assays and was reported by the vendors as occurring between H1047R and H1047L in certain contexts. 97 With respect to current clinical testing, it is at the discretion of health care providers and patients whether a CDx or NOTE. PIK3CA NCBI transcript NM_006218.4 used to determine position and codon Exon numbering: to be most consistent with the published literature, exons are numbered by coding exons. The true exon 1 constitutes the 5 -UTR of the gene and has historically not been included in the exon count. As such, the exons listed above correspond to one exon higher when the noncoding exon is included (eg, coding exon 1 above corresponds to true exon 2; coding exon 20 corresponds to true exon 21).
Abbreviations: HER2, human epidermal growth factor receptor 2; NCBI, National Center for Biotechnology Information; UTR, untranslated region. non-CDx testing approach is optimal, as variables such as testing accessibility, insurance coverage, and pricing may influence the benefits and limitations of the two approaches. Ordering physicians should be mindful that the resulting formats of NGS-based assays are variable, and CDx and non-CDx mutations may be differentially emphasized depending on the specific mutations/disease combinations.
Data on whether specific PIK3CA mutations render differential responses to alpelisib are limited. 89,98,99 Early evaluations have suggested that tumors with c-terminus H1047* mutations respond to alpelisib better compared with tumors with other PIK3CA mutations. 89,99 In the SOLAR-1 study, patients whose tumors have PIK3CA mutations in the helical domain (E542*, E545*) or kinase domain (H1047*) had similar PFS benefits. 98 In vitro analyses suggest that double PIK3CA-mutant tumors harbor increased sensitivity to PI3Ki including alpelisib. 91 Such double PIK3CA mutations predominantly occur in the cis-position and have been observed in 15% of HR+/ HER2-and 5% of HER2+ and triple-negative tumors. 91 Intriguingly, HER2+ breast tumor cell lines with and without PIK3CA mutation are more sensitive to inhibition by PI3Ki including alpelisib than cell lines without HER2 amplification. 79,80 Notably, two of eight patients successfully treated for more than two years with alpelisib on SOLAR-1 trial had HER2+ disease. 86 Extrapolation of the findings of any one study to infer overall clinical significance must be tempered, as response to alpelisib may vary depending on numerous clinical and genetic factors including breast cancer subtype, tumor burden, prior therapy, and the presence of concomitant mutations. 74,89,100,101 Over time, a more nuanced approach is likely to emerge that incorporates other alterations known to influence the PI3K pathway such as PTEN, AKT1, and RAS aberrations. 74,89,102,103 Indeed, the target-and/or pathwayspecific focus can inadvertently neglect the influence of other pathways, rendering an oversimplified understanding of the clinical utility of targeted agents and their biomarkers. To this point, current preclinical data suggest combination therapy of alpelisib with other targeted inhibitors including neratinib, vistusertib, and OSI-027 (mTOR), erlotinib (epithelial growth factor receptor), erdafitinib (fibroblast growth factor receptor), AEW541 (insulin-like growth factor 1 receptor), and ribociclib (CDK4/6). Additionally, SGI-1776 and AZD-1208 (PIM kinase inhibitors) may overcome/delay resistance to alpelisib and produce additive/synergistic effects in tumor inhibition. [104][105][106][107][108][109][110][111][112][113] As it is pertinent to breast cancer, alpelisib was synergistic with the CKD4/6 inhibitor ribociclib in triple-negative cell lines and PDX models, 108 and showed synergy with the pan-HER TKI neratinib in inhibiting growth of HER2+ cell lines. 113 In summary, development of a biomarker for PI3Ki is an evolving field. Currently, it is prudent to consider all non-CDx PIK3CA mutations on a case-by-case basis, recognizing that not all mutations identified by NGS assays have confirmed deleterious effects on protein function and a verified ability to respond to alpelisib. In the setting of certain PIK3CA alterations, it is biologically feasible that HER2+ disease may derive a unique benefit from alpelisib or other PI3Ki.
In conclusion, the PI3K pathway is fundamentally important for tumorigenesis of HER2+ tumors and escape from subset of known or likely PIK3CA alterations. Alterations tested either preclinically or clinically with alpelisib (bottom portion of gene) include the amplified gene (not depicted) and span the full length of the gene. These mutations involve the p85-binding, C2, helical, and kinase domains as well as interdomain regions and the c-terminus. To date, 37 known or likely oncogenic non-CDx mutations have been tested in the context of alpelisib administration/exposure (for more information, see Table 3). Red: known activating alteration. Yellow: likely activating alteration. Orange: known and likely activating alteration reported for given amino acid position. a Mutations reported to co-occur with a known activating mutation: K111R, D350N, E453Q, and P539R co-occur with H1047R; M1043L co-occurs with E454K. P85_bd: p85 binding domain; PI3K_rbd: RAS binding domain; PI3K_C2: C2 domain; PIK_helical: helical domain; PI3_PI4_kinase: kinase domain. Activating status determined using OncoKB curated information as annotated in the cBioPortal database (last accessed February 2021). Hotspots E545 and H1047 (marked by broken line in the lollipop) have a much higher frequency of mutations compared with other loci. Lollipop length does not correspond to relative mutation frequency. Gene structure adapted from cBioportal. 94,95 CDx, companion diagnostic; PI3K, phosphatidylinositol 3-kinase.
HER2i that limits the survival of patients with HER2+ disease. Despite the challenges faced in the clinical testing for PIK3CA mutations and the high toxicities and abandonment of many PI3Ki, clinical development of PI3Ki and their predictive biomarkers in HER2+ breast cancer is ongoing. Effective PI3K blockade paired with a sensitive and specific biomarker of response may improve outcomes of thousands of patients with HER2+ breast cancer. | 2022-04-02T06:23:33.654Z | 2022-03-01T00:00:00.000 | {
"year": 2022,
"sha1": "592bbb718a58b22577ae527bf355d3bc565bd632",
"oa_license": "CCBYNCND",
"oa_url": "https://ascopubs.org/doi/pdfdirect/10.1200/PO.21.00370",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "89fd0b44b95c7a577d32745a24075d25182f2bb5",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
55997737 | pes2o/s2orc | v3-fos-license | Peptic ulcers: causes, prevention, perforation and treatment
Peptic ulcer disease (PUD) is a sore on the lining of stomach or duodenum. In some cases, a peptic ulcer may develop just above stomach in esophagus, i.e. esophageal ulcer. Peptic ulcer has always been the most common etiology underlying upper gastrointestinal perforation, while gastric perforation represents 10-15% of all peptic ulcers. Alternation in balance between aggressive and protective factors at the luminal surface of the epithelial cells, leads to development of peptic ulcer. Aggressive factors include Helicobacter pylori, hydrochloric acid, pepsins, nonsteroidal anti-inflammatory drugs (NSAIDs), bile acids, ischemia, hypoxia, smoking and alcohol. On the other hands, bicarbonate, mucus layer, mucosal blood flow, prostaglandins (PGs) and growth factors are the most known defensive factors. The shared symptoms include a dull or burning pain in middle or upper stomach between meals or at night, bloating, heartburn, nausea and vomiting are among the most common symptom of peptic ulcer. Some other symptoms in severe cases are dark or black stool (due to bleeding), vomiting blood, weight loss and pain in the mid to upper abdomen. An ulcer may or may not have all or some of the symptoms. In general, ulcers are best diagnosed by instrumental procedures depending on the case and the available facilities in the centre. The most reliable tests include, I) sophagogastroduodenoscopy (EGD), a thin tube with a camera inserted through the mouth into GIT. A biopsy is also taken during an EGD to be examined for H. pylori, II), X-ray for the upper GIT taken after drinking a thick barium salt, III) blood test to check if there is anemia and IV) detection of blood in the stool. Perforation is the major complication of PUD and the perforated peptic ulcer needs the use of surgical intervention. About 80% of the indications for peptic ulcer surgery are related to perforate peptic ulcers. Most cases are treated by simple suture of the perforated intestinal wall. In most cases, the best treatment is surgical or laparoscopic suture closure of the perforated ulcer. Natural products have found a special place in the treatment plans for peptic ulcer disease. They exhibit their antiulcerogenic activities by various mechanisms either prophylactic or therapeutic or both. The prophylactic products have considerable antioxidant and anti-inflammatory activities. On the other hand, therapeutic agents possess antisecretory or healing effects. The anti-H.pylori activity of some plant extracts, however, may explain their antiulcerogenic activity.
INTRODUCTION
Ulcers in the gastrointestinal tract (GIT) are commonly divided into two types according to their location, i.e. being ulcerative colitis (lower) and peptic ulcer (upper).
Ulcerative colitis (UC) is an inflammatory bowel disease primarily affecting the colonic mucosa.In its most limited form it may be restricted to the distal rectum, while in its most extended form, the entire colon could be involved. 1UC can occur in both sexes and in any age group but most often begins in people between 15 and 30 years of age.The exact causes of UC are still not clear but different factors have been postulated as possible etiologic agents.They are genetic factors, infective agents, immunological basis, smoking, medications and pathological factors. 2 Peptic ulcer disease (PUD), known as a sore on the lining of stomach or duodenum, affects a considerable number of people worldwide.In some cases, a peptic ulcer may develop just above stomach in esophagus, i.e. esophageal ulcer.Peptic ulcer has always been the most common etiology underlying upper gastrointestinal perforation 3 , while gastric perforation represents 10-15% of all peptic ulcers. 4lternation in balance between aggressive and protective factors at the luminal surface of the epithelial cells, leads to development of peptic ulcer.Aggressive factors include Helicobacter pylori, hydrochloric acid, pepsins, nonsteroidal antiinflammatory drugs (NSAIDs), bile acids, ischemia, hypoxia, smoking and alcohol.On the other hands, bicarbonate, mucus layer, mucosal blood flow, prostaglandins (PGs) and growth factors are the most known defensive factors. 5lthough in many cases peptic ulcers heal in the absence of special treatment, their warning signs should not be ignored.When the ulcer is not properly treated serious health problems may occur, including bleeding, perforation (a hole through the wall of the stomach) and gastric outlet obstruction from swelling or scarring that blocks the passageway from stomach to the small intestine Taking NSAIDs can lead to an ulcer without any warning.The risk is especially concerning for the elderly and for those with a prior history of having peptic ulcer disease.
It has been reported that the incidence of PUD have fallen globally since the beginning of 21 st century. 6Many advances have taken place in both diagnosis and management of peptic ulcer disease, including improvements in endoscopic diagnostic and therapeutic facilities, the increased use of proton pump inhibitors and Helicobacter pylori eradication therapies.In spite of all these, peptic ulcer perforation rate has remained unchanged and, therefore, is a major health challenge
CAUSES
A number of factors contribute to the prevalence of peptic ulcer and increasing the risk.The most important causes are classified into the following categories.
• Long-term use of nonsteroidal antiinflammatory drugs (NSAIDs) including aspirin and ibuprofen.
• Helicobacter pylori (H. pylori)infection
• A family history of ulcers As there are some evidences that H. pylori could be transmitted either from person to person or through food and water, the first step to avoid peptic ulcer is preventing infection by improving personal hygiene and eating well cooked foods.
NSAIDs are a class of analgesic medicines including acetyl salicylic acid and ibuprofen.It is a well-known fact that long-term use of NSAIDs can cause peptic ulcer disease.Development NSAIDinduced peptic ulcer is the result of regular use of NSAIDs.It is best to avoid taking medicines that are prescribed by physician with meals and avoid drinking alcohol when taking medicines.
Stress may worsen the signs and symptoms of a peptic ulcer.Smoking interferes with the protective lining of the stomach and increases the secretion of stomach acid leading to higher risk of peptic ulcer.Different foods do not cause or prevent peptic ulcers, but drinking alcohol does make ulcers worse.Excessive use of alcohol can irritate and erode the mucous lining in stomach and intestines, causing inflammation and bleeding.
SYMPTOMS
Small ulcers may not cause any symptoms, while some big ulcers can cause serious bleeding.However, some symptoms are common and general for both small and big ulcers. 7The shared symptoms include a dull or burning pain in middle or upper stomach between meals or at night, bloating, heartburn, nausea and vomiting are among the most common symptom of peptic ulcer.Some other symptoms in severe cases are dark or black stool (due to bleeding), vomiting blood, weight loss and pain in the mid to upper abdomen.An ulcer may or may not have all or some of the symptoms.
DIAGNOSIS
The diagnosis of PUD caused by H. pylori infection is either through invasive or non-invasive methods including endoscopy with biopsy, a urease breath test and detection of antibodies in salivary fluid, serum and urine.The diagnosis of H. pylori infection, should then be confirmed using two tests, one is based on the outcomes of biopsied tissue, either resulting in a culture, or urease test.A protonpump inhibitor and two antibiotics for a period of 14 days, are recommended as the medical treatment.This gives rise to clearance of about 90% of the bacteria and prevents the development of resistant strains. 8n general, ulcers are best diagnosed by instrumental procedures depending on the case and the available facilities in the center.The most reliable tests include, I) sophagogastroduodenoscopy (EGD), a thin tube with a camera inserted through the mouth into GIT.A biopsy is also taken during an EGD to be examined for H. pylori, II), X-ray for the upper GIT taken after drinking a thick barium salt, III) blood test to check if there is anemia and IV) detection of blood in the stool.
PREVALENCE
Despite the development in life style and personal care, the prevalence of H. pylori infection is still quite high in the developing world.However, its prevalence has declined considerably in the western world.The bacterial infection can be transmitted from person-to-person, and children most commonly acquire infection from mothers.Most published studies demonstrate household crowding, sharing a bed with children, and sharing plates, spoons, or tasting food before feeding a child are related to infection in children. 9emorrhagic peptic ulcer could be considered as one of the highest incidences for emergency in many hospitals. 10Although the number of hemorrhagic peptic ulcer cases is decreased worldwide, the overall incidence in elderly patients has remained quite high. 11,12 ignificantly higher incidences of hemorrhagic peptic ulcers in elderly patients than in younger patients have been reported in Norway. 12It is worth indicating that according to literature, elderly patients are defined as individuals 80 years or older in previous studies.
Bleeding of peptic ulcers is mainly due to H pylori infection and high intake of drugs.However, the role of each factor may change based on the prevalence of H pylori and use of NSAIDs.For example, in a prospective survey on 204 patients with peptic ulcer bleeding, it has been demonstrated that the use of NSAIDs, aspirin, and clopidogrel are the most important cause of peptic ulcer bleeding in southern Taiwan. 13
PERFORATION
About 7-10 in 100,000 cases of gastroduodenal peptic ulcer are perforated over time. 14erforation is the major complication of PUD and the perforated peptic ulcer needs the use of surgical intervention.About 80% of the indications for peptic ulcer surgery are related to perforate peptic ulcers. 15he role of non-operative management, laparoscopic versus open laparotomy approach and the type of procedure to be used in emergency situations are still under debate. 16Alternative to surgery include placement of drains or maintenance of naso-gastric decompression which are only mentioned in the literature without indication of some scientific evidences.
The widespread clinical use of H2 receptor antagonists and proton-pump inhibitors, have resulted in a marked reduction of surgery, especially in the case of uncomplicated peptic ulcer.Common complications associated with perforation of peptic ulcer, however, are remained problem for a long period. 17,18 bout 70 years ago, the first report on nonoperative management of PPU was published.It was performed on a series of 28 patients and indicated the mortality rate of 14%, as compared to direct simple closure with omental patch (approximately 20%). 3However, the mortality of PPU treated by the nonoperative approach has significantly reduced since then. 4,19 U is among complicated surgical emergencies and appropriate early management is essential for reducing subsequent problems such as gastrectomy.While the risk of cancer is almost zero for duodenal ulcers, about 6-15% of perforated gastric ulcers (PGU) will have a malignant aetiology. 19,21 Athough the figure seems small, it is important in decision making for treatment to patch or resect.
A research conducted in Royal Infirmary of Edinburgh during 2007-2011, has concluded that almost all perforated gastric ulcers could effectively be managed by laparotomy and omental patch repair. 21However, initial biopsy and follow-up endoscopy with repeat biopsy was recommended as an essential task to avoid missing an underlying malignancy.
However, it is emphasized that urgent repair of perforation has remained the standard approach for PPU in most cases. 22he incidence of perforated PUD depends on geographical area, socio-demographic and possibly some environmental factors. 23It has been reported that in developing countries, younger male patients suffer from perforated PUD, in contrast to developed countries were the patients with perforated PUD are mainly the elderly. 24,25 eavy smoking and alcohol among the young male people in developing countries could be one of the factors causing the higher incidence of PUD.While in the more developed countries the drug ingestion is mostly among older people. 26It is also noted that in the developing countries, the patients with perforated PUD present late to definitive management centres. 27any patients first sought medical assistance from traditional healers and unauthorized medical personnel prevalent in developing countries. 27or children suffering from PPU, the use of an open surgery is preferred as a surgical management. 28On the other hand, treatment of children with complicated peptic ulcer disease, the use of laparoscopy is both effective and safe. 29rescribing an antacid procedure for a stable patient at the time of initial surgery in the case of an adult patient with perforated peptic ulcer is in debate.However, the role of acid lowering medications in children has not been studied extensively.It has been reported that from 29 pediatric patients with complicated PUD, 5 have been managed with an antacid procedure at initial operation. 30In a successful rare case of PPU diagnosed in an Asian nine-year-old, based on a literature review, it was decided to treat the perforation with primary repair and omental buttress 31 .
PREVENTATION
Some important precautions are recommended by specialist in order to reduce the risk of developing ulcer.Some precautions are recommended to prevent the disease such as quieting smoke, avoiding alcohol drink, reducing intake of aspirin and/or NSAIDs.In many cases, taking NSAIDs can cause ulcer without any warning.The risk is especially concerning for the elderly and for those with a prior history of having peptic ulcer disease.Although the peptic ulcer is rare in children and young adults, especially in well-developed societies, they may rarely be at risk by NSAIDs Helicobacter pylori induced ulcers.Some cases have been reported even in western medical reports for developed countries 31 .In fact, acute surgical abdomen in infants and children are relatively few.The diagnosis of possible etiologies needs, therefore, very specialist attention and various diagnostic tests to be examined.However, literature reports about peptic ulcers in children, especially in western countries, is rare and goes back to decades ago. 32,33,34Change of diet and reducing the wrong eating habits such as fast food and various salty and fried snacks are best recommended precautions for children to reduce the risk of peptic ulcer in modern societies.Considering the advanced world and its related busy minds, the risk of peptic ulcer could highly be reduced by lowering the oxidative stress, especially in aged population.A more relaxed and quite life and taking natural antioxidants are always helpful to prevent ulcers.For example, it has been reported that oral use of ethanol extracts from bark of Combretum leprosum Mart.& Eiche (Combretaceae) could have a gastroprotective and anti-ulcerogenic effect due to inhibition of the gastric acid secretion and increase of some mucosal defensive factors. 35urmeric (Curcuma longa Linnaeus, Zingiberaceae) is one of the main spices broadly used in Asian foods, its most important constituent, curcumin, has a considerable inhibitory effect on H. pylori growth.Its anti H. pylori effect is tested both in vitro and in vivo. 36Its use in cooked foods could, therefore, effectively prevent H. pylori induced peptic ulcer.
TREATMENT OF PEPTIC ULCER DISEASE
It is reasonable that hospital volume could affect outcomes of various endoscopic treatments. 37,39,39Hospitals with large case volumes may have more experienced endoscopists who resulting in fewer complications and shorter length of hospital stay (LOS).It has been demonstrated that a higher hospital volume leads to a significantly shorter LOS, i.e. lower medical costs of hospitalization.However, the mortality was not altered when the volumes of cases in various hospital were compared. 40
Synthetic drugs
A severity score method has been purposed to assess quantitatively the PPU.This could help monitoring the evolution of a patient's condition after admission, especially when nonoperative management is the initial action for a patient. 22owever, it should always be remembered that the test does not fully replace the clinical approach, but may supplement individual clinical judgment.It is suggested that the proposed model should be more validated using a larger population for being applicable finally to any PPU case.
It has been reported that about 72% of PPU can be successfully treated by nonoperative models with morbidity and mortality similar to the immediate surgical treatment. 4However, the indication of nonoperative management for PPU has not been well established.The decision making for therapy is correlated to four parameters: age ≥70 years, fluid collection detected by ultrasound, contrast extravasation detected by gastroduodenal imaging, and acute physiology and chronic health evaluation II (APACHE) II ≥8.The score obtained by combining these parameters could be an accurate measure for prediction the need for surgery. 4he severity of peritonitis should be considered when deciding the treatment approach, because secondary peritonitis is more life threatening for PPU patients than peptic ulcer itself.
The water-soluble contrast imaging is important in determining if the perforation is fully closed.When the contrast agent continues to spillover it is suggested that perforation is not well closed leading to continued leakage of gastrointestinal contents.On the other hand, positive results for the two radiological examinations reflect that peritoneal contamination is severe and aggravating.In addition, age and APACHE II score, which reflect the general condition of patients, are also the important parameters suggesting surgical therapy. 4everal clinical strategies are performed for treatment ulcers, including a change in lifestyle, using specialist recommended synthetic or, in some cases, natural medicines and surgery.The specialist is the best person to find out the most suitable treatment for each individual depending on the cause, symptoms and severity of the peptic ulcer.A number of pharmacological agents have proven to be effective in the management of the acid peptic disorders.These groups include: • Antacids, control the pH by neutralization the excess acid, such as aluminum hydroxide and magnesium trisilicate.There are also specific treatments using a combination of various mechanisms, for example, triple therapy consists of a treatment plan for one week.A proton pump inhibitor such as omeprazole and the antibiotics clarithromycin and amoxicillin are prescribed together for a week. 41,42he most common and accepted treatment of peptic ulcers depends on using a number of synthetic drugs that reduce the rate of stomach acid secretion (antiacids), protect the mucous tissues that line the stomach and upper portion of the small intestine (demulcents) or to eliminate Helicobacter pylori (H.pylori).
A considerable number of ulcers are well treated using inhibitors of proton pump (PPIs).On the other hand, the NSAIDs induced peptic ulcer could be best treated if taking those medicines is stopped.
In the case of ulcer caused by H. pylori infection, special antibiotics are prescribed.The multiple mixture choice of antibiotics are recommended to be taken for one to two weeks together with a PPI.Bismuth is also part of some treatment regimens.If the ulcer is bleeding, it can be best treated using an endoscopic treatment.
Combination therapy using two or more specific synthetic drugs or prescription of a synthetic medicine administrated with a natural pharmacological plant part has received special attention recently and the efficacy of each medication is improved. 43,44owever, inhibition of acid secretion is always the first common and most important strategy in treatment plan for a peptic ulcer.During the recent three decades public and scientific understanding from the physiology of gastric acid secretion and its effects on peptic ulcer disease has improved considerably.The design of a possible therapeutic plan to inhibit acid secretion is the main goal of physicians following this understanding. 14During 1970s, histamine-2 receptors (H-2R) the proton pump, that regulates the secretion of gastric acid, were discovered.This important improvement in the field of internal medicine led to the development of H-2R antagonists during the same decade and the design of proton pump inhibitors (PPIs) in the following 1980s.More recent scientific discoveries have confirmed the importance of and the role played by H.pylori infection in pathogenesis of peptic ulcer disease.PPI's have been found well effective to suppress gastric acid secretion.It has been reported that PPIs are considerably more effective than H-2RAs for decreasing morbidity and mortality during nonsurgical management of perforated peptic ulcer. 15
Natural products
In most cases, relapses and adverse reactions is observed following synthetic antiulcer therapy.Therefore, finding and examining a safe medication for management of ulcers with minimum side effects is the serious concern of researchers in pharmacetutical and natural products industries.On the other hand, a considerable number medicinal plants and their secondary metabolites with antiulcer potential have been reported in the scientific literature.
However, some new anti-ulcer agents from natural sources have been introduced during the last 30 years.It is believed that many medicinal plants, herbs and spices, vegetables and fruits are potential sources for control of various diseases including gastric ulcer and ulcerative colitis.A number of medicinal plants and their secondary metabolites with potential anti-ulcer activities have been reported in scientific literature. 46According to the traditional experiences and modern investigations by different scientists, treatment with natural products could result in promising results and perhaps fewer side effects.
Different parts of medicinal plants are traditionally used in various forms, i.e, freshly obtained plant parts, frozen or dried, extracts using different solvents and essential oils.The most acceptable and commonly prescribed medicinal form of natural sources is solvent extracted plant part.A number of polar and less polar solvents have been used to extract secondary metabolites from plants.The most widely used solvents, however, are, methanol, ethanol, diethyl ether, chloroform, ethyl acetate, n-butanol and water. 46Although the choice of solvent is determinant in the active constituents, in preparations for medicinal purposes their toxic side effects should be considered carefully.A very scientific knowledge of the biochemical to extracted, its medical benefits is recommended for these preparations.
It has been confirmed that ethanol extract of turmeric (Curcuma longa Linnaeus, Zingiberaceae) when administrated orally produced significant antiulcerogenic activity in rats.The effect was related to increase gastric wall mucus leading to restoration of the non-protein sulfhydryl (NP-SH) content in the glandular stomachs of the rats. 47 is well documented that green tea extract (Camellia sinensis L.) can effectively treat ulcerative colitis. 48Drinking green tea at a daily dose can also reduce body weight and act control diarrhea.It is purposed that it can inhibit the disruption of the colonic architecture, reduction of myeloperoxidase (MPO).
Most apple species have been traditionally used for their gastrointestinal effect and their role in digestion, which is now known to be due to the presence of cartenoids.For example, seven carotenoids have been identified in the peel extract of Malus domestica Borkh.Rosaceae (Golden delicious apple).These secondary metabolites have shown to exhibit a potent anti-H.pylori activity. 49he aqueous extract of Enantia chlorantha Oliv.(Annonaceae) stem bark possesses both in vitro and in vivo activities against H. pylori. 50inger is a common spice used in traditionally in many foods deserts and drinks, especially in Asian countries.It has been shown that aqueous extract of ginger roots possess strong antiulcer properties.
This property is related to augmentation of mucin secretion and decreased cell shedding rather than offensive acid and pepsin secretion. 51he gastroprotective effect of methanolic extract of Terminalia arjuna (TA) has been reported on diclofenac sodium induced gastric ulcer of rats with effective dose of 400 mg/kg body weight. 52The results showed a considerable reduction in lesion index of ulcer induced animals treated with TA compared to ulcerated rats not treated using TA.They also reported a significant increase in pH, NP-SH, GSH and a number of enzymic antioxidants.On the other hand, a significant decrease in volume of gastric juice, free and total acidity, pepsin concentration, acid output, LPO levels and MPO activities in TA treated rats was observed as compared to non-treated animals.It was, therefore, concluded that the free radical scavenging activity of T. arjuna makes it a suitable gastroprotective agent. 52ymenaea stigonocarpa Mart.ex Hayne (MHs) is a medicinal plant found in the Brazilian savannah.Its anti-ulcer effect has been investigated in experimental rodent models. 53It has been found that MHs could display a considerable gastroprotective effect on experimental gastric and duodenal ulcers.It is suggested that these effects are related to the presence of condensed tannins and flavenoids in the plant extracts.
The ethyl acetate extract of Saussurea lappa C.B. Clarke (Asteraceae) has shown to exhibit antiulcerogenic properties.It also possesses strong activity against peptic ulcer through a cytoprotective effect. 54he ethanol extracts of Encholirium spectabile Mart.(Bromeliaceae) aerial parts has a considerable protection effect on gastric mucosa against ulceration. 55he ethanolic extract of coconut seed (Cocus nucifera L., Arecaceae) when used orally could treat peptic ulcer as compared to control subjects. 56ral administration of Erythrina indica L. (Febaceae) extract has shown to possess significant antiulcer properties in rats with peptic ulcer induced by indomethacin.The observed effect is partly due to the presence of polyphenolic compounds in methanolic leaf extracts. 57he extracts of some medicinal plants have antibacterial activity which could be effective on peptic ulcers induced by H. pylori.Antibacterial and antioxidant activity of medicinal plants can be examined in vitro.The in vitro assay of essential oil obtained from Apium nodiflorum L. (Apiaceae) has shown strong antibacterial activity against H. pylori, resulting in minimum inhibitory concentration (MIC) value of 12.5 μg/ml. 58he oral administrated of an ethanol extract from Combretum leprosum Mart.has shown to have gastroprotective and anti-ulcerogenic effects.It is suggested that the extract could inhibit secretion of gastric acid and increase factors that provide defensive activity of mucosal system. 35he wound healing mechanisms of Rhizophora mangle L. extract is based on the formation of a thick coating of plant extract which adheres macroscopically to the gastric mucosa.It then forms a physical barrier the same as the topical wounds. 59hile aloe vera, ginger and honey are best active against both peptic ulcer and ulcerative colitis, licorice, turmeric, Al-Hagnah (Desmostachia bipinnata), Catinga de Bode (Ageratum conyzoides), Chamomile and ginger have been found to be considerably effective in the treatment of peptic ulcer through their cytoprotective effect in addition to their anti H. pylori effect. 60
Psychological Intervention
It has been noted recently that the psychological conditions play a key role in the development of PUD and recovery from it.The PUD patients usually suffer from emotional disorders including depression and anxiety.Some of the patients suffering from PUD have experienced emotional disorders such as anxiety and depression.Therefore, the exclusive drug therapy may not be well effective and could result in recurrent.
It is known that behavior cognitive therapy is a reliable method leading to improvement of the patient's reality cognition, anxiety, depression and discontent emotions.Psychological intervention can eliminate the irrational way of thinking, as well as the mental and behavior disorders.On the other hand, it is a universally accepted fact that a highly stressful life may lead to acute peptic ulcer. 43he efficacy of psychological intervention against ulcer has been studied on 96 PUD patients using a mental intervention together with drug therapy. 61In the mentioned research, the patients were divided into two groups, trial and control.To the control group the medical advice and Tagamet (800 mg daily) for 6 weeks.The trial group also received the same drug therapy together with the psychological interventions.Their results indicated that healing rate of trial group was significantly higher than control.
Last and not least, it should always be considered that many factors may lead to important and, sometimes dangerous, side effects.The possibility of adverse drug resistances (ADRs) to occur is also of prime importance when designing a treatment plan or a drug regim.A number of factors are associated with ARD occurrence, some of which are patient related, drug related or socially related.Age, race, alcohol intake, smoking, kidney and liver problems, drug dose and frequency, gender, pregnancy, breast feeding, and many other factors may contribute.
SUMMARY
Based on literature review performed in this study, the following conclusions are made.Peptic ulcer disease secondary to H. pylori infection is important to be diagnosed due to the high reported incidence of recurrence.
Perforated gastro-duodenal ulcer is normally managed from conservative non-operative therapy to radical surgical treatment (gastrectomy).Most cases are treated by simple suture of the perforated intestinal wall.In most cases, the best treatment is surgical or laparoscopic suture closure of the perforated ulcer.The first step should be treatment of H.pylori infection regardless of chosen intervention.Perforated gastric cancer can be treated urgently by simple suture closure, followed by a second-stage carcinologic gastrectomy without compromising survival.Perforation from peptic ulcer disease is adequately treated with primary closure, omental buttress, and medical management of the underlying etiology.
The incidence of perforated peptic ulcer in children has been decreasing in industrialized, well developed countries.Laparoscopy is a safe and effective tool in the surgical management of complicated peptic ulcer disease in children.
Anxiety and depression can postpone gastric digestion and emptying process.The overloaded gastrointestinal function could make the base for prevalence of peptic ulcer.They also aggravate the somatic symptoms of the ulcer leading to the negative emotions reversely in a vicious circle.Therefore, a psychological intervention amending the circle is an important step for ulcer therapy.In sum, psychological intervention combined treatment is suggested to be considered with higher emphasis in the future medical care.
Natural products have found a special place in the treatment plans for peptic ulcer disease.They exhibit their antiulcerogenic activities by various mechanisms either prophylactic or therapeutic or both.The prophylactic products have considerable antioxidant and anti-inflammatory activities.On the other hand, therapeutic agents possess antisecretory or healing effects.The anti-H.pyloriactivity of some plant extracts, however, may explain their antiulcerogenic activity.
Consumption of a healthy diet composed of fruit pulp and various vegetables could give rise to duodenal healing effects.The observed effects by natural plant extracts may be due to their antioxidant effect and the presence of condensed polyphenols and flavonoids present in their various parts. | 2018-12-07T21:13:50.086Z | 2016-12-22T00:00:00.000 | {
"year": 2016,
"sha1": "2743c9faddd49671362f2fe2e4d45dac181089fe",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.15562/ijbs.v10i2.121",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "2743c9faddd49671362f2fe2e4d45dac181089fe",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
258588285 | pes2o/s2orc | v3-fos-license | A Neural Emulator for Uncertainty Estimation of Fire Propagation
Wildfire propagation is a highly stochastic process where small changes in environmental conditions (such as wind speed and direction) can lead to large changes in observed behaviour. A traditional approach to quantify uncertainty in fire-front progression is to generate probability maps via ensembles of simulations. However, use of ensembles is typically computationally expensive, which can limit the scope of uncertainty analysis. To address this, we explore the use of a spatio-temporal neural-based modelling approach to directly estimate the likelihood of fire propagation given uncertainty in input parameters. The uncertainty is represented by deliberately perturbing the input weather forecast during model training. The computational load is concentrated in the model training process, which allows larger probability spaces to be explored during deployment. Empirical evaluations indicate that the proposed model achieves comparable fire boundaries to those produced by the traditional SPARK simulation platform, with an overall Jaccard index (similarity score) of 67.4% on a set of 35 simulated fires. When compared to a related neural model (emulator) which was employed to generate probability maps via ensembles of emulated fires, the proposed approach produces competitive Jaccard similarity scores while being approximately an order of magnitude faster.
Introduction
Wildfires are a destructive phenomenon affecting natural flora and fauna as well as posing many risks to human life and property. Response and management efforts are critical to mitigating the damage caused by wildfires [7,8,40]. To this end, in silico wildfire simulations are often performed in order to aid managers in making informed choices for fire risk management and mitigation efforts, as well as evacuation planning.
There are a number of computational fire spread models that are currently deployed for simulating and predicting the spreading behaviour of fires [17,33,45]. These models typically take into account underlying empirical observations of fire spread given specific conditions, such as fuel class (eg. vegetation type), terrain characteristics (eg. slope), as well as weather (eg. temperature, wind speed and direction). These simulations are usually deterministic [33,45], and predictions depend on static initial conditions (initial fire location, fuel class, terrain maps) and estimates of future weather conditions. As wildfire spread is inherently a stochastic process, a single simulation of fire progression is unlikely to account for uncertainty in the underlying variables driving the model. For example, there are variabilities within fuel classes, uncertainty in classifying a fuel, and uncertainty in weather forecasting. As such, it is desirable to produce likelihood estimates of fire location (probability map) at a given point in time, in order to provide a form of uncertainty quantification.
One approach for dealing with uncertainty in a given variable is to run an ensemble of simulations, where each simulation draws the value of the variable from a distribution. Ensemble predictions are a standard approach in areas such as weather and flood forecasting [19,30,48]. However, a major limitation of ensemble-based simulations is that the parameter space increases exponentially with the number of variables. The size of the ensembles is hence constrained by the computational demands of the simulations, which in turn can limit the scope of uncertainty analysis to a subset of variables.
To address this shortcoming, in this paper we explore the use of a spatio-temporal neural-based model to directly estimate the output of an ensemble of simulations. The proposed model is trained via probability maps, which are generated from ensembles of fire-fronts produced by the traditional SPARK simulation platform [33]. Within the training ensembles, weather parameters are perturbed according to a noise model. The computational load is hence concentrated in the model training process, rather than during deployment. Empirical evaluations show that the proposed model generalises well to various locations and weather conditions, and generates probability maps comparable to those indirectly generated via ensembles of SPARK simulations.
The paper is continued as follows. Section 2 provides a brief overview of related work. Section 3 describes the proposed model architecture. Section 4 provides comparative evaluations. We conclude by discussing the results, limitations and future work in Section 5.
Related Work
Recent work has investigated use of machine learning methods for wildfire related applications, covering topics such as susceptibility prediction, fire spread prediction, fire ignition detection and decision support models [2,7,24]. Deep learning architectures have also been used to develop models related to wildfires [4,9,22,37]. Burge et al. [9] and Hodges et al. [22] present convolutional neural network (CNN) emulators that estimate fire spread behaviour, while Allaire et al. [4] present a CNN emulator for hazard assessment within a target region.
Neural networks have also been used to predict the likelihood of a region being burned, given the confidence the model has in its own prediction. Radke et al. [37] propose a CNN based emulator that predicts the likelihood a pixel is burned within a 24 hour period. Sung et al. [44] also develop a model that predicts the likelihood that a pixel is burned; the model is trained using daily fire perimeters, rather than synthetic data. Dabrowski et al. [12] combine a Bayesian physics informed neural network with level-set methods [31] for modelling fire-fronts. Outside of the fire space, Gunawardena et al. [20] deploy a hybrid linear/logistic regression based surrogate model to estimate the behaviour of ensemble weather modelling. Egele et al. [16] propose an automated approach for generating an ensemble of neural networks to aid uncertainty quantification.
In a separate line of research, surrogate models (also known as model emulators) can be used as more computationally efficient representations of complex physical process models [3,39]. Such models are often based on statistical or machine learning approaches. Models of various physical systems have been achieved using techniques such as logistic regression [20], Gaussian processes [5,11], random forests [18,29] and deep neural networks [25,26,39,42].
Potential advantages of neural network based model emulators include accelerated computation through the use of Graphics Processing Units (GPUs) via open-source packages such as TensorFlow [1] and PyTorch [36]. Furthermore, since neural network models are in essence a large set of numerical weight parameters, they are portable across computing architectures, easily deployed in cloud-based computing environments, and can be integrated into larger frameworks with relative ease [39].
Model Architecture
We extend the spatio-temporal neural network model architecture presented in [6]. As illustrated in Fig. 1, the architecture comprises an Encoding block, an Inner block, and a Decoding block. Compared to [6], the inner block is expanded with two layers: layer C implements edge detection, and layer G incorporates update constraints.
There are three types of inputs to the model: (i) initial probability map, (ii) spatial data and forcing terms, (iii) weather time series. The probability map consists of an image where pixels values represent the likelihood of the fire reaching a location at a given point in time. The spatial data are images representing height maps and land classes (eg. forest, grassland). The forcing terms are curing and drought factors. The weather time series consists of the mean and standard deviations of temperature, wind (x and y components), and relative humidity. These are summary statistics of the values used in the ensemble of simulations used for training.
The encoder and decoding steps are as described in [6]. The encoding step is used to create a more compact latent representation of features. This is illustrated by the light blue shading in Fig. 1. The probability map is encoded using an autoencoder, illustrated in Fig. 2. The relationship between pixel value and probability is preserved in the encoding process. The autoencoder component of the model is trained separately and its weights are frozen during the main training step. This transfer learning approach ensures that dynamics are not inadvertently learned by the encoding/decoding layers, as well as reducing the number of weights that need to be trained by the full model. The height map and landclass map features are encoded using alternating convolutional and strided convolutional layers which reduce the spatial extent.
The inner model is designed to handle the dynamics of the system, incorporating a shallow U-Net structure [38,46]. See Fig. 3 for details. New latent probability map estimates are generated, which are then propagated forward in time as new weather values are processed. The updates are also informed by the latent spatial and climate features.
Within the inner model, layer C applies a Sobel edge filter [13], where the output is x and y edge components. The edge detection aims to make it easier for the neural network to identify where fire fronts are within the probability map. This approach has parallels with saliency guided training [23]. Fig. 4 shows examples of edge magnitudes.
Layer G implements constrained updates to the latent probability map. A new map P t+1 at time t + 1 is generated using the previous map P t and update terms from F t at time t, where F is the output layer from the U-Net component of the model (see Fig. 3). F uses a sigmoid activation function so that values lie on the [0, 1] interval [10].
More specifically, layer G acts to update the latent probability map as follows. Let F t ij denote the value of the i-th kernel of F at pixel location j at time t. Let P t j ∈ [0, 1] be the probability that pixel j has been burned by time t. The model is trained such that F t ij terms estimate the likelihood that fire spreads from a surrounding local region to location j at time t + 1. The spread characteristics and likelihoods are learned by the model in the U-Net component. The probability that pixel j has been burned at time t + 1 is estimated using: In the above equation, the updated likelihood is given by 1 minus the probability that the pixel is not burned by an exterior source 1 − F t ij , and is not already burned 1 − P t j . The update satisfies the following three constraints: 1. The likelihood a pixel is burned is non-decreasing and bounded by 1, ie., 1 ≥ P t+1 j ≥ P t j . 2. The likelihood a pixel is burned is unchanged if there are no external sources, ie., P t+1 If there is at least one source that is very likely to burn a pixel, then the likelihood the pixel is burned approaches 1, ie., P t+1 We have empirically observed that in order for the model to converge well in training, it is necessary to ensure that the F i terms are sufficiently small so that the latent probability map is gradually updated between iterations. More specifically, preliminary experiments indicated that the model is unlikely to converge during training when using the default value of 0 for the initial bias of layer F . As such, we have empirically set the initial bias to −5.0. [6]. Key additions are in the inner component: edge detection (layer C) and constrained update layer (layer G). The initial probability map and features are encoded via down-sampling. The inner component of the model updates the latent probability map using spatial features and weather variables (see Fig. 3 for details). Once all weather inputs have been used a final latent probability map estimate is produced and decoded via up-sampling. The layers used to encode and decode probability maps are trained separately as an autoencoder (see Fig. 2 for details). The latent arrival state is concatenated with this stack and is passed through a shallow U-Net [38]. The resulting output is used as a residual component. This residual is added to the latent arrival state, which is then used in the next time step. When the weather inputs are exhausted, the final latent arrival state is the output.
Data Preparation
We use a dataset comprised of 180 fire probability maps, split into 145 training and 35 validation samples. Likelihood estimates are given in half hour intervals throughout the duration of the sample. To generate each probability map, we produce an ensemble of 50 fire perimeter simulations using the SPARK platform [33]. For each time interval, the fire perimeters are calculated and the ensemble is flattened into a probability map. See Fig. 4 for an example.
The location of the fires is representative of regional South Australia. Realistic land classification maps and topology data are used. There is a large bias in land class distribution towards temperate grassland (55.4%), sparse shrubland (11.3%), water/unburnable (10.6%) and mallee heath (10.1%). Other classes represent 12.6% of the total area. Weather conditions are sampled from days where the forest fire danger index is severe [14,32]; the temperatures are typically high and the relative humidity is low.
The fires within an ensemble are identical, except that each time series of weather values is perturbed using the following noise model. For proof-of-concept we use a straightforward random walk noise model. At each time step (30 min interval), ±3% of the mean variable value is added. We choose 3% based on preliminary analysis; this value is large enough so that simulations within an ensemble display distinctly different features, while being small enough that the general shape of each fire is similar. Let x i be the noise added at time i, V j the noiseless variable from the time-series V at time j. We then define The weather variables form the time series and are wind speed, wind direction, temperature and relative humidity. Due to use of degrees for wind direction within the weather data, the wind direction was perturbed by ±2°at each step, rather than 3% of the mean value .
There are N = 50 unique time series for each variable (eg. temperature) of length T . Rather than designing the model to use N inputs at each time step for each variable, we use their mean and standard deviations, taken across the ensemble at a given time. This produces an input time series of size (2, |V |, T ), where |V | = 4 is the number of weather variables Several pre-processing steps are applied to the features before they are presented to the model. The sample bounds, coordinate reference system, and resolutions (30m) of all spatial data are all made identical. Height maps are converted into gradient maps. The wind components of the weather data are converted from wind speed and direction to x and y components. Normalisation of weather and forcing terms is performed so that values fall within the range −1 to +1.
Land classification datasets derived from Department of Agriculture and Water Resources Land Use of Australia 2010-11 dataset. Topography data sets derived from Geoscience Australia SRTM-derived 1 Second Digital Elevation Models Version 1.0. Meteorological time series data sets derived from Australian Bureau of Meteorology automated weather station data. Percentage of the mean value cannot be used here, due to the nature of the measurement used for wind direction (degrees). If the mean value was used, winds with mean direction of 0°would never be perturbed, while winds with mean direction of 359°would have a large perturbation.
Training
The model training process performs further modification to the data, as follows. When a sample is selected, a random time within the burn duration is chosen as the initial probability map. The final probability map is taken D intervals later, and the corresponding weather variables are extracted.
Once a starting time is selected, a random square of the probability map is cropped, such that the central pixel j has a value, x j ∈ (0, 1), ie., excluding 0 and 1. We center the cropping on this pixel so that the cropped region is likely to contain pixels that have an intermediate probability of being burned. Such regions are likely to undergo dynamic changes within the time interval D used for training. We have empirically observed that in regions where pixel values are dominated by 0 or 1, the values are less likely to change during the time interval D. Fig. 5 shows an example of a cropped initial probability map suitable for training.
Each sample then undergoes a random rotation and/or reflection operation. This, as well as randomly choosing a starting time and a cropping location, is used for data augmentation with the aim of preventing over-fitting. The reduction in size reduces memory requirements, while the standardisation of sizes allows training to be batched.
The autoencoder component is trained using mean squared error (MSE) loss. For training the model, a loss function is used that evaluates how well the model reduces the MSE between predicted (y p ) and target (y t ) fire shapes compared to the MSE loss between target and initial (y 0 ) fire shapes. Let MSE(a, b) be the pixelwise mean squared error between images a and b. The loss L of fire state P is defined as L(P) = log 10 MSE(y p , y t ) + τ / MSE(y 0 , y t ) + τ . The term τ = 10 −12 is used to remove singularities that arise if either MSE value approaches zero.
Models were implemented using TensorFlow [1], with the Adam optimiser [28] and a batch size of 16. The autoencoder was trained for 25 epochs and returned an MSE loss of 2.9 × 10 −4 . The model was trained for 50 epochs. Unless explicitly stated otherwise, all layer activation functions in the model are Leaky Rectified Linear Units [10,27].
Evaluation
We investigate training the model using D = 1 (30 min) and D = 4 (2 hours) difference between initial and final probability maps. In the former case, individual fires within the probability map may not significantly diverge, and the behaviour may be difficult for the model to learn. In the latter case, the dynamics are allowed more time to develop, but this comes at the cost of more invocations of the inner model. This recursive behaviour may lead to issues regarding vanishing/exploding gradients [21]; we found that the model weights failed to converge when using D = 8.
We also evaluate two configurations (A and B) of the inner model (shown in Fig. 3). In configuration A, layer C acts as an identity operation, simply passing through the latent probability map. In configuration B, layer C applies a Sobel edge filter with kernel size 3. We also train the model under various crop window sizes c. The amount of padding p is increased in proportion with the window size. For example a cropped square of 256 pixels has a padding size of 32 pixels, and a cropped square of 512 pixels has a padding size of 64 pixels.
We use two metrics to evaluate the performance: weighted MSE (WMSE) and Jaccard index. The WMSE is the average of MSE weighted by the amount of area that is burned in each sample. The area of land likely to be burned between initial and final states is given by A(y k t , y k 0 ) = y k t − y k 0 for a sample k. We then define WMSE(y t , y p , y 0 ) = 1/( k A(y k t , y k 0 )) k MSE(y k t , y k p ) × A(y k t , y k 0 ), where we sum over all k samples. Lower values of WMSE indicate better performance.
The Jaccard index (aka Jaccard similarity coefficient) is defined as the intersection between two sets divided by their union [15]. We use the Jaccard index to gauge how similar the probability maps directly generated by the proposed model are to the probability maps indirectly generated via SPARK [33] (ie., where an ensemble of simulations is flattened into a probability map), with SPARK derived data treated as the ground-truth. Higher values of the Jaccard index indicate higher similarity. However, direct use of the Jaccard index on probability maps can be problematic, as there is no implicitly defined boundary on which to evaluate the index. As such, we use a threshold of 10% to define the boundary of the target and prediction sets (ie., probabilities less than 10% are ignored). This was chosen in order to evaluate the model's ability to capture the behaviour of reasonably likely events, while not penalising the model for missing highly unexpected behaviour. Furthermore, the 10% threshold may meet the needs of fire risk management (eg., decision planning and responsiveness), where low likelihood events are less likely to be considered as actionable.
Results presented in Table 1 show that D = 4 leads to better performance than D = 1. Furthermore, we observe a general reduction in the WMSE as the size of the cropped region is increased from 256 to 512 pixels. In terms of the Jaccard index, there is a notable improvement with configuration B compared to configuration A, which indicates that fire front detection via Sobel edge detection is a useful feature for determining ensemble fire dynamics. Fig. 6 provides a graphical example of the performance on a sample fire over a long duration (14 hours). This example was produced using configuration B and a cropping size of 512 pixels during model training. Likelihoods below the 10% threshold have been truncated and are not shown.
Comparison
As the proposed model is an evolution of the model presented in [6], in this section we evaluate the quality of probability maps directly generated by the proposed model (as per Section 4.3) against the quality of probability maps indirectly generated from ensembles of emulated fires produced via the model in [6]. In the latter case, the model is trained using noiseless weather variables on the same training dataset as the proposed model.
The quality is evaluated in terms of the Jaccard index (with contours defined at the 10% likelihood level), where we measure how similar the probability maps are to those indirectly generated via SPARK [33], which are treated as ground-truth. We also measure the time taken (in seconds) for each model to generate the probability maps, running on the same hardware (a modern computer with a multicore CPU and high-end GPU) and using the same software stack as per Section 4.2. For the direct method we choose the configuration with the lowest training and evaluation loss, which corresponds to setting shown in the last row of Table 1. For the indirect method, we vary the ensemble size for generating the probability maps; three ensemble sizes are evaluated: 10, 20 and 50.
The results shown in Table 2 indicate that for an ensemble size of 50 (the same size as used for training the direct approach), the indirect approach (ensembles of emulated fires produced via the model in [6]) produces somewhat higher quality maps. However, this comes at the cost of requiring considerably longer execution time, where the indirect approach is about 19 times slower than the direct approach. For an ensemble size of 20, the indirect approach produces probability maps which are roughly comparable in quality to the direct approach, but is about 7 times slower. For an ensemble size of 10, the indirect approach produces considerably lower quality maps and is still noticeably slower.
Overall, for the indirect approach there is a correlation between execution time and the quality of the resultant probability maps. However, increasing the size of the ensemble from 20 to 50 only leads to a minor increase in quality, at the cost of a large increase in the execution time. The considerably lower execution time of the proposed direct approach (while still producing reasonable quality probability maps) allows for the evaluation of more fire progression scenarios within a given time budget. This in turn can be useful for fire risk management, where large ensembles are typically used to explore large variable spaces. fire probability maps in terms of the Jaccard index (with contours defined at the 10% likelihood level) on the validation set, using SPARK [33] derived maps as ground-truth. Two methods are evaluated: ensembles of emulated fires produced via the model in [6], and the proposed model which directly generates probability maps. The proposed model uses the same setting as shown in the last row of Table 1. Execution time is reported in two forms: number of seconds, and relative to the execution time of the direct approach.
Concluding Remarks
In this paper we have demonstrated that deep neural networks can be constructed to directly generate probability maps for estimating wildfire spread. This direct approach can be used instead of indirectly obtaining probability maps via ensembles of simulations, which are typically computationally expensive. In comparison to ensembles of simulated fires obtained via the traditional SPARK platform [33], the proposed approach achieves reasonable quality probability maps when weather forecasts are exposed to random walk noise. On an evaluation set of 35 fires, the overall Jaccard index (similarity score) is 67.4%. Qualitatively, the proposed approach reproduces the correct general response to fuel class and wind direction, and the shape of the probability maps is comparable to those produced by traditional simulation.
The computational load for the proposed approach is concentrated in the model training process, rather than during deployment. When compared to a related neural model (emulator) which was employed to generate probability maps via ensembles of emulated fires, the proposed model produces competitive Jaccard similarity scores while being considerably less computationally demanding.
We observe that the model does not generate a perfectly smooth transition of probabilities from high to low, as observed in probability maps from ensembles of simulated fires. This will likely limit the capability for the model to predict accurately over long durations. Further refinement of the architecture, model restraints, summary variables and incorporation of variable distributions may yield better results.
A further area of research would be the incorporation of better noise models for weather variables. In this work we have used a straightforward noise model, aimed for demonstrating proof-of-concept. This noise model was not developed to be representative of the complexities in weather forecasting. A more detailed discussion of more comprehensive weather models that incorporate uncertainty can be found in [34,35,43]. In our noise model, any two fires within an ensemble are likely to smoothly diverge from each other over time, except in cases where natural barriers cause bifurcating behaviour. In more realistic settings, bifurcation may occur due to rapid changes in wind direction that are predicted to occur at various times within an ensemble of simulations. It may be possible to capture the behaviour of more complex noise models by using binning methods to generate summary variables, rather than simply using mean and standard deviation.
More sophisticated fire spread models (for example ones that incorporate ember attacks [41,47]) may require very large ensembles in order to reliably gauge the behaviour of fires. Generating likelihood estimates through very large ensembles is likely to be computationally taxing. Direct generation of probability maps, as proposed in this work, may be very beneficial in these contexts. Finally, we note that the proposed approach is not specific to wildfire prediction, and other similar spatio-temporal problems may benefit from direct likelihood estimation, such as diffusion of pollution, pest and disease spread, and flood impact models. | 2023-05-11T01:16:23.972Z | 2023-05-10T00:00:00.000 | {
"year": 2023,
"sha1": "68f6121cd4f813ea6c65d207d0d8e214a69ef776",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "68f6121cd4f813ea6c65d207d0d8e214a69ef776",
"s2fieldsofstudy": [
"Environmental Science",
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Engineering"
]
} |
268276528 | pes2o/s2orc | v3-fos-license | Damage sensing through TLR9 Regulates Inflammatory and Antiviral Responses During Influenza Infection
Host response aimed at eliminating the infecting pathogen, as well as the pathogen itself, can cause tissue injury. Tissue injury leads to the release of a myriad of cellular components including mitochondrial DNA, which the host senses through pattern recognition receptors. How the sensing of tissue injury by the host shapes the anti-pathogen response remains poorly understood. In this study, we utilized mice that are deficient in toll-like receptor-9 (TLR9), which binds to unmethylated CpG DNA sequences such as those present in bacterial and mitochondrial DNA. To avoid direct pathogen sensing by TLR9, we utilized the influenza virus, which lacks ligands for TLR9, to determine how damage sensing by TLR9 contributes to anti-influenza immunity. Our data show that TLR9-mediated sensing of tissue damage promotes an inflammatory response during early infection, driven by the epithelial and myeloid cells. Along with the diminished inflammatory response, the absence of TLR9 led to impaired viral clearance manifested as a higher and prolonged influenza components in myeloid cells including monocytes and macrophages rendering them highly inflammatory. The persistent inflammation driven by infected myeloid cells led to persistent lung injury and impaired recovery in influenza-infected TLR9−/− mice. Further, we show elevated TLR9 activation in the plasma samples of patients with influenza and its association with the disease severity in hospitalized patients, demonstrating its clinical relevance. Overall, we demonstrate an essential role of damage sensing through TLR9 in promoting anti-influenza immunity and inflammatory response.
Introduction:
Toll-like receptors (TLRs) are ubiquitously present innate sensors of specific molecular patterns that are associated with pathogens (PAMPs) or tissue damage (DAMPs) 1 .Sensing pathogens by TLRs leads to an anti-pathogen response while damage sensing is often associated with tissue reparative phenotype 1,2 .However, how the damagesensing contributes to the anti-pathogen response remains incompletely understood.In a clinically relevant infection, both pathogen-and damage-associated molecular patterns are present simultaneously, making it difficult to dissect the effect of damage sensing on the anti-pathogen response from those generated by pathogen sensing.
Toll-like receptor-9 (TLR9) specifically detects bacterial DNA due to the presence of highly unmethylated CpG domains, enabling it to distinguish it from host DNA [3][4][5] .In host cells, while nuclear DNA undergoes methylation, mitochondrial DNA (mtDNA) remains highly unmethylated and is detectable by TLR9 6 .During infection-induced tissue injury, mtDNA is often released from dying cells, making mtDNA an important DAMP sensed by TLR9.To understand how the damage signaling contributes to the host defense independent of pathogen sensing, we utilized influenza viral infection, which lacks TLR9 ligands.
Influenza infection is one of the leading causes of infectious disease-induced mortality worldwide with an annual death toll of approximately 500,000 7,8 .Influenza-mediated pandemics have led to some of the biggest loss of human life 9 and seasonal influenza infections remain a leading cause of morbidity and mortality 7 despite the widespread availability of vaccines and various antiviral agents.Influenza A often has a zoonotic origin including avian and swine.It is interesting to note that birds lack TLR9 10 despite influenza being a prominent avian pathogen, emphasizing the dispensable role of TLR9 for influenza sensing.
In light of this knowledge, we sought to determine how tissue damage sensing during influenza infection through TLR9 shapes host response against the influenza virus.In this study, we investigated the role of TLR9 in influenza infection using TLR9-/-mice, meyloid specific TLR9-/-mice, an ex vivo cell culture system, human TLR9-/-cells, and peripheral blood samples from patients infected with influenza.Our data show that TLR9 deficiency limits the early inflammatory response and lung injury during influenza infection which comes at the cost of impaired viral clearance.The absence of TLR9 rendered lung cells, especially myeloid cells, susceptible to influenza infection.Furthermore, influenza infection of myeloid cells rendered them highly inflammatory, promoting persistent inflammatory response even after viral clearance.The persistent inflammatory response prolonged the tissue injury and impaired recovery of the host.Deletion of TLR9 in myeloid specific manner preserved early inflamamtory resopnse but ameliorated inflamamtory resposne during recovery phase without affecting viral clearance.Taken together, our findings highlight the key role of damage sensing through TLR9 in promoting pathogen clearance and limiting inflammatory response.
Results: TLR9 promotes early inflammatory and injury responses during influenza infection:
To understand the role of TLR9-mediated damage sensing in the host response to influenza infection, we infected TLR9-/-and wild type mice (Both B6 background) with 10 PFUs of PR8 11 , a strain that causes significant lung tissue damage 11 .Mice were euthanized at days 4, 7, and 14 post-infection to understand the impact of damage sensing on the time course of infection, including repair.Our data show that WT and TLR9-/-mice had similar inflammatory and injury responses on day 4 (Sup Fig. 1).However, on day 7, relative to WT mice, TLR9 deficient mice demonstrated substantially lower white blood cell (WBC) counts in their bronchoalveolar lavage (BAL, Fig 1A), suggesting a decrease in the overall inflammatory response in the lung.Next, using multicolor flow cytometry (Sup Fig. 2), we found that this reduced WBC count was largely due to decreased macrophage and monocyte recruitment in TLR9-/-mice (Fig. 1B-E).In parallel with this inflammatory response on day 7, TLR9 -/-mice showed a significantly decreased lung injury response as shown by lower BAL total protein content (Fig 1F ), RBC counts (Fig 1G), platelet counts (Fig. 1H), and RAGE levels (Fig. 1I).
To further characterize mediators of these ameliorated inflammatory and injury responses, we profiled cytokines associated with inflammatory response (IL-1b, IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p70, IL-15, IL-27p28, IL-33, IP-10, KC/GRO, MCP-1, MIP-2a, MIP-2, IFNg, and TNFa) in the BAL on post-infection days 4 and 7. Our data show that on day 4, relative to WT mice, there was a reduction in the macrophage/monocyte cytokines such as MIP-1a and MIP-2, along with a decrease in IL-6 and IL-9; both mice exhibited similar levels of these soluble mediators on day 7 (Fig. 1J-M).In contrast, we observed a significant increase in IL-2 levels on day 7 (Fig. 1N).We observed no significant changes in the other cytokine concentrations between WT and TLR9 -/-mice on day 4 or 7.These findings suggest that injury sensing through TLR9 mediates both inflammatory and injury responses in the lung in the early stages of influenza infection.
Damage signaling through TLR9 promotes viral clearance independent of interferon response:
Having found the contribution of TLR9-mediated damage sensing in lung inflammation and injury response to influenza, we then endeavored to identify the role of TLR9 in viral clearance.On day 7, relative to their WT counterparts, TLR9 deficient mice exhibited higher viral loads by rt-qPCR (Fig. 2A) and by Western blotting of influenza NS1 protein (Fig. 2B and C).However, we did not find any difference in replicating virus as measured by TCID50 assay (Fig. 2D), indicating a role of TLR9 signaling in promoting non-productive viral replication.The elevated viral protein signature persisted even at day 10 post-infection before getting cleared by day 14, a time point when the replicating virus was cleared from both genotypes (Sup Fig. 3).Interestingly, this increased viral load was not related to an impairment of type I and type III interferon response.We observed similar early interferon response at early time point of day 4 (Sup Fig. 1D and E).On day 7, we observed similar levels of IFNl (Fig. 2E), but an enhanced type 1 interferon response in the BAL of TLR9 deficient mice (Fig 2F).The elevated levels of type I interferons in the BAL on day 7 are likely a consequence of elevated viral fragments present in immune cells.This is of significance since the interferon levels are elevated despite the decreased inflammatory cell recruitment on day 7 (Fig. 1A-1E).Further, despite increased viral load, histological analysis did not show increasaed lung pathology in TLR9 KO mice on day 7 (Sup Fig. 1F).
TLR9 deficiency leads to infection of innate immune cells:
To gain mechanistic insights into the contribution of damage sensing through TLR9 in nonproductive viral clearance during influenza infection, we performed single-cell RNA-sequencing of lung tissue.Lung samples were obtained on day 4 post-infection to determine cell-specific differences in early infection and antiviral responses.Our single-cell approach yielded significant populations of both immune cells including macrophages, monocytes, neutrophils, T cells, B cells, basophils, dendritic cells, and epithelial cells including ciliated, AT1, AT2, goblet, and secretory epithelial cells (Fig. 3A and Sup Fig. 4A).The cellular specificity was determined based on the canonical gene signatures 12,13 as indicated in Fig. 3B and Sup Fig. 4B.We determined the presence of viral infection in these cells by influenza transcripts to understand the differences in the cellular targets of influenza infection in TLR9-/-lungs.A minimal infection was observed across the immune cell types in wild-type mice on day 4 post-infection.In contrast, TLR9-/-immune cells demonstrated significantly more infections which were spread across all the immune cells including innate immune cells such as neutrophils, monocytes, and macrophages (Fig. 3C and D).A distinct monocyte population, that expanded during viral infection, had high influenza infectivity denoted as iMon_high infection (Fig. 3A).These cells expanded during influenza infection in both wild type and TLR9-/-mice, however, a more robust expansion was observed in TLR9-/-mice (Sup Fig. 5).The increased viral burden in TLR9-/-deficient cells were not limited to only immune cell compartment, we found elevated infection in the epithelial compartment as well (Sup Fig. 4 C & D).
To better understand the impact of damage sensing on the host antiviral response, we investigated differentially expressed genes in wild-type vs TLR9-/-immune cells.TLR9-/monocytes upregulated chemokines and cytokines and associated receptors including Ccl3, Ccl4, Ccr5, IL-15, and IL-18.In contrast, we observed an increase in Bcl2, Cd36, and fibronectin-1 in the wild-type monocytes (Sup Fig. 6A).To determine whether these high expression of cytokines and chemokines interact with epithelial cells to have functional consequences at later time points, we generated Circos plots to understand cell-cell communications.Using iMON as signal senders (ligands) and epithelial cells such as Secretory, AT2, and Ciliated cells as signal receivers (receptors), we identified several ligand-receptor pairs such as Ccr2, Ccr10, Il1r, Il-12rb, Il15ra with their multiple chemokine and cytokine receptors (Sup Fig. 6B).Diff-Connectome analysis demonstrate ligand-receptor interactions are upregulated in TLR9_KO cells.
To understand pathways associated with these gene changes, we performed IPA analysis based on differentially expressed gene signatures in inflammatory monocytes.Wild-type monocytes demonstrated enrichment of tissue reparative pathways such as "interleukin 4 and interleukin-13 signaling" and "Binding and uptake of ligands by scavenger receptor".Further, as expected "Regulation of TLR by endogenous ligand" was the top upregulated pathway in the wild type monocytes (Fig. 3E).In contrast, the type I interferon pathway was the most significantly enriched in TLR9-/-monocytes (Fig. 3F).This observation was in agreement with our data demonstrating elevated type I interferon levels in TLR9-/-mice in vivo (Fig. 2F).Interestingly, these enriched pathways were unique to monocytes and not shared by epithelial cells such as ciliated epithelium (Sup Fig. 4 E & F)).To understand whether TLR9-/-deficiency leads to increased susceptibility to influenza infection in immune cells, we directly infected wild-type and TLR9-/-alveolar macrophages ex vivo.Our data show that TLR9-/-alveolar macrophages had inherent defects in the viral control as higher viral replication was observed in these cells (Fig. 3G).A similar finding was observed in human cells TLR9-deficient peripheral blood mononuclear cells (PBMCs).Infection of these cells demonstrated impaired viral control in PBMCs with TLR9 deficiency (Fig. 3H).
Immune cells concurrently upregulate influenza and inflammatory gene signature:
To understand how extensive infection of TLR9-/-immune cells affect their inflammatory response, we determined both the influenza response and inflammatory gene signature in these cells.For the measurement of inflammatory gene signature in the immune cells, we used "hallmark_inflammatory_response" module that includes 197 mouse genes (GSEA M5932) to show that elevated inflammatory response is present in the immune cells, especially that of myeloid origins such as monocytes, macrophages and neutrophils, which was further upregulated by influenza infection (Fig. 4A).Similarly, we explored the "influenza response" signature using 36 genes to demonstrate that influenza upregulated this gene signature in both wild-type and to a higher extent in TLR9-/-myeloid cells (Fig. 4B).This was supported by a signficant upregulation of STAT1 and NFkB pathways in the immune cells in absence of TLR9 (Sup Fig. 7A and B).Furthermore, we performed a crossover analysis between inflammatory and influenza gene signatures to show a significant overlap, especially in myeloid cells including monocytes, macrophages, and neutrophils in both these signatures (Fig. 4C).Next, we investigated whether similar elevation of STAT1 and NFkB was observed in epithelial cell compartment.Our data show that similar to immune cells, TLR9 KO epithelial cells upregulated STAT1 compared to wild type cells (Sup Fig. 7C), a likely consequence of elevated viral burden and subsequent interferon responses in these cells.However, in contrast to immune cells, we observed a signficant downregulation of NFkB pathway in TLR9 KO epithelial cells (Sup Fig. 7D), potentially contributing to dampened early cytokine and inflamamtory response in TLR9 KO mice during influenza infection (Fig. 1).
Next, to find the consequences of these upregulated inflammatory pathway in immune cells, we investigated cell death pathways.Our data show that influenza infection upregulated multiple cell death pathways including apoptosis, necroptosis, and pyroptosis pathways in immune cells, implicating excessive infection in immune cell death in absence of TLR9 (Sup Fig. 8).
To understand how human myeloid cells respond to infection, we performed time course of inflammatory response in human PBMC cell line.Our data show that while TLR9 deficient cells had ameliorated inflammatory responses during early time points (2 or 6 hours), the inflammatory response in TLR9-/-cells increased dramatically as the infection progresses (Sup Fig. 6C).These data support the hypothesis that TLR9 deficiency ameliorates the early inflammation through dampened inflamamtory response by epithelial and immune cells, however, this advantage wanes due to excessive replication of influenza in absence of TLR9, especially in immune cells.
Elevated viral load impaired recovery in TLR9-/-mice:
We observed a reduction in early inflammation and injury response in TLR9-/-mice (Fig. 1), however, it came at a cost of impaired viral control (Fig. 2 and 3).Further, we observed an elevated inflammatory response in the myeloid cells that had influenza gene signatures (Fig. 4).We next sought to determine how this impaired viral clearance contributes to the recovery in mice from influenza infection.We followed wild-type and TLR9-/-mice for up to 14 days post influenza infection.Our data show that TLR9-/-mice had significantly impaired recovery manifested as delayed body weight gains (Fig. 5A).The impaired recovery in TLR9-/-mice was associated with a persistent inflammatory response in the lung, manifested as elevated immune cell numbers in the BAL (Fig. 5B).Significantly elevated levels of monocytes and neutrophils persisted in the TLR9-/-mice at this time, while the number of macrophages in the BAL were similar at this time point (Fig. 5C-E).We also saw an increased recruitment of CD4+ and CD8+ T cells in the airspaces of TLR9-/-mice on day 14 (Fig. 5F and G).This persistent inflammatory response was also associated with increased tissue injury manifested as elevated total protein content (Fig. 5H), along with increased RBC (5I), platelet counts (5J), and RAGE in the BAL (5K).However, at this time, despite elevated inflammatory cell recruitment in these mice, we observed a significant reduction in both type I and type III interferon response (Fig. 5L and M) indicating that TLR9 signaling controls interferon signaling in the absence of a replicating virus (Fig. 3C).Despite elevated inflamamtory response, we did not observe an elevated fibrotic response at this time point assessed by trichrome staining of lung tissues (Sup Fig. 9A).Lack of increase fibrosis in TLR9 KO mice despite elevated inflamamtory resopnse is a likely consequence of dynamic changes in inflamamtory injury during the course of influenza infection.
To determine whether this impaired viral recovery is due to elevated viral load or TLR9 have independent role in promoting host recovery in viral infection, we stimulated mice with TLR9 agonist on day 8 post viral infection.Our data show that treatment with TLR9 agonist ODN2006 at a time point when most of the viral load is cleared, impaired host recovery (Supplemental Fig. 9B) indicating that impaired recovery in TLR9-/-mice is likely due to persistent presence of viral infection.This was further supported by experiments utilizing antiviral drug oseltamivir in mice, which blocked the impaired recovery phenotype in TLR9-/-mice (Sup Fig. 9C), emphasizing a role of impaired viral control in recovery.
The contribution of impaired viral control was further supported by utilization of myeloid specific TLR9 knockout mice.In contrast to the global deficiency, myeloid specific deletion of TLR9 did not lead to impaired viral control (Sup Fig. 10A).However, it ameliorated inflammatory response during the recovery phase on day 14 (Sup Fig. 10B), further supporting the notion that elevated viral load in global TLR9 KO contributes to the impaired recovery.The ameliorated inflammatory response in LysM Cre+TLR9-/-mice was mainly attributed to the myeloid cells including macrophages and monocytes (Sup Fig. 10C).These data demonstrate that TLR9 signaling acts on multiple cells to shape host response to influenza infections.
Increased TLR9 ligands in patients with influenza infection:
Finally, to determine the clinical relevance of our findings, we sought to measure the levels of TLR9 ligand mtDNA in the clinical samples of patients with influenza infection.We recruited influenza-infected hospitalized patients at Yale New Haven Hospital and obtained the age and sex-matched control samples.The detailed demographics are presented in Fig. 6A.Our data show that compared to plasma samples from healthy controls, those from individuals infected with influenza displayed significantly higher TLR9 activation as observed with a human reporter cell line (Fig. 6B).To further confirm the TLR9 activity of plasma samples, we measured the levels of mtDNA in these samples (Fig. 6C).To investigate whether elevated TLR9 signaing or mtDNA concentrations are relevant to the clinical disease, we compared these levels between those who were admitted to the general medical ward versus those who required a higher level of care in the medical intensive care unit (MICU).Our data show that the overall TLR9 activation by patient plasma was significantly elevated in those who were admitted to MICU as compared to those who did not (Fig. 6D).We did not observe similar effect with mtDNA, indicating that receptor activation is more relevant than ligand levels.To further investigate this biology, we measured the TLR9 activity of mouse serum during the time course of influenza infection using a mouse TLR9 reporter cell line.Our data showed that a signficant elevation in TLR9 activity was observed on day 7 (peak inflammation) (Sup Fig. 11), indicating a potential correlation between mouse model and human disease.Together these data indicate that circulating TLR9 activity serves as a biomaker for influenza infection and associated disease severity.
Discussion: In this study, we identified an essential role of damage sensing through TLR9 on host defense mechanisms during influenza infection.During an infection, the host balances its response to eliminate the pathogen (host resistance) while limiting the collateral tissue damage (host tolerance).Host defense and host tolerance are distinct strategies that a host employs to survive an infection 14,15 .However, how the host finetunes the inflammatory response during an infection and how the tissue damage shapes the antipathogen response still needs to be completely understood.In this study, we demonstrate that the absence of damage sensing through TLR9 deficiency improved the host tolerance to influenza infection at the early stage of the infection by limiting early inflammatory response by epithelial and immune cells (Fig. 1).However, this absence of damage sensing impaired the host defense, manifested as elevated and prolonged presence of viral remnants (Fig. 2), which ultimately leads to persistent injury and inflammatory response (Fig. 5), compromising the disease recovery.
Other damage-sensing receptors such as TLR4, which senses oxidized phospholipids, contribute to pathogenic inflammation and tissue injury, including in influenza infection 16,17 .However, unlike our observations for TLR9, the absence of TLR4 did not impair the host defense against influenza infection, leading to overall better outcomes 16 .This highlights the complexity of damage sensing where specific DAMP molecules contribute differentially to the host resistance and host tolerance.A complete understanding of damage sensing through various receptors might provide us with therapeutic targets that don't have potential deleterious effects on host immunity.
Our data also shed critical insights into how TLR9 signaling contributes to the host ability to limit viral replications, especially those by immune cells where nonproductive viral replication takes place.TLR9 deficiency led to extensive infection of immune cells such as macrophages, inflammatory monocytes, and neutrophils.The increase in the viral load in the immune cells appeared to be independent of conventional antiviral pathways mediated by type I interferons 18 .We observed a dichotomy between wild-type and TLR9-/-immune cells in their response to influenza infection where wild-type cells upregulated genes that were required for homeostasis and tissue regeneration such as "binding and uptake of scavenger receptors" 19 and "retinoid metabolism" 20 .In contrast, TLR9-/-cells upregulated the genes that were associated with an inflammatory milieu such as "hypercytokinemia/hyperchemokinemia response" or "pathogen-induced cytokine storm".It is critical to note, that these changes, especially in long-living cells such as monocytes and macrophages could drive the persistence of inflammatory milieu, as observed in TLR9-/-mice.These data shed light on the clinical manifestations of influenza infections where pathological inflammation persists even after the clearance of actively replicating virus and contributes to the lethality.This hypothesis is further supported by prior studies in influenza virus where presence of viral remnants contributes to the persistent inflammation at the site of infection much after the replicating virus is cleared 21 .Although it is possible that extensive infection in the myeloid cells may not lead to effective replication and release of infectious virus, it may be sufficient to promote persistent inflammation and delayed recovery, as observed in TLR9-/-mice.
There are limited studies that have been performed to investigate the role of damagesensing through TLR9 in influenza infections.Previous studies have indicated that stimulation of TLR9 along with TLR2/6 can provide robust protection against influenza infection by promoting viral clearane 22 , which seemed to be independent of the immune cell compartment 23 and dependent on the cell's ability to generate reactive oxygen species 24 .The potential interaction of TLR9 with TLR2 has also been demonstrated by their synergistic role in detecting certain strains 25 of herpes simplex virus-1 through plasmacytoid dendritic cells.Similarly, the absence of TLR9 renders the host susceptible to certain viruses such as cytomegalovirus 26 , however, initial evidence indicated that it does not affect the clearance of influenza infection 27 but controls post-viral bacterial infections.The role of TLR9 in recognizing the bacterial DNA and mounting a protective host response during pulmonary Streptococcus infection has been demonstrated previously 28 .However, these studies either boost TLR9 signaling before the tissue damage [22][23][24] or utilize a secondary bacterial infection that can be detected by TLR9 through bacterial DNA 27 .Thus, our study is the first to demonstrate the indispensable role of damage signaling through TLR9 in limiting viral replication and persistence.
To ensure the translational aspects of our findings, we utilized human peripheral blood mononuclear cells that were deficient in TLR9 to demonstrate the role of TLR9 in controlling influenza infection in immune cells.Our human data also corroborated mouse data demonstrating that increased viral load in the TLR9-/-PBMCs led to elevated inflammatory response.Further, we demonstrated the enhanced activity of TLR9 among patients who were hospitalized with influenza infection, especially those admitted to the MICU (Fig. 6).TLR9 activation was similiarly observed in mouse model during peak disease on day 7 (Sup Fig. 11), indicating its potential to serve as a biomaker for influenza disease severity.We also observed a dichotomy between mtDNA and overall TLR9 activity, which can be explained by other DAMPs present during the disease state.Prior studies have shown that other DAMP molecule HMGB1 has been known to modulate the TLR9 activating potential of mtDNA 29 .
Overall, our study demonstrated that damage sensing through TLR9 plays a key role in the anti-influenza immune response.This anti-influenza response appeared to be independent of type I interferon signaling, a well-established pathway downstream of TLR9 signaling.
Material and Methods:
Mouse model of influenza infection: Wild type and TLR9-/-mice were obtained from Jackson laboratory on C57 B/6 background and bred in house in the same room at the Yale University.TLR9 flox mice were generated and provided by Dr. Mark Shlomchik at Yale University (currently at University of Pittsburgh).Mice were infected with influenza virus (H1N1, PR8) by intranasal route as previously described.Mice were anesthetized using ketamine/xylazine solution and an infection inoculum of 10 PFUs in 50 µl of PBS was administered by intranasal route.The control infection mice received the same amount of PBS.ODN2006 was administered by intranasal route.All the animal work performed was approved by Institutional Animal Care and Use Committee at Yale School of Medicine (Protocol #20044).
Human Subjects: All the human research was approved by the Institutional Review Board at Yale School of Medicine with HIC# 0901004619.All the patient samples and health volunteers were recruited after obtaining written or verbal (in presence of a witness) informed consent.
In vitro culture of murine alveolar macrophage and human peripheral blood monocytes: Alveolar macrophages were cultured from bone marrow cultures of wild type and TLR9-/-mice in the presence of granulocyte-macrophage colony-stimulating factor, TGFβ, and the PPARγ activator as described recently 30
Cell counting and flow cytometry:
White blood cell counts (WBC), red blood cells (RBCs), and platelets were counted using a Coulter counter.Flow cytometry was performed with multicolor flow cytometry protocols similar to those published previously.
Western Blot: Western blot was performed as described previously by our lab 31 .At indicated time points, mice were euthanized and gathered whole lungs.Tissues were placed in a lysis buffer and homogenized with PRO25D Digital homogenizer.Tissue lysates were cleared of debris by centrifugation (15min, 3000g) and mixed with 2x Laemmli Sample Buffer.Equal amounts of protein (30 μg) were resolved by SDS-PAGE using 4-20% Mini-PROTEAN Ò TGX Stain-Free PrecastÔ Gels gels and transferred onto Trans-Blot Ò TurboÔ membranes (Bio-Rad).Membranes were blocked with 5% non-fat milk in TBS-T for 1 hour at room temperature and probed with primary antibodies against (Influenza A virus NS1) overnight at 4°C.After incubation with HRP-conjugated secondary antibodies, protein bands were visualized using an enhanced chemiluminescence detection kit (Thermo Fisher Scientific).qPCR: Total RNA was extracted from whole lungs from IAV non-infected / infected mice and human PBMCs using (RNeasy mini kit, Qiagen).RNA quantity and purity were assessed using a spectrophotometer (NanoDrop).Subsequently, 1μg of total RNA was reverse transcribed into cDNA using a high-capacity cDNA reverse transcription kit (iScript cDNA Synthesis Kit, Bio-Rad) according to the provided protocol.Specific targeting primers were designed using Primer3 software and validated for efficiency and specificity.
Detection of mitochondrial DNA: DNA was extracted from plasma using the QiaAMP DNA Mini-Kit (Qiagen) as previously described 32 .Briefly, 200 µl of plasma sample was used to elute 100 µl of DNA.The presence of human mtDNA in each sample of DNA was assayed by qPCR for the MT-ATP6 gene, using primers for MT-ATP6 (ThermoFisher) and probes (SsoAdvanced Universal Probes Supermix, Bio-Rad Laboratories).The ViiA 7 Real-Time PCR System (ThermoFisher) was employed with the following cycling conditions: incubation at 95°C for 10 minutes; 40 amplification cycles at 95°C for 10 seconds, 60°C for 30 seconds, and 72°C for 1 second; and cooling at 40°C for 30 seconds.The number of MT-ATP6 copies per microliter was determined based on a standard curve developed using serial dilutions of a commercially available DNA plasmid (OriGene) with complementary DNA sequences for human MT-ATP6.
ELISA and Multiplex: BAL from IAV non-infected / infected mice were assayed for cytokine levels using commercially available sandwich ELISA kits for RAGE, IFN-b, and IFN-l(R&D Systems).Multiplex kits from MSD were used as per the manufacturer's instructions.
Statistics: Data were analyzed using either student's t-test or one-way ANOVA with Tukey's multiple comparison test.Statistical significance was assumed when p values were <0.05.
10x scRNAseq Library Preparation, and Sequencing
Lung samples across different conditions were digested with collagenase type 4 (2mg/ml) and DNase (20U/ml) in DMEM with 10% FBS to prepare single-cell suspension for an expected cell recovery population of 10,000 cells per lane.scRNAseq libraries were generated with the Chromium Single Cell 3' assay (10x Genomics) and were sequenced on the NovaSeq platform per library in Yale Center for Genomics Analysis.Downstream processing was conducted with Cellranger 3.0.2with the default parameters.
RNA-Seq Data Filtration, Normalization, Scaling, Integration, and Clustering: Aligned data was processed using Seurat v3.1.2under R environment.Default parameters were used for each function unless otherwise specified 33 .After creating Seurat objects for each sample, we filtered on the percent mitochondrial genes to exclude damaged and dying cells and on the number of genes and transcripts to remove debris and doublets.All samples after filtration were integrated by the IntegrateData function in Seurat.The gene counts in the integrated object were then normalized to the total counts and multiplied by a 10,000 scaling factor.We used Variance Stabilizing Transformation (VST) to select highly variable genes.The top 200 genes within the selected genes were retained for downstream analysis.Variable genes were scaled using Seurat's ScaleData function and subsequently were used for principal component analysis (PCA) to generate clusters.The numbers of clusters produced using PCA analysis are controlled by a resolution parameter with higher values giving more clusters and by the number of PCs.All cells across different samples were assigned into two-dimensional Uniform Manifold Approximation and Projection (UMAPs).
To sub-cluster immune cells and epithelial cells, we first used subset function in Seurat to select cell clusters that are positive for Ptprc, and Epcam while negative for Col1a1, and Cdh5 to create a new lung immune cell or epithelial cell Seurat objects, respectively.We then re-normalized, re-scaled, and re-clustered these objects for downstream analysis.
Cluster Identification by Differentially Expressed Genes (DEGs): The FindAllMarkers function in Seurat was applied to each sample to identify differentially expressed genes for each cluster relative to all the other clusters in the object.We only chose genes that were expressed in at least 10% of cells in one of these clusters to ensure the quality of genes.These marker lists were used to apply cluster labels based on top defining genes of canonical cell types in the literature 12,13 .
To identify viral reads in the sequencing data, the influenza A genomic reference was created with CellRanger mkref.The sample fastq files were then aligned to this reference genome using CellRanger count.Barcodes corresponding to reads that aligned with the influenza virus were considered "infected."This information was transferred to the singlecell object's metadata.
After pre-processing and quality control, two gene module scores were calculated: (1) influenza-related genes, and (2) hallmark inflammatory genes.Influenza gene lists were generated by searching the top 30 genes associated with the keywords "influenza" in Geneshot Pubmed.The hallmark inflammatory genes were obtained from the GSEA gene list `hallmark_inflammatory_response` (GSEA M5932).For each cell, the composite gene module score was calculated.These two scores were then plotted, split by condition.The overlap between gene lists (1) and ( 2) was plotted using the Seurat FeaturePlot blend option.
To compare TLR9 KO and WT infected mice, log-fold changes in expression were evaluated for both ligand and receptor sides of all edges using FindMarkers.A perturbation score was calculated to facilitate differential edge plotting, reflecting the logfold change and incorporating directional information from ligands and receptors.Edges meeting criteria (at least 10% involvement of sending and receiving clusters, adjusted pvalue < 0.05 via Wilcoxon rank sum test) were visualized using CircosPlot.
Data Availability: Raw data used to generate the figures of the manuscript has been provided as an excel file (Supporting Table 1).Single cell RNA seq data has been deposited to Gene Expression Omnibus under the accession number GSE271505.
Funding: This work was supported by funding from American Lung Association Catalyst Award (LS) and the Parker B Francis Fellowship (LS).YY is supported by NIH R00HL159261.ELH is supported by 5R01HL163984-02 and 5R01HL152677-04.CSDC is supported by VA Merit Grant and Department of Defense grants.CR is supported by NIH-NHLBI K08HL151970-01.The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Figure Legends:
Fig. 1.TLR9 signaling promotes early inflammation and injury response during influenza infection: Wild type and TLR9-/-mice were infected with influenza virus by the intranasal route and euthanized on day 7 post infection to harvest broncho-alveolar lavage fluid (BAL).Total white blood cells were counted using a coulter counter (A).The composition of the BAL cells was determined using multicolor flow cytometry.The number of macrophages (B), monocytes (C) neutrophils (D), and CD4+ T cells (E) were obtained.Lung injury was measured by quantifying total protein in the BAL (F) using BCA assay, by measuring the levels of RBCs (G) or platelets (H) in the BAL of mice using a coulter counter and RAGE using ELISA (I).The specific cytokines were measured in the BAL samples using mesoscale discovery multiplex assays (J-N).Two groups were compared with Mann-Whitney analysis, while the cytokine data were analyzed using two-way ANOVA.* = P<0.05,** = P<0.01,*** = P<0.005,**** = P<0.001,ns = not significant.N = 18 WT and 19 TLR9 KO for A, G and H, pooled from three independent experiments of which flow cytometic analysis was performed for two experiments (N= 13 WT and 14 TLR9 KO) for Fig. B, C, D, E, F, G. Multiplex assays were performed from two independent experiments, N = 13 WT and 9 TLR9 KO for day 4 and N=12 each for WT and TLR9 for day 7.
Fig. 2. TLR9 signaling limits non-productive viral growth in influenza infection:
Viral load was measured in the lung tissue of wild type and TLR9-/-mice on day 7 post-infection using the qPCR of viral matrix gene (A) where 18S was used as housekeeping control gene.The viral load was confirmed using Western blotting of the lung tissue using antibodies against the N protein of influenza.Western blot analysis (B) and densitometric analysis (C) are shown.Beta-actin was used as housekeeping control.Productive viral replication was measured using TCID50 using MDCK cell culture (D).Levels of interferon b (E) or interferon l (F) were measured using sandwich ELISA.Data were analyzed using Mann-Whitney analysis.* = P<0.05,** = P<0.01,***, ns = not significant.Each dot represents a mouse, N = 12 WT and 11 KO for A and C pooled from two experiments and 18 and 19 from three experiments.
Fig. 3. TLR9 deficiency renders immune cells susceptible to influenza infection:
Uniform Manifold Approximation and Projection (UMAPs) plots of immune cells present in the lungs based on the known gene expression markers (A). iMON_Highinfection was the subset of inflammatory monocytes that were infected with the influenza virus and were exclusively present in the TLR9-/-mice.Among the immune cells, the influenza-infected cells were identified using influenza RNA expression in the cell and separated by the genotype (B).The infection status was quantified based on the genotype to compare wild type vs TLR9-/-cells (C).Top pathways enriched in the wild type (D) or TLR9-/-(E) inflammatory monocytes are demonstrated.Bone marrow-derived alveolar macrophages were infected with influenza virus (MOI of 1) for 24 hours and the RNA was isolated to perform qPCR of influenza matrix protein (F).TLR9-/-human peripheral blood monocyte cells were generated and infected with the influenza virus similarly to measure the viral load (G).Data are pooled from 3 independent experiments.Data were analyzed using Mann-Whitney analysis.* = P<0.05,** = P<0.01,***.The expression of the "hallmark_inflammatory_response" module that includes 197 mouse genes (GSEA M5932) in the immune cells (A) and influenza response (B).The crossover between "hallmark_inflammatory_response" and "influenza response" in the immune cells (C).Fig. 5. TLR9-/-mice have impaired recovery during influenza infection: Wild type and TLR9-/-mice were infected with influenza virus and body weight was measured every day for 14 days (A).Mice were euthanized on day 14 to harvest broncho-alveolar lavage.Total WBCs in the BAL were counted using coulter counter (B).The number of macrophages (C), monocytes (D), neutrophils (E), CD4+ (F), and CD8+ T (G) cells was determined using multi-color flow cytometry.Lung injury was measured by measuring total protein content (H), RBC counts (I), platelets (J) and RAGE (K) in the BAL.Levels of interferon b (L) and interferon l (M) were measured in the BAL obtained on day 14 post infection.Data are pooled from two independent experiments.Weight loss curves were analyzed using two-way ANOVA with Šídák's multiple comparisons test.Two groups were compared with Mann-Whitney analysis, * = P<0.05,** = P<0.01,*** = P<0.005,**** = P<0.001,ns = not significant, N=11 each group.
Fig. 6. Human influenza infection upregulates TLR9 ligands in circulating blood:
Plasma samples were obtained from patients with influenza infection and control subjects without any respiratory infections.The total number of subjects in each group and demographics are shown in Fig. 6A.The overall TLR9 activity of plasma was measured using reporter cell line (B).The levels of mtDNA were measured using qPCR-based assay (C).Plasma TLR9 activity (D) and mtDNA levels (E) were compared between those admitted to medical intensive care unit (MICU) and those remained hospitalized without the need of MICU, as marker of disease severity.*, P<0.05, ** = P<0.01,**** = P<0.001using Mann-Whitney test.
B
Supplemental Fig. 6.Violoin plots showing the expression levels of specific gene (A).Cell-cell communication analyses were performed using the R package Connectome.Average expression levels of ligands and receptors per cell type were computed across experimental groups.Connectomes were constructed for each experimental group, containing unfiltered lists of edges linking ligand-expressing cells to receptor-expressing cells.Diff-connectome was generated to demonstrated the upregulated receptor-ligand interactions in TLR9 KO cells (B).Human PBMC cell lines were infected with influenza (MOI of 1) and cytokine expressions were measured over the time course using qPCR (C).*, P < 0.05, **, P<0.01, and ***, P < 0.005 using two-way ANOVA.
Fig. 4 .
Fig. 4. Influenza infection simultaneously upregulates both influenza and inflammatory response in immune cells:The expression of the "hallmark_inflammatory_response" module that includes 197 mouse genes (GSEA M5932) in the immune cells (A) and influenza response (B).The crossover between "hallmark_inflammatory_response" and "influenza response" in the immune cells (C).
1 .. 2 . 3 .
Levels of total WBC, RBC, and platelets were counted using a Coulter counter on day 4 post infection in wild type and TLR9-/-mice (A-C).Interferon levels were measured using ELISA (D and E) Lung sections were stained with hematoxylin and eosin on day 7. Sup Fig.1Fshows representative pictures on day 7. Flow strategy to detect immune cell populations in the BAL.Western blot analysis of influenza NS1 protein on day 10 post-infection (A and B) and day 14 post infection (C).
. 4 .
UMAP of epithelial cells in the lung (A) and markers to identify specific cell types (B).The identity of infected cells and associated genotype (C) and quantification (D).IPA pathway analysis showing enriched pathway in wild type and TLR9-/-ciliated epithelium Supplemental Fig.5.UMAP showing an expansion of iMON_Highinfection in infected mice, especially in TLR9-/-lungs. | 2024-03-09T14:10:44.269Z | 2024-03-06T00:00:00.000 | {
"year": 2024,
"sha1": "89af35cfddd14898f91bc3d0cbc1025e8f4f1f8d",
"oa_license": "CCBYNCND",
"oa_url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10942338",
"oa_status": "GREEN",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "89af35cfddd14898f91bc3d0cbc1025e8f4f1f8d",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
738764 | pes2o/s2orc | v3-fos-license | Liquid Cloud Storage
A liquid system provides durable object storage based on spreading redundantly generated data across a network of hundreds to thousands of potentially unreliable storage nodes. A liquid system uses a combination of a large code, lazy repair, and a flow storage organization. We show that a liquid system can be operated to enable flexible and essentially optimal combinations of storage durability, storage overhead, repair bandwidth usage, and access performance.
INTRODUCTION
Distributed storage systems generically consists of a large number of interconnected storage nodes, with each node capable of storing a large quantity of data. The key goals of a distributed storage systems are to store as much source data as possible, to assure a high level of durability of the source data, and to minimize the access time to source data by users or applications.
Many distributed storage systems are built using commodity hardware and managed by complex software, both of which are subject to failure. For example, storage nodes can become unresponsive, sometimes due to software issues where the node can recover after a period of time (transient node failures), and sometimes due to hardware failure in which case the node never recovers (node failures). The paper [8] provides a more complete description of different types of failures.
A sine qua non of a distributed storage system is to ensure that source data is durably stored, where durability is often characterized in terms of the mean time to the loss of any source data (MTTDL) often quoted in millions of years. To achieve this in the face of component failures a fraction of the raw storage capacity is used to store redundant data. Source data is partitioned into objects, and, for each object, redundant data is generated and stored on the nodes. An often-used trivial redundancy is triplication in which three copies of each object are stored on three different nodes. Triplication has several advantages for access and ease of repair but its overhead is quite high: two-thirds of the raw storage capacity is used to store redundant data. Recently, Reed-Solomon (RS) codes have been introduced in production systems to both reduce storage overhead and improve MTTDL compared to triplication, see e.g., [10], [8], and [19]. As explained in [19] the reduction in storage overhead and improved MTTDL achieved by RS codes comes at the cost of a repair bottleneck [19] which arises from the significantly increased network traffic needed to repair data lost due to failed nodes.
The repair bottleneck has inspired the design of new types of erasure codes, e.g., local reconstruction codes (LR codes) [9], [19], and regenerating codes (RG codes) [6], [7], that can reduce repair traffic compared to RS codes. The RS codes, LR codes and RG codes used in practice typically use small values of (n, k, r), which we call small codes. Systems based on small codes, which we call small code systems, therefore spread data for each object across a relatively small number of nodes. This situation necessitates using a reactive repair strategy where, in order to achieve a large MTTDL, data lost due to node failures must be quickly recovered. The quick recovery can demand a large amount network bandwidth and the repair bottleneck therefore remains.
These issues and proposed solutions have motivated the search for an understanding of tradeoffs. Tradeoffs between storage overhead and repair traffic to protect individual objects of small code systems are provided in [6], [7]. Tradeoffs between storage overhead and repair traffic that apply to all systems are provided in [14].
Initially motivated to solve the repair bottleneck, we have designed a new class of distributed storage systems that we call liquid systems: Liquid systems use large codes to spread the data stored for each object across a large number of nodes, and use a lazy repair strategy to slowly repair data lost from failed nodes. 1 In this paper we present the design and its most important properties. Some highlights of the results and content of paper include: -Liquid systems solve the repair bottleneck, with less storage overhead and larger MTTDL than small code systems. -Liquid systems can be flexibly deployed at different storage overhead and repair traffic operating points. -Liquid systems avoid transient node failure repair, which is unavoidable for small code systems.
-Liquid systems obviate the need for sector failure data scrubbing, which is necessary for small code systems. -Liquid system regulated repair provides an extremely large MTTDL even when node failures are essentially adversarial. -Simulation results demonstrate the benefits of liquid systems compared to small code systems.
-Liquid system prototypes demonstrate improved access speeds compared to small code systems.
-Detailed analyses of practical differences between liquid systems and small code systems are provided. -Liquid systems performance approach the theoretical limits proved in [14].
Objects
We distinguish two quite different types of data objects. User objects are organized and managed according to semantics that are meaningful to applications that access and store data in the storage system. Storage objects refers to the organization of the source data that is stored, managed and accessed by the storage system. This separation into two types of objects is common in the industry, e.g., see [3]. When user objects (source data) are to be added to the storage system, the user objects are mapped to (often embedded in) storage objects and a record of the mapping between user objects and storage objects is maintained. For example, multiple user objects may be mapped to consecutive portions of a single storage object, in which case each user object constitutes a byte range of the storage object. The management and maintenance of the mapping between user objects and storage objects is outside the scope of this paper.
The primary focus of this work is on storing, managing and accessing storage objects, which hereafter we refer to simply as objects. Objects are immutable, i.e., once the data in an object is determined then the data within the object does not change. Appends to objects and deletions of entire objects can be supported. The preferred size of objects can be determined based on system tradeoffs between the number of objects under management and the granularity at which object repair and other system operations are performed. Generally, objects can have about the same size,
Storage and repair
or at least a minimum size, e.g., 1 GB. The data size units we use are KB = 1024 bytes, MB = 1024 KB, GB = 1024 MB, TB = 1024 GB, and PB = 1024 TB.
Storage and repair
The number of nodes in the system at any point in time is denoted by M , and the storage capacity of each node is denoted by S, and thus the overall raw capacity is D raw = M · S. The storage overhead expressed as a fraction is β = 1 − Dsrc Draw , and expressed as a per cent is β · 100%, where D src is the aggregate size of objects (source data) stored in the system.
When using an (n, k, r) erasure code, each object is segmented into k source fragments and an encoder generates r = n − k repair fragments from the k source fragments. An Encoding Fragment ID, or EFI, is used to uniquely identify each fragment of an object, where EFIs 0, 1, . . . , k − 1 identify the source fragments of an object, and EFIs k, . . . , n − 1 identify the repair fragments generated from source fragments. Each of these n = k + r fragments is stored at a different node. An erasure code is MDS (maximum distance separable) if the object can be recovered from any k of the n fragments.
A placement group maps a set of n fragments with EFIs {0, 1, . . . , n − 1} to a set of n of the M nodes. Each object is assigned to one of the placement groups, and the selected placement group determines where the fragments for that object are to be stored. To ensure that an equal amount of data is stored on each node, each placement group should be assigned an equal amount of object data and each node should accommodate the same number of placement groups.
For the systems we describe, Dsrc Draw = k n and thus β = r n , i.e., the amount of source data stored in the system is maximized subject to allotting enough space to store n fragments for each object. Object loss occurs when the number of fragments available for an object is less than k, and thus the object is no longer recoverable. A repairer is an agent that maintains recoverability of objects by reading, regenerating and writing fragments lost due to node failures, to ensure that each object has at least k available fragments. In principle the repairer could regenerate r missing fragments for an object when only k fragments remain available. However, this repair policy results in a small MTTDL, since one additional node failure before regenerating the r missing fragments for the object causes object loss.
A repairer determines the rate at which repair occurs, measured in terms of a read repair rate, since data read by a repairer is a fair measure of repair traffic that moves through the network. (The amount of data written by a repairer is essentially the same for all systems, and is generally a small fraction of data read by a repairer.) Of primary interest is the amount of network bandwidth R peak that needs to be dedicated to repair to achieve a given MTTDL for a particular system, as this value of R peak determines the network infrastructure needed to support repair for the system. Generally, we set a global upper bound R peak on the allowable read repair rate used by a repairer, and determine the MTTDL achieved for this setting of R peak by the system. The average read repair rate R avg is also of some interest, since this determines the average amount of repair traffic moved across the network over time.
The fixed-rate repair policy works in similar ways for small code systems and liquid systems in our simulations: The repairer tries to use a read repair rate up to R peak whenever there are any objects to be repaired.
Small code systems
Replication, where each fragment is a copy of the original object is an example of a trivial MDS erasure code. For example, triplication is a simple (3, 1, 2) MDS erasure code, wherein the object can be recovered from any one of the three copies. Many distributed storage systems use replication.
A Reed-Solomon code (RS code) [2], [17], [12] is an MDS code that is used in a variety of applications and is a popular choice for storage systems. For example, the small code systems described in [8], [10] use a (9, 6, 3) RS code, and those in [19] use a (14, 10, 4) RS code.
When using a (n, k, r) small code for a small code system with M nodes, an important consideration is the number of placement groups P to use. Since n M, a large number P of placement groups are typically used to smoothly spread the fragments for the objects across the nodes, e.g., Ceph [22] recommends P = 100·M n . A placement group should avoid mapping fragments to nodes with correlated failures, e.g., to the same rack and more generally to the same failure domain. Pairs of placement groups should avoid mapping fragments to the same pair of nodes. Placement groups are updated as nodes fail and are added. These and other issues make it challenging to design placement groups management for small code systems.
Reactive repair is used for small code systems, i.e., the repairer quickly regenerates fragments as soon as they are lost from a failed node (reading k fragments for each object to regenerate the one missing fragment for that object). This is because, once a fragment is lost for an object due to a node failure, the probability of r additional node failures over a short interval of time when r is small is significant enough that repair needs to be completed as quickly as practical. Thus, the peak read repair rate R peak is higher than the average read repair rate R avg , and R avg is k times the node failure erasure rate. In general, reactive repair uses large bursts of repair traffic for short periods of time to repair objects as soon as practical after a node storing data for the objects is declared to have permanently failed.
The peak read repair rate R peak and average read repair rate R avg needed to achieve a large MTTDL for small code systems can be substantial. Modifications of standard erasure codes have been designed for storage systems to reduce these rates, e.g., local reconstruction codes (LR codes) [9], [19], and regenerating codes (RG codes) [6], [7]. Some versions of LR codes have been used in deployments, e.g., by Microsoft Azure [10]. The encoding and decoding complexity of RS, LR, and RG codes grows non-linearly (typically quadratric or worse) as the values of (n, k, r) grow, which makes the use of such codes with large values of (n, k, r) less practical.
Liquid systems
We introduce liquid systems for reliably storing objects in a distributed set of storage nodes. Liquid systems use a combination of a large code with large values of (n, k, r), lazy repair, and a flow storage organization. Because of the large code size the placement group concept is not of much importance: For the purposes of this paper we assume that one placement group is used for all objects, i.e., n = M and a fragment is assigned to each node for each object.
The RaptorQ code [20], [13] is an example of an erasure code that is suitable for a liquid system. RaptorQ codes are fundamentally designed to support large values of (n, k, r) with very low encoding and decoding complexity and to have exceptional recovery properties. Furthermore, RaptorQ codes are fountain codes, which means that as many encoded symbols n as desired can be generated on the fly for a given value of k. The fountain code property provides flexibility when RaptorQ codes are used in applications, including liquid systems. The monograph [20] provides a detailed exposition of the design and analysis of RaptorQ codes.
The value of r for a liquid system is large, and thus lazy repair can be used, where lost fragments are repaired at a steady background rate using a reduced amount of bandwidth. The basic idea is that the read repair rate is controlled so that objects are typically repaired when a large fraction of r fragments are missing (ensuring high repair efficiency), but long before r fragments are missing (ensuring a large MTTDL), and thus the read repair rate is not immediately reactive to individual node failures. For a fixed node failure rate the read repair rate can in principle be fixed as described in Appendix A. When the node failure rate is unknown a priori, the algorithms described in Section 9 and Appendix A should be used to adapt the read repair rate to changes in conditions (such as node failure rate changes) to ensure high repair efficiency and a large MTTDL.
We show that a liquid system can be operated to enable flexible and essentially optimal combinations of storage durability, storage overhead, repair bandwidth usage, and access performance, exceeding the performance of small code systems.
Information theoretic limits on tradeoffs between storage overhead and the read repair rate are introduced and proved in [14]. The liquid systems for which prototypes and simulations are described in this paper are similar to the liquid systems described in Section 9.A of [14] that are within a factor of two of optimal. More advanced liquid systems similar to those described in Section 9.B of [14] that are asymptotically optimal would perform even better. Fig. 1 shows a possible liquid system architecture. In this architecture, a standard interface, such as the S3 or SWIFT API, is used by clients to make requests to store and access objects. Object storage and access requests are distributed across access proxy servers within the liquid access tier. Access proxy servers use a RaptorQ codec to encode objects into fragments and to decode fragments into objects, based on the flow storage organization. Access proxy servers read and write fragments from and to the storage servers that store the fragments on disks within the liquid storage tier. Lazy repair operates as a background process within the liquid storage tier. Further details of this example liquid system architecture are discussed in Section 8.
Since distributed storage systems consist of a large number of components, failures are common and unavoidable. The use of commodity hardware can additionally increase the frequency of failures and outages, as can necessary maintenance and service operations such as software upgrades or system reconfiguration on the hardware or software level.
Different failure types can affect the system in various ways. For example, a hard drive can fail in its entirety, or individual sectors of it can become unreadable. Node failures can be permanent, e.g. if the failure results in the entire node needing to be replaced, or it can be transient, if it is caused issues such as network outages, reboots, software changes. Network failures can cause subsets of many nodes to be unreachable, e.g. a full rack or even a complete site.
Moreover, distributed storage systems are heterogeneous in nature, with components that are commonly being upgraded and replaced. Device models do not remain the same; for example, a replacement hard drive will often be of a different kind, with different and possibly unknown expected lifetime. Even within a device model, different manufacturing runs can result in variation in expected lifetimes. Further, upgrades may require a large set of machines to be replaced in a short time frame, leading to bursts of increased churn. Changes to the underlying technology, such as a transition from hard drives to SSDs, are also likely to profoundly impact the statistics and nature of storage failures.
For these reasons, the failure behavior of a distributed storage system is difficult to model. Results obtained under complicated realistic error models are difficult to interpret, and hard to generalize. The results in this paper are obtained under a fairly straightforward failure model which focuses on two types of failures: Node failures and sector failures.
Node failures
Although node failure processes can be difficult to accurately model, analysis based on a random node failure models can provide insight into the strengths and weaknesses of a practical system, and can provide a first order approximation to how a practical system will operate. The time till failure of an individual node is often modeled by a Poisson random variable with rate λ, in which case the average time till an individual node fails is 1/λ. Nodes are assumed to fail independently. Since failed nodes are immediately replaced, the aggregate node failure process is Poisson with rate λ · M . Our simulations use this model with 1/λ = 3 years, and also use variants of this model where the node failure rate bursts to 3 times this rate (1/λ = 1 year) or 10 times this rate (1/λ = 0.3 years) for periods of time.
Nodes can become unresponsive for a period of time and then become responsive again (transient node failure), in which case the data they store is temporarily unavailable, or nodes can become and remain unresponsive (node failure), in which case the data they store is permanently unavailable. Transient node failures are an order of magnitude more common than node failures in practice, e.g., see [8].
Unresponsive node are generally detected by the system within a few seconds. However, it is usually unknown whether the unresponsiveness is due to a transient node failure or a node failure. Furthermore, for transient node failures, their duration is generally unknown, and can vary in an extremely large range: Some transient node failures have durations of just seconds when others can have durations of months, e.g., see [8].
Section 9 discusses more general and realistic node failure models and introduces repair strategies that automatically adjust to such models.
Sector failures
Sector failures are another type of data loss that occurs in practice and can cause operational issues. A sector failure occurs when a sector of data stored at a node degrades and is unreadable. Such data loss can only be detected when an attempt is made to read the sector from the node. (The data is typically stored with strong checksums, so that the corruption or data loss becomes evident when an attempt to read the data is made.) Although the sector failure data loss rate is fairly small, the overhead of detecting these types of failures can have a negative impact on the read repair rate and on the MTTDL.
Data scrubbing is often employed in practice to detect sector failures: An attempt is made to read through all data stored in the system at a regular frequency. As an example, [4] reports that Dropbox scrubs data at the frequency of once each two weeks, and reports that read traffic due to scrubbing can be greater than all other read data traffic combined. Most (if not all) providers of data storage systems scrub data, but do not report the frequency of scrubs. Obviously, scrubbing can have an impact on system performance.
The motivation for data scrubbing is that otherwise sector failures could remain undetected for long periods of time, and thus have a large negative impact on the MTTDL. This impact can be substantial even if the data loss rate due to sector failures is relatively much smaller than the data loss rate due to node failures. The reason for this is that time scale for detecting and repairing sector failures is so much higher than for detecting and repairing node failures.
We do not employ a separate scrubbing process in our simulations, but instead rely on the repair process to scrub data: Since each node fails on average in 1/λ years, data on nodes is effectively scrubbed on average each 1/λ years. Sector failures do not significantly affect the MTTDL of liquid systems, but can have a significant affect on the MTTDL for small code systems.
We model the time to sector failure of an individual 4 KB sector of data on a node as a Poisson random variable with rate λ SF , and thus the average time till an individual 4 KB sector of data on a node experiences a sector failure is 1/λ SF , where sector failures of different sectors are independent. Based on sector failure statistics derived from the data of [25], our simulations use this model with 1/λ SF = 5 · 10 8 years. Thus, the data loss rate due to sector failures is more than 10 8 times smaller than the data loss rate due to node failures in our simulations.
REPAIR
Although using erasure coding is attractive in terms of reducing storage overheads and improving durability compared to replication, it can potentially require large amounts of bandwidth to repair for fragment loss due to node failures. Since the amount of required repair bandwidth can be high and can cause system-wide bottlenecks, this is sometimes referred to as a repair bottleneck. For example, [19] states the goal of their work as follows: Our goal is to design more efficient coding schemes that would allow a large fraction of the data to be coded without facing this repair bottleneck. This would save petabytes of storage overheads and significantly reduce cluster costs.
The repair bottleneck sometimes has proven to be an obstacle in the transition from simple replication to more storage efficient and durable erasure coding in storage clusters. It simply refers to the fact that, when using a standard erasure code such as a RS code, the loss of a fragment due to a node failure requires the transfer of k fragments for repair, thus increasing k-fold the required bandwidth and I/O operations in comparison to replication. For example, [19] states the following when describing a deployed system with 3000 nodes: . . . the repair traffic with the current configuration is estimated around 10-20% of the total average of 2 PB/day cluster network traffic. As discussed, (14,10,4) RS encoded blocks require approximately 10x more network repair overhead per bit compared to replicated blocks. We estimate that if 50% of the cluster was RS encoded, the repair network traffic would completely saturate the cluster network links.
The recognition of a repair bottleneck presented by the use of standard MDS codes, such as RS codes, has led to the search for codes that provide locality, namely where the repair requires the transfer of < k fragments for repair. The papers [9], [10], [19] introduce LR codes and provide tradeoffs between locality and other code parameters. We discuss some properties of small code systems based on LR codes in Section 5.2.
The following sub-sections provide the motivation and details for the repair strategies employed in our simulations.
Repair initiation timer
In practice, since the vast majority of unresponsive nodes are transient node failures, it is typical to wait for a repair initiation time T RIT after a node becomes unresponsive before initiating repair. For example, T RIT = 30 minutes for the small code system described in [10], and thus the policy reaction is not immediate but nearly immediate.
In our simulations repairers operate as follows. If a node is unresponsive for a period at most T RIT then the event is classified as a transient node failure and no repair is triggered. After being unresponsive for a period of T RIT time, a node failure is declared, the node is decommissioned, all
Repair strategies
fragments stored on the node are declared to be missing (and will be eventually regenerated by the repairer). The decommissioned node is immediately replaced by a new node that initially stores no data (to recover the lost raw capacity).
Transient node failures may sometimes take longer than 30 minutes to resolve. Hence, to avoid unnecessary repair of transient node failures, it is desirable to set T RIT longer, such as 24 hours. A concern with setting T RIT large is that this can increase the risk of source data loss, i.e., it can significantly decrease the MTTDL.
Since a large value for T RIT has virtually no impact on the MTTDL for liquid systems, we can set T RIT to a large value and essentially eliminate the impact that transient node failures have on the read repair rate. Since a large value for T RIT has a large negative impact on the MTTDL for small code systems, T RIT must be set to a small value to have a reasonable MTTDL, in which case the read repair rate can be adversely affected if the unresponsive time period for transient node failure extends beyond T RIT and triggers unnecessary repair. We provide some simulation results including transient node failures in Section 5.4.
Repair strategies
In our simulations, a repairer maintains a repair queue of objects to repair. An object is added to the repair queue when at least one of the fragments of the object have been determined to be missing. Whenever there is at least one object in the repair queue the repair policy works to repair objects. When an object is repaired, at least k fragments are read to recover the object and then additional fragments are generated from the object and stored at nodes which currently do not have fragments for the object.
Objects are assigned to placement groups to determine which nodes store the fragments for the objects. Objects within a placement group are prioritized by the repair policy: objects with the least number of available fragments within a placement group are the next to be repaired within that placement group. Thus, objects assigned to the same placement group have a nested object structure: The set of available fragments for an object within a placement group is a subset of the set of available fragments for any other object within the placement group repaired more recently. Thus, objects within the same placement group are repaired in a round-robin order. Consecutively repaired objects can be from different placement groups if there are multiple placement groups.
Repair for small code system
In our simulations, small code systems (using either replication or small codes) use reactive repair, i.e., R peak is set to a significantly higher value than the required average repair bandwidth, and thus repair occurs in a burst when a node failure occurs and the lost fragments from the node are repaired as quickly as possible. Typically only a small fraction of objects are in the repair queue at any point in time for a small code system using reactive repair.
We set the number of placement groups to P = 100·M n to the recommended Ceph [22] value, and thus there are 100 placement groups assigned to each of the M nodes, with each node storing fragments for placement groups assigned to it. The repair policy works as follows: -If there are at most 100 placement groups that have objects to repair then each such placement group repairs at a read repair rate of R peak /100. -If there are more than 100 placement groups that have objects to repair then the 100 placement groups with objects with the least number of available fragments are repaired at a read repair rate of R peak /100.
This policy ensures that, in our simulations, the global peak read repair rate is at most R peak , that the global bandwidth R peak is fully utilized in the typical case when one node fails and needs repair (which triggers 100 placement groups to have objects to repair), and that the maximum traffic from and to each node is not exorbitantly high. The need for small code systems to employ reactive repair, i.e., use a high R peak value, is a consequence of the need to achieve a large MTTDL. Small codes are quite vulnerable, as the loss of a small number of fragments can lead to object loss. For instance, with the (14, 10, 4) RS code the loss of five or more fragments leads to object loss. Fig. 2 illustrates the drastically different time scales at which repair needs to operate for a small code system compared to a liquid system. In this example, nodes fail on average every three years and Fig. 2 shows the probability that N ≤ k − 1 of n nodes are functioning at time t when starting with N = n nodes functioning at time zero. As shown, a failed node needs to be repaired within two days to avoid a 10 −10 probability object loss event for a (14, 10, 4) RS code, whereas a failed node needs to be repaired within 10 months for a (3010, 2150, 860) large code. Note that the storage overhead β = r/n = 0.286 is the same in these two systems. A similar observation was made in [21].
Exacerbating this for a small code system is that there are many placement groups and hence many combinations of small number of node failures that can lead to object loss, thus requiring a high R peak value to ensure a target MTTDL. For example, P = 100·M n = 21500 for a (14,10,4) small code in a M = 3010 node system, and thus an object from the system is lost if any object from any of the 21500 placement groups is not repaired before more than four nodes storing fragments for the object fail. Furthermore, a small code system using many placement groups is difficult to analyze, making it difficult to determine the value of R peak that ensures a target MTTDL.
Although small codes outperform replication in both storage overhead and reliability metrics, they require a high R peak value to achieve a reasonable MTTDL, which leads to the repair bottleneck described earlier. For example, R peak used by small code system can be controlled to be a fraction of the total network bandwidth available in the cluster, e.g., 10% of the total available bandwidth is mentioned in [10] and 10% − 20% was estimated in [19] to protect a small fraction of the object data with a small code. However, using these fractions of available bandwidth for repair may or may not achieve a target MTTDL. As we show in Section 5 via simulation, small code systems require a large amount of available bandwidth for bursts of time in order to achieve a reasonable target MTTDL, and in many cases cannot achieve a reasonable MTTDL due to bandwidth limitations, or due to operating with a large T RIT value.
Repair for fixed rate liquid system
The simplest repair policy for a liquid system is to use a fixed read repair rate: the value of the peak read repair rate R peak is set and the read repair rate is R peak whenever there are objects in the repair queue to be processed. A liquid system employs a repairer that can be described as lazy: The value of R peak is set low enough that the repair queue contains almost all objects all the time. Overall, a liquid system using lazy repair uses a substantially lower average repair bandwidth and a dramatically lower peak repair bandwidth than a small code system using reactive repair.
There are two primary reasons that a liquid system can use lazy repair: (1) the usage of a large code, i.e., as shown in Fig. 2, the (3010, 2150, 860) large code has substantially more time to repair than the (14, 10, 4) small code to achieve the same MTTDL; (2) the nested object structure.
A liquid system can store a fragment of each object at each node, and thus uses a single placement group for all objects. For example, using a (3010, 2150, 860) large code, an object is lost if the system loses more than 860 nodes storing fragments for the object before the object is repaired.
Since there is only one placement group the nested object structure implies that if the object with the fewest number of available of fragments can be recovered then all objects can be recovered. This makes it simpler to determine an R peak value that achieves a target MTTDL for a fixed node failure rate. The basic intuition is that the repair policy should cycle through and repair the objects fast enough to ensure that the number of nodes that fail between repairs of the same object is at most r. Note that D src /R peak is the time between repairs of the same object when the aggregate size of all objects is D src and the peak global read repair bandwidth is set to R peak . Thus, the value of R peak should be set so that the probability that there are more than r node failures in time D src /R peak is extremely tiny. Eq. (3) from Appendix A provides methodology for determining a value of R peak that achieves a target MTTDL.
Unlike small code systems, wherein T RIT must be relatively small, a liquid systems can use large T RIT values, which has the benefit of practically eliminating unnecessary repair due to transient failures.
REPAIR SIMULATION RESULTS
The storage capacity of each node is S = 1 PB in all simulations. The system is fully loaded with source data, i.e., D src = S · M · (1 − β), where β is the storage overhead expressed as a fraction.
There are node failures in all simulations and node lifetimes are modeled as independent exponentially distributed random variables with parameter λ. Unless otherwise specified, we set 1/λ to 3 years. When sector failures are also included, sector lifetimes are also exponentially distributed with 1/λ SF fixed to 5 · 10 8 years.
At the beginning of each simulation the system is in a perfect state of repair, i.e., all n fragments are available for all objects. For each simulation, the peak global read repair bandwidth is limited to a specified R peak value. The MTTDL is calculated as the number of years simulated divided by one more than the observed number of times there was an object loss event during the simulation. The simulation is reinitialized to a perfect repair state and the simulation is continued when there is an object loss event. To achieve accurate estimates of the MTTDL, each simulation runs for a maximum number of years, or until 200 object loss events have occurred. Fig. 3 presents a comparison of repair traffic for a liquid system and a small code system. In this simulation, lazy repair for the liquid system uses a steady read repair rate of 704 Gbps to achieve an MTTDL of 10 8 years, whereas the reactive repair for the small code system uses read repair rate bursts of 6.4 Tbps to achieve an MTTDL slightly smaller than 10 7 years. For the small code system, a read repair rate burst starts each time a node fails and lasts for approximately 4 hours. One of the 3010 nodes fails on average approximately every 8 hours.
In Fig. 4, we plot R peak against achieved MTTDL for a cluster of M = 402 nodes. The liquid systems use fixed read repair rate equal to the indicated R peak and two cases of β = 33.3% and As can be seen, the bounds and the simulation results agree quite closely. Fig. 4 also shows simulation results for a small code system with storage overhead 33.3%, which illustrates a striking difference in performance between small code systems and liquid systems.
Fixed node failure rate with 402 nodes
Fig. 5 plots the simulation results for R peak against MTTDL for a wide variety of M = 402 node systems. Each simulation was run for 10 9 years or until 200 object loss events occurred. The number of nodes, the node failure rate, and the small codes used in these simulations are similar to [10]. The square icons correspond to 16.7% storage overhead and the circle icons correspond to 33.3% storage overhead. For the small code systems, a (18, 15, 3) small code is used for 16.7% storage overhead, and a (9, 6, 3) small code is used for 33.3% storage overhead. For the liquid systems, a (402, 335, 67) large code is used for 16.7% storage overhead, and a (402, 268, 134) large code is used for 33.3% storage overhead. The unshaded icons correspond to systems where there are only node failures, whereas the shaded icons correspond to systems where there are both node failures and sector failures.
The small code systems labeled as "SC, 30min" and "SC, 24hr" use the read repair rate strategy described in Section 4.3 based on the indicated R peak value, with T RIT set to 30 minutes and 24 hours respectively. The triplication systems labeled as "Trip, 30min" and "Trip, 24hr" use the read repair rate strategy described in Section 4.3 based on the indicated R peak value with T RIT is set to 30 minutes and 24 hours respectively.
The liquid systems labeled as "LiqF, 24hr" use a fixed read repair rate set to the shown R peak , which was calculated according to Eq. (3) from Appendix A with a target MTTDL of 10 7 years. The value of T RIT is set to 24 hours As would be expected since Eq. (3) is a lower bound, the observed MTTDL in the simulations is somewhat above 10 7 years. The liquid systems labeled as "LiqR, 24hr" have T RIT set to 24 hours and use a version of the regulator algorithm described in Section 9 to continually and dynamically recalculate the read repair rate according to current conditions. The regulator used (per object) node failure arrival rate estimates based on the average of the last 7 6 · r − F node failure inter-arrival times where F denotes the number of missing fragments for an object at the time it forms the estimate. The maximum read repair rate was limited to the shown R peak value, which is about three times the average rate. In these runs there were no object loss events in the 10 9 years of simulation when there are both node failures and sector failures. This is not unexpected, since the estimate (using the MTTDL lower bound techniques described in Section 9 but with a slightly different form of node failure rate estimation) for the actual MTTDL is above 10 25 years for the 33.3% storage overhead case and above 2.5 · 10 12 years for the 16.7% storage overhead case. These estimates are also known to be conservative when compared to simulation data. The same estimate shows that with this regulator an MTTDL of 10 9 years would be attained with the smaller overhead of r = 115 instead of r = 134. As shown in Table II, the average read repair rate R avg is around 1/3 of R peak , and the 99% read repair rate R 99% is around 1/2 of R peak . The large MTTDL achieved by the regulator arises from its ability to raise the repair rate when needed. Other versions of the regulator in which the peak read repair rate is not so limited can achieve much larger MTTDL. For example, a version with a read repair rate limit set to be about five times the average rate for the given node failure rate estimate achieves an MTTDL of around 10 39 years for the 33.3% storage overhead case.
As can be seen, the value of R peak required for liquid systems is significantly smaller than that required for small code systems and for triplication. Furthermore, although regulated rate liquid system can provide an MTTDL of greater than 10 25 years even when T RIT is set to 24 hours and there are sector failures in addition to node failures, the small code systems and triplication do not provide as good an MTTDL even when T RIT is set to 30 minutes and there are no sector failures, and provide a poor MTTDL when T RIT is set to 24 would be further degraded for small code systems if both T RIT were set to 24 hours and there were sector failures. Fig. 6 shows detailed simulation results, plotting the value of R peak and the resulting value of MTTDL for various M = 3010 node systems. Each simulation was run for 10 8 years or until 200 object loss events occurred. The number of nodes, the node failure rate, and the small code systems used in these simulations are similar to [19]. The square icons correspond to 14.3% storage overhead and the circle icons correspond to 28.6% storage overhead. For the small code systems, a (24,20,4) small code is used for 14.3% storage overhead, and a (14, 10, 4) small code is used for 28.6% storage overhead. For the liquid systems, a (3010, 2580, 430) large code is used for 14.3% storage overhead, and a (3010, 2150, 860) large code is used for 28.6% storage overhead. The remaining parameters and terminology used in Fig. 6 are the same as in Fig. 5. There were no object loss events in the 10 8 years of simulation for any of the liquid systems shown in Fig. 6.
Fixed node failure rate with 3010 nodes
The systems "LiqR, 24hr" liquid systems used a regulated repair rate similar to that described for the 402 node system described above. The target repair efficiency was set at 2 3 r. In this case the average repair rate is higher than for the fixed rate case because the fixed rate case was set using the target MTTDL which, due the larger scale, admits more efficient repair than the 402 node system for the same target MTTDL. Estimates of MTTDL for the regulated case using techniques from Section 9 indicate that the MTTDL is greater than an amazing 10 140 years for the 28.6% storage overhead case and 10 66 years for the 14.3% storage overhead case even without the benefit of node failure rate estimation. As before, the regulated repair system could be run at substantially lower overhead while still maintaining a large MTTDL. For the "LiqR, 24hr" liquid systems, the average read repair rate R avg is around 1/3 of R peak , and the 99% read repair rate R 99% is around 1/2 of R peak .
The conclusions drawn from comparing small code systems to liquid systems shown in Fig. 5 are also valid for Fig. 6 but to an even greater degree. The benefits of the liquid system approach increase as the system size increases.
The paper [19] describes a repair bottleneck for a small code system, hereafter referred to as the Facebook system, which has around 3000 nodes and storage capacity per node of approximately S = 15 TB. The Facebook system uses a (14, 10, 4) RS code to protect 8% of the source data and triplication to protect the remaining 92% of the source data. The amount of repair traffic estimated in [19] for the Facebook system is around 10% to 20% of the total average of 2 PB/day of cluster network traffic. The paper [19] projects that the network would be completely saturated with repair traffic if even 50% of the source data in the Facebook system were protected by the RS code.
Taking into account the mix of replicated and RS protected source data and the mixed repair traffic cost for replicated and RS protected source data, there are around 6.5 to 13 node failures per day in the Facebook system. An average of 6.5 to 13 node failures per day for a 3000 node system implies that each individual node on average fails in around 115 to 230 days, which is around 2.5 to 5 times the node failure rate of 1/λ = 3 years we use in most of our simulations. If all source data were protected by a (14, 10, 4) RS code in the Facebook system then the repair traffic per day would average around 1 to 2 PB. However, the node failure per day statistics in [19] have wide variations, which supports the conclusions in [19] that protecting most of the source data with the RS code in the Facebook system will saturate network capacity.
Note that 1 to 2 PB/day of repair traffic for the Facebook system implies an average read repair rate of around 90 to 180 Gbps. If the storage capacity per node were S = 1 PB instead of S = 15
Varying node failure rate
TB then the Facebook system average read repair rate R avg would be approximately 6 to 12 Tbps, which is around 2.5 to 5 times the average read repair rate R avg for the very similar "SC, 30min" small code system shown in Fig. 6. This relative difference in the value of R avg makes sense, since the Facebook system node failure rate is around 2.5 to 5 times larger than for the"SC, 30min" small code system.
The "SC, 30min" small code system shown in Fig. 6 achieves an MTTDL of just under 10 7 years using a peak read repair rate R peak = 6.4 Tbps and using T RIT = 30 minutes. Assume the node failure rate for the Facebook system is 5 times the node failure rate for the "SC, 30min" small code system shown in Fig. 6, and suppose we set T RIT = 30/5 = 6 minutes for the Facebook system. The scaling observations in Section 5.5 show that this Facebook system achieves a MTTDL of just under 10 7 /5 = 2 · 10 6 years when R peak = 6.4 · 5 = 32 Tbps.
The average node failure rates reported in [19] for the Facebook system are considerably higher than the node failure rates we use in our simulations. One possible reason is that the Facebook system uses a small T RIT value, e.g., T RIT = 10 minutes, after which unresponsive nodes trigger repair. This can cause unnecessary repair of fragments on unresponsive nodes that would have recovered if T RIT were larger.
It is desirable to set T RIT as large as possible to avoid unnecessary repair traffic, but only if a reasonable MTTDL can be achieved. The results in Fig. 6 indicate that a small code system cannot offer a reasonable MTTDL with T RIT = 24 hours. On the other hand, the results from Section 5.4 indicate that the average read repair rate increases significantly (indicating a significant amount of unnecessary repair) if a smaller value of T RIT is used for a small code system when there are transient node failures. In contrast, a regulated liquid system can provide a very large MTTDL operating with T RIT = 24 hours and with sector failures, thus avoiding misclassification of transient node failures and unnecessary repair traffic.
Results for small code systems using LR codes [9], [10], [19] can be deduced from results for small code systems using RS codes. A typical (14, 10, 2, 2) LR code (similar to the one described in [10] which defines a (16, 12, 2, 2) LR code) partitions 10 source fragments into two groups of five, generates one local repair fragment per group and two global repair fragments for a total of 14 fragments, and thus the storage overhead is the same as for a (14, 10, 4) RS code. At least five fragments are read to repair a lost fragment using the LR code, whereas 10 fragments are read to repair a lost fragment using the RS code. However, there are fragment loss patterns the LR code does not protect against that the RS code does. Thus, the LR code operating at half the R peak value used for the RS code achieves a MTTDL that is lower than the MTTDL for the RS code.
Similarly, results for [19] which use a (16, 10, 4, 2) LR code can be compared against the (14, 10, 4) RS code that we simulate. The LR code operating at half the R peak value used for the RS code achieves an MTTDL that may be as large as the MTTDL for the RS code. However, the LR code storage overhead is β = 0.375, which is higher than the RS code storage overhead of β = 0.286.
Varying node failure rate
In this subsection we demonstrate the ability of the regulated repair system to respond to bursty node failure processes. The node failures in all simulations described in this subsection are generated by a time varying periodic Poisson process repeating the following pattern over each ten year period of time: the node failure rate is set to 1/λ = 3 years for the first nine years of each ten year period, and then 1/λ = 1 year for the last year of each ten year period. Thus, in each ten year period, the node failure rate is the same as it was in the previous subsections for the first nine years, followed by a node failure rate that is three times higher for the last year.
Except for using variable λ for node failures, Fig. 7 is similar to Fig. 5. The small code systems labeled as "SC, 30min" in Fig. 7 use the read repair rate strategy described in Section 4.3 based on the shown R peak value, and T RIT is set to 30 minutes. Thus, even when the node failure rate is lower during the first nine years of each period, the small code system still sets read repair rate to R peak whenever repair is needed. The liquid systems labeled as "LiqR, 24hr" in Fig. 7 use the regulator algorithm described in Section 9 to continually and dynamically recalculate the read repair rate according to current conditions, allowing a maximum read repair rate up to the shown R peak value, and T RIT is set to 24 hours. In these runs there were no object loss events in the 10 9 years of simulation with both node failures and sector failures. [[We may want to compute something here TBD?]] Fig. 18 is an example trace of the read repair rate as function of time for the liquid system labeled "LiqR, 24hr" in Fig. 7, which shows how the the read repair rate automatically adjusts as the node failure rate varies over time. In these simulations, the average read repair rate R avg is around 1/9 of R peak during the first nine years of each period and around 1/3 of R peak during the last year of each period, and the 99% read repair rate R 99% is around 1/6 of R peak during the first nine years of each period and around 1/2 of R peak during the last year of each period. Thus, the regulator algorithm does a good job of matching the read repair rate to what is needed according to the current node failure rate.
Except for using variable λ for node failures, Fig. 8 is similar to Fig. 6. The conclusions comparing small code systems to liquid systems shown in Fig. 7 are also valid for Fig. 8.
Transient node failures
In this subsection we demonstrate the ability of the liquid systems to efficiently handle transient node failure processes. Fig. 10 plots the simulation results for R avg against MTTDL for a number of M = 402 node systems. Like in previous subsections, the node lifetimes for node failures are modeled as independent exponentially distributed random variables with parameter λ and is set to 1/λ = 3 years. The occurrence times of transient node failures are modeled as independent exponentially distributed random variables with parameter λ TF and is set to 1/λ TF = 0.33 years. Thus, transient node failures occur at 9 times the rate at which node failures occur, consistent with [8]. The durations of transient node failures are modeled with log-logistic random variables, having a median of 60 seconds and shape parameter 1.1. These choices were made so as to mimic the distribution provided in [8]. shows the graph from [8] with the log-logistic fitted curve overlaid. With this model, less than 10% of the transient node failures last for more than 15 minutes.
The unshaded markers mark simulations with just node failures and the shaded markers mark simulations with both node failures and transient node failures.
The R peak values in all simulations is the same as R peak value used correspondingly in Fig. 5. The small code systems labeled as "SC, 30min" and "SC, 24hr" in Fig. 10 use the read repair rate strategy described in Section 4.3, with R peak value of 6400 Gbps and T RIT is set to 30 minutes and 24 hours respectively. The liquid system labeled as "LiqR, 24hr" in Fig. 10 uses the regulator algorithm described in Section 9, with R peak value of 311 Gbps and T RIT is set to 24 hours. As evident, there is no difference in the R avg for the liquid system between simulations with transient node failures and those without. No object loss events were observed for the liquid system in the 10 9 years of simulation with both node failures and transient node failures. The R avg for small code system labeled as "SC, 30min" is however higher for the simulation with transient node failures and achieves an MTTDL that is less than half of what is achieved when there are no transient node failures. The R avg for small code system labeled as "SC, 24hr" is the same between simulations with transient node failures and those without, but they fail to achieve a high enough MTTDL and do not provide adequate protection against object loss.
Thus, the liquid systems do a better job of handling transient node failures without requiring more effective average bandwidth than small code systems.
Scaling parameters
The simulation results shown in Section 5 are based on sets of parameters from published literature, and from current trends in deployment practices. As an example of a trend, a few years ago the storage capacity of a node was around S = 16 TB, whereas the storage capacity of a node in some systems is S = 1 PB or more, and thus the storage capacity per node has grown substantially.
Repair bandwidths scale linearly as a function of the storage capacity S per node (keeping other input parameters the same). For example, the values of (R peak , R avg ) for a system with S = 512 TB can be obtained by scaling down by a factor of two the values of (R peak , R avg ) from simulation results for S = 1 PB, whereas the MTTDL and other values remain unchanged.
Repair bandwidths and the MTTDL scale linearly as a function of concurrently scaling the node failure rate λ and the repair initiation timer T RIT . For example, the values of (R peak , R avg ) for a system with 1/λ = 9 years and T RIT = 72 hours can be obtained by scaling down by a factor of three the values of (R peak , R avg ) from simulation results for 1/λ = 3 years and T RIT = 24 hours, whereas the MTTDL can be obtained by scaling up by a factor of three, and other values remain unchanged.
ENCODING AND DECODING OBJECTS
In a block storage organization commonly used by small code systems, objects are partitioned into k successive source fragments and encoded to generate r = n − k repair fragments (we assume the small code is MDS). In effect a fragment is a symbol of the small code, although the small code may be intrinsically defined over smaller symbols, e.g., bytes. A high level representation of a block storage organization for a simple example using a (9, 6, 3) Reed-Solomon code is shown in Fig. 11, where the object runs from left to right across the source fragments as indicated by the blue line.
A relatively small chunk of contiguous data from the object that resides within a single source fragment can be accessed directly from the associated node if it is available. If, however, that node is unavailable, perhaps due to a transient node failure or node failure, then k corresponding chunks of equal size must be read from other nodes in the placement group and decoded to generate the desired chunk. This involves reading k times the size of the missing chunk and performing a decoding operation, which is referred to as a degraded read [19], [18]. Reading k chunks can incur further delays if any of the nodes storing the chunks are busy and thus non-responsive. This latter situation can be ameliorated if a number of chunks slightly larger than k are requested and decoding initiated as soon as the first k chunks have arrived, as described in [11].
A flow storage organization, used by liquid systems, operates in a stream fashion. An (n, k, r) large code is used with small symbols of size Ssize, resulting in a small source block of size Bsize = k · Ssize. For example, with k = 1024 and symbols of size Ssize = 64 Bytes, a source block is of size Bsize = 64 KB. An object of size Osize is segmented into N = Osize Bsize source blocks, and the k source symbols of each such source block is independently erasure encoded into n symbols. For each i = 0, . . . , n − 1, the fragment with EFI i consists of the concatenation of the i th symbol from each of the N consecutive source blocks of the object. An example of a flow storage organization showing a (1536, 1024, 512) large code with N = 24, 576 is shown in Fig. 12, where each row corresponds to a source block and runs from left to right across the source symbols as indicated by the dashed blue lines, and the object is the concatenation of the rows from top to bottom.
A chunk of data consisting of a consecutive set of source blocks can be accessed as follows. For a fragment with EFI i, the consecutive portion of the fragment that corresponds to the i th symbol from each of consecutive set of source blocks is read. When such portions from at least k fragments are received (each portion size a 1/k-fraction of the chunk size), the chunk of data can be recovered by decoding the consecutive set of source blocks in sequence. Thus, the amount of data that is read to access a chunk of data is equal to the size of the chunk of data. Portions from slightly more than k fragments can be read to reduce the latency due to nodes that are busy and thus non-responsive. Note that each access requires a decoding operation.
Let us contrast the two organizations with a specific example using an object of size Osize = 1.5 GB. Consider an access request for the 32 MB chunk of the object in positions 576 MB through 608 MB, which is shown as the shaded region in Fig. 13 with respect to both a block storage organization and a flow storage organization. With the block storage organization shown in Fig. 11, the 32 MB chunk is within the fragment with EFI = 2 as shown in Fig. 13. If the node that stores this fragment is available then the chunk can be read directly from that node. If that node is unavailable, however, then in order to recover the 32 MB chunk the corresponding 32 MB chunks must be retrieved from any six available nodes of the placement group and decoded. In this case a total of 192 MB of data is read and transferred through the network to recover the 32 MB chunk of data.
With the flow storage organization shown in Fig. 12, the chunk of size 32 MB corresponds to 512 consecutive source blocks from source block 9216 to block 9727 as shown in Fig. 13, which can be retrieved by reading the corresponding portion of the fragment from each of at least 1024 fragments. From the received portions, the chunk can be recovered by decoding the 512 source blocks. Thus, in a flow storage organization a total of 32 MB of data is read and transferred through the network to recover the original 32 MB chunk of data.
Naturally, the symbol size, source block size, and the granularity of the data to be accessed should be chosen to optimize the overall design, taking into account constraints such as whether the physical storage is hard disk or solid state.
PROTOTYPE IMPLEMENTATION
Our team implemented a prototype of the liquid system. We used the prototype to understand and improve the operational characteristics of a liquid system implementation in a real world setting, and some of these learnings are described in Section 8. We used the prototype to compare access performance of liquid systems and small code systems. We used the prototype to cross-verify the lazy repair simulator that uses a fixed read repair rate as described in Section 4.4 to ensure that they both produce the same MTTDL under the same conditions. We used the prototype to validate the basic behavior of a regulated read repair rate as described in Section 9.
The hardware we used for our prototype consists of a rack of 14 servers. The servers are connected via 10 Gbit full duplex ethernet links to a switch, and are equipped with Intel Xeon CPUs running at 2.8 GHz, with each CPU having 20 cores. Each server is equipped with an SSD drive of 600 GB capacity that is used for storage. The liquid prototype system software is custom written in the Go programming language. It consists of four main modules: -Storage Node (SN) software. The storage node software is a HTTP server. It is used to store fragments on the SSD drive. Since a liquid system would generally use many more than 14 storage nodes, we ran many instances of the storage node software on a small number of physical servers, thereby emulating systems with hundreds of nodes. -An Access Generator creates random user requests for user data and dispatches them to accessors.
It also collects the resulting access times from the accessors. -Accessors take user data requests from the access generator and create corresponding requests for (full or partial) fragments from the storage nodes. They collect the fragment responses and measure the amount of time it takes until enough fragments have been received to recreate the user data. There are in general multiple accessors running in the system. -A Repair Process is used to manage the repair queue and regenerate missing fragments of stored objects. The repair process can be configured to use a fixed read repair rate, or to use a regulated read repair rate as described in Section 9.
In addition, the team developed a number of utilities to test and exercise the system, e.g., utilities to cause node failures, and tools to collect data about the test runs.
Access performance setup
The users of the system are modeled by the access generator module. We set a target user access load ρ (0 < ρ < 1) on the system as follows. Let C be the aggregate capacity of the network links to the storage nodes, and let s be the size of the user data requests that will be issued. Then C/s requests per unit of time would use all available capacity. We set the mean time between successive user requests to t = s ρ·C , so that on average a fraction ρ of the capacity is used. The access generator module uses an exponential random variable to generate the timing of each successive user data request, where the mean of the random variable is t.
As an example, there is a 10 Gbps link to each of the six storage nodes in the setup of Fig. 14, and thus C = 60 Gbps. If user data requests are of size s = 10 MB = 80 · 2 20 bits, and the target load is ρ = 0.8 (80% of capacity), then the mean time between requests is set to t = 80 · 2 20 0.8 · 60 · 10 9 ≈ 1.75 milliseconds. The access generator module round-robins the generated user data requests across the accessor modules. When an accessor module receives a generated user data request from the access generator module, the accessor module is responsible for making fragment requests to the storage nodes and collecting response payloads; it thus needs to be fairly high performance. Our implementation of an accessor module uses a custom HTTP stack, which pipelines requests and keeps connections statically alive over extended periods of time. Fragment requests are timed to avoid swamping the switch with response data and this reduces the risk of packet loss significantly. The server network was also tuned, e.g., the TCP retransmission timer was adjusted for the use case of a local network. See [5] for a more detailed discussion of network issues in a cluster environment. These changes in aggregate resulted in a configuration that has low response latency and is able to saturate the network links fully.
Access performance tests
The prototype was used to compare access performance of liquid systems and small code systems. Our goal was to evaluate access performance using liquid system implementations, i.e., understand the performance impact of requesting small fragments (or portions of fragments) from a large number of storage servers. For simplicity we evaluated the network impact on access performance. Fragment data was generally in cache and thus disk access times were not part of the evaluation. We also did not evaluate the computational cost of decoding. Some of these excluded aspects are addressed separately in Section 8. Figure 14 displays the setup that we use for testing access speed. Each physical server has a full duplex 10 GbE network link which is connected over a switch to all the other server blades. Testing shows that we are able to saturate all of the 10 Gbps links of the switch simultaneously. Eight of our 14 physical servers are running accessors, and one of the eight is running the access generator as well. The remaining six servers are running instances of the storage node software. This configuration allows us to test the system under a load that saturates the network capacity to the storage nodes: The aggregate network capacity between the switch and the accessors is 80 Gbps, which is more than the aggregate network capacity of 60 Gbps between the switch and the storage nodes, so the bottleneck is the 60 Gbps between the switch and the storage nodes.
During access performance tests there are no node failures, all n fragments for each object are available, and the repair process is disabled. For all tests, 67 storage node instances run on each storage server, which emulates a system of 402 nodes. We operate with a storage overhead of 33.3%, a (402, 268, 134) large code is used for the liquid system, and a (9, 6, 3) small code is used for the small code system. We tested with 10 MB and 100 MB user data request sizes.
We run the access prototype in three different configurations. The "Liq" configuration models liquid system access of user data. For a user data request of size s, the accessor requests k + E fragment portions, each of size s/k, from a random subset of distinct storage nodes, and measures the time until the first k complete responses are received (any remaining up to E responses are discarded silently by the accessor). Each fragment portion is of size around 25.5 KB when s = 10 MB, and around 255 KB when s = 100 MB. We use E = 30, and thus the total size of requested fragment portions is around 10.75 MB when s = 10 MB, and around 107.5 MB when s = 100 MB. See [11] for an example of this access strategy.
The "SC" configuration models small code system normal access of user data, i.e., the requested user data is stored on a storage node that is currently available. For a user data request of size s, the accessor requests one fragment portion of size s from a randomly selected storage node, and measures the time until the complete response is received. The total size of the requested fragment portion is 10 MB when s = 10 MB, and 100 MB when s = 100 MB.
The "SC-Deg" configuration models small code system degraded access of user data, i.e., the requested user data is stored at a storage node that that is currently unavailable, e.g., see [19], [18]. For a user data request of size s, the accessor requests k fragment portions of size s from a random subset of distinct storage nodes and measures the time until the first k complete responses are received. The total size of requested fragment portions is 60 MB when s = 10 MB, and 600 MB when s = 100 MB. Most user data is stored at available storage nodes, and thus most accesses are normal, not degraded, in operation of a small code system. Thus, we generate the desired load on the system with normal accesses to user data, and run degraded accesses at a rate that adds only a nominal aggregate load to the system, and only the times for the degraded accesses are measured by the accessors. Similar to the "Liq" configuration, we could have requested k + 1 fragment portions for the "SC-Deg" configuration to decrease the variation in access times, but at the expense of even more data transfer over the network. Figures 15 and 16 depict access time results obtained with the above settings under different system loads. The average time for liquid system accesses is less than for small code system normal accesses in most cases, except under very light load when they are similar on average. Even more striking, the variation in time is substantially smaller for liquid systems than for small code systems under all loads. Liquid system accesses are far faster (and consume less network resources) than small code system degraded accesses.
Our interpretation of these results is that liquid systems spread load equally, whereas small code systems tend to stress individual storage nodes unevenly, leading to hotspots. With small code systems, requests to heavily loaded nodes result in slower response times. Hotspots occur in small code systems when appreciably loaded, even when the fragment requests are distributed uniformly at random. It should be expected that if the requests are not uniform, but some data is significantly more popular than other data, then response times for small code systems would be even more variable.
Repair tests and verification of the simulator
We used the prototype repair process to verify our liquid system repair simulator. We started from realistic sets of parameters, and then sped them up by a large factor (e.g., 1 million), in such a way that we could run the prototype repair process and observe object loss in a realistic amount of time.
We ran the prototype repair process in such a configuration, where node failures were generated artificially by software. The resulting measured statistics, in particular the MTTDL, were compared to an equivalent run of the liquid system repair simulator. This allowed us to verify that the liquid system repair simulator statistics matched those of the prototype repair process and also matched our analytical predictions.
IMPLEMENTATION CONSIDERATIONS
Designing a practical liquid system poses some challenges that are different from those for small code systems. In the following sections we describe some of the issues, and how they can be addressed.
Example Architecture
As discussed in Section 2.4, Fig. 1 shows an example of a liquid system architecture. This simplified figure omits components not directly related to source data storage, such as system management components or access control units. Concurrent capacity to access and store objects scales with the number of access proxy servers, while source data storage capacity can be increased by adding more storage nodes to the system, and thus these capacities scale independently with this architecture. Storage nodes (storage servers) can be extremely simple, as their only function is to store and provide access to fragments, which allows them to be simple and inexpensive. The lazy repair process is independent of other components; thus reliability is handled by a dedicated system.
Metadata
The symbol size used by a liquid system with a flow storage organization is substantially smaller than the size used by a small code system with a block storage organization. For a small code system each symbol is a fragment, whereas for a liquid system there are many symbols per fragment and, as described in Section 6, the symbol to fragment mapping can be tracked implicitly. For both types of systems, each fragment can be stored at a storage node as a file. The names and locations of each fragment are explicitly tracked for small code systems. There are many more fragments to track for a liquid system because of the use of a large code, and explicit tracking is less appealing. Instead, a mapping from the placement group to available nodes can be tracked, and this together with a unique name for each object can be used to implicitly derive the names and locations of fragments for all objects stored at the nodes for liquid systems.
Network usage
The network usage of a liquid system is fairly different from that of a small code system. As described in Section 7.2, liquid systems use network resources more smoothly than small code systems, but liquid systems must efficiently handle many more requests for smaller data sizes. For example, a small code system using TCP might open a new TCP connection to the storage node for 8.4 Storage medium each user data request. For a liquid system, user data requests are split into many smaller fragment requests, and opening a new TCP connection for each fragment request is inefficient. Instead, each access proxy can maintain permanent TCP connections to each of the storage nodes, something that modern operating systems can do easily. It is also important to use protocols that allow pipeline data transfers over the connections, so that network optimizations such as jumbo packets can be effectively used. For example the HTTP/2 protocol [1] satisfies these properties.
Technologies such as RDMA [16] can be used to eliminate processing overhead for small packets of data. For example, received portions of fragments can be placed autonomously via RDMA in the memory of the access proxy, and the access proxy can decode or encode one or more source blocks of data concurrently.
Storage medium
A key strength of liquid systems is that user data can be accessed as a stream, i.e., there is very little startup delay until the first part of a user data request is available, and the entire user data request is available with minimal delay, e.g., see Section 7.2. Nevertheless, random requests for very small amounts of user data is more challenging for a liquid system.
The storage medium used at the storage nodes is an important consideration when deciding the large code parameters (n, k, r). Hard disk drives (HDDs) efficiently support random reads of data blocks of size 500 KB or larger, whereas solid state drives (SSDs) efficiently support random reads of data blocks of size 4 KB or larger. We refer to these data block sizes as the basic read size of the storage medium. The parameters (n, k, r) should be chosen so that if s is the typical requested user data size then s/k is at least the basic read size, which ensures that a requested fragment portion size is at least the basic read size. In many cases this condition is satisfied, e.g., when the requested user data size is generally large, or when the storage medium is SSD.
In other cases, more advanced methods are required to ensure that a liquid system is efficient. For example, user data request performance can be improved by assigning different size user data to different types of objects: Smaller user data is assigned to objects that are stored using smaller values of k, and larger user data is assigned to objects that are stored using larger values of k, where in each case s/k is at least the basic read size if s is the requested user data size. In order to achieve a large MTTDL for all objects, the storage overhead r n is set larger for objects stored using smaller values of k than for objects stored using larger values of k. Overall, since the bulk of the source data is likely to be large user data, the overall storage overhead remains reasonable with this approach.
Only a small proportion of user data is frequently accessed in many use cases, e.g., in the use case described in [15], only about 10% of the user data is accessed frequently, and this user data is easy to identify based on the amount of time it has been stored in the storage system. Thus, a caching layer can be used to store and make such user data available without having to request the user data from the storage system. For a liquid system, it makes most sense to cache user data that is both small and frequently accessed.
Another strategy to minimize requests for small user data is to store related user data together in objects and coalesce accesses. Suppose for example, the storage system stores web content. To load a web page, dozens of small files of user data are typically loaded, such as HTML content and small images. Since the same content is accessed whenever such a web page is displayed, the system can store and access this user data together.
RaptorQ software performance
With small code systems, user data can usually be accessed directly without requiring erasure decoding. With liquid systems, each access of user data typically requires erasure decoding.
RaptorQ codes are the large codes of choice for liquid systems. The encoding and decoding complexity of RaptorQ codes is inherently linear by design [20], i.e., for a (n, k, r) RaptorQ code, the encoding complexity is linear in n, and the decoding complexity is linear in k, which makes the use of a RaptorQ code with large values of (n, k, r) practical. Software implementations of RaptorQ codes have been deployed in a number of applications, including real-time applications that require low encoding and decoding complexity. Encode and decode speeds of over 10 Gbps are achieved using a software implementation of RaptorQ running on one core of a 20 core Intel Xeon CPU ES-2680 server running at 2.8 GHz, with code parameters (n, k, r) = (1500, 1000, 500) and 256 byte symbols. Aggregate encode and decode speeds of around 150 Gbps are achieved using all 20 cores of the same server, with the same code parameters and symbol size.
These numbers suggest that RaptorQ decoding would add a fractional extra cost to the data processing pipeline: It is expected that other transformations are performed when accessing user data, such as for example checksumming, decompression and decryption. Of those MD5 checksums can be computed on the same machine at a rate of at most about 4 Gbps. As for decompression, Google's run-time performance oriented Snappy codec achieves reportedly about 4 Gbps on a Core i7 processor [23], suggesting that RaptorQ coding is not a bottleneck.
RaptorQ hardware performance
Hardware based implementations have the potential to further cut down on power usage by the RaptorQ codes. Based on a hardware RaptorQ code chip design by Qualcomm for a different application, our estimate is that the power usage in an ASIC implementation would be about 3 Watts per 100 Gbps encoding or decoding speed.
To put this into perspective, we compare this to the power use of an SSD to access data. For example, the Samsung SM863a 2 TB enterprise SSD drive [24], achieves a data transfer rate of up to about 4.1 Gbps and uses about 2.5 W of power when reading. This translates to a power use of about 61 W to achieve 100 Gbps. Thus, the power needed for the RaptorQ code is about 5% of the power needed to read the data off a drive. If other necessary aspects such as network transfer, etc, were incorporated as well, the RaptorQ code share drops well below 5%.
RaptorQ decode reliability
Extremely high durability is required for storage systems. MDS codes have the property that an object can be recovered from any k fragments for the object. RaptorQ codes are essentially MDS, but not literally MDS: it is not necessarily the case that an object can be recovered from any k fragments. We present simulation results showing the probability of object loss due to RaptorQ codes not being MDS can be essentially ignored. Consider a liquid system using a (1200, 1000, 200) RaptorQ code, where the target w = 50, i.e., the number of available fragments for any object is always at least k + w = 1050. If 1/λ = 3 years, then on average there are 400 node failures per year. We group node failures together into consecutive batches of w − 4 = 46, and consider testing for decoding using the set of k + w EFIs for available fragments at the beginning of the batch minus the w − 4 = 46 EFIs for fragments lost within the batch, and thus we decode from k + w − (w − 4) = k + 4 = 1004 EFIs. Note that if decoding is possible from this set of 1004 EFIs then decoding is possible for all objects at each point during the batch.
As shown in Table I, there were no RaptorQ decoding failures in an experiment where 10 11 random sets of 1004 EFIs were tested for decodability. Since there are on average 400 node failures per year and the batch size is 46, there will be on average 400 46 < 10 such batches per year, and thus these 10 11 decoding tests cover 10 billion years of simulation time. Thus there would be no decoding failures for RaptorQ within 10 billion years, which is a thousand times a target MTTDL of 10 million years.
READ REPAIR RATE REGULATION
The simple Poisson process model of node failures discussed in Section 3.1 is useful to benchmark the reliability of a distributed storage system. However, node failures in real systems are not necessarily faithful to a simple statistical model.
Detailed statistical assumptions of the node failure process do not significantly affect reactive repair decisions for a small code system, i.e., the actions taken by the reactive repair process are largely event driven and immediate. For example, repairs are executed as soon as fragments are lost due to node failures, with no dependency on assumptions on future node failures. The anticipated node failure rate is only germane to the initial provisioning of repair resources (e.g., total repair bandwidth). This is in sharp contrast to lazy repair for a liquid system, where repairs are postponed to achieve greater repair efficiency. The degree to which repairs can be postponed is clearly dependent on some expectation of future node failures. The tradeoff between repair efficiency and MTTDL rests on assumptions about the long term node failure rate. In the liquid system with fixed read repair rate this dependence appears prominently in the choice of the read repair rate. Lower read repair rate results in greater repair efficiency but higher probability of object loss. This is clearly evident in the (tight) MTTDL estimate for the fixed rate system derived in Appendix A.
For example, consider a liquid system using a (402, 268, 134) large code. If the node failure rate λ is 1/3 per year and the read repair rate is set so that λ · T = 0.21 then the achieved MTTDL (estimate) is 3.6 · 10 9 years. If instead λ is 10% higher then the MTTDL is 1.0 · 10 7 years, and if λ is 20% higher then the MTTDL is 8.7 · 10 4 years. Thus, the MTTDL of a fixed read repair rate design is sensitive to error in knowledge of λ. If there is uncertainty in the node failure rate then either some repair efficiency must be sacrificed or a reduction in MTTDL must be tolerated for a fixed read repair rate design.
Fortunately, liquid systems can be fitted with an adjustment algorithm that smoothly and automatically adjusts the read repair rate in response to node failure conditions so that accurate a priori knowledge of node failure rates is not required. Suppose for example that the node failure rate increases at some point in time, possibly due to component quality issues. This will result in a relative increase of the number of missing fragments among objects in the repair queue. Moreover, over time the increase in node failure rate can be detected by a node failure rate estimator. Both of these occurrences can signal to the lazy repair process a need to increase the read repair rate. Properly designed, a regulated read repair rate process can achieve significantly larger MTTDLthan a fixed read repair rate process running at the same average read repair rate, at the cost of some variation in read repair rate.
Even in the case where the node failure process is a Poisson process of known rate, regulation of the read repair rate can bring significant benefit. To develop a good regulation method it is helpful to consider in more detail the behavior of a fixed read repair rate process. We assume a liquid system with n nodes and a node failure rate λ. We assume a fixed read repair rate that we characterize using T, the time required to repair all objects in the system. We work in the large number of objects limit where the size of a single object is a negligible fraction of all source data. Objects are repaired in a fixed periodic sequence. We view all objects as residing in a repair queue and we use x to denote the relative position in the queue. Thus, an object at position x = 1 is under repair and the object in position x = 0 has just been repaired. An object in position x will be repaired after a time (1−x)·T and was repaired a time x · T in the past.
For an object currently in position x let f (x) · n denote the number of its fragments that have been erased due to node failures. 2 Assume 0 ≤ x ≤ y ≤ 1 and 0 ≤ s ≤ t ≤ 1, and that the read repair rate is characterized by T. The number of erased fragments f (y) · n the object will have when it reaches position y is given by the transition probability where B(n, m, q) is the probability mass at m of a binomially distributed random variable with parameters (n, q). That is, where we have introduced the notationq = 1 − q. Applying this to the particular case with initial condition f (0) · n = 0 we have Pr [f (x) · n = m] = B(n, m, e −λ·T ·x ) . Using Stirling's approximation ln(n!) = n · ln(n) + O(ln(n)) we can write ln( where where H(f ) = −f · log 2 (f ) −f · log 2 (f ) is the entropy function. As a function of f, the exponent E(f, λ · T · x) is minimized at f = 1 − e −λ·T ·x , which is also the expected value of f (x), being the solution to the differential equation which governs the expected value. Thus, in the large system limit the fraction of erased fragments as function of repair queue position concentrates around the function 1 − e −λ·T ·x . In particular we identify n · (1 − e −λ·T ) as the number of fragments that are typically repaired. We can interpret the quantity f tar := 1 − e −λ·T as a target fraction of repair fragments. Note that f tar and λ · T are in a one-to-one relationship so, given the node failure rate λ, we can view f tar as a specification of the read repair rate. This perspective on specification of system behavior is a convenient one for consideration of regulated read repair rate systems.
To gain a further understanding of the fixed read repair rate system behavior it is helpful to consider typical behavior under object loss. We assume that object loss occurs when an object has more than r = β · n fragments erased and consider the number of erased fragments as a function of queue position for such an object. Thus, we consider the distribution Pr [f (x) · n | f (1) · n > β · n] , which, using Bayes rule, can be written as Using (1) and Stirling's approximation as above, we find that the in the large n limit the solution concentrates around f (1) · n = β · n, and, more generally, f (x) = β ftar · (1 − e −λ·T ·x ) which is simply a scaled version of the nominal queue function 1 − e −λ·T ·x . Note that this solution satisfies the equation Thus, for small x, where f (x) is small, the solution corresponds to a typical behavior for the system with node failure rate λ · β ftar . Another likely contribution to typical object loss is a skewing of node failures to those nodes which, for the given object in the given repair cycle, have not previously failed. This effect can explain the second factor 1 − ftar β · f (x) which replaces 1 − f (x) in the equation for the expected value. Thus, we observe that object loss typically involves a sustained increase in the observed node failure rate over an entire repair cycle and possible skewing of the node failure selection. The fluctuation in node failure rate can be detected by a node failure rate estimator. Both the increased node failure rate and the node selection skewing can be detected from the atypical behavior of the repair queue. We shall first consider a regulator that assumes a known fixed node failure rate λ and responds only to the repair queue, and later extend the design to incorporate estimators of node failure arrival rates.
Under regulated repair we would expect that if the repair queue state is typical, i.e., the fraction of erased fragments for an object in position x is near its nominal value of 1 − e −λ·T ·x then the read repair rate should be kept near its nominal value. If, however, at some points in the repair queue the number of missing fragments is larger than the nominal amount then the read repair rate should be increased. As a general philosophy we adopt the view that each object in the repair queue takes responsibility for itself to ensure that the probability of object loss for that object is kept small. Each object, knowing its position in the queue and knowing its number of erased fragments, desires a certain system read repair rate such that, if adopted, would adequately protect that object against object loss. The repair system will apply the highest of the desired read repair rates thus satisfying all read repair rate requests for all the objects.
The regulator design defines a function φ(f, x) such that an object in position x with a fraction f missing fragments requests a read repair rate corresponding to a system repair time of λ −1 φ(f, x). Recall that a read repair rate corresponds to time required to repair all object in the system T and that λ · T is the expected number of times a single node fails during the time it takes to repair all objects. In this form, φ(f, x) expresses a desired read repair rate in terms of the desired value for λ · T. In general φ(f, x) will be monotonically decreasing in f and increasing in x. Under these assumptions, as well as some assumptions on node failure rate estimators, it follows that the read repair rate will always be determined by certain critical objects in the repair queue. The critical objects are those that were at the tail of the queue when a node failure occurred (see Appendix A for more detail). Note that the fixed read repair rate case corresponds to setting φ constant.
As a basis for designing the function φ(f, x) we adopt the notion that objects desire to recover certain properties of the nominal trajectory. One quantity of particular interest is the probability of object loss for an object. Given a position x in the queue and a fraction f (x) < β erased fragments the probability that the object will experience object loss can be expressed using (1). For an object on the nominal trajectory, f (x) = 1 − e λ·T ·x = 1 − (1 − f tar ) x , this probability is easily seen to be decreasing in x. Thus, if the object enjoys typical behavior early in the repair queue then its probability of object loss decreases. As a design principle, therefore, we can stipulate that an object in position x with a fraction f of erased fragments will request a system repair parameter λ · T, as represented by the function φ(f, x), such that the probability of object loss for that object would be the same as an object at that position on the nominal trajectory. Under this stipulation, we obtain φ(f, x) implicitly as the solution to Note that in this context the quantity β ·n represents a particular threshold that the regulator attempts avoid reaching and need not necessarily correspond with the erasure correcting limit of the used erasure code. For that reason we introduce a different notation f T for the target hard threshold. In practice one will typically have f T β.
Simpler expressions of the above principle for deriving appropriate φ functions can be obtained in a number of ways. One possibility is note that the above sums are dominated by the term with s · n = β · n and then, using the Stirling approximation, to equate the rate functions on both sides. Another approach is to use a Gaussian approximation of the above binomial distributions and equate the integrals that correspond to the above sums. Under the Gaussian approximation we obtain the equation where φ nom = − ln(1 − f tar ) and f e denotes the expected fraction of erased fragments that will be repaired for an object in position x with f erased fragments assuming that the read repair rate 9.1 Regulated read repair rate simulation is subsequently fixed according to φ(f, x), i.e., we havef e =f e −φ(f,x)·(1−x) . Substituting this into the above we obtain a quadratic equation for f e whose solution then determines φ. A good approximation is under which the level curves of φ(f, x) correspond exactly to asymptotic trajectories for the regulated system that can be parameterized by their implied target fragment repair fraction f e . This implies that an object that is controlling the repair rate expects on average to hold the repair constant until it is repaired. For the simulation results given below we used the Gaussian approximation approach to obtain φ. It is not hard to show that φ is bounded above (read repair rate never goes to 0). A lower bound is 0 (infinite read repair rate) which is requested if the f (x) is sufficiently close to f T . In practice the read repair rate will obviously be limited by available resources, and such a limit can be brought into the definition of φ by limiting its minimum value and for simulations we typically saturate φ at some postive value, often specified as fraction of φ nom .
Recall that λ −1 φ represents desired time to repair all objects precisely once. This indicates how to incorporate estimates for the node failure rate into the operation of the regulator: one simply replaces λ in the expression λ −1 φ with a suitable estimate of λ. In general, an estimator of the node failure rate will be based on past node failures . One pertinent quantity in this context is the appropriate time scale to use to form the estimate. We may choose to vary the time scale as a function of the number of erased fragments the object experiences, or other aspects of the node failure process.
Let us assume a system with a total of O objects of equal size. Then each object has a position in the queue x with x · O ∈ {0, ..., O − 1}. When an object is repaired the position x of each object is incremented by 1/O. According to the regulated read repair rate process the repair time for the next object is then given by λ · O where f (x) denotes the fraction of erased fragments for the object in position x. Here we tacitly assume f (x) < f T for all x but for analytical purposes we may extend the definition of φ to all f and x in [0, 1]. In practice φ is effectively bounded from below by the repair speed limitations of the system. In our simulations we typically assume a floor for φ of γ · φ nom with γ set to a small fraction, e.g. γ = 1 3 . Note further that we always have φ(0, 0) = φ nom = − ln(1 − f tar ) and so this gives a lower bound on the read repair rate and we can impose a ceiling on the function φ of value φ nom without altering system behavior.
The choice of f tar involves a tradeoff: lower values of f tar lead to higher read repair rates, but also a more reliable system, and one that is potentially less prone to large read repair rate variations, whereas a larger value of f tar gives higher repair efficiency at the price of larger read repair rate fluctuations and a reduced object loss safety margin of the system. For simplicity, all the results that we present in this section with the repair adjustment algorithm were obtained by running the algorithm with f tar = 2·r 3 , which is a reasonable choice in all but some contrived circumstances. For certain forms of failure rate estimators it is feasible to compute lower bounds on MTTDL. The details of the models and the calculations can be found in Appendix A. The bounds can be used to designing system parameters such as f T and more generally φ so as to maximize repair efficiency while ensuring a desired MTTDL. Fig. 17 illustrates how the regulated read repair rate automatically varies over time when the node failure rate is 1/3 years for a liquid system using a (402, 268, 134) large code with storage overhead 33.3%, and the first row of Table II shows a comparison between the regulated repair rate with an average of R avg = 107 Gbps and a fixed read repair rate of R peak = 104 Gbps (set so that the MTTDL is ≈ 10 7 years). In this table, the MTTDL of 10 9 for the regulated read repair rate indicates there were no object losses in 10 9 years of simulation (based on estimates using the results from Appendix A, the MTTDL is orders of magnitude larger than 10 9 years), and R peak is a hard upper limit on the used read repair rate. The variability of the regulated read repair rate is moderate, e.g., 99.99% of the time the read repair rate is less than 231 Gbps. The second row of Table II 18 shows the behavior of regulated read repair rate when the node failure rate is varied as described in Section 5.3. The upper portion of Fig. 18 displays the average node failure rate varying over time, and the lower portion of Fig. 18 displays the corresponding read repair rate chosen by the regulator. This is a stress test of the regulator, which appropriately adjusts the read repair rate to the quickly changing node failure rate to avoid object losses while at the same time using an appropriate amount of bandwidth for the current node failure rate. There were no object losses in 10 9 years of simulation using the regulator running in the conditions depicted in Fig. 18, and based on estimates using the results from Appendix A the MTTDL is orders of magnitude larger than 10 9 years. The results from Appendix A show that the regulator is effective at achieving a large MTTDL when the node failure rate varies even more drastically.
Regulated read repair rate simulation
A repair adjustment algorithm could additionally incorporate more advanced aspects. For example, since the bandwidth consumed by the read repair rate is a shared resource, the read repair rate may also be adjusted according to other bandwidth usage, e.g., the read repair rate can automatically adjust to higher values when object access or storage rates are low, and conversely adjust to lower values when object access or storage rates are high. Furthermore, the read repair rate could be regulated to take into account known future events, such as scheduled upgrades or node replacements.
CONCLUSIONS
We introduce a new comprehensive approach to distributed storage, liquid systems, which enable flexible and essentially optimal combinations of storage reliability, storage overhead, repair bandwidth usage, and access performance. The key ingredients of a liquid system are a low complexity large code, a flow storage organization and a lazy repair strategy. The repair regulator design we present provides further robustness to liquid systems against varying and/or unexpected node failure. Our repair and access simulations establish that a liquid system significantly exceeds performance of small code systems in all dimensions, and allows superior operating and performance trade-offs based on specific storage deployment requirements. We address the practical aspects to take into consideration while implementing a liquid system and provide an example architecture. A liquid system eliminates network and compute hot-spots and the need for urgent repair of failed infrastructure.
While not detailed in the paper, we believe a liquid system provides superior geo-distribution flexibility and is applicable to all kinds of Object, File, Scale-out, Hyper-converged, and HDD/SSD architectures/use-cases. There are also optimizations that the team has worked on that involve pregeneration of EFIs and result in better storage efficiencies at the storage nodes.
properties presented herein. This prototype software is used to cross validate the simulation results presented in Section 5.
Christian Foisy helped to produce the RaptorQ performance results presented in Section 8.5, and also helped cross verify the prototype software for the liquid system. Menucher Menuchehry and Valerie Wentworth provided high level comments that greatly improved the presentation.
We would like to thank Rick, Jim, Todd, Christian, Menucher, and Valerie for their valuable contributions.
A. MTTDL ANALYSIS OF LIQUID SYSTEM
In this section we analyze regulated repair of the liquid system and derive some bounds on MTTDL. In the case of constant repair rate the bound is particularly tight. Although we typically assume that the node failure process is a time invariant Poisson process the analysis can be applied to cases where the rate is not known and time varying. We assume throughout that the number of nodes M is the same as the code length n, and we use n to denote this number.
The framework for the analysis ignores the possibility that an object may be unrepairable. In effect we suppose that as each object leaves the repair queue (because it reaches the head of the queue and, normally, is repaired) a new object enters at the tail. This notion is an analytical convenience that takes advantage of the relative ease of determining or estimating the asymptotic steady state distributions of this system. We assume a large number of objects O, and for some of the analysis we ignore effects of a finite number of objects and consider the infinite object limit. This has no significant effect on the results but simplifies the analysis considerably.
For the analysis we generally assume that the repair rate is updated upon object repair. In the limit of infinitely many objects this amounts to continuous update. With a finite number of objects the queue state, as sampled at object repair times, the system forms a Markov chain with a unique invariant distribution. In the case where the node failure rate is known it is a finite state chain. Of particular interest is the marginal distribution of the number of erased fragments for the object at the head of the queue. In the case of fixed repair rate we can express this distribution in closed form, and for the regulated case we will derive a bound on the cumulative distribution.
A.1. MTTDL Estimates
We define the MTTDL as the expected time given a perfect initial queue state (all objects have zero erased fragments) until an object achieves r + 1 missing fragments prior to completing repair. For the system with node failure rate estimation, we also need to specify the initialization of the estimator. The system is typically designed to achieve a very large MTTDL. We will show in the constant repair rate case that we can obtain a sharp closed-form lower bound on the MTTDL. The bound can be applied even in the case of a mismatch between an assumed node failure rate and an actual node failure rate. In the regulated repair case the analysis is more difficult, but we obtain an efficiently computable generalization of the lower bound for the fixed node failure arrival rate case. We further extend the analysis to cases where the node failure rate is estimated based on the node failure realization.
Between node failure events the system evolves in a purely deterministic fashion; objects simply advance in the repair queue according to the repair rate policy. Under a time invariant Poisson model the node failure interarrival times are independent exponentially distributed random variables and the selection of the failing node is uniformly random over all of the nodes. Thus, the repair can be viewed as a random process consisting of a sequence of samples of the queue state taken upon node failure. The process is generally Markov where the state space consists of a pair of queue states, that just prior to the node failure and that after, and, when there is failure rate estimation, a state representation of the failure rate estimator. The transition to the next state has two components. First, there is the time until the next node failure, which is an exponential random variable. Given that time the queue state advances deterministically from the state after the current node failure to the state just prior to the next. The second component is the selection of the failing node. That choice determines which objects in the queue experience additional fragment loss and which do not. The update of the state of the failure rate estimators can also occur at this point. Because of the nested nature of the process there will be a point x in the queue such that objects in positions less than x experience fragment loss and those more than x do not, because they were already missing the fragment associated to the failing node. Note that the queue remains monotonic: the set of erased fragments for an object in position x is a subset of the erased fragments for an object in position y whenever x ≤ y. In the case of node failure rate estimation, those estimates will be updated upon node failure.
A.1.1. The Constant Repair Rate Case. When the repair rate is held constant the deterministic update portion of the process is particularly simple. Let q denote the distribution of the number of erased fragments for the object at the head of the repair queue immediately prior to node failure. Because node failure is a time invariant Poisson process the distribution q is also the distribution of the number of erased fragments for the object at the head of the repair queue at an arbitrary time. In other words, q is the the distribution of the number of erased fragments upon repair under continuous (steady state) operation. We therefore have q(s) = B(n, s, e −λT ).
Let F (t) denote the number of erased fragments for the object at the head of the repair queue at time t. If t is a node failure time let F (t−) and F (t+) denote value of F (t) immediately prior to and after the node failure. The probability of simultaneous node failures is 0 so we assume F (t−) ≤ F (t+) ≤ F (t−) + 1. Let t be a node failure time, then we have Note that this event F (t+) = r + 1 ∧ F (t−) = r corresponds to a data loss event and that the first time this occurs, starting from the perfect queue state, is the first data loss event. We will refer to this event as a data loss transition. Thus, the MTTDL is by definition the expected time to the first data loss transition starting from a perfect queue state. Applying Kac's theorem to the Markov chain we have ( n−r n · q(r)) −1 is the expected number of node failure events between data loss transitions under steady state operation. The expected time between such events is given by (λ·(n−r)·q(r)) −1 since λ·n is the rate of node failure. To obtain a rigorous bound on MTTDL some further analysis is needed because of differing assumptions on the initial condition. The rigorous correction is negligible in regimes of interest so we use the estimate 1 λ · (n − r) · q(r) MTTDL as a basis for choosing repair rates (T ) as a function of MTTDL. Starting at time 0 let T S denote the expected value of the first data loss transition time where the initial queue state is distributed according to the the steady state distribution conditioned on F (0) ≤ r. Let us also introduce the notation q(> r) = s>r q(s).
LEMMA A.1. Assume a constant repair rate with system repair time T. Then we have the following bound PROOF. We prove the Lemma by establishing the following two inequalities Consider a fixed node failure process realization. Among all initial queue conditions with F (0) ≤ r the perfect queue will have a maximal time to data loss. Thus, T S ≤ MTTDL.
Assume the system is initialized at time 0 with an arbitrary initial queue state and let us consider the expected time until the next data loss transition. Note that k ·T is time when the object initially at the tail of the queue is repaired for the k-th time. Let k * denote the smallest k > 0 where upon repair that object has at most r erased fragments. Then the queue state at time t = k * · T is the steady state distribution conditioned on the object at the head of the queue having at most r erased fragments, i.e., conditioned on F (t) ≤ r. Hence, the expected time to the next data loss transition can be upper bounded by E(k * · T ) + T S . Now, the probability that k * = k is (1 − q(> r)) · q(> r) k−1 hence E(k * ·T ) ≤ k>1 (1−q(> r))·q(> r) k−1 ·k·T = T 1−q(>r) . It follows that MTTDL ≤ T 1−q(>r) +T S and that the expected time between data loss transitions, which is (λ · (n − r) · q(r)) −1 , is also upper bounded by T 1−q(>r) + T S .
A.2. Regulated Repair Rate
For the regulated repair rate case the memory introduced into the system by the selection of the maximum requested repair rate makes it quite difficult to analyze. In particular the rate of advancement of the repair queue depends on the entire state of the queue, so there is no convenient way to reduce the size of the Markov chain state space, as was effectively done in the fixed repair rate case. The addition of node failure rate estimation further complicates the problem. It turns out, however, that by approximating queue advancement with a greedy form of operation we can obtain a related lower bound on the MTTDL. When applied to the constant repair rate case the bound is smaller than the one above by a factor of (1 − f tar ).
Although we shall consider the case of known node failure arrival rate, much of the analysis is common with the case including node failure estimation. Thus, we will first develop some general results that admit node failure rate estimation.
We introduce the notation F Dis (s, t) to denote the number of distinct node failures in the interval (s, t). As before, if s is a node failure time we use s− and s+ to denote times infinitisemally before or after the node failure respectively. For convenience we assume that φ(f, x) is bounded below by a positive constant, is defined for all f and x and is non-increasing in f and non-decreasing in x. For computation we may also assume that φ(f, x) is bounded above by the nominal value λ · T so that, in the known node failure rate case, T is an upper bound on the time between repairs for a given object. Recall that since λ · T = φ(0, 0) we always have inf x∈[0,1] φ(f (x), x) ≤ λ · T so this restriction does not affect the operation of the system. In the case with node failure rate estimation we will typically assume some minimum postive value for the estimated node failure rate so that there will be some finite time T during which a complete system repair is guaranteed.
When we admit failure rate estimation we suppose that each object has it's own failure rate estimate. That estimateλ figures in the requested repair rate through the implied requested system repair time ofλ −1 · φ(f, x). For an object that entered the queue at time s we write its node failure rate estimate at time t asλ(s, t). The estimate is allowed to depend on past node failures. It can depend on s only through the partition of node failures into 'past' and 'future' implied by time s. Consequently, all objects that enter the queue between the same pair of successive node failures use the same value forλ. In other words, as a function of s, the estimateλ(s, t) is piecewise constant in s with discontinuity points only at node failure times. We remark that this can be relaxed to havinĝ λ(s, t) non-decreasing in s between node failure points. In practice the node failure rate estimate used will generally be a function of relatively recent node failure history which can be locally stored. Note that the estimator for an object may depend on its number of erased fragments.
We call an object in the repair queue a critical object if it was at the tail of the queue at a node failure point. Letting s(x, t) denote the queue entry time of the object in position x of the queue at time t, an object is critical if s(x, t) is a node failure time. To be more precise, a critical object is an object that was the last object completely repaired prior to a node failure. Such an object has an erased fragment while it is still at the tail of the queue.
Let χ(t) denote the set of critical object queue locations x ∈ [0, 1] at time t. Let f (x, t) = F Dis (s(x, t), t)/n and letλ x (t) denoteλ(s(x, t), t). Under the current assumptions the repair regulator need only track the critical objects in the system. LEMMA A.2. The repair rate is always determined by a critical object, i.e., PROOF. Of all objects that entered the queue between the same pair of successive node failures, the critical object has the minimal value of x and all such objects have the same value ofλ and f.
Since φ(f, x) is non-decreasing in x the result follows.
To obtain a bound on the MTTDL we will adopt a greedy approximation in which we effectively assume that objects control the repair rate during their time in the repair queue. To this end, for any time s set y(s, s) = 0 and for t ≥ s let us define y(s, t) as the solution to d dt y(s, t) =λ(s, t) · (φ(F Dis (s, t), y(s, t))) −1 which is well defined as long as y(s, t) ≤ 1. Further, let τ (s) be the minimal t such that y(s, t) = 1.
An interpretation of y is that it is the position in the queue of an object that was at the tail of the queue at time s and is designated to be the object that determines the system repair rate as long as it remains in the queue. Let x(s, t) denote the actual position of the object at time t given that it is at the tail of the queue at time s. Thus, x(s, t) − y(s, t) < 0 implies d dt (x(s, t) − y(s, t)) ≥ 0 and since x(s, s) = y(s, s) = 0 we see that x(t, s) − y(t, s) < 0 cannot occur by the mean value theorem.
For data loss to occur there must be node failure times s ≤ t such that both F Dis (s−, t+) ≥ r + 1 and x(s−, t) < 1. 3 In this event data loss has occured to the critical object that was at the tail of the queue immediately prior to time s. By Lemma A.3 this implies that F Dis (s−, τ (s)) ≥ r + 1. The condition F Dis (s−, τ (s)) ≥ r + 1 denotes a greedy data loss where the object arriving at the tail of the queue at time s greedily controls the repair rate and suffers data loss. We refer to the node failure at time s as a greedy data loss node failure. Note that the greedy data loss actually occurs after time s.
A.2.1. The Regulated Repair Case with Fixed Failure Rate. We now assume that the estimated repair rateλ is constant and may or may not be equal to the actual node failure rate λ. In this case there is no ambiguity in the intialization ofλ and the MTTDL can be lower bounded by the expected value of s where s is the time of the first greedy data loss node failure.
Assume s is a node failure time and let q now denote the distribution of F Dis (s−, τ (s−)). By Kac's theorem, 1/q(> r) is the expected number of node failures between greedy data loss node failures. Hence 1/(λ · n · q(> r)) is the expected time between greedy data loss node failures.
Let T G denote the expected time time until the first greedy data loss node failure assuming an unconditioned future node failure process. Note that the time to a greedy data loss is independent of the initial queue state. It follows that 1/(λ · n · q(> r)) is upper bounded by T + T G . This is because if we condition on 0 being a greedy data loss node failure time, then the node failure process after time T is independent of this condition. Recalling that T G ≤ MTTDL we obtain 1 λ · n · q(> r) − T ≤ MTTDL In regimes of interest we will have q(> r) q(r + 1) and assuming a fixed repair rate we have n · q(r + 1) = n · B(n − 1, r, e −λ·T ) = e λ·T · (n − r) · B(n, r, e −λ·T ) . Ifλ = λ then we see that the resulting MTTDL bound is approximately a factor e −λ·T = 1 − f tar worse than the previous one.
Given φ it is not difficult to quantize the system to compute the distribution q. To perform the computation we consider quantizing x with spacing δx. 4 This is essentially equivalent to assuming a finite number, O = δ −1 , of objects and we describe the quantization from that perspective. Assume that repair rates are updated upon object repair. Under regulated repair the repair system then selects the next object repair time, which is now given by the minimum of δ · φ(f (jδ), jδ) for j = 0, . . . , O − 1. Assume that the next repair time ∆ is given and that we are given that the object in position k · δ has a random number of erased fragments F that is distributed according to p k (F ). Then the distribution of the number of erased fragments for the object when it reaches position (k + 1) · δ is given by To compute the distribution q we proceed similarly but, in effect, track the greedy controlling object as it moves through the repair queue, assuming that that object is the one determining the repair rate. Letting q k (F ) denote the distribution of the number of missing fragments when the object is (greedily) in position kδ. We have q k+1 (F ) = 0≤F ≤F q k (F ) · B(n − F , F − F , e −δ·φ(F /n,kδ) ) .
Note that the distribution q is simply q O (·) assuming we initialize with q 0 (F ) = 1 F −1 (the distribution with unit mass at F = 1.) Using the above expression we see that the distribution q can be computed with complexity that scales as n · O.
Note that if the above computation we instead use the initial condition 1 F then the resulting distribution q provides an upper bound on the steady state distribution of the number of erased fragments upon repair. More precisely, the resulting q(> s) upper bounds the probability that more than s fragments of an object are erased when that object is repaired. Using this value of q we obtain n s=0 q(> s) as an upper bound on the expected number of repaired fragments, which represents the repair efficiency of the system. In Fig. 19 we plot the computed cumulative distribution of the number of missing fragments upon repair and the computed upper bound. The five shown bounds are different only in the limit placed on the maximum repair rate which is indicated as a factor above the nominal rate. The "1x" case corresponds to the fixed repair rate policy. Already by allowing a "2x" limit a dramatic improvement is observed. It is observed that the greedy bound tracks the true distribution, as found through simulation for the "3x" limited case, with some shift in the number of missing fragments. Although somewhat conservative, it is evident that the computed bound can provide an effective design tool and that rate regulation can provide dramatic improvements in MTTDL. Fig. 20 shows how the cumulative distribution of the number of missing fragments changes with node failure rate mismatch. We set the repair rate limit to 10 times the nominal rate in increased the node failure rate by factors of 1.25, 1.5, and 2.0 respectively. The regulator responds by increasing the average repair rate and reaching a new asymptotic queue state equilibrium with respectively larger repair targets. This is indicated in the shifts of the cumulative distribution. For very large shifts in failure rate this level of adaptation would be inadequate to ensure data durability. Fortunately node failure rate estimation is easily incorporated into the regulator design, leading to a significantly more robust system. A.2.2. The Regulated Repair Case with Failure Rate Estimation. The rate regulator as described above is designed to concentrate the repair queue near its nominal behavior when the actual node failure rate matches the design's assumed failure rate. As demonstrated, the regulator can compensate for some variation in the failure rate. If the actual node failure rate is persistently different from the design assumed rate then the regulator typical behavior is shifted relative to the nominal one. By using an estimator of the node failure rate the system can bring the typical behavior back toward the nominal one. Moreover, by using an estimate of the node failure rate the system can adapt to much larger variations in the actual node failure rate. Many forms of estimators may be considered. We shall consider some relatively simple forms that are amenable to computation analysis similar to the fixed arrival rate case. These estimates are based on first order filtering of some function of the node failure interarrival times. A typical update of the node failure interarrival time will take the form ξ(T ) ← α · ξ(T ) +ᾱ · ξ(T ) .
where T is a most recent node failure interarrival time and the functional ξ is an invertible function. A convenient case to consider is ξ(T ) = T but other potentially interesting forms are ξ(T ) = log T or ξ(T ) = 1/T, or regularized versions of these functions such as ξ(T ) = log(1+T ). For simplicity we will focus on the case ξ(T ) = T. The node failure rate estimateλ is simply given asT −1 .
Corresponding to our assumptions on failure rate estimators, we assume that the value of α used by an object is a function only of the number of erased fragments for that object and the total number of node failures that have occurred since that object entered the repair queue, for which we introduce the notation F All (s, t). Let us now consider how to compute the distribution of the greedy behavior. In this case we take a slightly different approach then previously. We aim to recursively compute the joint greedy queue position and failure rate estimate distribution at node failure times. Since a node failure may or may not result in a fragment erasure for the greedy object, we must also jointly consider the distribution of the number of erased fragments.
We will now briefly discuss how, give F + , we can computationally update a joint distribution of F, y,λ to the next node failure point. Assume that at time t 0 the greedy object has an estimatê λ(s, t 0 ) and greedy queue position y = y(s, t 0 ) and a certain value of F. Then the joint distribution of y andλ(s, t) at the next node failure point lies on a curve parameterized by t 1 , the next node failure time. Note that since we assume that no additional node failures occur, we have that λ(s, t), F (t), and F + (t) are all constant and we can drop the dependence on s and t. Note also that α is detemined and is independent of the next node failure time. For t ≥ t 0 such that y(s, t) < 1 we have y(s, t) = y(s, t 0 ) +λ t t0 φ −1 (F/n, u)du .
Assuming t = t 1 is the next node failure the update ofλ is given bŷ T (s, t) = α ·T (s, t 0 ) +ᾱ(t − t 0 ) and we obtain the curve y(s, t 1 ),T (s, t 1 ) for y(s, t 1 ) < 1. Note, however, that at y(s, t 1 ) = 1 we actually have the discontinuityT (s, t 1 ) =T (s, t 0 ) since in this case the object reached the head of the queue prior to additional node failures. To obtain the probability density on this curve observe that until the time of the next node failure t 1 we have t − t 0 =λ −1 y(s,t) y(s,t0) φ(F/n, u)du .
The distribution of t 1 − t 0 is exponential with rate λ · n so the probability that t − t 0 is greater than a quantity t > 0 is given by e −λ·n·t . Since y(s, t) is monotonic in t, the probability that y(s, t 1 ) is greater than z where y(s, t 0 ) ≤ z < 1 is given by Pr [y(s, t 1 ) > z] = e −λ·n·λ −1 z y(s,t 0 ) φ(F/n,u)du and this gives the distribution along the curve y(s, t 1 ),T (s, t 1 ). Let us introduce the notation Q F + (F, y,λ) to denote the conditional density of the extended definitions of F, y,λ conditioned on F + . Using the above described calculations we can update this distribution to the distribution at the next node failure point. We will use the notationQ F + (y,λ) to denote this updated distribution. Assuming y < 1, the next node failure is incrementing with probability (1 − F/n) thus for y < 1 we have Q F + (F, y,λ) = (F/n)Q F + −1 (F, y,λ) + (1 − (F − 1)/n)Q F + −1 (F − 1, y,λ) . and for y = 1 we have Q F + (F, y,λ) =Q F + −1 (F, y,λ) .
With suitable quantization of y andλ the above procedure can be used to compute the greedy distribution.
To obtain the greedy bound for an arbitrary initial time we initialize the distrbution Q 0 so that the marginal distribution of y is 1 y (and, necessarily, F = 0 with probability 1.) Given a Poisson node failure model and a particular estimator the distributionλ(s, s) may be determined. In general the distribution ofλ can be set according to the desired assumptions of the computation. The limit as F + tends to +∞ of Q F + (F, 1,λ) is the greedy approximation to the distribution upon repair of F andλ. The probability of large F + with y < 1 decays exponentially in F + , so this computation converges quickly in practice.
To obtain a bound on the MTTDL we need to perform the corresponding computation for critical objects. This requires only one additional initialization step in which the distribution Q 1 (F, y,λ) is modified to project out y and put all probability mass on y = 0. Note that when F + = 1 we have F = 1 with probability 1. With such a modification of the initialization process we can obtain a lower bound on the MTTDL as before. We remark, however, that in this case we take the definition of MTTDL to include a random initialization ofλ according to the distribution ofλ(s, s) and the above computation should also use this initial condition. Fig. 21 shows the comulative distribution of the number of missing fragments upon repair q as obtained from simulation and the computed bound. The simulation objects used for an estimate of node failure rate the average of the node failure interrarrival times for the last 155 − F node failures, where F denotes the number of erased fragments in a object. Note that the value of 155 was chosen slightly larger than r +1 = 135. The computed bound used a slightly different estimator consistent with the above analysis. The estimator used is a first order filter but, to best model the corresponding simulation, the time constant of the filter was adjusted as a function of F to match 155 − F. In both cases the peak repair rate was limited to 3 times the nominal rate as determined by the actual node failure rate. It can be observed from these the plots that even compared with know node failure arrival rate, the estimated case can give larger MTTDL. This is intuitive since data loss is generally associated to increases in node failure rate and the increase in the estimated node failure rate case gives rise to faster repair. In addition, as shown previously, a regulator with rate estimation can accomodate large swings in node failure rates. | 2017-05-22T20:21:09.000Z | 2017-05-22T00:00:00.000 | {
"year": 2017,
"sha1": "478b7374c7be7552644118b7b9282dd16521f684",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "38fdead9b98b8b9a5bcfaa147b9fcf2df3c016dd",
"s2fieldsofstudy": [
"Computer Science",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
1057579 | pes2o/s2orc | v3-fos-license | Prior-Cancer Diagnosis in Men with Nonmetastatic Prostate Cancer and the Risk of Prostate-Cancer-Specific and All-Cause Mortality
Purpose. We evaluated the impact a prior cancer diagnosis had on the risk of prostate-cancer-specific mortality (PCSM) and all-cause mortality (ACM) in men with PC. Methods. Using the SEER data registry, 166,104 men (median age: 66) diagnosed with PC between 2004 and 2007 comprised the study cohort. Competing risks and Cox regression were used to evaluate whether a prior cancer diagnosis impacted the risk of PCSM and ACM adjusting for known prognostic factors PSA level, age at and year of diagnosis, race, and whether PC treatment was curative, noncurative, or active surveillance (AS)/watchful waiting (WW). Results. At a median followup of 2.75 years, 12,453 men died: 3,809 (30.6%) from PC. Men with a prior cancer were followed longer, had GS 8 to 10 PC more often, and underwent WW/AS more frequently (P < 0.001). Despite these differences that should increase the risk of PCSM, the adjusted risk of PCSM was significantly decreased (AHR: 0.66 (95% CI: (0.45, 0.97); P = 0.033), while the risk of ACM was increased (AHR: 2.92 (95% CI: 2.64, 3.23); P < 0.001) in men with a prior cancer suggesting that competing risks accounted for the reduction in the risk of PCSM. Conclusion. An assessment of the impact that a prior cancer has on life expectancy is needed at the time of PC diagnosis to determine whether curative treatment for unfavorable-risk PC versus AS is appropriate.
Introduction
While favorable-risk (PSA ≤ 20; T2b category or less; Gleason score ≤ 7 [1]) prostate cancer (PC) can have a long natural history [2] and is often curable, unfavorable-risk PC (which comprises approximately 20% of cases) accounts for the majority of prostate cancer deaths [3]. Men of PC bearing age are also at risk for a metachronous cancer (i.e., history of or subsequent diagnosis of another cancer). When considering life expectancy in men with PC, competing risks are particularly relevant in men with favorable-risk disease [4][5][6][7][8][9][10], in order to avoid overtreatment of PC where the potential toxicities of treatment can be sustained with no prolongation in survival.
To our knowledge, no study has investigated the impact that the comorbidity of a prior cancer has on the risk of PCSM.
Therefore, we used a SEER population database registry to evaluate the impact that a prior cancer had on the risk of PCSM and all-cause mortality (ACM) in men with newly diagnosed, node negative, nonmetastatic PC, adjusting for age at and year of diagnosis, race, initial treatment (curative or noncurative) or active surveillance (AS) or watchful waiting (WW), and known PC prognostic factors.
Patients Selection and SEER Data Registry.
We used a population-based registry, SEER [11], in order to identify 166 preceding their PC diagnosis versus those without a prior cancer. The Wilcoxon rank-sum test [13] was used to compare the distribution of the continuous clinical characteristics and a Mantel-Haenszel chi square test [14] was used to compare the distribution of categorical covariates.
Risk of PCSM and ACM.
We used a Fine and Gray [12] competing risks and Cox regression [15] to assess whether there was an association between risk of PCSM and ACM, respectively, in men diagnosed with a prior cancer versus no prior cancer adjusting for known PC prognostic factors, curative, noncurative treatment or AS/WW, race, and age at and year of diagnosis of PC. Time zero was defined as the date of the PC diagnosis. Adjusted ( ) and unadjusted hazard ratios (HR) were calculated and are reported along with their 95% confidence intervals (CI) and associated values. A two-sided value <0.05 was considered statistically significant. was used for all calculations related to Fine and Gray and SAS version 9.3 (SAS Institute, Cary, North Carolina) was used for all remaining statistical analyses.
Estimates of PCSM and ACM.
The cumulative incidence method [16] was used to calculate estimates of PCSM stratified by whether the patient had a history of cancer or not at the time of his PC diagnosis. Age adjusted comparisons of the estimates were performed using Fine and Gray's regression. ACM was defined as one minus overall survival. Estimates of ACM were calculated using the method of Kaplan and Meier [17] and age adjusted estimates [18] stratified by whether the patient had a history of cancer or not at the time of his PC diagnosis were compared using an age adjusted log rank value. also less likely to be African American (7.4% versus 12.3%, < 0.001) and they were more likely to have clinical tumor stage T1c (42.0 versus 37.5%, < 0.001) and were more likely to be diagnosed earlier in time ( = 0.035). Table 2, for men with a history of cancer at the time of the PC diagnosis versus not having the risk of PCSM was significantly decreased (AHR: 0.66 (95% CI: (0.45, 0.97); = 0.033), whereas the risk of ACM was significantly increased (AHR: 2.92 (95% CI: 2.64, 3.23); < 0.001) as shown in Table 3. Given the differences shown in Table 1 that should have led to an increase in the risk of PCSM amongst men with versus without a prior cancer, these results suggest competing risks and non curative PC treatment accounted for the observed reduction in the risk of PCSM.
Estimates of PCSM and ACM Stratified by Prior History of
Cancer. As shown in Figure 1(a), the age-adjusted cumulative incidence of PCSM was significantly lower for men with a prior cancer as compared to those without a history of cancer prior to the PC diagnosis ( = 0.012) However, age adjusted estimates of all-cause mortality were significantly higher ( < 0.001) for men with a prior cancer as compared to no prior cancer, as shown in Figure 1(b).
Discussion
Overtreatment remains an issue in the United States for men with low-risk PC; a target population for whom greater use of AS may be more appropriate particularly in men with significant comorbidity [19]. Our results indicate that men who had a history of cancer prior to the diagnosis of PC were followed longer, were older, and were more likely to have GS 8 to 10 PC and undergo WW or AS compared with those who did not have a history of cancer. However, despite the less frequent use of curative treatment in these men who had more aggressive PC that should have led to an increased risk of PCSM, these men had a significant decrease in the risk of PCSM while their risk of ACM increased significantly, suggesting that competing risks (prior cancer and other comorbidities) and not curative PC treatment accounted for the reduction of PCSM. The clinical significance of these findings is that AS should be more judiciously employed in men with competing risks. Specifically, while men with low risk PC are offered AS if their life expectancy is less than 10 years as per the 2013 NCCN guidelines [20] AS would not be the preferred strategy by the existing NCCN guidelines, those who have had a prior cancer diagnosis should be considered for AS if the expected rate of cure of the prior CA is low. Several points require further discussion. First, a recent study has shown that men with low-risk PC are more likely to be offered AS when seen in concurrent multidisciplinary setting rather than sequentially [21]. These results suggest that when a multidisciplinary team of physicians consult on a patient AS is more likely to be recommended. Second, investigators have attempted to define novel assays such as circulating tumor cell burden [22] and gene profiling [23,24] that can stratify patients with castration resistant PC into cohorts with longer or shorter median survivals. Such tools are needed in men with low-risk PC in order to more appropriately select men for AS.
Third, men with a prior history of cancer as shown in Table 1 were more likely to be diagnosed with Gleason score 8 to 10 prostate cancer (19.2% versus 15.1%; < 0.001). This likely reflects the fact that they were also significantly older at prostate cancer diagnosis (median age: 72 versus 66; < 0.001) and advancing age has been shown to be associated with higher grade prostate cancer [25]. Also these men were more likely to be diagnosed at an earlier tumor stage (42% versus 37.5% T1c, < 0.001), which may reflect more active medical monitoring with PSA screening given the history of the prior malignancy.
Potential limitations of this study include the inherent limitations associated with a SEER analysis, including limited information about the biopsy specimen (number and extent of positive cores, tertiary Gleason grade 5, perineural invasion). Also, data on comorbidity other than a prior history of cancer is lacking. Nevertheless, the results of our analysis add to the ongoing dialogue about quantifying the impact competing risks can have on life expectancy when deciding on whether curative treatment is likely to benefit a patient with favorable-risk PC. Conversely, some men who are otherwise healthy with favorable-risk PC may require immediate curative treatment of PC as opposed to AS in order to avoid PCSM. In order to ascertain who these men are randomized controlled trials evaluating curative treatment compared to AS should employ a prerandomization stratification by comorbidity using a validated metric of comorbidity [9] in order to assess whether treatment compared to AS benefits all men or only those with no or minimal comorbidity.
In conclusion, while an attempt is being made to offer men AS, the degree to which this has been occurring in the United States between 2004 and 2007 in men with a prior cancer does not appear to be adequate to avoid overtreatment. Therefore, an assessment of the impact that the prior cancer has on life expectancy is needed at the time of PC diagnosis to determine whether curative treatment for unfavorable-risk PC versus AS is appropriate. | 2017-04-05T18:06:08.777Z | 2014-01-30T00:00:00.000 | {
"year": 2014,
"sha1": "0213432f4d5d83845bc0901ed808a1329e2002ce",
"oa_license": "CCBY",
"oa_url": "https://downloads.hindawi.com/archive/2014/736163.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "33ba8433cdbf09aac810662401006da133a942d2",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
88512553 | pes2o/s2orc | v3-fos-license | Decomposition tables for experiments. II. Two--one randomizations
We investigate structure for pairs of randomizations that do not follow each other in a chain. These are unrandomized-inclusive, independent, coincident or double randomizations. This involves taking several structures that satisfy particular relations and combining them to form the appropriate orthogonal decomposition of the data space for the experiment. We show how to establish the decomposition table giving the sources of variation, their relationships and their degrees of freedom, so that competing designs can be evaluated. This leads to recommendations for when the different types of multiple randomization should be used.
1. Introduction. The purpose of this paper, and its prequel [10], is to establish the orthogonal decomposition of the data space for experiments that involve multiple randomizations [9], so that the properties of proposed designs can be evaluated. In [10], this was done for randomizations that follow each other in a chain, as in Figure 1(a). Here, analogous results to those in [10] are obtained for experiments in which the randomizations are two-to-one, as in Figure 1(b). In such randomizations, two different sets of objects are directly randomized to a third, as in Figures 3, 5 and 6. The unrandomized-inclusive, independent and coincident randomizations from [9] are of this type. Also covered are experiments in which the randomization is two-from-one in that two different sets of objects have a single set of objects randomized to them; that is, experiments with double randomizations [9] [see Figure 1(c)].
As in [10], we always denote the set of observational units by Ω, so that the data space is the set V Ω of all real vectors indexed by Ω. This data space has an orthogonal decomposition into subspaces defined by inherent factors and managerial constraints. We call this decomposition the "structure" on Ω, and identify it with the set P of mutually orthogonal idempotent matrices which project onto those subspaces. Thus if P ∈ P then P is an Ω×Ω matrix, because its rows and columns are labelled by the elements of Ω [10].
In the setting of Figure 1(b), there are two other sets, Υ and Γ, which typically contain treatments of different types to be randomized to Ω. For example, in Figure 3, the set of treatments (Γ) and the set of rootstocks (Υ) are randomized to the set of trees (Ω). Then V Υ is the space of all real vectors indexed by Υ, and V Γ is defined similarly. Each of the sets Υ and Γ also has a structure defined on it, the structures being orthogonal decompositions of V Υ and V Γ , respectively. These are identified with complete sets Q and R of mutually orthogonal idempotent matrices.
There is an immediate technical difficulty. As first defined, a matrix Q in Q is not the same size as a matrix P in P. However, the outcome of the randomization of Υ to Ω is a function f which allocates element f (ω) of Υ to observational unit ω. This function defines a subspace V f Υ of V Ω isomorphic to V Υ . Similarly, the outcome of the randomization of Γ to Ω is a function g which allocates element g(ω) of Γ to observational unit ω. Thus we have a subspace V g Γ of V Ω isomorphic to V Γ . From now on, we identify V f Υ with V Υ , and V g Γ with V Γ . We also assume that equation (4.1) in [10] holds for both f and g, so that we may regard each matrix Q in Q and each matrix R in R as an Ω × Ω matrix without losing orthogonality or idempotence. This condition is satisfied for all equi-replicate allocations, and for many others.
In [10] it was seen that a standard two-tiered experiment has just two sets of objects, Ω and Υ say, typically observational units and treatments. To evaluate the design for such an experiment, one needs the decomposition of the data space V Ω that takes into account both P and Q. Brien and DECOMPOSITION TABLES II. TWO-ONE RANDOMIZATIONS 3 Bailey [10] introduced the notation P ⊲ Q for the set of idempotents for this decomposition, and established expressions for its elements under the assumption that Q is structure balanced in relation to P. They exhibited the decomposition in decomposition tables based on sources corresponding to the elements of P and Q.
For the idempotents for two sources from different tiers, such as P in P and Q in Q, we follow James and Wilkinson [15] in defining Q to have first-order balance in relation to P if there is a scalar λ PQ such that QPQ = λ PQ Q. If this is satisfied and λ PQ = 0, then P ⊲ Q is defined in [10] to be λ −1 PQ PQP, which is the matrix of orthogonal projection onto Im PQ, the part of the source P pertaining to the source Q. The scalar λ PQ is called the efficiency factor ; it lies in [0, 1] and indicates the proportion of the information pertaining to the source Q that is (partially) confounded with the source P. Furthermore, a structure Q is defined in [10] to be structure balanced in relation to another structure P if (i) all idempotents from Q have first-order balance in relation to all idempotents from P; (ii) all pairs of distinct elements of Q remain orthogonal when projected onto an element of P, that is, for all P in P and all pairs of distinct Q 1 and Q 2 in Q, the product Q 1 PQ 2 = 0. If Q is structure balanced in relation to P, and P ∈ P, then the residual subspace for Q in Im P is just the orthogonal complement in Im P of all the spaces Im PQ: its matrix of orthogonal projection P ⊢ Q is given by where ′ Q∈Q means summation over all Q in Q with λ PQ = 0. This notation was extended in [10] to describe the decomposition for three-tiered experiments where the two randomizations follow each other in a chain, as in composed and randomized-inclusive randomizations [9] [see Figure 1(a)]. This involved combining the three structures P, Q and R defined on three sets of objects to yield the two equivalent decompositions (P ⊲ Q) ⊲ R and P ⊲ (Q ⊲ R). It was seen that the idempotents of these decompositions could be any of the following forms: (P ⊲ Q) ⊲ R, P ⊲ (Q ⊲ R), (P ⊲ Q) ⊢ R, P ⊲ (Q ⊢ R), and P ⊢ Q, where P, Q and R are idempotents in P, Q, R, respectively. In some cases, some idempotents in (P ⊲ Q) ⊲ R may reduce to idempotents of the form P, P ⊲ Q, Q, R or Q ⊲ R.
In Sections 2-3 of this paper, corresponding results are obtained for the two-to-one randomizations: unrandomized-inclusive, independent and coincident randomizations. It is shown that, in addition to the decompositions above, the following decompositions occur: P ⊲ R, (P ⊲ R) ⊲ Q and (P ⊲ Q) (P ⊲ R), where " " denotes "the combination of compatible decompositions" in a sense defined in Section 3. Also, the list of forms of idempotents is expanded to include: P ⊲ R, P ⊢ R, (P ⊢ Q) ⊲ R and (P ⊢ Q) ⊢ R.
Section 4 deals with experiments having the only two-from-one randomization: double randomizations.
There are differences between different types of multiple randomization in the reduced forms for the above idempotents and in the efficiency factors. Section 5 gives recommendations for when the different types of multiple randomization should be used. How the results might be applied to experiments with more than three tiers is outlined in Section 6. We finish in Section 7 with a discussion of a number of issues that arise in the decompositions for multitiered experiments.
Unrandomized-inclusive randomizations.
In an experiment with unrandomized-inclusive randomizations, Υ is randomized to Ω in an initial two-tiered experiment. The unrandomized-inclusive randomization involves a third set, Γ, which is randomized to Ω taking account of the result of the first randomization. As for randomized-inclusive randomizations, the order of the two randomizations is fixed.
Two functions are required to encapsulate the results of these randomizations, say f : Ω → Υ and g: Ω → Γ. For ω in Ω, f (ω) is the element of Υ assigned to ω by the first randomization, and g(ω) is the element of Γ assigned to ω by the second randomization. The set-up is represented diagrammatically in Figure 2.
We consider experiments in which the structure Q on Υ is structure balanced in relation to the structure P on Ω, so that the first randomization gives the combined decomposition P ⊲ Q of V Ω described in [10]. The second randomization takes account of P ⊲ Q, both in the choice of systematic design and in restricting the permutations of Ω to preserve P ⊲ Q, so we assume that the structure R on Γ is structure balanced in relation to P ⊲ Q.
Put I Q = Q∈Q Q, which is the matrix of orthogonal projection onto V Υ . The condition for Q to be structure balanced in relation to P can be written as I Q PQ = λ PQ Q for all P in P and all Q in Q. Similarly, put I R = R∈R R, which is the matrix of orthogonal projection onto V Γ . Theorem 2.1. Let P, Q and R be orthogonal decompositions of the spaces V Ω , V Υ and V Γ , respectively, with V Υ ≤ V Ω and V Γ ≤ V Ω . If Q is structure balanced in relation to P with efficiency factors λ PQ , and R is structure balanced in relation to P ⊲ Q with efficiency factors λ P⊲Q,R and λ P⊢Q,R , then: (a) R is structure balanced in relation to P with efficiency matrix Λ PR whose entries are This proves that R is structure balanced in relation to P with the given efficiency matrix.
(b) Since R is structure balanced in relation to P ⊲ Q, we may apply the "⊲" operator to elements of P ⊲ Q and R, to obtain . Moreover, writing * R∈R to mean summation over R ∈ R with λ P⊲Q,R = 0, applying equation (1.1) to P ⊲ Q and R gives Similarly, Thus, using Definition 4 in [10], the decomposition (P ⊲ Q) ⊲ R is as given.
The expression for (P ⊲ Q) ⊲ R in Theorem 2.1(b) differs from that in equation (5.1) of [10] because (P ⊢ Q) ⊲ R is zero for composed and randomized-inclusive randomizations, but may not be zero for unrandomizedinclusive randomizations.
For simplicity, we write the one-dimensional space for the Mean as V 0 , with projector P 0 = Q 0 = R 0 = n −1 J, where n = |Ω| and J is the n × n all-1 matrix.
As Brien and Bailey [9] show, unrandomized-inclusive randomizations are common in superimposed experiments. In such an experiment, it may well be the case that V Γ ∩ V ⊥ 0 is orthogonal to every P ⊲ Q of the decomposition P ⊲ Q. In this case, the decomposition has the simpler form given by Corollary 2.2.
Corollary 2.2. Suppose that Q is structure balanced in relation to P and that R is structure balanced in relation to P ⊲ Q. If (P ⊲ Q)R = 0 for all P in P, all Q in Q and all R in R \ {R 0 }, then Proof. If λ P⊲Q,R = 0 for all Q ∈ Q, then λ PR = λ P⊢Q,R . If this is true for all R, then (P ⊲ Q) ⊢ R = P ⊲ Q. The result follows. Lemma 2.1. Suppose that Q is structure balanced in relation to P, and let P ∈ P \ {P 0 }. The following conditions are equivalent.
, and all combinations of elements of Υ with elements of Γ occur on Ω.
Proof. If λ PQ = 0 then QP = 0 so QPR = 0. If λ PQ = 0 then QPR = λ −1 PQ × I Q PQPI R R = I Q (P ⊲ Q)I R R. Condition (i) implies that all these terms are zero, which implies condition (ii). Summing QPR over all Q and all R gives I Q PI R , so condition (ii) implies condition (iii). Finally, if λ PQ = 0 then (P ⊲ Q)I R = λ −1 PQ PQPI R = λ −1 PQ PQ(I Q PI R ), so condition (iii) implies condition (i).
Summing condition (iii) over all P in P \ {P 0 } gives 0 = I Q (I P − P 0 )I R = I Q I R − I Q P 0 I R = (I Q − Q 0 )(I R − R 0 ), since P 0 = Q 0 = R 0 . This shows that [2] shows that the Universe is the only partition marginal to both Υ and Γ considered as factors on Ω. Then orthogonality and Proposition 3 of [2] show that all combinations of Υ and Γ occur on Ω.
The conditions in Lemma 2.1 are a general form of adjusted orthogonality [14].
Example 1 (Superimposed experiment in a row-column design). The initial experiment in Example 10 in [9] is a randomized complete-block design to investigate cherry rootstocks: there are three blocks of ten trees each, and there are ten types of rootstock. Many years later, a set of virus treatments is superimposed on this, using the extended Youden square in Table 1. This "square" is a 3 × 10 rectangle whose rows correspond to Blocks and columns to Rootstocks. Each of the five treatments occurs twice in each Block (row), while their disposition in Rootstocks (columns) is that of a balanced incomplete-block design. The sets of objects for this experiment are trees, rootstocks and treatments. Figure 3 shows both randomizations.
For this example, using the notation for sources in [10], but writing P Mean as P 0 , the three structures are See the first two columns of Table 2. Blocks The efficiency factors for the structure on treatments in relation to the joint decomposition of trees and rootstocks are derived from the extended Youden square. Viruses are orthogonal to Blocks, which means that shows that the structure on treatments is orthogonal in relation to the structure on trees since because the Viruses source is not orthogonal to Rootstocks. This leads to nonorthogonality between R and P ⊲ Q. In particular, the Viruses source is not orthogonal to Trees[Blocks] ⊲ Rootstocks. Consequently, the decomposition is given by Theorem 2.1(b) rather than Corollary 2.2. The full decomposition of V trees , that contains six elements, one for each line in the decomposition table, is in Table 2: As expected, this decomposition does contain a nontrivial idempotent of the form (P ⊲ Q) ⊲ R. Also, unlike the chain randomizations in [10], it contains an idempotent of the form (P ⊢ Q) ⊲ R.
The efficiency factors are recorded in the decomposition in Table 2, which shows that the Viruses source is partly confounded with both Rootstocks and the part of Trees[Blocks] that is orthogonal to Rootstocks. A consequence of this is that four Rootstocks degrees of freedom cannot be separated from Virus differences. However, there are five Rootstocks degrees of freedom that are orthogonal to Virus differences. Further, while the Viruses source has first-order balance in relation to Rootstocks, the reverse is not true.
3. Independent or coincident randomizations. For independent or coincident randomizations, two sets of objects are randomized to the third; thus we could have Γ and Υ randomized to Ω. Two functions are needed to encapsulate the results of these randomizations, say f : Ω → Γ and g: Ω → Υ. The set-up is represented diagrammatically in Figure 4. A particular feature of these randomizations is that there is no intrinsic ordering of Γ and Υ, because neither randomization takes account of the outcome of the other. Associated with Ω, Υ and Γ are the decompositions P, Q and R. We assume that Q and R are both structure balanced in relation to P. . Independent randomizations in Example 2: rootstocks are randomized to trees in such a way that all trees in each plot have a single type of rootstock; later, fertilizers are randomized to trees in such a way that each fertilizer is applied to one tree per plot; B, P denote Blocks, Plots, respectively.
The difference between coincident and independent randomizations is that, for coincident randomizations, there are sources from the two randomized tiers which are both (partly) confounded with the same source in the unrandomized tier. For independent randomizations this does not occur (apart from the Mean).
3.1. Independent randomizations. For a pair of independent randomizations, the two functions are randomized by two permutations chosen independently from the same group of permutations of Ω. The precise definition of independence, which we were unable to give in [9], is that the conditions in Lemma 2.1 are satisfied, for all P in P \ {P 0 }, for all possible outcomes of the two randomizations. If λ PQ λ PR = 0 then some outcomes will have QPR = 0, violating these conditions. Hence independent randomizations require that λ PQ λ PR = 0 for all Q in Q and all R in R unless P = P 0 . Lemma 2.1 shows that, if Q and R are both structure balanced in relation to P, then they are also structure balanced in relation to P ⊲ R and P ⊲ Q, respectively, with λ P⊲Q,R = λ P⊲R,Q = 0 unless P = P 0 , Q = Q 0 and R = R 0 . Therefore As outlined in [9], Section 8.5, wherever possible we reduce two independent randomizations to a single randomization. However, as noted in [9], Section 4.3, this is not always possible-for example, when it is not physically possible to do them simultaneously.
Example 2 (Superimposed experiment using split plots). Example 6 in [9] is a superimposed experiment in which the second set of treatments (fertilizers) is randomized to subunits (trees) of the original experimental units (plots). The randomizations are independent, being carried out at different times and with the later one taking no account of the earlier one except to force fertilizers to be orthogonal to rootstocks. See Figure 5. Table 3 shows the decomposition.
In this example the independence of the randomizations implies that (P P[B] ⊲ Q R ) ⊲ R F = 0 and so the conditions in Lemma 2.1 are satisfied.
and PR are both nonzero. If Im PQ and Im PR are both proper subspaces of Im P, then the relationship between Q and R depends on the choice of the two independent permutations used in randomizing Υ and Γ to Ω; restricting one of the randomizations to preserve the relationship would make the multiple randomizations unrandomized inclusive rather than coincident. On the other hand, if Im PQ = Im P then Im PR is always contained in Im PQ. If Q is structure balanced in relation to P and Im PQ = Im P, then P ⊲ Q = P and the two sources corresponding to Q and P have the same number of degrees of freedom. The condition for coincident randomizations hinted at in [9], Section 4.2, is precisely that for all P in P, Q in Q and R in R, if PQ and PR are both nonzero then one of P ⊲ Q and P ⊲ R is equal to P.
A special, commonly occurring, case arises when Q and R can be assigned to the two randomized sets of objects such that the following condition is satisfied: for all P in P and Q in Q, if PQ and PI R are both nonzero then Theorem 3.1. If Q and R are both structure balanced in relation to P and condition (3.3) is satisfied then R is structure balanced in relation to Proof. If λ PR = 0 then PR = 0, so (P ⊲ Q)R = 0 for all Q with λ PQ = 0, and hence (P ⊢ Q)R = 0. Suppose that PI R = 0. Then either PI Q = 0 or there is a unique Q in Q with λ PQ = 0, which satisfies P = P ⊲ Q. In the first case, P = P ⊢ Q: therefore Example 3 (A plant experiment). Example 5 in [9] is an experiment to investigate five varieties and two spray regimes. Each bench has one spray regime and two seedlings of each variety. See Figure 6. The sets are positions, seedlings and regimes. The diagram includes the pseudofactor S 1 for Seedlings [Varieties], which indexes the groups of seedlings randomized to the different benches. Although the factor Seedlings is nested in Varieties, S 1 is not, because each of its levels is taken across all levels of Varieties.
The Hasse diagrams displaying the structures for this experiment are in Figure 7. The decomposition is in Table 4, where the source Seedlings[Varieties] ⊢ S 1 is the part of Seedlings[Varieties] which is orthogonal to the source S 1 .
The full decomposition of V positions in this case contains five elements and is This experiment clearly meets condition (3.3), because the only source for positions which is nonorthogonal to sources from both of the randomized tiers is the Benches source, and the five-dimensional pseudosource S 1 is equal Fig. 6. Coincident randomizations in Example 3: seedlings and regimes are both randomized to positions; V denotes Varieties, B denotes Benches; S1 and S2 are pseudofactors for Seedlings. to Benches. That is, P B ⊲ Q S 1 = P B = Q S 1 . The other source nonorthogonal to Benches is the one-dimensional source Regimes, which is a proper subspace of the Benches source, and so (P Consequently, the elements of the full decomposition can be written as follows: Decomposition (3.4) is convenient for algorithms, because it is (P ⊲ Q) ⊲ R, like the decompositions in Theorem 5.1(d) in [10], Theorem 2.1(b), Corollary 2.2 and equation (3.1). However, it gives the false impression that the decomposition of V Ω must have P refined by Q, then P ⊲ Q refined by R, suggesting that Q and R have different roles. Moreover, condition (3.3) does not hold for all pairs of coincident randomizations. We therefore introduce another joint decomposition that emphasizes the symmetry between Q and R. [3] shows that if B and C are compatible then the nonzero products BC, for B in B and C in C, give another orthogonal decomposition of V Ω , which is a refinement of both B and C. Hence if B 1 , . . . , B m are pairwise compatible then there is no need for parentheses in defining B 1 B 2 · · · B m . This decomposition could be referred to as "B 1 combined with B 2 combined with · · · combined with B m ." Lemma 3.1. If PQPRP is symmetric for all P in P, all Q in Q and all R in R, then P ⊲ Q is compatible with P ⊲ R.
Proof. If PQPRP is symmetric then PQPRP = PRPQP. Hence if . Thus if P ⊲ Q is defined then it commutes with P and with every P ⊲ R, so it commutes with P ⊢ R. Similarly, if P ⊲ R is defined then it commutes with P ⊢ Q. Now the same argument shows that P ⊢ Q commutes with P ⊢ R. If P i and P j are different elements of P then P i ⊲ Q and P i ⊢ Q commute with P j ⊲ R and P j ⊢ R, because all products are zero. Hence P ⊲ Q is compatible with P ⊲ R. Proof. The first conditions imply that QPR = 0 or P = Q = R = P 0 . The second implies that QPR = 0 or PQP = λ PQ P or PRP = λ PR P. In each case, PQPRP is symmetric, so Lemma 3.1 completes the proof.
Thus the decomposition (P ⊲ Q) (P ⊲ R), which is symmetric in Q and R, can be used for coincident or independent randomizations, or for unrandomized-inclusive randomizations which satisfy the conditions in Lemma 2.1. It is the same as decomposition (3.4) for coincident randomizations when condition (3.3) holds, the same as the decomposition in Corollary 2.2 for unrandomized-inclusive randomizations when the conditions in Lemma 2.1 hold, and the same as decomposition (3.1) for independent randomizations. Condition (3.2) shows that, for a pair of coincident randomizations, each idempotent in (P ⊲ Q) (P ⊲ R) has one of the following forms: P, P ⊲ Q, P ⊲ R, P ⊢ Q or P ⊢ R.
If a pair of coincident randomizations does not satisfy condition (3.3), then it may be possible to refine R to, say, R 2 in such a way that R 2 is structure balanced in relation to P ⊲ Q, so that the decomposition in Theorem 2.1(b) can be used. It is possible if R = P whenever P ⊲ R = P.
Example 3 (Continued). As already noted, this example satisfies condition (3.3), so P ⊲ Q is compatible with P ⊲ R. Here 4. Double randomizations. Double randomization is the one known type of two-from-one randomizations. In an experiment with double randomization, one set of objects is randomized to two others; thus we could have Γ randomized to Υ and to Ω. We follow the convention that the set of observational units is designated as Ω. Two functions are needed to encapsulate the results of these randomizations, say f : Ω → Γ and g: Υ → Γ. These two functions are randomized independently using two different groups of permutations. The set-up is shown in Figure 8.
Now we obtain a subspace V f Γ of V Ω and a subspace V g Γ of V Υ , both isomorphic to V Γ . If |Υ| = |Γ| then V Υ = V g Γ , so we may effectively identify V Υ , V Γ and V f Γ . If |Υ| > |Γ| then we cannot identify V Υ with a subspace of V Ω without further information explicitly assigning an element of Υ to each observational unit in Ω. This may not be possible (see, e.g., Figure 28 in [9]). Thus we shall assume that |Υ| = |Γ|.
Associated with Ω, Υ and Γ are the decompositions P, Q and R. If R is structure balanced in relation to Q and |Υ| = |Γ|, then Lemma 4.2 in [10] shows that Q ⊲ R = R. Therefore it suffices to have R structure balanced in relation to P. Then the overall decomposition is P ⊲ R = P ⊲ (Q ⊲ R), which must be done from right to left.
Example 4 (An improperly replicated rotational grazing experiment). Example 8 in [9] is the rotational grazing trial shown in Figure 9, with Cows substituted for Animals. The double randomization of Availability results in the assignment of Cows to Paddocks, the Cows assigned to an Availability forming a single herd that is used to graze all Paddocks with the same level of Availability. The sets of objects are observational units, paddocks and treatments, and the numbers of paddocks and treatments are equal, as required. The Hasse diagrams for treatments and observational units are like the middle diagram in Figure 7; that for paddocks is trivial.
The structures on observational units, paddocks and treatments are P = {P 0 , P C , P R , P C#R }, Q = {Q 0 , Q P } and R = {R 0 , R A , R R , R A#R }, respectively. This leads to the decomposition P ⊲ (Q ⊲ R) in Table 5. It shows that there are no residual degrees of freedom for testing any treatment differences-hence the experiment being dubbed improperly replicated.
In this case, Also, PR is equal to either R or 0 for all P ∈ P and all R ∈ R. That is, R is orthogonal in relation to P. Therefore the complete decomposition for the experiment is In [9] this example was redone as a case of randomized-inclusive randomization, using two pseudofactors P A and P R for Paddocks, aliased with Availability and Rotations, respectively. These are required if Q itself is to be structure balanced in relation to P, giving a decomposition from left to right like the one in Section 6 in [10].
Summary.
We have shown in [10] and here that, under structure balance, the six different types of multiple randomization identified in [9] all lead to orthogonal decompositions of V Ω using some of the following idempotents: P, P ⊲ Q, P ⊲ R, (P ⊲ Q) ⊲ R, P ⊲ (Q ⊲ R), P ⊢ Q, P ⊢ R, (P ⊲ Q) ⊢ R, P ⊲ (Q ⊢ R), (P ⊢ Q) ⊲ R and (P ⊢ Q) ⊢ R. The differences between the different multiple randomizations lead to differences in the reduced forms for these elements and in the efficiency factors.
Composed randomizations. If each design is structure balanced then so is the composite; the decompositions P ⊲ (Q ⊲ R) and (P ⊲ Q) ⊲ R are equal, and so the decomposition may be done in either order; and there are no idempotents of the form (P ⊢ Q) ⊲ R or (P ⊢ Q) ⊢ R.
Randomized-inclusive randomizations. The structures Q 1 and R 1 for design 1 are refined to Q and R using the pseudofactors that are necessary for the second randomization, and then the results are the same as for composed randomizations.
Unrandomized-inclusive randomizations. We must have R structure balanced in relation to P ⊲ Q; use the decomposition (P ⊲ Q) ⊲ R, which is done from left to right; if the conditions in Lemma 2.1 hold then there are no idempotents of the form (P ⊲ Q) ⊲ R apart from the Mean, nor any of the form (P ⊲ Q) ⊢ R, the decomposition P ⊲ Q is compatible with P ⊲ R, and (P ⊲ Q) ⊲ R = (P ⊲ Q) (P ⊲ R).
Independent randomizations. The conditions in Lemma 2.1 must hold; if both designs are structure balanced then each remains structure balanced after the other has been taken into account; P ⊲ Q is compatible with P ⊲ R; the decompositions (P ⊲ Q) ⊲ R, (P ⊲ R) ⊲ Q and (P ⊲ Q) (P ⊲ R) are equal; and there are no idempotents of the form (P ⊲ Q) ⊲ R apart from the Mean, nor any of the form (P ⊲ Q) ⊢ R.
Coincident randomizations. Condition (3.2) must hold; P ⊲ Q is compatible with P ⊲ R; use the decomposition (P ⊲ Q) (P ⊲ R), whose idempotents have the form P, P ⊲ Q, P ⊲ R, P ⊢ Q or P ⊢ R; if condition (3.3) holds, this is the same as the decomposition (P ⊲ Q) ⊲ R, which is done from left to right; otherwise, there may be a refinement of R giving a leftto-right decomposition.
Double randomizations. We require that |Υ| = |Γ| and that R be structure balanced in relation to both Q and P, so that the decomposition is P ⊲ R = P ⊲ (Q ⊲ R), which is done from right to left. It appears that they can also be formulated as randomized-inclusive randomizations using pseudofactors to refine Q to Q 2 for which the left-to-right decomposition (P ⊲ Q 2 ) ⊲ R is correct.
6. Structure-balanced experiments with four or more tiers. Each experiment in Sections 2-4 involves only one type of multiple randomization, and so involves three tiers and three structures. However, multitiered experiments are not limited to this configuration. Examples 12-14 in [9] each have four tiers and involve more than one type of multiple randomization. In general, there is the set of observational units, Ω, and each randomization adds another set of objects with its associated tier. Section 7 in [10] shows how to deal with three or more randomizations which follow each other in a chain. Mixtures of other types of multiple randomization should be amenable to successive decompositions of the sort summarized in Section 5, so long as they are handled in the correct order. Thus we can use a recursive procedure in which each new structure refines the decomposition of V Ω obtained using structures accounted for previously. All that is required is that each successive structure should be structure balanced in relation to the previous decomposition.
One class of experiments with both two-one randomizations and chain randomizations consists of multiphase experiments in which different treatment factors are applied in different phases, as the following example demonstrates.
Example 5 (A two-phase corn seed germination experiment). Example 12 of [9] has the four tiers shown in Figure 10. Here we have taken the opportunity to correct the diagram given in [9]. The 36 Lots of grain within each Plot should be completely randomized to Plates ∧ Containers within each Interval. This will not be achieved by permuting Containers within Intervals and Plates within Intervals ∧ Containers, as implied in the rightmost panel of Figure 10. We introduce pseudofactors L 1 and L 2 for Lots, with nine and four levels, respectively, like the pseudofactors for Seedlings in Example 3. The 36 Lots must be randomly allocated to the combinations of levels of L 1 and L 2 , independently within each level of Sites ∧ Blocks ∧ Plots, so that neither pseudofactor corresponds to any inherent source of variation. At each randomization, an orthogonal design is used, so there is no difficulty in constructing the decomposition in Table 6 Bailey [5] suggests an analysis for this example which we reproduce in the first three columns of Table 7(a). In this, the 3-level factors Temperature and Moisture have been combined into a single 9-level Treatment factor, the intertier interactions [9] of Sites, Harvesters and Treatments have been included, and the notation × is used in place of #. We cannot be sure, but it is plausible that he based this decomposition on the crossing and nesting relationships summarized in the formula (T * H * (S/B))/Q, (6.1) where T, H, S, B and Q represent factors for Treatments, Harvesters, Sites, Blocks and Plates, with 9, 3, 3, 2 and 4 levels, respectively. The sources derived from this are in the final column of Table 7(a), with degrees of freedom matching those in the preceding column.
Revision of Table 6 along similar lines, and with pseudosources replaced with actual sources, yields the skeleton analysis-of-variance table in Table 7(b). Note that, given Step 4 in Table 1 of [8], an intertier interaction will Table 7 Skeleton analysis-of-variance tables for Example 5(a) given by Bailey [5] and (b) from Table 6 generally occur in the right-most tier that contains a main effect in the interaction. Table 7(b) differs from Table 7(a) in the following ways.
1. The rationale for the sources in Table 7(a) is unclear. We had to reverseengineer it by producing formula (6.1). On the other hand, the sources in Table 7(b) are based on the relationships between factors within each tier and on the confounding between sources from different tiers.
2. Table 7(a) does not show, as Table 7(b) does, the successive decomposition of the vector space indexed by the observational units. The impression given is that there is a set of sources that arise from the field phase and another set that arises from the laboratory phase. 3. Table 7(a) has four sources called "experimental error" and does not mention plates, containers, intervals, blocks, plots or lots. Hence, there is no indication of the sources of error variation. By contrast, each source called "Residual" in Table 7(b) is unambiguously identified; and the labelling shows that all terms are affected by variation from both phases. For example, the Residual for Plots[B ∧ S], labelled Experimental error (b) in Table 7(a), clearly arises from variability associated with Plots within the Sites-Blocks combinations and variability associated with Intervals. Similarly, it can be seen from Table 7(b) that the Residual in Table 7(a) arises from variability associated with Plates and Lots. 4. As discussed in [9], Section 7.1, the usual default is that there are no intertier interactions because such inclusions would mean that the analysis cannot be justified by the randomization used. It parallels the assumption of unit-treatment additivity in single-randomization experiments. The approach using Table 6 forces the statistician to to consult the researcher about whether intertier interactions should be included, and, if so, to justify them. Tables 7(a) and (b) include the intertier interactions of Sites, Harvesters and Treatments, which suggests that it is anticipated that Harvesters and Treatments will perform differently at different Sites. Table 7(b). To justify an analysis based on Table 7(a), one would need to argue that unit-treatment interaction of Treatments with Blocks within Sites can be anticipated in this experiment.
7.1. Implications of incoherent unrandomized-inclusive randomizations. The phenomenon of incoherent unrandomized-inclusive randomizations is described in [9], Section 5.2.1. Essentially, when there has been a randomization to factors that are crossed, one or more of these factors become nested in the second randomization.
Consider the cherry rootstock experiment in Example 1. The trees tier gives an orthogonal decomposition of V Ω into sources Mean, Blocks and Trees[Blocks] of dimensions 1, 2 and 27, respectively, in the left-hand column of Table 2. Similarly, the rootstocks tier decomposes V Υ into sources Mean and Rootstocks of dimensions 1 and 9. The result of the first randomization is to make the Mean sources equal and to place the Rootstocks source inside Trees[Blocks], thus giving the finer decomposition of V Ω shown in the middle column of Table 2.
The result of the unrandomized-inclusive randomization should be to further decompose the decomposition resulting from the first two tiers. In the extended Youden square, the source Viruses is orthogonal to Blocks but partially confounded with Rootstocks, so the Viruses source defines the decomposition in the right-hand column in Table 2. That is, the source Viruses further decomposes the sources Rootstocks and the Residual for Trees[Blocks], as required.
In [9] we discussed the possibility that the designer of the superimposed experiment ignores the inherent crossing of the factors Blocks and Rootstocks and randomizes Viruses to Blocks in Rootstocks in a balanced incomplete-block design. Then the randomizations are incoherent. The permutation group for the second randomization does not preserve the structure arising from the first two tiers, exhibited by the two left-most columns in Table 2. We can see immediately that this randomization is senseless because it destroys the Blocks subspace preserved by the first randomization. This randomization might have some appeal if no block effects had been detected during the 20 years of the original experiment, but then the analysis of the second experiment would be based on an assumed model rather than on the intratier structures.
Other examples of incoherent unrandomized-inclusive randomizations are more complicated, and perhaps less easily detected. One is the design proposed by several authors for a split-plot experiment in which the subplot treatments are to be assigned using a row-column design. Example 6 illustrates how consideration of the decomposition table for the proposed design facilitates the design process and helps the detection of incoherence.
Example 6 (Split-plots in a row-column design). Example 11 in [9] is based on the design with split-plots in a row-column design given in Cochran and Cox [13], Section 7.33. Diagrams for the two randomizations are given in Figure 11, with leaf treatments named as viruses for clarity, soil treatments designated as different soils for brevity, and Altitude substituted for Layer so that no two factors begin with the same letter. Two diagrams are needed, because the assumed structure on leaves changes between the randomizations, as shown in the two right-hand panels. At first sight, this experiment seems to involve unrandomized-inclusive randomizations, because soils are randomized to leaves in the first randomization, and then viruses are randomized to leaves, taking into account the location of the soils. However, the change in the assumed structure on the leaves between the two randomizations makes them incoherent rather than unrandomized-inclusive. Table 8 shows an attempt to build up a decomposition table for this design. The first two columns follow directly from the randomization in the top half of Figure 11. The third column corresponds to the leaves tier in the bottom half of Figure 11. When we use it to refine the decomposition given by the first two tiers, we find that the Soils source occurs in two tiers. Although this can happen in special circumstances like those in Example 4, this is already a signal that something may be wrong. We also find that If Altitudes # Benches is merged with Altitudes # Plants[Benches] in the decomposition table, then the analysis is orthogonal and is equivalent to that given in [13]. However, this does not allow for consistent Altitude differences across Plants, so it removes six spurious degrees of freedom from what Cochran and Cox call "Error (b)" in [13], page 310. The problem is that the design for the Viruses does not respect the factor relationships established in applying the Soils. As Yates showed in [19], if the randomization respects Benches and Altitudes then a randomization-based model must include their interaction.
What is needed is a design for a two-tiered experiment in which the twelve treatments (combinations of levels of Soils and Viruses) are randomized to leaves 1 in such a way that there is a refinement of the natural decomposition of the treatments space which is structure balanced in relation to Altitudes # Benches. For example, one might choose the systematic design in Table 9 and then randomize benches, altitudes, and plants within benches. In this design the twelve treatments are arranged in a (3 × 3)/4 semi-Latin square constructed from a pair of mutually orthogonal Latin squares of order 3. The Viruses are arranged according to one square for soils s 0 and s 1 , and according to the other square for s 2 and s 3 . Theorem 5.4 in [1] shows that this design is the most efficient with respect to Altitudes # Benches. Let S 1 Table 9 Proposed design for Example 6 (columns denote plants; s0-s3 are different soils; 0-2 denote viruses) Bench III Soils s0 s1 s2 s3 s0 s1 s2 s3 s0 s1 s2 s3 Altitude Top 0 0 0 0 1 1 1 1 2 2 2 2 Middle 2 2 1 1 0 0 2 2 1 1 0 0 Bottom 1 1 2 2 2 2 0 0 0 0 1 1 be a pseudofactor for Soils whose two levels distinguish between the first two and the last two levels of Soils. The design is structure balanced: Viruses and Viruses # S 1 have efficiency factor 1/2 in Altitudes # Benches, while the rest of the interaction Viruses # Soils is orthogonal to Altitudes # Benches. It has the advantage of having 10 degrees of freedom for the Residual for Altitudes # Plants[Benches], two more than for the Cochran and Cox [13] design. Thus construction of the decomposition table when designing the experiment can help to detect problems with a proposed design. In this case, it helped to draw attention to the incoherent randomizations, to highlight the associated problems and to give insight into how they might be redressed.
Other structures.
All the examples in [10] and this paper are poset block structures, being defined by some factors and their nesting relationships, as explained in [3,10]. More generally, a structure may be a Tjur structure that is defined by a family of mutually orthogonal partitions or generalized factors (see [17] or [2]). Again, the generalized factors are summarized in the Hasse diagram that depicts their marginality relations. There is one projector P for each generalized factor F , obtained from the Hasse diagram just as in Section 3 in [10], so that the effect of P on any vector is still achieved by a straightforward sequence of averaging operations and subtractions. It is possible for some of these projectors to be zero. Structures derived from tiers belong to this class.
Another common source of structure is an association scheme [3,7]: for example, the triangular scheme for all unordered pairs from a set of parental types, which is appropriate in a diallel experiment with no self-crosses when the cross (i, j) is regarded as the same as the cross (j, i). Then the matrices P are the minimal idempotents of the association algebra [6], and the corresponding subspaces are its common eigenspaces [3], Chapter 2. The effect of P is a linear combination of the operations of taking sums over associate classes. In the case of the triangular association scheme with n parental types, the subspaces have dimensions 1, n − 1 and n(n − 3)/2; they correspond to the Mean, differences between parental types and differences orthogonal to parental types, respectively. The decomposition R 3 in Example 5 in [10] comes from an association scheme with two associate classes.
The set of treatments in a rectangular lattice design exhibits yet another kind of structure [4]. Although this structure derives neither from partitions nor from an association scheme, the effect of each P is achieved by averaging and subtracting.
The results here and in [10] apply to any structure that is an orthogonal decomposition of the relevant vector space, so long as each structure can be regarded as a decomposition of V Ω . For a Tjur structure Q on a set Υ randomized to Ω, condition (4.1) in [10] must hold in order for Q to be regarded as an orthogonal decomposition of V Ω . For structures not defined by partitions, it seems that we need Q i X ′ XQ j to be zero whenever Q i = Q j , where X is the Ω × Υ design matrix. For an association scheme, this implies that the design must be equireplicate. The analogue of Theorem 5.1(a) in [10] for association schemes is given in [3], Section 7.7.
We admit that there are relevant experimental structures, such as neighbour relations in a field or increasing quantities of dose, that are not adequately described by an orthogonal decomposition of the space. Nonetheless, a theory which covers designed experiments where all the structures are orthogonal decompositions has wide applicability, and we limit ourselves to such structures here and in [10].
Multiphase experiments.
Multiphase experiments are one of the commoner types of multitiered experiment. As outlined in [9], Section 8.1, twophase experiments may involve almost any of the different types of multiple randomizations and, as is evident from Section 5, these differ in their assumptions.
If treatments are introduced only in the first phase, then the randomizations form a chain, as in [10]. In [18], Wood, Williams and Speed consider a class of such two-phase designs for which R is orthogonal in relation to the natural structure Q 1 on the middle tier, and there is a refinement Q 2 of Q 1 such that Q 2 ⊲ R is structure balanced in relation to P. The results there are less general than ours. First, the assumptions for the second phase are in the nature of those for randomized-inclusive randomizations only. Second, the designs are restricted to those for which the design for the first phase is orthogonal.
If treatments are introduced after the first phase, as in Example 5, then some form of two-to-one randomization is needed. Similarly, Brien and Demétrio [11] describe a three-phase experiment involving composed and coincident randomizations. 7.4. Further work. While obtaining mixed model analyses of multitiered experiments has been described in [9], Section 7, and [11], it remains to establish their randomization analysis. The effects of intertier interactions on the analysis need to be investigated. We would like to establish conditions under which closed-form expressions are available for the Residual or Restricted Maximum Likelihood (REML) estimates of the variance components [16] and Estimated Generalized Least Squares (EGLS) estimates of the fixed effects. Also required is a derivation of the extended algorithm described in [12] for obtaining the ANOVA for a multitiered experiment.
Furthermore, we have provided the basis for assessing a particular design for a multitiered experiment, yet general principles for designing them are still needed. | 2010-11-11T13:44:20.000Z | 2010-10-01T00:00:00.000 | {
"year": 2010,
"sha1": "c4f3fc49bc71814115451190f2e4914edc78ee41",
"oa_license": "implied-oa",
"oa_url": "https://doi.org/10.1214/09-aos785",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "c4f3fc49bc71814115451190f2e4914edc78ee41",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
219506777 | pes2o/s2orc | v3-fos-license | The Multisensory Experience of Handling and Reading Books.
The failure of e-books to take over from the traditional print format, as was so confidently predicted would happen only a few years ago, highlights how there is more to reading than merely the content of what we see. In fact, like any other object, the experience of interacting with a book, especially an old or historic volume, offers the reader the potential for a multisensory encounter. One that involves not only what the book looks and feels like, both the weight of the volume and the feel of the pages, but also the distinctive smell. In fact, one might also want to consider the particular sound made by the pages as they are turned over. However, it is the smell of older, and seemingly more olfactorily-redolent, works that appears to be especially effective at triggering nostalgic associations amongst readers. It is therefore only by understanding the multisensory nature of handling books, as stressed by this review, that one can really hope to fully appreciate the enduring appeal of the traditional format in the modern digital era. Several recent exhibitions that have attempted to engage their visitors by means of exploring the multisensory appeal of historic books or manuscripts in their collections are briefly discussed. While the multisensory mental imagery that is typically evoked by reading is unlikely to differ much between the print and e-book formats, there is nevertheless still some evidence to suggest that physical books can occasionally convey information more effectively than their digital counterparts.
Introduction
Given the 'sensory turn' that has been documented in so many fields of academic enquiry in recent years (e.g., Smith, 2007Smith, , 2015, it is natural to consider the multisensory attributes, not to mention appeal, of books in the same light too. Intriguingly, and contrary to so many predictions in the popular press, physical books still outsell e-books no matter whether they are read on Kindle or iPad (Preston, 2017). In 2017, for example, sales of e-books were predicted to fall by 1% whereas sales of paper books were set to rise by 6% (Bridge, 2017). According to Handley (2019): "Publishers of books in all formats made almost $26 billion in revenue last year in the U.S., with print making up $22.6 billion and e-books taking $2.04 billion, according to the Association of American Publishers' annual report 2019. Those figures include trade and educational books, as well as fiction." (see Note 1). According to Nielsen Book International: "genres that do well in print include nature, cookery and children's books, while people prefer to read crime, romantic novels and thrillers via e-reader" (quote from Handley, 2019).
Interestingly, though, despite the rapid growth in portable technologies, we have also witnessed a return, especially amongst younger people, to reading on paper rather than e-books (Cain, 2017). Even more pertinent is the fact that this preference would appear to be more pronounced when people read for pleasure rather than for their business or education (see Stoltzfus, 2016). For instance, the results of a 2017 survey commissioned by eBay revealed that just over half of the more than 2000 people questioned actually preferred physical over digital versions of books, CDs and records, and boxed video games (Bridge, 2017; see also Sax, 2019).
The question here has to be why? Why exactly has the sale of e-books plateaued, and why may it even be in decline? In part, the answer may simply be that electronic books do not have anything like the same multisensory (nor nostalgic emotional) appeal as traditional books (Ruecker, 2002(Ruecker, , 2006. At the same time, however, it is also worth highlighting the fact that 62% of the respondents in the eBay survey explicitly reported that they used physical books to help them to disconnect from the online world (Bridge, 2017). Indeed, one of the recently recognised problems with e-readers (Kindle) and tablets (iPad) is that that they can all too easily interfere with our sleep in a way that traditional books simply do not do. According to the results of a 2015 study by Chang et al., those reading an electronic book on a light-emitting device in the hours before bedtime take longer to fall asleep, feel less sleepy in the evening, secrete less melatonin, exhibit a later timing of their circadian clock, and show reduced next-morning alertness as compared to those who read a printed book instead. That said, the participants in this particular study had to read for four hours straight with the light-emitting device turned up to its maximum brightness.
Outline
This review first addresses the question of why e-books have not taken over from the traditional format for reading (Section 2). One important factor identified here may be the multisensory stimulation that interacting with a print book necessarily delivers, especially an older volume, that to date has simply not been captured by their electronic counterparts (see also Ruecker, 2002Ruecker, , 2006. Section 3 breaks down the individual sensory inputs -sight, smell, feel, sound, and even taste -that have been shown to influence the multisensory experience of handling a physical book, especially an older volume. Thereafter, several recent attempts to organise multisensory exhibitions around the theme of books are briefly reviewed (Section 4). The review then shifts to a consideration of the multisensory nature of reading, focusing on the multisensory mental imagery, including the inner voice (or voices), that so many of us hear when we read (Section 5). Finally, Section 6 questions the claim that reading a physical book is capable of conveying information more effectively than the digital version.
Why E-Books Have not Taken Over From the Traditional Physical Format
There are likely to be a number of reasons as to why e-books have not taken over from the traditional paper format, despite the many practical benefits that the former would appear to offer (Bilton, 2012;Tedeschi, 2011). One of the first things to notice here is that e-books do not smell. As Bilton (2012) puts it: "The scent of physical books -the paper, the ink, the glue -can conjure up memories of a summer day spent reading on a beach, a fall afternoon in a coffee shop, or an overstuffed chair by a fireplace as rain patters on a windowsill. iPads and Kindles, in comparison, don't necessarily smell like anything". It would seem likely that smell would presumably come low down the list if people were asked to rank the importance of the various senses as far as their enjoyment of books are concerned (Fenko et al., 2010; see also Ruecker, 2002Ruecker, , 2006. Nevertheless, it turns out that olfactory cues may play an important, if unacknowledged, role in the emotional (often nostalgic) response that many of us have to handling a real book, especially an older volume. According to Strlič et al. (2009, p. 8617): "The aroma of an old book is familiar to every user of a traditional library. A combination of grassy notes with a tang of acids and a hint of vanilla over an underlying mustiness, this unmistakable smell is as much part of the book as its contents" (see also Margolies, 2006;Rindisbacher, 1992).
The Emotional/Nostalgic Response to Handling Books
The handling of books, especially old and historic volumes, constitutes a genuinely multisensory experience, one that connects as much on an emotional/nostalgic as on a rational level (Borland, 2018;Griffiths and Starkey, 2018;cf. Neff, 2000). According to Meryl Halls, managing director of the Booksellers' Association in the UK, it is simply "harder to have an emotional relationship with what you're reading if it's on an e-reader" (quoted to Handley, 2019; see also Kaakinen et al., 2018and Ruecker, 2002. Halls continues: "The book lover loves to have a record of what they've read, and it's about signaling to the rest of the world. It's about decorating your home, it's about collecting, I guess, because people are completists aren't they, they want to have that to indicate about themselves". It is, then, the long-term record of what one has read, as well as the ability to share what one has read more easily, that also helps to explain the enduring popularity of the physical book (see Anon., 2015; Note 2).
One of the key reasons as to why e-books have not caught on may, in part, also be that they simply do not provide the same emotional multisensory appeal as do their physical print counterparts. Consider here only Neff's (2000, p. 22) suggestion concerning product innovation that: ". . . the most successful new products appeal on both rational and emotional levels to as many senses as possible". As Bridge (2017, p. 24) notes when describing the results of the eBay survey that was mentioned earlier: "Some 70% of those surveyed said they simply liked the look and feel of physical items". In his review of manuscript-historian Christopher De Hamel's (2016) book Meetings with Remarkable Manuscripts, Thomson (2016) highlights the latter's "excitement at holding the famous Chaucer manuscript at the National Library of Wales in Aberystwyth". Notice how it is the physical object, not the story itself, that is key here.
The nostalgia associated with real books and second-hand bookshops often merits a mention online (Bilton, 2012). For instance, one commentator captured the latter associations with the physical object thus: "Print books are physical reminders of your intellectual journeys. That beat-up copy of Catcher in the Rye on your bookshelf takes you back to sophomore year of high school. The Selected Poems of Pablo Neruda conjures up memories of late-night dorm room bull sessions. The food and wine-stained Lonely Planet Greece brings back that trip through the Greek Isles. A Kindle is just a Kindle" (Anon., 2015). Though this olfactory association is likely a missed opportunity for all those modern bookstores that now smell like coffee (Luttinger & Dicum, 2006, p. 164) or chocolate instead (Doucé et al., 2013).
Talking of nostalgic memories of books triggered by smell, my own experience as a child was with Desmond Marwood's (c. 1971), The Enchanted Marwood's (1971), The enchanted island: A young world scratch-and-sniff book, a rare early example of a scratch-and-sniff book. This most multisensory of volumes is more memorable to your present author than any of his other less olfactorilyinteresting books from childhood. This is perhaps an example of the Proustian moment (see Downes, 2000, 2002;Herz and Schooler, 2002;van Campen, 2014). Fig. 1). The multisensory experience has stayed with me for almost half a century. In fact, this volume constitutes an 'early and unusual example of a 'scratch-and-sniff' book' (https://childrensbookshop.com/book-103396.html) (Note 3). My own anecdotal experience, then, would certainly seem to agree with Laska's (2011) intuition that the distinctive smell of certain books may play an important role in aiding the recall of their contents.
The Multisensory Appeal of Books
Many scholars have acknowledged just how multisensory the experience of handling and reading a book really is. For instance, according to Baron (2015, p. 142), "Smell and sight are relevant senses when it comes to reading but touch may well be the most important". Along similar lines, Mc Laughlin (2015, p. 31) notes how "the feel of the book to the hand, the smell of the paper, the haptic pleasure of manipulating the screen [. . . ] reinforce and deepen the habit of reading" (see also Ruecker, 2002Ruecker, , 2006. In the sections that follow, I want to take a closer look at the distinctive sensory features that distinguish the experience of handling physical books from their digital counterparts.
Don't Judge a Book by Its Cover: the Visual Appearance of Books
The cover of a book often provides the first point of contact. And crucially, the visual impression created by the cover design is likely to set expectations in the mind of the reader about the nature of the contents. Visual imagery and/or art works constitute a distinctive feature of many book covers too. One might therefore want to consider the sensation transference that may occur, such that what one thinks about the cover art/design is likely to carry over to influence one's expectations of the contents (Duncan and Smyth, 2019). This links to what Hagtvedt and Patrick (2008) refer to as the art infusion effect. It is, however, important to note that the impression made by a book's cover need not always be positive. Relevant in this regard, novelist Joanne Harris (author of best-selling novel Chocolat) complained recently that the 'pink and frivolous' covers that are so often used for women's novels incorporating 'frivolous cursive fonts'. She argued that this may be dissuading competition judges from taking their writing seriously (Patel, 2020;Note 4). Much like the covers of records and CDs, then, the choice of colour and typeface for a physical book may serve to communicate symbolically, on an almost emotional level, about the tone of the contents (e.g., Carroll, 2016;Childers and Jass, 2002;Fox, 2005;Haverkamp, 2014;Lupton, 2018;Morrison, 1986;Van Egmond, 2004;see also 'Register', 2020;van Rompay and Pruyn, 2011).
The typeface in which the text itself is written provides an interesting aspect of the visual appearance of every book. Consider here only how the Comic Sans typeface (created in 1994 by the Microsoft Corporation) would be wholly inappropriate for conveying anything having a serious content/theme. Another intriguing contrast here is between serif and sans serif fonts. Serifed fonts apparently look better in print, whereas sans serif looks better on the web perhaps reflecting the lower resolution of screens (cf. Garfield, 2011;Note 5). The distinctive typeface that one finds in many old books and manuscripts is also interesting in this regard (e.g., De Hamel, 2016;Thomson, 2016). While the typefaces that are used have undoubtedly changed dramatically over the centuries, then, as now, there was likely an attempt to convey meaning through the connotative meanings associated with specific typeface (see Hyndman, 2015; Velasco and Spence, 2019, for reviews). It would be interesting to study the extent to which people's experience when reading an old volume is influenced by the particular typeface in which the text is presented. The lack of processing fluency that is typically associated with reading an unfamiliar old typeface today would also be worth considering here experimentally too (cf. Celhay et al., 2015;Venkatesan et al., in press;Winkielman et al., 2003).
Creating the Next 'Best Smeller': the Smell of Books
Old and historic books represent a core part of our material culture (e.g., Edwards et al., 2006; see also Thomson, 2016). However, while the content of these volumes is undoubtedly key to their enduring value, it is important to note that books are fundamentally multisensory objects, and that the multisensory characteristics often tend to be more pronounced than for the case of contemporary volumes. For instance, many old and historic books, in particular, have a distinctive smell (Bembibre and Strlič, 2017), be it from the binding (or glue), the paper (Hunter, 1987), or perhaps even from the ink itself.
According to Cambridge University don and librarian Christopher De Hamel: "I have no vocabulary to define this, but there is a curious warm leathery smell to English parchment, unlike the sharper, cooler scent of Italian skins" (quoted in Armitstead, 2017). Meanwhile, novelist Ray Bradbury has been quoted as saying that: "If a book is new, it smells great. If a book is old, it smells even better. It smells like ancient Egypt" (Madeline Slaven, pers. comm., December 9th, 2019). According to Claire Armitstead (2017), writing in the Guardian newspaper: "Cocoa, wood, rusks -every book has a distinctive smell. And each smell says something about how and when it was made, and where it has been". Meanwhile, Alberto Manguel, director of the National Library of Argentina, says that: "he was particularly partial to old Penguin paperbacks, which he loved for their odour of 'fresh rusk biscuits"' (Armitstead, 2017). Meanwhile, professor of heritage science, and trained chemist, Matija Strlič has noted that: "We know that books produced before approximately 1850 have a different smell to those produced between 1850 and 1990, and that's because late 19th-and most 20th-century printing was dominated by acid 'sizing' -the process to which pulp was subjected to reduce the water-absorbency of paper, so that it could then be written on" (quoted in Armitstead, 2017). E-books, by contrast, simply do not smell, as has already been noted (Note 6).
Anecdotally, when a book is brought to the conservators in Oxford's Bodleian Library, those working there will often comment that someone has brought in something from the Duke Humfrey's Library (https:// www.bodleian.ox.ac.uk/bodley/using-this-library/rooms/dh) based on nothing more than the smell that pervades the air when an item from this historic old reading room arrives. Once again, in this case it is unclear as to whether the smell emanates from the pages of the volumes themselves or from their bindings. Traditionally, of course, inks and dyes would have been much more olfactorily-redolent than they are today (e.g., see Blaszczyk and Spiekermann, 2017). To the knowledgeable conservator's nose, however, certain smells can also be indicative of degradation, and may signal the need to take prompt remediative action if the book or manuscript is to be saved (e.g., Adriaens et al., 2018;Strlič et al., 2009;Note 7). Recently, it was the terrible smell given by one particular volume that alerted one collector to the fact that the book in his possession had actually been covered in human skin (Wight and Dowell, 2020
. this decaying 'smell of history' is an insistent reminder of mortality and impermanence."'
A wide variety of volatile organic compounds (VOCs) are emitted from paper and other cellulose-based materials during degradation (Clark et al., 2011;Lattuati-Derieux et al., 2006;Strlič and Kolar, 2008), and their build-up in libraries and other book repositories can be measured (Fenech et al., 2010). The VOCs that are often detected, especially from older volumes, include organic acids such as acetic and formic acid, as well as aldehydes, including acetaldehyde and formaldehyde (Strlič et al., 2009). Furfural has been linked to the degradation of cellulose, and benzaldehyde and vanillin to the degradation of lignin (Bembibre and Strlič, 2017).
While on the topic of the smell of print books, one might also want to consider the scented inserts that used to be such a distinctive feature of many magazines and supplements a few decades ago (see Spence, 2002, for a review). For instance, in 2007, The Los Angeles Times ran a 'scratch-and-sniff' advertisement in a 32-page section about the season's forthcoming films. According to The New York Times, "The ad, for a family movie called 'Mr. Magorium's Wonder Emporium, 'used scented ink meant to evoke the olfactory pleasures of frosted cake"'. (Elliott, 2007). The suggestion in this case was that print publishers were looking for new ways to use fragrance in order to entice marketers to spend more. This was particularly important given the increasing loss of ad revenue to online media. Interestingly, though, in this case, in order to avoid complaints around unwanted exposure to fragrance, those paying for the ad wanted to ensure that 'you have to do something to the ad to make it smell' (Elliott, 2007). Going much further back, one finds Aikman (1951, pp. 545-546) writing about periodicals being printed on scented paper, with the article's author at one point, describing experiments to 'spice' a cookbook, hinting' at unexplored possibilities for best smellers'.
Handling: the Feel of Books
One of the most impressive aspects of many books, especially older volumes and manuscripts, is their heft and weight. Weight in the hand, as when we hold something heavy, is typically taken to denote quality and expense (e.g., Ackerman et al., 2010;Jostmann et al., 2009;Spence, 2019a;Spence and Piqueras-Fiszman, 2011). There is, after all, more than a grain of truth to the phrase 'A weighty matter', and as Linden (2015a, p. 15) observes of people interacting with books in the store: "once a book is selected, there is usually a brief moment of assessing its weight. Again, this is not always a conscious process". Ackerman et al. (2010) published the results of a fascinating study in Science in which job applicants were shown to be rated more highly when their curriculum vitae has a heavy rather than light cover (even if the CV itself remained unchanged). That is, certain tactile attributes of objects, even when not directly relevant to a given judgement, can nevertheless still affect people's perception and behaviour. At the same time, however, there would also appear to be a link between micro-books and devotional experience (Send, 2018), perhaps because at times of persecution such works can more easily be hidden and/or kept close to the body.
It is easy to imagine how the thickness of the paper may also cue quality, though I am not aware of anyone having studied this. The important role that touch has been argued to play in connoisseurship is perhaps also worth stressing here (MacDonald, 2007). The importance of handling to our understanding of material culture is undoubtedly a topic of growing academic interest (e.g., see Bacci and Pavani, 2014;Chatterjee, 2008;Pye, 2007). For instance, many of the respondents in a questionnaire study reported by Ruecker (2002Ruecker ( , 2006 also drew attention to how much they liked the feel and weight of real books, and how the quality of the haptic interaction changed as a book was repeatedly read or browsed through. Meanwhile, a recent article by Hernandez (in press) also stresses the importance of the feel of the weight of physical books. There is, then, a proprioceptive/kinaesthetic component to reading (McLaughlin, 2015).
One relevant notion here is that of affective ventriloquism, namely that what people think about a product, in this case the contents of a book, is likely influenced by what they think, or feel, about the packaging, in this case the cover of the book (see Spence and Gallace, 2011). Indeed, the subtle tactile cues have been shown to influence our purchasing decisions of even high-value products, such as the feel of the pocket lining when purchasing a fur coat (see Sheldon and Arens, 1932; and see Spence, 2019b, for a recent review).
The feel of the pages and their consistency throughout a volume also constitute salient sensory characteristics. For instance, the deliberate use of variation to contrast different textures is also an intriguing aspect of the feel and/or look of the pages of a physical volume (Kuitert, 2015). The contrast between different materials can, on occasion, be found in historic books (including a number of the volumes in the special collection of the Bodleian Library in Oxford). However, a similar approach is occasionally encountered in some contemporary volumes too. For instance, The Aviary Cocktail Book (Achatz et al., 2018) provides one such example, where the change in material is used to distinguish between the two main sections of the book. At this point, it is worth highlighting the fact that some differences in surface feel that are distinguishable by touch may not necessarily be visible to the naked eye (Skedung et al., 2013).
A recent trade publication was designed/launched specifically to draw attention to the joy of the physical in a digital age (Sappi Europe and Brown, 2019). Indeed, many books, very often those dealing with design or tactility have covers that have been treated in such a way as to give an unusual/distinctive feel. Lupton's (2007) Skin: Surface, Substance, and Design has a distinctive aerated smooth plastic cover. The unusually-textured finish is also a distinctive feature of Lupton and Lipps' (2018), The Senses: Design beyond Vision and of Hartmann and Haupt's (2016), Touch! Der Haptic-Effekt im multisensorischen Marketing [Touch! The haptic effect in multisensory marketing]. North American neuroscientist David Linden also chose to draw attention to the tactility of his book Touch: The Science of Hand, Heart, and Mind (Linden, 2015a, b). His book is covered in a thermally-responsive material in an attempt to entice more people to interact with it in the bookstore. This is likely to be a sensible idea as the likelihood of someone purchasing a product increases dramatically if they can be encouraged to pick it up in the store (see Spence, 2019b, for a review). The distinctive haptic experience associated with the handling of such volumes is obviously lost as soon as one switches to the e-book format.
At the same time that a customer's tactile interaction with a product increases the likelihood of purchase, it can also lead to worries about tactile contamination (see Argo et al., 2006;Linden, 2015a). This dislike of items that others have already touched presumably helping to explain why so many of us take a newspaper from somewhere other than the top of the pile! However, at the opposite extreme, items can also take on a special value if they happen to have been touched, used, or read by some figure of particular historic merit/relevance (see Gallace and Spence, 2014;Parisi, 2018;Thomson, 2016). For instance, take T. E. Lawrence's copy of the novel Ulysses, which has 'a sweet, somewhat smoky aroma that suffuses every bit of paper and leather' (Oram and Bishop, 2005). According to the latter commentators, researchers were sufficiently intrigued to find out more about the author's life experiences that may have underpinned the fragrant notes associated with the book (see also Bembibre and Strlič, 2017). Note again how this 'relic'-like response is lost when reading via a digital device. A little closer to home, one might think about handling books that meant something to now-deceased family members.
The Sound and Feel of Interacting With Books
The sound and feel of paper as the reader touches the pages, and the sound made by the latter as they are being turned over by the reader (e.g., Guest et al., 2002), also constitute part of the multisensory experience when reading a physical book. As one commentator put it: "Print books have pages that are nice and soft to the touch. Paper makes reading physically pleasurable.
Reading an e-book, on the other hand, feels like using an ATM. And after staring at a computer screen at work all day, how relaxing is it to curl up at home and stare at another screen?" (Anon., 2015). Meanwhile, in his essay on aesthetics, Japanese writer Junichirō Tanizaki (2001, pp. 17-18) notes that "Western paper turns away the light, while our paper seems to take it in, to envelop it gently, like the soft surface of a first snowfall. It gives off no sound when it is crumpled or folded, it is quiet and pliant to the touch as the leaf of a tree." Some of the respondents in Ruecker's study (2002Ruecker's study ( , 2006 did mention that they appreciated the crispness of the pages in a new book, others that they enjoyed the feel of the paper itself. That said, to the best of my knowledge there has not, as yet, been much research on the sound made by books as the pages are turned (see Stanton and Spence, 2020, for a recent review of audiotactile interactions in the context of action). In passing, consider only how some ereaders attempt to imitate the sound using an auditory page-turning icon (see Gaver, 1986).
And beyond the sound of the books themselves as they are interacted with, there is also the unique soundscape of the places in which those books are stored and read. In order to help people cope with pandemic lockdown, the Bodleian Library in Oxford recently released a stream of the sounds of the university, such creaks, rustles, coughs and traffic noises that you would normally hear in the libraries and the Radcliffe Camera; they even have a recording from the Duke Humfrey's library (see the Sounds of the Bodleian website; https://www.ox.ac.uk/soundsofthebodleian/#radcam; Kidd, 2020).
The Taste of Books
Taste (or gustation) is not a sense that one normally associates with handling books. In a historical context, though, it can be imagined how readers would once have licked/moistened their fingers in order to make it easier to turn the pages of a substantial volume (Note 8). This would have meant that they, in some sense at least, tasted, albeit indirectly, the books that they were examining. Indeed, a number of the historic illustrated manuscripts in the collection at the Bodleian Library in Oxford show evidence of former readers' saliva smudging certain of the images. One might also want to mention here Francis Bacon's (1601) very old adage that "Some books are to be tasted, others to be swallowed, and some few to be chewed and digested". On one documented occasion, such advice was taken literally: the guests at one party were instructed to chew pages from Clement Greenberg's Art in London, where John Latham was teaching at the time (John Latham with Barry Flanagan, Still and Chew, 1966-67; see also Stiles, 1998). This example is recounted together with the following quote in Kirshenblatt-Gimblett (1999, p. 9). "Examples abound of edible texts, from alphabet soup and birthday cakes to Ro Malone's cooked books" (Note 9).
The Multisensory Display of Books
The growing recognition of the importance of the multisensory, and specifically the non-visual, attributes of old books has led a number of museums, libraries, and archives to put on exhibitions that try to engage more of the visitor's senses (e.g., Bacci and Pavani, 2014;Classen, 2017;Classen and Howes, 2006;Levent and Pascual-Leone, 2014). Over the last few years, events have been organised at St. Paul's Cathedral Library (Bembibre and Strlič, 2017), the Birmingham Museum and Art Gallery, and, in late 2020, the Sensational Books exhibition at the Weston Library, part of Oxford's Bodleian Library (https://visit.bodleian.ox.ac.uk/events-exhibitions; see also De Bruxelles, 2020). Bembibre and Strlič (2017) analysed samples from an old book picked up in a second-hand shop (see Fig. 2). This enabled the researchers to develop a 'historic book odour wheel'. And, as for many other odour/flavour wheels, the idea here is that this helps connect identifiable chemicals with people's reactions to them. Thereafter, using fibres from the novel, an "extract of historic book" was produced, and presented to 79 visitors to Birmingham Museum and Art Gallery. The words that were most frequently used to describe the smell of a copy of Romanian writer Panait Istrati's 1928 novel Les Chardons du Baragan were 'chocolate'/'cocoa'/'chocolatey' followed by 'coffee', 'old', 'wood' and 'burnt'. "'From the analytical perspective, and given that coffee and chocolate come from fermented/roasted natural lignin and cellulose-containing product, they share many VOCs (volatile organic compounds) with decaying paper,' wrote the researchers" (quoted in Armitstead, 2017). Intriguingly, it is the smell of old books that one individual who had lost his sense of smell recounts as being a key scent that highlighted his loss (Classen et al., 1994).
The undoubted interest and olfactory appeal of old books and manuscripts also fits in with a more general fascination with historic scents and smells (e.g., Bembibre and Strlič, 2017;Blackson, 2008;Greenwood, 2017;Jenner, 2011;Jones, 2010;Reinarz, 2014). At the same time, however, it is important to note that there are practical curatorial challenges around the use of scents in heritage spaces (see Drobnick, 2006Drobnick, , 2014Henshaw et al., 2018). At this point, one might also ask how exactly those exhibitions and displays that have attempted to bring old and historic book to life have achieved their goals. While in some cases the multisensory angle relates specifically to the handling of the books themselves, on other occasions the aim may rather be to illustrate some aspect of the book's contents by means of enhanced multisensory engagement. Think here, for instance, how a curator might attempt to recreate the sights, sounds, and smells associated with the Canterbury Tales without it necessarily relating to the book itself (see also Aggleton and Waskett, 1999, on the use of smell in a museum setting). The Sensational Books exhibit to be held later in the year at the Weston Library in Oxford promises to capture, and synthetically reproduce, the scent of the historic Magna Carta. There is, of course, also an accessibility opportunity by extending museum book displays beyond the merely visual gaze (Graven et al., 2020).
Multisensory Mental Imagery Evoked by Reading
While the multisensory aspects of handling books mainly has to do with perception, the multisensory aspects of reading are concerned mostly with the rich mental imagery that is evoked by the written word. The multiple sensory aspects of mental imagery that may be at work during reading include the auditory 'inner voice' during silent reading (e.g., Moore and Schwitzgebel, 2018;Perrone-Bertolotti et al., 2012), and the complementary 'inner ear' imagining any non-speech sounds that may be evoked in a passage (e.g., Brunyé et al., 2010). There is even some evidence that readers may give different characters given different voices too (Alexander and Nygaard, 2008;Kurby et al., 2009). For many people, visual imagery is also elicited while reading (e.g., Boerma et al., 2016;Brosch, 2018).
Mental imagery may also be triggered in other sensory modalities while reading. For instance, olfactory and somatosensory mental imagery likely also occur, in at least some individuals. Indeed, reading olfactory descriptors, or words related to a distinctive smell, such as the word 'cinnamon', has been shown to give rise to increased activation in olfactory brain areas (González et al., 2006). Potentially relevant here, there has been something of a 'sensory turn' in literature in recent years. What this means, in practice, is that a number of novelists have started to increase the amount of sensory descriptive language in their works (see also Alryyes, 2006;Harris, 1997;Hertel, 2016;Vinge, 1975).
Australian novelist Tim Winton is one author whose work has been identified as being especially sensory in terms of the language used. According to one review of Winton's novel Breath (Winton, 2008), the author was quoted as saying that he wanted: "to give people an experience, taking the reader from their world and to achieve sensory momentum" (Eshelby, 2008, p. 69). Hence, when thinking of the olfactory imagery demonstrated by the laboratory research that was mentioned a moment ago, one can perhaps imagine how active the reader's olfactory cortex must be when reading the following sentences from the novel: "She was a foot away. She smelled of butter and cucumber and coffee and antiseptic." Ultimately, there would not seem to be much difference between physical and digital in their capacity to evoke mental imagery. However, this remains an empirical question that has yet to be addressed formally. What is more, it is currently unclear whether the enhancements (speaking characters, animations, etc.) that have been incorporated into some of the latest digital books might one day serve to replace the need (or opportunity) to engage the reader's mental imagery capacities so fully. Another intriguing, though as yet unanswered, question here concerns how any such reduction in evoked mental imagery affects comprehension/learning outcomes, enjoyment, engagement etc. Such questions, though, start to take us away from the key themes of this review, namely the multisensory experience of handling and reading books.
Do Physical Books Really Convey Information Better Than the Digital Version?
As one commentator noted: "You can write in the margins of a print book, dog-ear the important pages, and underline the key sentences with a pencil. Ebooks often allow the digital equivalents of these acts -but they just aren't the same. There is a link between physical gestures and cognition: the things we do to print books seem to help us to understand and remember better" (Anon., 2015). Indeed, longhand note taking has been shown to have advantages over laptop note taking (Mueller and Oppenheimer, 2014), perhaps connected to the superiority of multisensory to unisensory encoding of memory (cf. O'Connor, 1969). The reader's sense of immersion, or engagement, may also differ between physical and digital formats (Mangen and Kuiken, 2014). One has more of a sense how far one has progressed through a physical book, or as Anne Mangen puts it: "When you read on paper you can sense with your fingers a pile of pages on the left growing, and shrinking on the right. You have the tactile sense of progress, in addition to the visual... [The differences for Kindle readers] might have something to do with the fact that the fixity of a text on paper, and this very gradual unfolding of paper as you progress through a story, is some kind of sensory offload, supporting the visual sense of progress when you're reading. Perhaps this somehow aids the reader, providing more fixity and solidity to the reader's sense of unfolding and progress of the text, and hence the story" (as quoted in Flood, 2014).
Print books might also be more effective at conveying information to the reader. This, at least, was the claim made in a study by Mangen et al. (2019) that was first reported in the Guardian newspaper in 2014 (see Flood, 2014). Mangen and colleagues apparently found that those readers using a Kindle were less likely to recall events in a mystery novel, a 28-page short story of 10 800 words by Elizabeth George, than those who read the same novel in print format. The sample size in this particular study was, though, quite small (n = 25 in each group, and only two were experienced Kindle users). Nevertheless, similar claims concerning format effects on the reading experience and/or on the recollection of what has been read have also been made by a number of other researchers over the years (e.g., see Ackerman and Goldsmith, 2011;Delgado et al., 2018;Dell'Antonia, 2011;Grothaus, 2019;Halamis and Elbaz, 2019;Hou et al., 2017;Mangen et al., 2013;Parish-Morris et al., 2013;Roseberry et al., 2009;Singer and Alexander, 2017;Troseth et al., 2020;Wolf, 2018;Woody et al., 2010; though see also Köpper, Mayr and Buchner, 2016). Although there is not space to discuss these studies in detail here, in general, there would appear to be some agreement for a modest, albeit nuanced, influence of reading format on various outcome measures (see also Walsh, 2016, for a review).
There is undoubtedly also an important accessibility aspect to the question of book format too. For many years, a number of famous authors have resisted the move to e-books. For instance, J. D. Salinger famously refused to release digital versions of his books, as did Maurice Sendak and Ray Bradbury. According to recent reports, though, the late author's estate have now decided to make J. D. Salinger's books available in e-book format (Cain, 2019). In this case the reason was of improving accessibility for those suffering from a visual impairment. Audiobooks should probably also be mentioned here (cf. Graven et al., 2020). In relation to the latter case, though, Rogowsky et al. (2016) reported that comprehension in those who just read a couple of chapters of a novel, just listened to it, or read the e-text while listening to the digital audiobook audio show no differences in comprehension when tested immediately after reading, nor when tested again two weeks later (see also Heid, 2018).
Finally here, it is interesting to take a historical perspective on reading and technology for a moment. Hernandez (in press) has recently drawn attention to a couple of intriguing examples showing how people at the end of the 19th century were worried that people would listen to books rather than read (Uzanne, 1895). However, once again this predicted change to our reading behaviour (see Fig. 3) never took off.
Conclusions
Reading a physical book is a much more multisensory experience than reading the same content digitally. First-person reports clearly highlight the emotional, and very often nostalgic, associations that are triggered by the smell of, especially older, books. The feel and sound of interacting with (i.e., handling) a physical book is also very different from the digital version of the same experience (Ruecker, 2002(Ruecker, , 2006. The limited success of e-readers can therefore presumably be explained, at least in part, by their failure to acknowledge the importance of the non-visual senses to the reader's experience. Indeed, the evidence outlined here might be taken to suggest that digital innovation has not, so far, been able to successfully reproduce or substitute for the pleasures that are associated with tactile and haptic contact with the objects that we use on an everyday basis (and perhaps, as Jacques Derrida (2005) notes in his book Paper Machine, the "e-book, is just a phase in the evolution of reading technologies"; see also Hernandez, in press). Recognizing the importance of multisensory stimulation, a number of recent exhibitions have started to foreground how the experience of interaction with old books can stimulate all of the senses. At the same time, however, when it comes to the multisensory imagery that is triggered while reading, there is far less evidence thus far for a convincing difference between formats. Differences in comprehension and memorability between formats, digital, audiobook, and traditional printed format also appear to be fairly modest.
Notes
1. Though given the substantial price differential between print and e-books, actual sales figures would presumably be more informative than the financial figures reported here.
2. Another, less prosaic, reason for the revival of the physical book may be to do with the rise of the 'shelfie' (see Bridge, 2017). According to newspaper reports, it would appear that these days twentysomethings increasingly want to share 'shelfies' -that is, pictures of shelves of physical books, vinyl records, DVDs, etc. on sites such as Instagram.
3. The good news, as I discovered recently, is that the odours are non-toxic too (see Alderson, 1973).
4. According to Harris: "Books are subconsciously perceived to be more literary, and therefore of more value, if they have male-coded packaging. That means neutral imagery; male figures; a 'serious' font" (quoted in Patel, 2020).
5. It would be interesting to know whether printed and e-book versions of the same text use different fonts, though I have not been able to find any information on this particular score (though see https://designshack.net/articles/ typography/serif-vs-sans-serif-fonts-is-one-really-better-than-the-other/).
6. Though presumably it cannot be too long before someone spots a gap in the market and comes along with a plug-in old book scent to match the new car smell that exerts such a profound influence over people's perception of their own car (e.g., Hamilton, 1966;Moran, 2000a, b, c;Van Lente and Herman, 2001; see also Braun et al., 2016). The British Library shop sells a candle that is supposed to smell of 'library' (Armitstead, 2017; see also Bembibre and Strlič, 2017;Kuitert, 2015).
7. Another scent that signals danger is the recently-developed 'smell of data', which is designed to be released by a plug-in scent-dispenser when one's digital technology is in danger of being compromised over the internet (see Jones, 2017; https://smellofdata.com/; and see Whitelocks, 2012, for an attempt to synthetically recreate the distinctive smell of a computer).
8. This scene was captured in the movie version of Umberto Eco's (1983) The Name of the Rose. Until recently, one would also see bank tellers wetting their fingers before counting a stack of bank notes. Something that is no doubt going to become an even more distant memory in the era of COVID-19, and the weaponisation of saliva and other bodily fluids.
9. The scratch-and-sniff aromas in The Enchanted Island (Marwood, 1971), mentioned earlier, were all associated with edible products such as citrus, chocolate, strawberries, and peppermint sweets. Thus, it can be said that olfactory cues were being used to evoke specific tastes, or better said, flavours. | 2020-05-28T09:15:41.936Z | 2020-09-15T00:00:00.000 | {
"year": 2020,
"sha1": "a57ad6b5fa3cb6aa1f63c1233c4d420b16372d8f",
"oa_license": "CCBY",
"oa_url": "https://brill.com/downloadpdf/journals/msr/33/8/article-p902_4.pdf",
"oa_status": "HYBRID",
"pdf_src": "Adhoc",
"pdf_hash": "a8a486b1e081186a5fa59f3a1dcba590602c8fee",
"s2fieldsofstudy": [
"Art"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
58887863 | pes2o/s2orc | v3-fos-license | Industrialized building system - an innovative construction method
: Industrialized building system (IBS) has been introduced in construction industry as early as 1960’s. The challenges facing our construction industry are due to shortage of skill labour and high cost of construction. Therefore, the concept of construction using IBS is proposed to reduce our dependency on intensive labor works and to reduce the cost of construction. However, the process of changing the traditional construction method to IBS has been delayed due to the reluctant to accept the changes and even resistant to changes. Therefore this paper presents the concept of Industrialized Building System, the scope of IBS, challenges and way forward to inculcate the use of IBS construction in Indonesia.
Introduction
Housing needs continue to rise very rapidly in Indonesia, particularly in urban areas.Back log on housing needs up to 2016 has reached up to 7.6 million units, while demand for new homes every year has reached up to 800,000 units [1].There are about 1,350,000 units gap that must be resolved in order to fulfill the housing need in Indonesia [1].One of the ways to speed up the construction of these affordable housing is to use Industrialized Building System (IBS) in construction.Industrialized Building System is usually associated with sustainable construction which is described as the ability of the construction system to consider the environmental impact of a building over its entire lifetime, while optimizing its economic viability and the comfort and safety of its occupants [2].Typical standard building practices are guided by short term economic considerations; while IBS construction is based on best practices which emphasize on long term affordability, quality and efficiency.At each stage of life cycle of the building, it increases comfort and quality of life, while decreasing negative environmental impacts and increasing the economic sustainability of the project [3].The concept of IBS usually associates with the preservation of the environment which also concern on related issues such as the efficient use of resource, continual social progress, promising economic growth, and improve standard of living.The main goals of IBS construction are to meet present day needs for housing, working environments and infrastructure without compromising the ability of future generations to meet their own needs in times to come.In Indonesia, typical building construction is usually associates with construction system based on typical reinforced concrete design where formwork is formed using timber and plywood while the concrete is cast in-situ.As IBS system is quite new in Indonesia, a way forward to motivate and promote the system is a big challenge to the Indonesia construction industry.
This paper generally discussed on issues facing our construction industry and current researches related to the use of Industrialized Building System (IBS).One of the main problems facing our construction industry is a serious shortage of skill construction workers.Therefore, the introduction of Industrialized Building System (IBS) by Indonesia Government is a good alternative to reduce the dependency on skill workers.However, the use of IBS as an alternative to conventional system (reinforced concrete) has not widely accepted by construction industry players.Some of negative perceptions given by construction industry players on the use of IBS are listed as not ready towards new concept, insufficient information and lack of knowledge to understand changes in IBS.Hervas revealed that construction sector is known as traditional sector that can be identified as not willing to accept changes or even resist changing [4].This paper also discusses the current implementation of IBS, its shortcomings and the way to enhance the system forward.It also elaborates the current issues by focusing on stability, build-ability and cost efficiency and research done in IBS.
Definition and types of IBS
IBS is defined as a construction system which components are manufactured in a factory, on or off site, positioned and assembled into structure with minimum additional site work [5].Dietz defined IBS as total integration of all sub-system and components into overall process fully utilizing industrialized production, transportation and assembly techniques [6].According to Parid (1997), IBS is a system which use industrialized production technique either in the production of component or assembly of the buildings or both [7].Lessing et al defined IBS as an integrated manufacturing and construction process with well-planned organization for efficient management, preparation and control over resources used, activities and results supported by the used of highly developed component [8].
Fig. 1. Pre-cast concrete system
Nowadays, there is an increase in demand for housing due to an increase in population.In most countries including Indonesia, the demand is exceeding the supply.In order to solve the inadequacies of housing, Industrialized Building System (IBS) is introduced to replace conventional construction method.Basically, CIDB has classified IBS into five groups [3] These systems are used due to several advantages.It saves a lot of time as the casting of precast structure can be done in factory and installed the construction components on site.[9] Besides, the delay in construction can be minimized as the construction components are fabricated in the factory with controlled environment [9].Industrialized building system has the benefits when compared to the conventional construction method in the aspects of cost savings, less relied on labor, quality control, environmental friendly and less dependent on weather problem.
Cost
The implementation of the IBS in construction will reduce the construction time.CIDB [3][4] stated that the casting of the precast element in the factory and foundation at site can be carried out simultaneously.As the construction time decreases, the overall cost of the construction will also be reduced.Besides, the repetitive use of system formwork made up of steel and scaffolding can save the cost [11].
Labour
Since IBS on mechanical means to produce prefabricated elements and only minimal in-situ construction, it reduces the requirement of labors for prefabrication of element and erection at site.Skilled and semi-skilled labors are needed for IBS and unskilled workers are mostly not needed.Countries like Malaysia where the construction industry is the host to a large number of foreign workers from Indonesia, Bangladesh and Pakistan.It is important to stress on that the labors are still have to send for training for the skills appropriate with the IBS.Although the training may cause some cost, it is believed that trained labor would be more quality than the untrained labor.
Quality control
The quality of the prefabricated concrete components in IBS is much better than the conventional method due to the quality in terms of the selection of materials, manufactured under controlled environment and inspection before the components send to the site.The materials used are based on the quality standard.Besides, the prefabricated concrete components are normally carried out in the factory which is indoor environment and it will not affected by the raining season in Malaysia.Inspection will be done to make sure the quality of the components is achieving the standard or the requirement needed by the clients.
The overall advantages of the using IBS system be summarized as shown in Fig. 6. [12].The roadmap listed five main focus areas that are needed to drive forward the use of IBS in our construction industry in Malaysia.The five areas are listed as Manpower, Materials, Management, Monetary, and Marketing [13].IBS Steering Committee was formed to enhance further the concept of IBS in the construction system [14].The functions of the Steering Committee were to develop standard drawings for common use, to provide expertise in the process of development and production of IBS components, to carry out test and evaluation and to provide assistance in analysis and design procedures [14].
Challenges and delays
After all efforts done by the steering committee, shortcomings and delays of IBS implementation can be listed as follows [15]: • IBS is not popular choice by consultant and developers due to lack of knowledge on the performance of the IBS system.
• The needs to change the mind-set of construction industry player that for long run IBS is much better than the traditional construction.• Insufficient push factors from governments and policy makers such as Public Works Department and local authority.• Lack of technical know-how from production to erection.• Volume and initial cost of production are also issues to be addressed in the supplier chain.Besides all the shortcomings and negative remarks mentioned above the way forward to implement the use of IBS are very much needed in order to benefit the advantages developed from IBS.A series of workshops to enhance awareness have been carried out by CIDB to all construction industry players.The awareness in using IBS has also been initiated by consultants, universities, and companies produce the IBS components.
Planning and implementation
The establishment of steering committee by Indonesia government should be seen as a very good step to inculcate IBS system in construction industry.One of the ways forward to enhance the effective planning and implementation of IBS is to promote research and development (R&D) of IBS in all related parties in construction.The R&D and workshops organized by Indonesia Construction Board should be carried out to the public so that the awareness of the IBS can be well accepted and understood.The initiatives to accept the use of IBS should be done by all parties such as contractors, consultants, universities, developers and research institutes.The main objectives of promoting the use of IBS is to ensure that factors such as improve the performance, quality, and minimize the dependency of unskilled workers and reduced cost can be well explained and understood.A complete and comprehensive study on IBS solutions which take into consideration the role of the entire value chain should be emphasized.A technology transfer model from other parts of developed countries should be done through smart collaboration.
5 Some experimental works 5.1 Influence of longitudinal spaced bolted shear connectors on composite beam integrated with concrete and cold-formed steel [16] have experimented composite construction with conventional hot rolled steel (HRS) sections has been known to perform much better than Cold-formed steel (CFS) sections for decades.However, the composite action of CFS with an in-situ concrete, especially Self-Compacting Concrete (SCC) and bolted shear connector 'haven't been investigated'.This study attempted to investigate the behavior of bolted shear connector used with SCC and CFS as 'shear connector' to form a composite beam system at designated longitudinal spacing.Push-out test specimens of shear connector longitudinal spacing of 300 mm, 250 mm and 150 mm with bolted shear connector of grade 8.8.Steel-concrete composite beam has being in use as a structural member in building and bridges for decades.The shear connection between steel and concrete slab in composite beams is inherently significant, as it resists separation between the two components as well as enhance longitudinal shear transfer.
Fig. 7. Test of shear composite
The paper has come to summary that the effect of shear connector longitudinal spacing was investigated, and its influence on the ultimate strength capacity of the shear connector was established.The connection was classified as ductile shear connector since it attained a characteristic slip capacity.The longitudinal spacing of bolted shear connector had shown the influenced the strength capacity of the shear connector.
Performance of varying bolted shear connectors in cold-formed steel composite beam
Anis Saggaff study a suitable bolt and nut type of shear connection mechanism has been proposed.The Experiment has presented the possibility of using a type of bolt and nut as shear connector.The experimental tests results were conducted to ascertain the performance of the proposed connectors.The aim of the research was to investigate the structural performance of bolted shear connectors in providing composite action between Selfcompacting concrete (SCC) and CFS section integrated as composite beam system.
Fig. 8. Push-out test
The push-out test results have shown the load-slip relationships of the tested specimens.The specimens with M14 and M16 bolt diameters could be attributed to the failure of CFS section by flange local buckling.Perhaps high resistance of the applied load by the concrete slabs and bolted shear connectors could be responsible for it.The failure due to steel buckling occurred at the top shear connector position of specimens which was close to the load application position.The failure then extended upwards from the initial position where it had occurred to the part where the CFS was not covered by the concrete slab.
The Remarkable shear resistance and slips at ultimate loads were attained by the specimens with all the bolted shear connectors.It has proved that the maximum design capacity of the specimens with M14 and M16 bolted shear connectors is based on the tested value obtained due to CFS failure.
The study has shown the influence of shear connectors.The shear strength capacity of bolted shear connectors was noted as follows: 40.6% between M12 and M14 41.0% between M12 and M16 and 0.3% between M14 and M16.It was concluded that the shear connectors with single embedded nut and washer influence on the ultimate strength capacity of bolted shear connectors.The bolted shear connectors used in this study possessed good shear resistance capacity.The formation of the formwork for each specimen was done using the lipped cold-formed steel section with stiffeners positioned at right angle to shape the footing as a boxed type specimen and to avoid changes in shape during concreting.The reinforcement bars used in the LCCFSS specimens were A10 which was bent downward about the length of the cover needed for the footing and sit on the LCCFSS specimen as shown in.This was purposely done so that no spacer is needed underneath the A10 wire mesh to meet the requirement of the concrete cover.For typical reinforced concrete pad footing the A10 wire mesh was bent upward which result to the need of the spacer underneath the reinforcement to meet the requirement of concrete cover.Test results has shown the maximum axial load applied until failure based on the theoretical calculation from BS 8110 part 1:1997 [12].The load capacity of the footing also depends on the length of the footing besides the thickness of the footing.This study has maintained the thickness is kept constant.The load capacity of the footing has been reduced by up to 30% as the length of the footing increased from 1000mm to 1750mm.The reduced in load capacity is very much related to the failure of the footing which is shifted from punching shear failure to bending failure.
Conclusions
The use of IBS system needs all construction players to adopt or willing to accept the way forward in construction industry.It is hoped that the finding addressed in this paper could highlight the benefits of using IBS in actual construction.It is also hoped that the IBS can be used to solve the problem of housing needs in Indonesia.The delays in implementation should not hinder the way forward in construction as the benefits of using IBS are too many to be ignored.Indonesia government should take a drastic measures and regulation in order to prosper the success of IBS in Indonesia.
The author of this paper would like to acknowledge Specials thanks addressed to which are • Precast Concrete System (see Fig. 1) • Steel Formwork Systems (see Fig. 2) • Steel Framing Systems (see Fig. 3) • Prefabricated Timber Framing Systems (see Fig. 4) • Block Work Systems.
Construction Research Center UTM, Prof Ir.Dr Mahmood Md Tahir., and Structure and Construction Research Center also Sriwijaya University that make this paper 101 5 10.Layout of test setting | 2018-12-18T20:07:07.432Z | 2017-01-01T00:00:00.000 | {
"year": 2017,
"sha1": "475b93245a37e25130a1a4b3f9979d3651a06bd6",
"oa_license": "CCBY",
"oa_url": "https://www.matec-conferences.org/articles/matecconf/pdf/2017/15/matecconf_sicest2017_05001.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "475b93245a37e25130a1a4b3f9979d3651a06bd6",
"s2fieldsofstudy": [
"Economics",
"Computer Science"
],
"extfieldsofstudy": [
"Engineering"
]
} |
237327634 | pes2o/s2orc | v3-fos-license | Belminus santosmalletae (Hemiptera: Heteroptera: Reduviidae): New Species from Panama, with an Updated Key for Belminus Stål, 1859 Species
Simple Summary A new species of Belminus, discovered during the study of unidentified Triatominae specimens from the Hemiptera collection of the National Museum of Natural History, Smithsonian Institution, Washington DC, USA is described here. After comparison with previously described species, significant morphological and morphometric differences were observed, confirming the discovery of a new triatomine species, Belminus santosmalletae. Abstract Belminus santosmalletae, a new triatomine species, is described based on a specimen from Panama, deposited in the collection of the National Museum of Natural History (NMNH), Smithsonian Institution, Washington, DC, USA. Attempts failed to identify this specimen using the keys by Lent and Wygodzinsky (1979) and Sandoval et al. (2007). A comparison was made with specimens of Belminus Stål, 1859 specimens deposited at the Triatominae collection at the Oswaldo Cruz Institute (CTIOC), Rio de Janeiro, Brazil; and with previous descriptions of Belminus species. These comparisons showed the specimen represents a new species, described in the present paper. It differs from other species of the genus mainly by the grainy tegument, scarce pilosity along the body, and the number of tubercles observed on the pronotum.
Introduction
The subfamily Triatominae (Hemiptera: Reduviidae) comprises 153 extant and three fossil species assigned to five tribes [1][2][3][4] with all the extant species being considered potential Chagas disease vectors. The tribe Bolboderini has been considered a monophyletic group, and includes the genera Bolbodera Valdés, 1910; Belminus Stål, 1859; Parabelminus Lent, 1943; Microtriatoma Prosen and Martínez, 1952 [4] (Figure 1). The genus Belminus was described by Stål [5] based on a single species, Belminus rugulosus, from Colombia [6]. The group is well characterized and can be easily differentiated from other triatomines by the small total body length (8.5-12 mm); elongate and fusiform head; dorsoventrally compressed labium, with the first and second visible segments elongate and subequal in length and longer than the third segment; the presence of 6-7 small discal tubercles on the anterior lobe of pronotum and connexivum with a dorsal longitudinal ridge [7,8]. It is the most diverse genus of Bolboderini, with eight previously described valid species: Belminus corredori Galvão (Figure 2A-H).
During a study of the Hemiptera collection at National Museum of Natural History, Smithsonian Institution (NMNH), Smithsonian Institution, Washington DC, USA, a specimen belonging to the genus Belminus was found. Upon comparison with other specimens of the genus deposited in the Triatominae collection of the Oswaldo Cruz Institute (CTIOC), and with the descriptions of the other Belminus species, it was clear that such specimen represents a new species. Here, we describe Belminus santosmalletae sp. n. (Figure 3) based on this single female specimen.
Materials and Methods
Measurements and observations were made using a Dino-Lite Edge digital microscope. Dorsal habitus images and detailed photos were taken using the wide zoom microscope OlympusDSX100 camera. The following characters and terminology used for description based on Lent and Wygodzinsky [8] and Sandoval et al. [9].
Coloration
Overall color brown to yellowish-brown, with light and dark brown areas on the pronotum, hemelytra, connexivum, legs, and abdomen; with a few spots paler than the integument. Visible labial segments brown to yellowish-brown. Pronotum predominantly yellow with brown stripes and spots. Scutellum brown with the apex of scutellar process yellow. Femora with a subapical yellow ring. Corium yellow with three dark brown areas, one external, almost straight, the other internal. Membrane of hemelytron dark brown, veins darker with secondary venation within cells visible but not conspicuous. Connexival segments with transverse marks, anterior dark brown marks wider than posterior dark yellow marks ( Figure 3A,B).
Morphological Features
Female. Total length 11.45 mm, width of pronotum 2.75 mm, and width of abdomen 4.63 mm. The integument of the entire body was very grainy with short pilosity, except for the hemelytra.
Head. Elongated, fusiform, very granulose, three times as long as wide (1:0.31), slightly longer than pronotum (1:0.84). Clypeus truncated on the apex. The anteocular region was more than twice as long as postocular region (1:0.41). The postocular region was sub-circular, with sides which were convex and convergent posteriorly. The genae was compressed laterally, with the apices, considerably, surpassing anteclypeus. The external spinelike projection of the antenniferous tubercle was short, and barely extending beyond the base of the first antennal segment. Antennae were inserted apically, a third of the way through the anteocular, towards the postocular region. Antennal segments were missing, except in segment I. Eyes in the lateral view reached the level of the lower but not the upper surface of the head (Figure 4). The ratio between eye width and synthlipsis was 1:2.1. Ocelli were very small, not elevated, laterally oriented, and situated at the level of the integument. The visible labial segments of the first segment was longer than second, not reaching the anterior border of the eyes, while the third visible segment barely reached the anterior portion of the stridulatory sulcus; the length ratio of visible labial segments was 1:0.8:0.4 ( Figure 4). Pronotum. The width of the pronotum was slightly bigger than the length (1:0.83). The anterior lobe was narrow, granulose, with 7 + 7 conspicuous discal tubercles and 2 + 2 conspicuous lateral tubercles ( Figure 5), sides forming a conspicuous angle at the junction of the posterior lobe sides. The anterolateral processes were short, subtriangular, and rounded apically. Th posterior lobe was granulose with submedian carinae almost attaining the posterior border of the pronotum. The scutellum was triangular, with the apex of the scutellar process being long, subcylindrical, and not pointed apically. Further scutellum characters are not available for this specimen, as it was primarily pinned through the scutellum. Prosternum with 1 + 1 projections, lateral to the stridulatory sulcus. Legs. Short and stout. Femora were sulcate at the venter, and with 2 + 2 subapical denticles, one being distinctively larger ( Figure 6). Tibiae were slender and compressed laterally with short pilosity. The specimen was without tarsi. The hemelytra fell distinctly short of the apex of abdomen, with venations consistent with the genus description. Abdomen. Abdominal venter was flattened, and the connexivum was wide dorsally with a dorsal ridge. The spiracles were close but not adjacent to the connexival suture ( Figure 7). Remarks on the conditions of the specimen. The specimen was found without the last three antennal segments, without tarsi, and with scutellum damaged by the pin.
Diagnosis. According to the comparative morphology of the genus, Belminus, B. santosmalletae, it seems to be closer to the species B. ferroae and B. pittieri, and can be distinguished by the yellow rings on all femora (these rings can be found in B. herreri and B. laportei, but the morphology of the wing is distinct), grainer integument, scarce pilosity, lighter and larger yellow spots on the pronotum, 7 + 7 discal tubercles on the pronotum, rather than 6 + 6 and darker membrane, with little color contrast between the cells and veins.
Etymology
This species is dedicated to Dr. Jacenir dos Santos-Mallet, a Brazilian researcher at the Fundação Oswaldo Cruz, in Rio de Janeiro, Brazil. She has dedicated over 30 years to the research of tropical disease vectors.
Discussion and Conclusions
The genus Belminus currently comprises nine valid species, with distribution ranging through Central America, Colombia, Peru, Venezuela, and northern Brazil. The genus, previously thought to be exclusively sylvatic (arboreal), was recently found inside human dwellings in some locations in Colombia and Peru [10]. To this date, however, there has been no report of infection by Trypanosoma cruzi (Chagas 1909), the etiological agent of Chagas disease.
A unique feature within Triatominae is also the fact that Belminus species have been found feeding on other insects, especially cockroaches, showing a broader host preference. Wheeler and Pennak [11] considered Belminus corredori one of the 100 most interesting species discovered in this century, and according to Gil-Santana [4] this fact launched "an unprecedented focus on the uniqueness of the representatives of this tribe".
Belminus santosmalletae is the ninth species described to this understudied genus, and the second from Panama. In addition to the original descriptions [8,12,13], taxonomic knowledge on the genus is restricted to a few morphological papers reflecting the fact that this is the one of the least known triatomine genera. Herrer et al. [7] made the first descriptions of eggs and nymphs of first and fifth instars of B. peruvianus, and only after 31 years, Rocha et al. [14] described all instars of one Belminus species: B. herreri. They provide a complete description, as well as ontogenetic morphometrics of post-embryonic head development. Gil-Santana and Galvão [15] described the male genitalia of B. rugulosus and B. corredori and provided a discussion on previous data in the literature on male genitalia of other species. It is impossible to make inferences on the biology of B. santosmalletae because, like B. pittieri, the description is based on a single female specimen. Additionally, a search on GenBank shows a single sequence (GenBank access AJ421964) from this group, highlighting the lack of data available to further studies on integrative taxonomic approaches for the genus. | 2021-08-28T06:17:19.429Z | 2021-07-30T00:00:00.000 | {
"year": 2021,
"sha1": "d5f518152c83e8b87094e4ee783196b1f0bd3c9c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2075-4450/12/8/686/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "69cb3f490c4e62f88ae524ee375aa405226b23df",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
11862147 | pes2o/s2orc | v3-fos-license | Ankyrin repeat domain-encoding genes in the wPip strain of Wolbachia from the Culex pipiens group
Background Wolbachia are obligate endosymbiotic bacteria maternally transmitted through the egg cytoplasm that are responsible for several reproductive disorders in their insect hosts, such as cytoplasmic incompatibility (CI) in infected mosquitoes. Species in the Culex pipiens complex display an unusually high number of Wolbachia-induced crossing types, and based on present data, only the wPip strain is present. Results The sequencing of the wPip strain of Wolbachia revealed the presence of 60 ankyrin repeat domain (ANK) encoding genes and expression studies of these genes were carried out in adult mosquitoes. One of these ANK genes, pk2, is shown to be part of an operon of three prophage-associated genes with sex-specific expression, and is present in two identical copies in the genome. Another homolog of pk2 is also present that is differentially expressed in different Cx. pipiens group strains. A further two ANK genes showed sex-specific regulation in wPip-infected Cx. pipiens group adults. Conclusion The high number, variability and differential expression of ANK genes in wPip suggest an important role in Wolbachia biology, and the gene family provides both markers and promising candidates for the study of reproductive manipulation.
Background
Wolbachia are obligate endosymbiotic bacteria that are maternally transmitted through the egg cytoplasm and are responsible for several reproductive disorders in arthropods, such as cytoplasmic incompatibility (CI) in infected Culex mosquitoes [1,2] and many other insects. Although Wolbachia are not found in mature sperm, they can modify developing sperm, possibly via chromatin binding proteins [3], such that when they fertilise an uninfected egg embryonic development is arrested. The reciprocal cross between infected females and uninfected males is, however, compatible; Wolbachia-infected females therefore produce a higher mean number of offspring than uninfected females. This unidirectional CI enables Wolbachia to rapidly invade uninfected populations [4], and provides a mechanism for driving anti-pathogen transgenes or a lifespan-shortening phenotype into mosquito populations [5,6]. Bidirectional CI can also occur between insect populations, usually when they are infected with different strains of Wolbachia.
The genome sequence of the wMel strain [7], a CI-inducing Wolbachia strain found in Drosophila melanogaster, revealed an unusually high number of ankyrin repeat domain (ANK) encoding genes. Ankyrin repeats, consisting of around 33 residues, have been identified in a large number of proteins [8]. Ankyrin repeats are known to mediate protein-protein interactions in eukaryotes and are present in proteins involved in very different functions including cell cycle regulation, mitochondrial enzymes, cytoskeleton interactions, signal transduction and toxins [9]. Although ankyrin repeats are common in both eukaryotic and viral proteins they are relatively rare in bacteria and their function remains largely unknown. A protein containing ankyrin repeats in the bacterium Ehrlichia phagocytophila was detected in the host cytoplasm and found to be associated with chromatin suggesting a possible role in host cell gene expression [10]. ANK proteins have also been shown to mediate protein-protein interactions in cyclin-dependent kinase (CDK) inhibitors. In Nasonia wasps, the control of host cell cycle timing at karyogamy appears to be disrupted in CI and inhibition of CDK1 has been proposed as a possible mechanism [11,12]. Taken together this has led to the suggestion that ANK genes could play a role in Wolbachia-induced CI [7].
Species in the Cx. pipiens complex display an extremely high number of Wolbachia-induced crossing types between populations, with a high frequency of uni-or bidirectional incompatibilities [13][14][15]. Despite the complexity of crossing types, no polymorphism in the wPip strain of Wolbachia, responsible for CI in Cx. pipiens mosquitoes, has been found in the nucleotide sequences of ftsZ [15] and 16S rRNA [16] or in the highly variable wsp (surface protein)gene [17]. Sequencing of the wPip genome was undertaken partly in order to attempt to resolve this discrepancy. Interestingly, sequence analysis of some ANK genes found in wPip revealed variation in both nucleotide sequence and predicted amino acid sequence for two prophage associated ANK genes, pk1 and pk2, between wPip-infected Cx. pipiens colonies [17]. The wAu strain of Wolbachia, found in Drosophila simulans, is closely related to the wMel strain but does not normally induce CI [18]. The homolog of pk2 in wAu contains a premature stop codon not present in the wMel homolog, which suggests it could be a candidate gene for involvement in CI in Drosophila [19].
Variable expression between sexes and strains of Cx. pipiens was detected for the pk2 gene, a characteristic that might be expected for genes involved in the specific modification and rescue functions between incompatible strains. Any differential expression of ANK genes between male and female wPip infected adult Cx. pipiens mosquitoes would suggest an important function of these genes in the interaction between Wolbachia and its insect host. How Wolbachia differentially modify sperm in males as well as rescue in females is as yet unknown, but could potentially involve variability in the expression and activity of Wolbachia genes in male and female insect hosts. Variable gene expression in Wolbachia is not thought to occur at a high rate, as only a small number of regulatory genes have been identified in the Wolbachia genomes sequenced to date [20]. In this study, we analysed the expression profile of all ANK genes in wPip in Cx. pipiens adult mosquitoes.
Number and distribution of wPip ANK genes
Analysis of the wPip genome revealed 60 ANK genes, which are numbered sequentially in Table 1. Several ANK proteins have predicted signal peptides and transmembrane domains. Thirteen of the wPip ANK genes are contained in several chromosomally integrated prophage regions, similar in sequence to the wMel WO-B prophage region [7]. The ANK genes pk1 and pk2, homologues of the wMel genes WD0596 and WD0636 respectively and previously shown to vary between incompatible Culex strains [17], are here shown to be present in multiple identical copies in different prophage regions: wPip_ANK8, wPip_ANK14 and wPip_ANK56 in the case of pk1 and wPip_ANK12 and wPip_ANK25 in the case of pk2. Two sequence variants of pk2 in wPip from different Cx. pipiens group colonies have been previously described and were named a and b. The wPip_ANK16 gene is also homologous to the pk2 genes/WD0636 in the wMel strain and is present in wPip in all the infected Cx. pipiens group colonies listed. A third pair of identical prophage-associated genes are also present, wPip_ANK13 and wPip_ANK26, which are homologues of WD0637. Thus in total there are 56 unique ANK genes present in the wPip genome.
Only 15 of the 23 wMel ANK genes have clear homologues in the wPip genome, which might reflect the high degree of heterogeneity in this group of genes. Thus, when likely paralogous groups (three non-identical homologues of WD0566 and two each of WD0636 and WD0637) and identical copies are taken into account, 37 of the 60 identified wPip ANK genes in the wPip genome do not have any clear homologues in the wMel genome. By way of comparison, the wBm strain of Wolbachia, thought to be a nutritional mutualist in the filarial nematode Brugia malayi, encodes only five ANK proteins [21], three of which are homologous to the wPip ANK encoding genes. The number of ANK domains as identified by Pfam, gene length (bp), the wMel and wBm homologous gene where there this can be clearly determined. Symbols # (pk1), *(pk2) and + denote groups of prophage-associated genes with identical sequences.
ANK gene expression
Transcripts were detected for all of the ANK encoding genes. For the majority, expression in adult males and females of the Pel colony was not obviously different based on agarose gel electrophoresis of RT-PCR products.
The wPip_ANK57 gene showed very low expression in Pel female extracts and no detectable expression in Pel male RNA extracts. wPip_ANK2 and wPip_ANK49 showed low levels of expression in both Pel male and female RNA extracts. RT-PCR analysis also suggested that wPip_ANK38 is highly expressed in both sexes.
wPip_ANK12 and wPip_ANK25 The identical prophage associated ANK encoding genes wPip_ANK12 and wPip_ANK25, previously together named pk2 [17], showed the greatest difference in expression between sexes, with no detectable RT-PCR products in the males of the Pel and Mol colonies. Expression of these genes was also not detected in males for an additional Cx. pipiens colony from Sri Lanka (Sumo Cyppe). Quantification of expression by quantitative reverse transcription (qRT-PCR) was carried out and the mean male expression of the pk2 gene in the Pel colony in comparison to female expression was 1.6% ( Figure 1). However, expression of pk2 was observed at similar levels in males and females of the Col colony. Primers were designed to discriminate between pk2 sequence variants pk2a present in the Pel, Sumo and Mol colonies and pk2b present in the Col colony and confirmed no detectable expression of pk2a from male RNA extracts of the Pel, Sumo Cyppe and Mol colonies using RT-PCR ( Figure 2). The pk2b gene variant was expressed at similar levels in Col colony adult females and males. Further RT-PCR analysis showed pk2 gene expression in both preblastoderm embryos and pooled 4th instar larvae (sex undetermined) of the Pel colony. pk2 expression in pooled testes from 20 Pel males was just detectable but the RT-PCR product was very weak compared to those for wsp and pk1 (not shown).
Differential expression between sexes was also observed for two genes directly downstream of pk2 (Figure 3). pk2-1 encodes a hypothetical protein present in identical copies in the two pk2 associated prophage regions. pk2-2 encodes a site-specific recombinase present in almost identical copies in the two prophage regions. Primers used for expression studies could not discriminate between the pk2-2 copies. For the gene upstream of pk2, also an ANK encoding gene present in two identical copies (wPip_ANK13 and wPip_ANK26), RT-PCR followed by agarose gel electrophoresis revealed similar expression levels in both female and male RNA extracts of the Pel colony. Primers designed to span the intergenic regions of pk2/pk2-1/pk2-2 produced RT-PCR products from females but no detectable products from males of the Pel colony. Primers spanning the intergenic region between pk2 and pk2+1 (primers 5 and 6) produced no amplification of a transcript from either female or male RNA extracts of the Pel colony. However, using the same primers, a product of correct size (733 bp) was amplified in both male and female DNA extracts of the Pel colony.
wPip_ANK16
The pk2/WD0636 homolog wPip_ANK16 was present in all the Culex strains tested based on PCR amplification. In the Pel, Mol and Sumo Cyppe colonies, wPip_ANK16 was expressed equally in males and females; however for the Col colony no expression could be detected in males, and only a weak RT-PCR product could be detected in females ( Figure 2).
wPip_ANK1
Standard RT-PCR analysis followed by agarose gel electrophoresis revealed much lower expression levels of wPip_ANK1 from pooled male RNA extracts of all wPipinfected Culex colonies in comparison to female RNA extracts. Quantification of expression by qRT-PCR was undertaken and the mean male normalized expression of the wPip_ANK1 gene in Pel males relative to Pel female expression was 4.7% ( Figure 1). Expression levels were similar in Pel females and males for both genes flanking wPip_ANK1, based on standard RT-PCR followed by agarose gel electrophoresis. wPip_ANK52 Expression analysis using RT-PCR followed by gel electrophoresis revealed lower expression levels of wPip_ANK52 from pooled male RNA extracts of all wPip-infected Cx. pipiens colonies in comparison to female RNA extracts. Quantification of expression by qRT-PCR was undertaken and the mean male normalized expression of the wPip_ANK52 gene in Pel males relative to Pel female expression was 19% ( Figure 1). Reduced expression levels were also observed in Pel male RNA extracts for three additional genes flanking wPip_ANK52 based on standard RT-PCR followed by agarose gel electrophoresis. However, although standard PCR using primers to span the intergenic regions of these genes resulted in an amplified product of approximately 1.5 Kb, no products were amplified using RT-PCR from either Pel female or male RNA extracts.
Discussion
The presence of 60 ANK genes is significantly more than the 23 identified in the wMel genome [7]; in fact, the number and density of ANK genes is the highest reported for any prokaryotic genome. The expansion of ANK genes in the wPip strain, the degree of sequence variability and sex-specific expression in adult Cx. pipiens mosquitoes suggests an important biological role in parasitic strains of Wolbachia. The RT-PCR analysis provides strong evidence for a single transcriptional unit (operon) produced from three prophage associated genes including pk2. However, there was no evidence that wPip_ANK1 and wPip_ANK52 are part of sex-specifically regulated operons.
The quantitative RT-PCR analysis in this study represents only an estimation of differences in relative ANK gene expression. The accurate quantification of RNA expression in bacteria has been limited due to the absence of reliable standardization. In eukaryotic cells, stably expressed housekeeping genes can be used as standards to perform relative quantification of gene expression. For an endosymbiotic bacterium such as Wolbachia, comparing the expression of ankyrin genes to the surface protein encoding gene (wsp) was used to normalize for variation in Wolbachia density but any differences in levels of wsp expression between sexes and stages could be a confounding factor. The wsp gene was previously shown to be expressed in all Cx. pipiens life stages including male and female adults [22].
As some of the ANK proteins have predicted signal peptides and transmembrane domains, it is possible that they are secreted into the mosquito cytoplasm or presented on the surface of the bacterium, which could suggest that they pk2 gene variants/homologs in wPip-infected Cx. pipiens group colonies are involved in Wolbachia's interaction with the host. A proteomics analysis including experiments such as immunolocalisation studies could be used to characterize the function of ANK proteins in Wolbachia. Current limitations to such studies include the difficulty of obtaining epitope specificity and the absence of a transformation system for Wolbachia. As co-regulated genes are highly likely to show functional interactions, studies to examine the role of the co-expressed prophage-associated genes adjacent to pk2 are also needed.
Associations between ANK gene sequence variants and particular crossing types have previously been reported [17], enabling use of these variants as markers to further investigate Wolbachia-induced CI in the Cx. pipiens group. The significance of sex-specific expression patterns in the pk2 genes in some host strains but not others is not yet understood, but its occurrence did not correlate with the crossing patterns described in Table 2. The Mol and Pel colonies are bidirectionally incompatible with each other but both show the same sex-specific expression of the pk2 genes in adult mosquitoes. Given the complexity of the phenotype in the Cx. pipiens group, it seems plausible or even probable that the genetic basis for these crossing type differences involves multiple Wolbachia genes, and factors such as the mosquito nuclear background interacting with Wolbachia variants can also contribute [17]. A hypothesis that variation at just one 'CI gene' could explain all the crossing type variation observed seems increasingly unlikely. Given the rapid evolution of ANK genes, sequence differences at particular ANK loci between crossing types does not necessarily mean that there is a causal link. However the differential expression between sexes of several ANK genes (including non prophage-associated genes) in wPip does provide further support for adaptations to sex-specific interactions with its host.
Conclusion
The number of ANK genes in the wPip genome is the highest yet reported in a prokaryote. The sex-specificity observed in patterns of expression for some of these genes and the differential expression between mosquito strains are also very unusual features, particularly given the generally very high level of sequence conservation between wPip variants. The elucidation of the functional roles and mechanisms of evolution of this family of genes will provide many insights into the biology of reproductive parasites.
Identification of ANK genes and primer design
Putative protein-encoding genes were identified in the wPip genome using ORPHEUS [23], followed by manual curation. The translated gene sequences were searched against the Interpro database using Interproscan [24] in order to locate ankyrin repeats and other protein motifs such as signal peptides and transmembrane domains. The protein sequences containing ANK domains were compared to the protein sequences of Wolbachia strain wMel using blastp in order to identify possible homologs. Identification of a putative origin of replication and the assignment of ANK gene numbers was based on the location of the dnaA gene. Gene specific primers with an annealing temperature ranging between 50-55°C were designed for all unique ANK genes (Additional file 1) using Primer Select 5.06 (DNAstar, Madison, WI, USA) and Primer3 [25]. The unfinished sequence of the wPip genome and the corresponding preliminary annotation of the ANK genes are available from the Wellcome Trust Sanger Institute website, and will be updated as the sequence is completed [26].
Mosquito colonies
Colonies of wPip-infected Cx. pipiens mosquitoes were selected for the study. Table 3 lists the colonies used in addition to the location of where the colonies originated. All mosquito colonies were reared using standard rearing procedures at low larval densities in insectary conditions (26°C and 70% relative humidity) with a 12:12 h light/ dark circadian cycle. Mass crossing experiments between Cx. pipiens colonies were carried out using 50 virgin individuals of each sex. The F 1 generation progeny from the crosses was analysed by calculating the percentage of hatched embryos from a minimum of eight egg rafts, each containing between 50-110 eggs per raft, as a measure of the CI phenotype. Female spermathacae were examined for the presence of sperm if the hatch rate was low to confirm insemination.
PCR
All
RNA extraction
Total RNA was extracted from young (1-2 days post eclosion) adult mosquitoes using Tri Reagent (Sigma-Aldrich) followed by chloroform extraction and isopropanol precipitation. RNA extracts were treated with DNase I (Sigma-Aldrich) to remove any contaminating DNA. As the density of Wolbachia is significantly lower in adult male Culex mosquitoes compared to females, three adult Cx. pipiens male mosquitoes were pooled prior to RNA extraction to increase the amount of Wolbachia RNA present for analysis. RNA extraction of testes was carried out by dissection of 20 Pel colony males under a dissecting microscope in 0.1% saline after immobilising adult mosquitoes on ice. Dissected testes were rinsed in PBS and then pooled in 1.5 mL microcentrifuge tubes in RNAlater (Ambion, Austin, TX, USA) to prevent RNA degradation. The quality and yield of total RNA was measured using a Nanodrop ND 100 spectrophotometer.
RT-PCR
Reverse transcription (RT) PCR analysis was performed using the Qiagen Onestep RT-PCR kit (Hilden, Germany). RNase-free water, Onestep RT-PCR buffer (1×), 400 µM dNTPs and Onestep RT-PCR enzyme mix were combined with gene specific primers (0.6 µM) to amplify 2.0 µL of template RNA in 50 µL reactions. Reverse transcription was carried out at 50°C for 30 min followed by 95°C for 15 min. Samples were PCR amplified by denaturing for 5 min at 94°C, cycled 35 times at 94°C (1 min) variable annealing temperature (1 min) and 72°C (1 min each), followed by a 10 min extension at 72°C using an Applied Biosystems GeneAmp PCR system 2700. A total of 10 µL of RT-PCR products and a 100 base-pair marker (Sigma-Aldrich) was electrophoresed on 1% agarose gels stained with ethidium bromide and visualized under ultraviolet illumination. To examine for false positives that might result from amplification of DNA, parallel reactions without adding the reverse transcriptase (Taq polymerase only, Sigma-Aldrich) to the reaction mixture were included.
Quantitative RT-PCR
Quantification of gene expression was carried out using the Qiagen Onestep SYBR green RT-PCR kit and the Opticon 2 Continuous Fluorescence Detection System Primers were designed to amplify ANK gene fragments of less than 250 bp. Standard curves were produced using serial dilution of RNA extracted from adult female mosquitoes and relative male RNA extract expression of ANK genes measured in comparison. Quantitative RT-PCR cycling conditions were 50°C for 30 min followed by 95°C for 15 min. Samples were cycled 40 times at 94°C (15 s), 55°C (30 s) and 72°C (30 s) followed by a read step. A melting curve was constructed between 50°C and 90°C. Quantitative RT-PCR assays were carried out on six male RNA extracts in two separate assays. Comparing the concentration of cDNA amplified from ankyrin genes to the wsp gene was used for normalization of the data, to control for both differences in extraction efficiency and also the higher Wolbachia density that occurs in adult female mosquitoes compared to males. The mean relative expression levels of the wsp gene in Pel males, used to normalize for differential Wolbachia density in individual adult mosquitoes, was found to be 44.2 ± 9.6% compared to expression levels in Pel female RNA extracts. | 2014-10-01T00:00:00.000Z | 2007-09-20T00:00:00.000 | {
"year": 2007,
"sha1": "417b5c195bc223390620422568c7019f95ef7159",
"oa_license": "CCBY",
"oa_url": "https://bmcbiol.biomedcentral.com/track/pdf/10.1186/1741-7007-5-39",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "f3b22fa9ff38dc587c175716b983ee73a4e3d187",
"s2fieldsofstudy": [
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
125825313 | pes2o/s2orc | v3-fos-license | XIMPOL: a new x-ray polarimetry observation-simulation and analysis framework
We present a new simulation framework, XIMPOL, based on the python programming language and the Scipy stack, specifically developed for X-ray polarimetric applications. XIMPOL is not tied to any specific mission or instrument design and is meant to produce fast and yet realistic observation-simulations, given as basic inputs: (i) an arbitrary source model including morphological, temporal, spectral and polarimetric information, and (ii) the response functions of the detector under study, i.e., the effective area, the energy dispersion, the point-spread function and the modulation factor. The format of the response files is OGIP compliant, and the framework has the capability of producing output files that can be directly fed into the standard visualization and analysis tools used by the X-ray community, including XSPEC which make it a useful tool not only for simulating physical systems, but also to develop and test end-to-end analysis chains.
-Introduction
XIMPOL is a new simulation framework specifically designed for X-ray polarimetry. It is based on the python programming language and the associated scientific ecosystem and it is publicly available under the GNU General Public License [1]. XIMPOL is able to produce realistic observation-simulations given as inputs a source model and the detector response functions. The generated output FITS files can be analyzed with standard analysis tools, such as XSPEC [2], and HEASARC ftools [3]. In addition, dedicated tools for the analysis of polarimetric informations are directly available within XIMPOL.
In this contribution we will give an overview of the basic architecture of the software and, although the framework is not tied to any specific mission or instrument design, we will present a few physically interesting case studies in the context of the the X-ray Imaging Polarimetry Explorer (XIPE ) mission [4], currently in phase study for the M4 ESA call.
-Architectural overview
The vast majority of the simulation and data preparation algorithms implemented in XIMPOL are available in four main executables, as illustrated in the block diagram of fig. 1. These consist in: • xpobssim: given a source model, a set of instrument response functions, and a given observation time, it produces a photon list corresponding to a given observation time in the form of a binary FITS table.
• chandra2ximpol : given a Chandra event list and a set of instrument response functions, it converts the Chandra observation into a XIMPOL observation-simulation.
• xpselect: it allows to select subsamples of photons in a given event file, based on the event energy, direction, time or phase, producing a new (smaller) event file.
• xpbin: it allows to bin the data using several different algorithms, producing counts maps and spectra, light curves, phasograms and modulation cubes (i.e., histograms of the measured azimuthal distributions in multiple energy layers).
Where applicable, the data formats are consistent with the common display and analysis tools used by the community, e.g., the binned count spectra can be fed into XSPEC, along with the corresponding response functions, for standard spectral analyses. All the simulation and analysis tools are fully configurable via command-line interface. Also, XIMPOL provides a "pipeline" facility to easy concatenate all the aforementioned functionalities in python scripts.
-Source model definition
The basic setup for a generic simulation is specified via a python configuration file. In order to define a source model, three functions are needed: 2) Polarization degree: P (E, t, ra, dec).
3) Polarization angle: φ(E, t, ra, dec). There is a great flexibility in defining these functions: they can be defined analytically (e.g., constants, power-laws,. . .), can be the result of interpolations built from a series of data points or, for the energy spectrum, can be a XSPEC spectral model. The source morphology can be defined as a variety of predefined structures: point sources, Gaussian or uniform disks, or an arbitrary extended sources (from FITS images). In addition, XIMPOL fully supports phase-dependent periodic sources and an arbitrary number of components can be easily overlaid in the same input model. For example, fig. 2 shows the Chandra image used to define the morphology of the source (Cas A, in this case) and the simple polarization pattern used in the simulation. Figure 3 shows the spectral model used. In this particular simulation, we have defined a thermal non-polarized spectral component overlaid with a polarized non-thermal component.
-Observation-simulation
The main purpose of the observation-simulation tool, called xpobssim, is to simulate a realistic observation of a source taking as inputs the source model and a set of suitable detector response functions. In the simulation work-flow, the first step consists in calculating the expected number of events. This is done convolving the provided source spectrum with the effective area of the instrument and integrating the curve in energy and in time. Then, using the count spectrum and the light curve as one-dimensional probability density functions, the event times (or phase) and the true energies are randomly sampled. The energies are also smeared with the energy dispersion. According to the source morphology and the instrument point spread function (PSF), the incoming photons direction in the sky is determined ( fig. 4, left panel). In the last step, the tool generates the angular distribution of the photoelectron emission angles based on the given polarization model ( fig. 4, right panel). The output of the simulation is a photon list that is saved in a FITS file containing all the information about the photons.
-Chandra2ximpol converter
The chandra2ximpol converter is a tool alternative to xpobssim, specifically designed to simulate a source starting from Chandra observations (for example, fig. 5, left panel). This technique has the advantage to preserve the full correlation between the source morphology and the spectrum, especially important for extended sources. The starting point of the conversion is the Chandra photon list, publicly available from the on-line database [5], which includes energies and spatial information of every event. The Chandra angular and energy resolutions are much better than those expected for the XIPE mission (angular resolution: ∼ 1 arcsec vs. ∼ 30 arcsec, energy resolution: 5% vs. ∼ 20% at 5.9 keV [7]), so that we can take the Chandra measured energies and positions as Monte Carlo truth. Based on the ratio of the effective areas and the given observation time, the events are down-or over-sampled and then smeared with the response functions (in this case XIPE : fig. 5, right panel). Finally, the emission angle is generated accordingly with the specified polarization model and the output event list is saved in a FITS file.
-Analysis tools
XIMPOL also comes with a basic set of command-line analysis tools. These allow the user to select subsamples of photons based on event energy, direction, time or phase (xpselect), or bin and fit the simulated data, producing count maps, spectra, phasograms, light and modulation curves (xpbin). The resulting spectra and the response functions are compatible with XSPEC, which can be therefore used to perform spectral analysis. All the tools, including XSPEC, can be seamlessly combined into analysis pipelines. As an illustrative example, fig. 6 shows the results of a 100 ks XIPE simulation of an observation of the Crab nebula (plus pulsar), based on [8], with superimposed the polarization model given as input. These figures show the result of the complete simulation-analysis pipeline. In each pulsar-phase bin, the data have been analyzed and the polarization degree and angle have been obtained. The comparison with the input model is also shown.
-Conclusions
In summary, XIMPOL is a newly developed observation-simulation and analysis framework that allows the user to easily produce fast and realistic simulation of astronomical sources as well as analyze and display the simulation results. The entire software package is not tied to any specific mission and the user simply provides the instrument response functions to use. Thanks to its capabilities and flexibility it can play a relevant role in a context of growing interest for the X-ray polarimetry. | 2019-04-22T13:07:43.230Z | 2016-07-01T00:00:00.000 | {
"year": 2017,
"sha1": "334ed3a7e0043d49b488a68578502341a81b0298",
"oa_license": "CCBY",
"oa_url": "http://eprints.bice.rm.cnr.it/19924/1/ncc11456.pdf",
"oa_status": "GREEN",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "59a1e8ae7d1fabc5576f481cc97ebc55dd5df766",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Engineering",
"Physics"
]
} |
103476681 | pes2o/s2orc | v3-fos-license | Correlation of concentration of modified cassava flour for banana fritter flour using simple linear regression
The purpose of this study was to determine the correlation of different concentrations of modified cassava flour that was processed for banana fritter flour. The research method consists of two stages: (1) to determine the different types of flour: cassava flour, modified cassava flour-A (using the method of the lactid acid bacteria), and modified cassava flour-B (using the method of the autoclaving cooling cycle), then conducted on organoleptic test and physicochemical analysis; (2) to determine the correlation of concentration of modified cassava flour for banana fritter flour, by design was used simple linear regression. The factors were used different concentrations of modified cassava flour-B (y1) 40%, (y2) 50%, and (y3) 60%. The response in the study includes physical analysis (whiteness of flour, water holding capacity-WHC, oil holding capacity-OHC), chemical analysis (moisture content, ash content, crude fiber content, starch content), and organoleptic (color, aroma, taste, texture). The results showed that the type of flour selected from the organoleptic test was modified cassava flour-B. Analysis results of modified cassava flour-B component containing whiteness of flour 60.42%; WHC 41.17%; OHC 21.15%; moisture content 4.4%; ash content 1.75%; crude fiber content 1.86%; starch content 67.31%. The different concentrations of modified cassava flour-B with the results of the analysis provides correlation to the whiteness of flour, WHC, OHC, moisture content, ash content, crude fiber content, and starch content. The different concentrations of modified cassava flour-B does not affect the color, aroma, taste, and texture.
Introduction
Indonesia is one of the tropical countries. The country of Indonesia has abundant natural wealth, especially on local food crops of tubers. One of the most recognized tuber crops is cassava. Utilization of cassava is still very limited. Therefore, a series of studies are required to increase the potential of cassava as an alternative food source of carbohydrate tubers that are in demand by the community.
Program and activity of agricultural development of food crops in West Java Province in 2015-2019 is -Pajale‖ is one program that aims to develop cultivation of crops, one of which is cassava. Development sites will be conducted in several areas such as in Bandung, Bogor, Cianjur, Subang, Sukabumi, Sumedang, and Tasikmalaya.
Subang regency produces superior varieties of cassava. One of them is in Gandasoli Village, Tanjung Siang District. This region is one of the potential areas of cassava producers, and one of the most sought after varieties is manggu. The average of Tanjung Siang-Subang district's produce reaches 114.400 quintals per year [1]. Utilization of cassava into flour can increase the economic value of cassava. Cassava flour can be processed into a variety of processed food products one of them as a mixture of coated flour or seasoned flour for fried products. The National Standarization Agency defines seasoned flour as a food ingredient in the form of a mixture of flour and spices with or without the addition of other foodstuffs and allowable food additives [2].
The modified starch is a particular treated flour to provide better properties for improving the prior nature, particularly its physicochemical and functional properties or to alter some other properties [3]. The food industry has now begun to utilize the use of modified flour as a food aid for certain food products. The addition of modified flour to food products can enhance the superiority of the quality of food products. The advantage of modified flour that can make the product more crispy, better in terms of mouthfeel, color and flavor when compared with traditional ingredient products such as insoluble fiber [4].
The purpose of this research is to know the different types of processed cassava flour (cassava flour, modified cassava flour-A, and modified cassava flour-B) and the exact modified cassava flour of concentration against banana fritter flour which are expected to have color, aroma, taste, and texture favored by consumers.
Materials, chemicals, and tools
The study is conducted at the Laboratory of Post-Harvest Development, Center for Appropriate Technology Development, Indonesian Institute of Sciences, Subang. The materials used in this study are cassava flour, modified cassava flour-A (using the method of the lactid acid bacteria), and modified cassava flour-B (using the method of the autoclaving cooling cycle), derived from cassava Manggu variety with harvesting age of 9 months, obtained from cassava farmer in the village district of Gandasoli, Tanjung Siang, Subang district. Supporting ingredients include: sago flour, corn starch, sugar, salt, baking soda, gum arabic, vanilla, kepok banana, and coconut oil. The chemicals used for chemical analysis are 1.25% H 2 SO 4 , 3.25% NaOH, 95% ethanol, distilled water, Luff Schoorl's Solution, HCl 9.5 N, concentrated HCl, universal pH indicator, H 2 SO 4 6 N, KI, Na 2 S 2 O 3 , and starch.
The process of making for banana fritter flour through the stages: weighing the material according to the formulation (50 g types of flour, 20 g sago flour, 15 g corn starch, 10 g sugar, 1.2 g salt, 0.7 g baking soda, 0.9 g gum arabic, 2.2 g vanilla), mixing ingredients, applying to fried bananas, mixing with water, banana coating, frying, and incision.
Main research is the determination of different concentrations of the selected flour to be used as banana fritter flour. Then performed the analysis of WHC, OHC, whiteness of flour, moisture content, ash content, crude fiber content, starch content [5], and organoleptic test [6] applied to banana fritter flour. The process of making for banana fritter flour through the stages: weighing the material according to the formulation (selected flour 40%; 50%; 60%), sago flour 20%, corn starch 15%, sugar 10%, salt 1.2%, baking soda 0.7%, gum arabic 0.9%, vanilla 2.2%), mixing of ingredients, applying to fried bananas, mixing with water, banana coating, frying, and incision.
Treatment and data analysis
The treatment design used consisted of two variables, those are the independent variable and the dependent variable. The independent variable (X) of this experiment consisted of flour concentration with three levels: (X 1 ) 40%, (X 2 ) 50%, and (X 3 ) 60%. The dependent variable (Y) of this experiment consisted of seven treatment methods (Y 1 ) WHC, (Y 2 ) OHC, (Y 3 ) whiteness of flour, (Y 4 ) moisture content, (Y 5 ) ash content, (Y 6 ) crude fiber content, and (Y 7 ) starch content.
The research design model used is simple linear regression with two variables measured, ie selected flour concentration to: WHC, OHC, whiteness of flour, moisture content, ash content, crude fiber content, and starch content. Furthermore each data obtained results plotted to the curve, then performed a simple linear regression analysis to determine the relationship between variables measured by processing methods, the equation is: Y = a + bX Y = Concentration of selected flour a = Constants b = Regression coefficient (increase if positive value or decrease if negative value) X = Measured variable (WHC, OHC, whiteness of flour, moisture content, ash content, crude fiber content, and starch content). Physical responses include: water holding capacity (WHC), oil holding capacity (OHC), and whiteness of flour. Chemical responses include: moisture content, ash content, crude fiber content, and starch content. Organoleptic response using hedonic method [6]. Assessment is given to the color, aroma, taste, and crisp texture of 30 somewhat trained panelists. Quantitative qualitative data obtained are performed using numerical scales: (1) strongly dislikes, (2) somewhat dislikes, (3) dislikes, (4) rather likes, (5) likes, (6) very likes.
Preliminary research result on organoleptic response
Banana fritter flour is made from different types of flour but uses the same concentration, then organoleptic testing on color, aroma, taste, and texture. Results on organoleptic response from the panelists were presented in Table 1. The result of the test of the color of banana fritter flour get a high value of 4.76 from sample modified cassava flour-B, with the response from the panelist stated rather like. The color produced from the banana fritter flour as a coating is influenced by the banana color used, resulting in a non enzymatic browning reaction with maillard reaction. When aldose /ketose is exposed to heat and reacts with the amino group, there is the production of various components such as flavour and dark-colored polymers [7]. According to [8], the reaction of carbohydrates, especially reducing sugar with the primary amino group, produces a brown material, for example in the process of frying banana fritter flour.
The result of organoleptic tests on the aroma of banana fritter flour showed a high value of 5.22 from sample modified cassava flour-B, with a response from the panelists states like. The aroma contained in the banana fritter flour is affected by the distinctive aroma of vanilla addition. Aroma of food is caused by the formation of volatile compounds [9] from vanilla.
The result of organoleptic test on banana fritter flour taste showed a high value 4.98 from sample modified cassava flour-B, with criterion of response from panelist states rather like. The taste found in banana fritter flour is influenced by the addition of sugar, which provides a savory taste and the use of oil in the frying process. The taste is influenced by aromas, foodstuffs, crispness, and food maturity levels [9].
The result of organoleptic test on banana fritter flour texture shows a value of 5.19 from sample modified cassava flour-B, with criteria of response from panelists express likes. Crispy texture on bananas fritter flour is influenced by the addition of corn starch, so banana fritter flour tend to be more crunchy and easily broken when bitten. In line with the results of research [10] that in the process of making wet noodles by using modified cassava flour, need the addition of corn flour to get a good noodle texture.
The result of organoleptic testing showed that the selected type of flour is modified cassava flour-B (using the method of the autoclaving cooling cycle), the results of physicochemical analysis of modified cassava flour-B are presented in Table 2 below.
Main research results against physicochemical response of banana fritter flour
The results of physicochemical analysis of different concentrations of modified cassava flour-B to be used as banana fritter flour are presented in Table 3 below. WHC is used to measure the ability of flour to retain water absorbed, the WHC value is influenced by the water content in the food. Water added to flour for doughs of coating, will affect physical properties and processing. The absorption of water in flour is also influenced by the size and structure of the starch granules, wherein the smaller starch granules will increase solubility and increase water absorption [12]. An increase in WHC value in the use of more concentration of flour (60%), because with the increasing use of modified cassava flour, the moisture content of banana fritter flour will be lower, causing the flour to absorb more water so that the water holding capacity also higher. The ability of flour to absorb and retain water is not only influenced by the moisture content in the ingredients, it is also influenced by several factors, those are the amylose content, the size of the starch granules, and the fat content of the ingredients. Water absorbed in starch molecules is caused by the physical properties of the granules as well as intramolecularly bound [13].
The result of linear regression analysis shows the correlation of the difference of flour concentration to WHC from banana fritter flour, which can be seen in Figure 1
Oil holding capacity.
OHC is used to measure the ability of flour to retain the oil it absorbs. This ability is determined by the presence of fat and fiber content. Fat can form a layer that is hydrophobic on the surface of the fiber network while the fiber has the ability to absorb oil. The low fat content in the flour will make the flour absorb more oil from the outside. High fiber content in flour will make the flour has the ability to absorb and retain more oil.
Oil holding capacity is due to oil trapped in a porous starch matrix or in the structure of the amylose helix or amylopectin due to the formation of the amylose-lipid complex. The changes of hydrophobic groups are more amyloid-lipid complex due to high temperature [14]. The decline in OHC value in the use of more concentration of flour (60%), this occurs because the ability of banana fritter flour to absorb and retain oil is affected by the frying process. Flour that has a greater OHC value will absorb more oil and hold the oil used for frying, this causes the cooking oil used quickly run out.
The result of linear regression analysis showed the correlation of different concentration of flour to OHC from banana fritter flour, which can be seen in Figure 2 The perfect linear relationship with slope is indicated by a negative slope (-0.9907), meaning that the higher the flour concentration, the lower the resulting OHC value. The results of the 60% flour concentration analysis method resulted in lower average OHC value compared with 40% concentration resulting in a high average OHC. According to [15 and 16] the ability to absorb high oils on flour shows the flour has a part that is lipophilic.
Whiteness of banana fritter flour.
The whiteness of flour is strongly influenced by the process of starch extraction, the more pure starch extraction process then the resulting flour will be white [17]. The white density of modified cassava flour as raw material is 60.42%, after going through the formulation process with the supporting material to become banana fritter flour has increased the whiteness of flour, which can be seen in Table 3.
The result of linear regression analysis shows the correlation of the difference of flour concentration to the whiteness of banana fritter flour, which can be seen in Figure 3
Moisture content.
Water is a major component of food, which plays an important role in determining the various reactions and quality of food. The moisture content in food is related to food damage by microorganisms [18]. High moisture content in flour product will greatly disrupt the stability of the product, which may result in flour agglomeration during storage [19]. The result of the 60% flour concentration analysis method resulted in average low moisture value compared with a concentration of 40% resulting in a high average moisture content. This is influenced by the addition of modified cassava flour (4.4% moisture content) added to the formulation. The increasing concentration of modified cassava flour used, the lower the moisture content of the banana fritter flour formulation. According to [20] that increasing of moisture content can cause the product loss of crispness.
Ash content.
The content of starch ash is related to the mineral content in it. Ash content is strongly influenced by the type of materials, age of materials, and others. According to [21], the content of flour ash varies from 0.30 -1.40% (ww). The greater the ash content of a food, the greater the mineral content contained in the food. In the formulation of banana fritter flour is added salt and baking soda which is inorganic salt, so it is still left as ash when it is done. The result of linear regression analysis shows the correlation of flour concentration to ash content of banana fritter flour, which can be seen in Figure 5.
The correlation between different flour concentration on banana fritter flour processing, shown by r value of linear regression equation, is Y = 4.59+(-0.043)x with correction coefficient (r) is 0.9613. This shows that between different flour concentration and ash content on banana fritter flour have a very strong correlation, so that the concentration of flour from banana fritter flour affects the content of ash produced, because if the correlation coefficient value of 0.8-1 indicates correlation very strong.
The perfect linear relationship with slope is indicated by a negative slope of -0.9613 meaning that the higher the concentration of flour the lower the ash content of the flour produced. The result of the 60% flour concentration analysis method yielded an average low ash content value compared to a concentration of 40% resulting in a high average ash content. This is influenced by different concentrations in the addition of modified cassava flour (ash content of 1.75%). The addition of salt and baking soda which is inorganic salt does not give a real effect, because it is added at the same concentration. The addition of sago flour containing ash content of 4.86% in banana fritter flour formulation has an effect on ash content. Due to the lower concentration of modified cassava flour, the addition of sago flour is higher. [22]. The crude fiber is determined from the residue after the food is treated with acids and strong bases. The result shows that with the increasing concentration of modified cassava flour, the crude fiber content is increasing. The result of linear regression analysis shows the correlation of the difference of flour concentration to the crude fiber content of banana fritter flour, which can be seen in Figure 6. The correlation between different flour concentration on banana fritter flour processing, shown by r value of linear regression equation, ie Y = (-0.23+0.047x) with correction coefficient (r) is 0.9999. This shows that between different flour concentration and crude fiber content in banana fritter flour has a very strong correlation, so that the concentration of flour from the banana fritter flour affects the crude fiber content produced, because if the correlation coefficient value is 0.8-1 shows very strong correlation.
The perfect linear relationship with slope is indicated by a positive slope of 0.9999 meaning that the higher the concentration of flour the higher the crude fiber content produced. The result of the 60% flour concentration analysis method yielded a mean value of high crude fiber content compared with a concentration of 40% yielding a low average crude fiber content. This is influenced by different concentrations in the addition of modified cassava starch (1.90% fiber content).
Starch content.
The composition of starch is an important factor in determining the texture and characteristics of the coating flour. Starch is composed of two main fractions, those are amylose and amylopectin [23]. Amylose plays a role in the absorption of oil during frying, high flour amylose content can be used to reduce oil absorption due to its ability to form film [24].
The The perfect linear relationship with slope is shown with a positive slope of 0.9588 means the higher the concentration of flour, the higher the starch content produced. The result of the 60% flour concentration analysis method yielded an average high starch content value compared to a concentration of 40% resulting in a low average starch content. This is influenced by different concentrations in the addition of modified cassava flour (67.31%), and addition of corn starch containing 88% starch content. According to [25] starch with high amylopectin content tend to give fragile and crispness product characters, while amylose give of texture and violence more resistant. Banana fritter flour color after the coating process on banana of kepok and frying process has a golden yellow color. The result of organoleptic testing on the color of banana fritter flour from sample modified cassava flour B (60%), gets value of 4.49, with the response from the panelists say rather like. Different concentrations of modified cassava flour have no effect on color parameters. According to [26] decreasing of brightness and brown color in product due to the frying process.
The aroma found in banana fritter flour is influenced by the distinctive aroma that arises from banana and the addition vanilla of banana fritter flour formulation. The result of organoleptic testing on the aroma of banana fritter flour from sample modified cassava flour B (60%), gets value of 5.22, with the response from the panelists say like. Different concentrations of modified cassava flour have no effect on flavour parameters. According to [27] vanilla aroma is included into a group of volatile compounds that are not heat resistant.
The result of organoleptic testing on the taste of banana fritter flour from sample modified cassava flour B (60%) gets value of 4.33, with the response from the panelists say rather like. Different concentrations of modified cassava flour have no effect on taste parameters. Sweet and savory of banana fritter flour flavors are influenced by the addition of sugar in the formulation of banana fritter flour and the use of cooking oil in the frying process. When flour is fried then the water molecules will evaporate and be replaced by oil that makes air cavities in the food, so resulting in the product of swelling and crunchy [28] The result of organoleptic testing on the texture of banana fritter flour from sample modified cassava flour B (60%) gets the value of 4.41, with the response from the panelists say rather like. Different concentrations of modified cassava flour have no effect on texture parameters. The crunchy texture of banana fritter flour is influenced by the addition of corn starch and baking soda (NaHCO 3 ) to banana fritter flour formulation. According to [29] corn starch when the frying process provides a more crunchy texture and easily broken when bitten, corn starch is recommended for use as a coating material (fried products). The addition of baking soda will make the banana fritter flour expand and increase the crispness, because at the time of heating/frying, baking soda will release carbon to form a fragile and crunchy structure [19]. | 2019-04-09T13:07:38.450Z | 2017-12-01T00:00:00.000 | {
"year": 2017,
"sha1": "741a5f9b842414851878849352f6e708adde680a",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1755-1315/101/1/012022",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "4f0cb4442badfbd23eca5a1ba2b9a93070f5caf3",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Physics",
"Mathematics"
]
} |
257780191 | pes2o/s2orc | v3-fos-license | Opportunistic Schedule Selection for Multiuser MIMO FSO Communications: A Security- Reliability Trade-Off Perspective
The security and reliability of free space optical (FSO) communications and the trade-off between the two are the most critical characteristics to emphasize in FSO systems, especially as optical wireless communications continue to evolve. In this work, we investigate the impact of opportunistic transmit aperture selection (TAS) on the security-reliability trade-off (SRT) of the multiuser multiple-input multiple-output (MU-MIMO) free space optical system. We consider a general scenario in which the FSO system consists of a transmitter side with <inline-formula><tex-math notation="LaTeX">$K$</tex-math></inline-formula> users and a single receiver with <inline-formula><tex-math notation="LaTeX">$N$</tex-math></inline-formula> apertures communicating over generalized Malaga-M turbulence channels. We derive the TAS combined signal-to-noise ratio (SNR) distributions for the FSO MIMO channel between the transmitting users and the receiver. A statistical analysis of the TAS scheme is performed, where the <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula>th best user with the highest SNR is selected. Based on this analysis, the closed-form expressions for the outage probability, average bit error rate, intercept probability, and SRT are derived. Then, the diversity advantages provided by the channel variations considering the atmospheric turbulence and generalized nonzero boresight pointing errors are further discussed. The results show that the considered TAS significantly improves the security, reliability, and SRT performance and has the potential to solve existing FSO communication problems. Monte Carlo simulations are used to verify the correctness of the numerical results.
I. INTRODUCTION
G LOBAL mobile traffic is expected to increase 670 times by 2030, driven by advanced technological devices and real-time applications with massive bandwidth [1]. In this way, security issues in wireless communication systems have attracted considerable attention at present and in the future. Moreover, this great interest in security issues of communication networks is expected to increase in the coming years with the emerging implementation of 5G and beyond networks. Secure transmission in the presence of an external eavesdropper is a significant research problem in communication networks. In this scenario, legitimate users want to send confidential messages to the legitimate recipient without being overheard by the eavesdropper [2]. Compared to the already congested radio frequency (RF) spectrum, free space optical (FSO) communication enables license-free operation, low costs, and high-data-rate transmission, all of which are critical in various application scenarios. The absence of licensing fees and regulatory restrictions on bandwidth consumption highlights the effectiveness of FSO communications as a backup solution for disaster recovery and military applications [3]. FSO systems initially attracted attention as an efficient solution to the "last mile" problem, bridging the gap between the end user and the existing fiber optic infrastructure. Connectivity between buildings, video surveillance, backhaul for cellular networks, IoT, and satellite communications are some attractive applications of FSO communications [4]. For this reason, there is growing interest in privacy and security issues of FSO communication systems in the presence of an external eavesdropper that can extract information from the legitimate transmission. In general, FSO is more secure than RF communications due to the high directionality of the laser beam. However, according to recent research on wireless optical systems, FSO communications are not interception-free at the physical layer, especially when the laser beam's main lobe is significantly larger than the receiver's size [5], [6], [7]. A plausible mechanism for interception is the reflection of some of the beam radiation by small particles and subsequent detection by an eavesdropper not in the line-of-sight (LOS) of both communication peers. In another scenario, the eavesdropper would block the laser beam to collect some of the optical energy. Since the laser beam diverges due to optical diffractions, one strategy for successful eavesdropping is to position the eavesdropper This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ in the beam's divergence region. Regarding long-distance FSO communication, the eavesdropper can intercept the FSO link better by collecting the power that legitimate peers do not capture [2]. An optical beam transmitted with a small divergence angle of 1 mrad might have a radius of 1 m over a 1 km transmission distance. Consequently, physical layer security (PLS) is emerging as a promising secure wireless communications paradigm relying on exploiting the physical characteristics of wireless channels for protection against eavesdropping attacks. Contrary to the cryptographic techniques with highly complex encryption methods, which require more processing resources for encryption and decryption and increase the latency imposed, PLS takes advantage of the inherent randomness of wireless channels to secure communications between legitimate peers [8], [9]. PLS can be considered the most secure method of communication since security in this context is provable (in the information-theoretic sense). For example, in a typical wiretap model, PLS theory states that secure communication is possible if the capacity of the legitimate channel is higher than that of the eavesdropping channel. Motivated by the PLS schemes of RF systems, various PLS techniques have been proposed to secure FSO communications, such as diversity approaches [10], [11], transmit aperture selection (TAS) [12], artificial noise injection [13], and generalized chaotic modulation [14]. However, many characteristics of FSO systems differ from RF systems (e.g., the channel, and the physical characteristics of the transmitting and receiving devices). Therefore, these differences must be taken into account when adapting PLS techniques to FSO systems. A number of studies have considered these unique properties. For example, Lopez-Martinez et al. [15] characterized the PLS in single-input single-output (SISO) FSO transmissions in the presence of an eavesdropper. Due to the effects of both laser beam divergence and turbulence-induced fading, they discussed possible eavesdropping mechanisms. They concluded that the eavesdropper can interfere with communications when it is close to the legitimate receiver or transmitter. Saber and Sadough [16] evaluated the secure performance of the SISO FSO link over Malaga-M turbulence channels. Later, Verma et al. [17] evaluated the impact of nonzero boresight pointing errors on secrecy performance for the same system in [16]. Furthermore, Monteiro et al. evaluated the effective secrecy throughput (EST) of an FSO system consisting of a multiple-aperture transmitter, a multiple-aperture receiver, and a multiple-aperture eavesdropper in [18]. They showed that the use of multiple apertures on the transmitter is crucial to approach the optimal EST. Moreover, Han et al. in [19] analyzed the secrecy performance of the SISO FSO link over the F-distribution turbulence with pointing error effects.
On one hand, there is no doubt that the reliability of FSO systems is vulnerable to path loss, misalignment, and atmospheric turbulence, which can seriously limit FSO transmissions at shorter distances [20]. On the other hand, despite capacity increments created by new technologies including FSO, mobile operators still struggle to meet user demands. A viable alternative to bandwidth expansion is to allocate the available bandwidth optimally, or at least more efficiently. The basic idea of opportunistic scheduling selection (OSS) is to exploit channel variations to schedule a user with the best channel conditions at a given time [21]. The contemporary standard that employs OSS includes WiMAX [22]. OSS is the function responsible for resource allocation between users, and its role is to decide which user should transmit/receive and when, and therefore affects the efficiency of bandwidth utilization. The gain is primarily due to multiuser diversity, which views channel fading as an opportunity rather than an obstacle: When selecting one user out of many, it is almost always possible to select one that is above the "average." Therefore, users in fading areas are typically not served [21]. An example of an opportunistic method is the transmit aperture selection, which is based on selecting a path with the highest value of irradiance or fading gain to counteract fading and improve performance [23].
Additionally, it is shown in several studies that as the intercept probability (P int ) (which considers a system's security metric) requirement is relaxed, the outage probability (P out ) (considered as reliability metric) performance improves, and vice versa [24], [25]. This implies a trade-off between the security and reliability of the wireless transmission in the presence of eavesdropping attacks, which is referred to as the security-reliability trade-off (SRT). Although the notion of SRT was studied in the context of FSO transmission [26], [27], these contributions were mainly focused on the employment of encryption algorithms to protect against eavesdropping attacks. By contrast, in our work, PLS-rather than encryption techniques -is invoked for characterizing the SRT attained in MIMO environments with the help of OSS schemes for enhancing the multiuser FSO system performance.
A. Related Work
Next, we are going to elaborate on the applications of opportunistic scheduling selection in FSO communication systems according to several fields that researchers have identified. These fields fall broadly into the following categories: 1) Diversity Techniques: Scanning the literature, there have been studies on employing opportunistic selection scheduling to improve the FSO systems' diversity gain and reduce the influence of wireless channel impurities. In this context, Abouei and Plataniotis investigated several OSS schemes for varying degrees of the trade-off between throughput and fairness between users across a log-normal (weak) turbulence channels for multiuser MIMO (MU-MIMO) FSO system assuming nonequidistant user placement [28]. Yang et al. evaluated the performance of a multiuser single-input multiple-output (MU-SIMO) FSO system with the best user selection (BUS) scheme over weak (log-normal) and strong (Gamma-Gamma) turbulence channels [29]. They do, however, assume a constant average signal-to-noise ratio (SNR) for all users. This implies that path loss effects are either disregarded or that each user is positioned equidistant from the central node, resulting in identical path loss. Such an assumption is unrealistic because users in a real-world system are likely to be spread out over a broad area. Furthermore, unlike RF systems, the fading variance in FSO systems is distance-dependent [12]; hence user placement has a significant influence on multiuser system performance.
Zhalehpour and Uysal developed an accurate analysis for the outage capacity of several scheduling schemes including in an MU-MIMO FSO system over a weak turbulence channel [30]. Later, they extended the analysis to the outage throughput of the system [31] considering the same channel model in [30]. However, the consequences of path loss and pointing errors are ignored in [28], [29], [30], and [31].
2) Reliability Performance Enhancement: Another group of studies investigated OSS for improving the FSO systems' reliability [32], [33], [34], [35], [36], [37], [38], [39], [40]. Qin et al. analyzed the outage probability of the MU-MIMO FSO transmission over Gamma-Gamma atmospheric turbulence using repetition coding (RC) and transmit laser selection (TLS) schemes [32]. To circumvent this, the end-to-end SNR of the MIMO FSO channel is approximated since the exact closedform expression for such a channel is remarkably complex and not easily tractable due to the complication of its statistics. This approximation method was first considered in [41] to approximate the MIMO FSO transmission over the Gamma-Gamma environment. Cherif et al. examined the effect of interference on the performance of multi apertures-multiuser (MA-MU) mixed FSO/RF relay networks with Malaga-M/Nakagami-m distributions [33]. To select between FSO apertures at the relay node, transmit aperture selection scheme was used. While the BUS was used on the RF link to select between users over the RF link. However, the pointing errors and path loss did not consider in [32], [33], and [41]. Salhab et al. employed MU-MISO mixed RF/FSO relay network with a BUS scheme in RF link to enhance the system reliability over Rayleigh/Gamma-Gamma fading channels [34]. Similar studies for mixed RF/FSO relay systems can be found in [35], [36], [37], [38], [39]. The Gamma-Gamma distribution characterized the SISO FSO link in [34], [35], [36], [37], [38], [39], while the Malaga-M distribution is chosen to represent the multi aperture FSO links with asymptotic analysis in [33]. Michailidis et al. introduced a hovering unmanned aerial vehicle (UAV) that served as a decode and forward relay (DF) between the ground central unit (CU) and multiple ground users (GU) for mixed FSO/RF system over log-normal/Nakagmi-IG fading channels in another study [40]. For the FSO, the TAS scheme is explored for opportunistic selection, while opportunistic GU scheduling (GUS) is used in the RF links to improve the system's outage probability [40].
3) Security Performance Enhancement: An additional group of studies explored the OSS for improving the security performance of the mixed RF/FSO and FSO systems against eavesdropping attacks [12], [42], [43], [44]. For example, Abd El-Malek et al. investigated the security of MU-SIMO mixed RF-FSO relay networks by employing the BUS scheme in the multiuser RF link [42]. They also looked at the effect of eavesdropper attacks in RF links on the security of the system. Later, in another study, they evaluate the influence of RF cochannel interference (CCI) on the SRT for the same system [43]. Additionally, they employed power allocation and cooperative jamming techniques to enhance the system's security performance [43]. However, the SISO FSO link with Gamma-Gamma distribution is considered as an extension for the multiuser RF link in all these studies [42], [43], [44]. Finally, Shakir [12] examined the security performance of the MISO FSO networks adopting the BUS scheme. The Malaga-M turbulence is used to characterized the FSO channel with the same approximation method for the channel model adopted in [33].
Analyzing the above and related literature, we find some research gaps in the study of the security-reliability trade-off of the FSO system based on OSS, which need to be addressed. Firstly, the pointing error loss was ignored or limited to zero boresight in FSO transmission. The nonzero boresight model generalizes the effect of pointing errors on the FSO transmission. However, although the statistical analysis of nonzero boresight pointing errors for the FSO system has been investigated extensively in the literature, it has yet to be addressed in the context of SRT for the MU-MIMO FSO systems.
Second, for atmospheric turbulence, lognormal models for weak turbulence or Gamma-Gamma models for strong and moderate turbulence were mostly used. It is worth noting that the SISO FSO transmission utilizing the general Malaga-M distribution to characterize atmospheric turbulence, has been intensively studied in many works in the literature to evaluate the performance of the FSO system over a wide range of turbulence conditions [45], [46]. However, very few works [12] have used the Malaga-M distribution in the multiuser FSO environment in the context of SRT due to the high complexity of the mathematical formulation of such systems.
According to the Malaga-M model, the small-scale fading characteristics of the atmospheric channel are modeled by three different signal components. Thus, the received irradiance contributes to a LOS field component and two scattered optical field components due to the small-scale fluctuations. The first of these two scattered components is the quasi-forward optical signal scattered by eddies on the propagation axis, which is assumed to be coupled to the LOS component. The second component is the classical optical scattered field, which is due to the energy scattered by the off-axis eddies and is statistically independent of the other two components. The inclusion of the component coupled to the LOS scattering component is the main novelty of the model and can be justified by the high directivity and narrow beamwidths of laser beams in atmospheric optical communications [45].
As mentioned earlier, the Malaga-M distribution is a significant model for the FSO channel because it accurately models the irradiance fluctuations in a closed mathematical expression over a wide range of turbulence conditions, i.e., from weak to strong. Another significant advantage of the Malaga-M distribution is that it generalizes other known distributions to adequately model the ever-changing atmospheric turbulence conditions, as it agrees well with experimental results [47]. In Table I of [47], it was shown that various distributions such as the Rice-Nakagami, the Gamma-Gamma, the Shadowed-Rician, the K-distribution, the homodyned-K, the exponential, the Gamma-Rician distribution, etc., can be constructed from the Malaga-M distribution. However, the Malaga-M distribution is represented by the modified Bessel function of the second kind. Therefore, it is very challenging and complicated to analyze the SRT of the MIMO FSO systems. To the best of the authors' knowledge, there are no analyses for the trade-off between security and reliability for the MU-MIMO FSO system considering the Malaga-M atmospheric turbulence.
B. Motivation and Contributions
Despite their considerable potential as excellent candidates for future backhaul networks and various other applications, the SRT of MU-MIMO FSO systems has not yet been studied in depth in the open literature, using an OSS scheme to improve both the security and reliability of these systems. At the same time, most related research focuses on implementing the OSS to improve the performance of the multiuser RF link's performance of the mixed RF/FSO systems regardless of the FSO link. In contrast to previous studies of comparable systems [42], [43], [44], we develop an SRT based on the TAS scheme for the MU-MIMO FSO systems using a generalized channel model. The chosen channel model makes it easy to understand how different system and channel parameters affect the systems' security, reliability, and SRT performance. Additionally, this paper presents a comprehensive analysis of the TAS-based opportunistic scheduling to improve the SRT performance of MU-MIMO FSO systems. The main contributions of this paper can be summarized as follows: 1) We first develop the probability density function (PDF) and cumulative distribution function (CDF) of the MU-MIMO FSO system with the TAS and consider the generalized Malaga-M distribution with nonzero boresight pointing errors to make the study more realistic. 2) The exact and asymptotic analytical expressions of the outage probability are obtained to provide insightful information about the reliability performance of the system under investigation.
3) The exact and asymptotic analytical expressions of the average bit error rate (BER) are obtained to provide insightful information about the general performance of the considered system. 4) An exact and asymptotic analysis of the intercept probability, the primary metric for system security and SRT in the presence of an eavesdropper, was evaluated. To obtain the SRT of the system, we adopt a definition that also takes into account the predetermined outage threshold for the first time in such systems. 5) These expressions are used to generate numerical results with specified figures. In addition, Monte Carlo simulations verify the accuracy of the analytical results. The following is the outline for this paper: Section II presents models of the MU-MIMO FSO system and channel in consideration, while Section III develops an analytical expression for the statistical analysis of TAS opportunistic scheduling for the considered system. Section IV considers the outage probability and the average BER analysis of the investigated system. Section V presents the secrecy performance analysis based on the intercept probability. The SRT of the considered system is discussed in Section VI. Section VII has several interesting numerical examples as well as informative discussions. Finally, Section VIII brings the paper to a conclusion.
II. SYSTEM AND CHANNEL MODELS
We consider an MU-MIMO FSO communication system consisting of a transmitter T with K user groups equipped with M i transmitting apertures that communicate with a receiver R equipped with N i apertures over legitimate T → R links, where i ∈ {1, 2, . . . , K}. A single unauthorized passive eavesdropper E equipped with P t transmit apertures attempts to intercept confidential information over the unauthorized T → E links, where t ∈ {1, 2, . . . , P }, E is located in the same receiving plane as R in this work. We assume that the transmit apertures of the ith user on the transmitter T are directed to the ith apertures of the receiver R, whose receive apertures are also directed to the transmit apertures of the ith user. Therefore, the transmitting users communicate with the ith receiver aperture through M i × N i MIMO FSO links, as shown in Fig. 1 where η is the effective optical-to-electrical conversion ratio, I mn ir is the irradiance of the link between the mth transmitting user and the nth receiving aperture, x i is the optical signal transmitted from the ith user to R, and w mn ir is the additive white Gaussian noise with zero mean and variance N o /2. In this work, we assume the distance d i between the ith user and the receiver R, whiled represents the average distance between all users and R.
The path gain factor can be written as where α a is the atmospheric attenuation coefficient [32]. Moreover, atmospheric attenuation is caused by both molecular absorption and scattering of aerosols, suspended in the air. The total channel attenuation is given as where A = πD 2 /4, ϑ and β v are the area of the receive aperture (with D denoting the receiver aperture diameter), the divergence angle of the optical beam in radian, and the atmospheric extinction coefficient, respectively [48].
According to the Malaga-M model, the probability density function of the normalized irradiance f I mn ir is derived in [47] as k α r −j (·) is the modified Bessel function of the second kind and order α r−j , Γ(·) is the gamma function, g = 2b o (1 − ρ) denotes the average power of the scattering component received by off-axis eddies, 2b o is the average power for the component, which is quasi-forward scattered by the eddies on the propagation axis, and the component due to energy which is scattered to the receiver by off-axis eddies. Moreover, 0 ≤ ρ ≤ 1 is the factor that represents the scattering power associated with the LOS component. The parameter Ω indicates the average power of the LOS component. The parameters α r and β r are two parameters related to the atmospheric conditions on the legitimate T → R links and can be written as [49] α r = exp 0.49δ 2 where δ 2 r is the Rytov variance, which is the function of the link distance d i , the optical wave number k, and the refractive index structure constant C 2 n and can be given by
A. Channel Assumptions
If E is close to R on the same receiving plane, it is reasonable to assume that the optical irradiance I and atmospheric attenuation α a parameters are the same for the links of R and E over a long transmission distance. To ensure that there is no turbulence-induced fading correlation, i.e., independence between the legitimate and intercepting links, the distance between the receivers of R and E should be sufficiently large. In this work, with the help of [49,Eq. (4)] that separation distances of at least 11, 17, and 20 cm are required to ensure channel independence corresponding to the weak, moderate, and strong turbulence, respectively. The values of C 2 n for the weak, moderate, and strong turbulence, are chosen as 3 × 10 −16 , 10 −15 , and 5 × 10 −15 m −2/3 , respectively, for an FSO link with d i = 4 km, where λ is the optical wavelength of λ = 1550 nm and α a = 0.43 dB/km.
III. STATISTICS ANALYSIS OF TAS OPPORTUNISTIC SCHEDULING FOR MU-MIMO FSO SYSTEM
In this section, we analyze the statistics of the MU-MIMO FSO system with the TAS opportunistic scheme. The goal of the TAS is to activate the user transmit lasers to ensure maximum irradiance at the receiver. Therefore, a feedback link and channel state information (CSI) are required. The combined irradiance over the legitimate T → R links received by R can then be expressed as follows [33] I r = max (6) According to (1), the combined received signal y ir of the ith user over the T → R links can therefore be expressed as follows where w ir is the sum of w mn ir ; and its variance is equal to N i N 0 /2. Thus, the combined SNR between the ith user and R can be expressed as γ r = 2(ηG ir I r x i ) 2 /N i N 0 , and its average is given asγ r = 2(ηG ir E[I r ]x i ) 2 /N i N 0 = 2(ηG ir x i ) 2 /N i N 0 since I r is normalized to unity, where E[.] denotes the expectation operator.
Traditionally, the PDF and the CDF of the Malaga-M distribution are usually expressed by Meijer's G-function. However, the function is too complex to express the performance in closed form when applied to a MIMO FSO environment. To overcome this inconvenience, the PDF of the Malaga-M distribution is represented by a power series expansion of the modified Bessel function [50] based on the fact that the behavior of the PDF near the origin dominates the asymptotic behavior of the system performance. Based on the above, we next approximate the PDF expression of (2) as follows Here, we consider the first two terms of the power series expansion as The crucial step in our subsequent calculations is to write (8) as f I mn ir (I) ≈ a 0 I β r −1 ( a 1 a 0 I + 1), which can be further approximated by f I mn ir (I) ≈ a 0 I β r −1 e a 1 a 0 I , adopting the approximation method proposed in [51]. This method approximates the exponential function near the origin by the first two terms of the Taylor series expansion so that the obtained function can be related to the gamma PDF. Replacing a 0 and a 1 by their values from (9), the PDF and CDF of the irradiance along T → R links around the origin can be advantageously approximated by where, f M (I, k, θ) = 1 Γ(k)θ k I k−1 e −x/θ stands for the gamma PDF with shape parameter k and scale parameter θ. In (10) and (11), α r β r , are two constants related to the α r , and β r parameters of the legitimate T → R links, γ inc (a, x) = x 0 t a−1 exp(−t)dt is the lower incomplete gamma function [32]. A plot of the exact PDF of (2) and the derived approximate PDF of (8) is shown in Fig. 2 for C 2 n = 3 × 10 −16 . From Fig. 2, it can be seen that a closed match between the exact PDF and the approximate PDF in Fig. 2 can be observed, which shows the correctness of our derived results.
Based on the theory of order statistics [52], the CDF of T → R links combined irradiance I r can be approximated as follows According to the TAS, R is assumed to select the kth best user per time slot, assuming it knows all the information about the channel state. Such a TAS scheme was originally applied in multiuser RF systems to provide a multiuser scheduling solution with low complexity [53]. The SNRs of the ith users γ i are arranged in descending order of feedback value, γ 1 ≥ γ 2 ≥ · · · ≥ γ K , and γ k denotes the kth best user SNR. In this case, the TAS-based SNR is given by [54, Eq. (7)] where represents the impairments due to the pointing errors, which are assumed to be the same for all T → R links, s is the parameter defining the type of detection technique in the FSO receiver, i.e., s = 1 represents heterodyne detection, and s = 2 represents intensity modulation/direct detection (IM/DD).
To simplify the calculation, we assume that each link has the same number of transmit apertures (M i = M ) and receive apertures (N i = N ), for i = {1, · · ·, K}, and also has the same average link distanced. Based on the assumptions, the SNRs are independent and identical distributions (i.i.d). Based on this assumption and using [55, Eq. (6)], the CDF of the T → R links in terms of the kth best user's SNR can be reduced to where F r (x) is the CDF of legitimate T → R links defined by (12). Applying (12) to (14), the CDF of the MIMO FSO links concerning the SNR of the kth best user is obtained as follows where L = M (i + j). Replacing the lower incomplete Gamma function in (15) and applying the multinomial expansion, the CDF of the T → R links with respect to γ k can be represented as follows where Recall the PDF of generalized nonzero boresight pointing errors, which can be accurately approximated by a modified Rayleigh distribution [57] as follows Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
where ϕ r = w L( eq ) 2σ mod is the ratio of the equivalent beam radius, w L( eq ) , and the standard deviation (jitter) at the receiver plane, (2σ mod ), ϕ , a is the aperture radius of the photodetector, and G f = . From (16), the CDF of TAS-based SNR, taking into account the effect of the generalized pointing errors, is given as where (18)
IV. PERFORMANCE ANALYSIS
In this section, we evaluate the outage probability, and the average BER of the considered system.
A. Outage Probability
The outage probability is defined as the probability that the SNR at the receiver falls below a given outage threshold γ th , i.e., P out = P r [γ r ≤ γ th ], where P r [.] denotes the probability operation, γ r is the SNR at R, and γ th = 2 R − 1, where R denotes the target data rate. Thus, the outage probability can be obtained by replacing γ with γ th in the CDF expressions (19) as follows where where (ζ 1 , ζ 2 ) = (1 − ϕ 2 r + δ l , s) , (1, s) and (φ 1 , φ 2 ) = (1 − ϕ 2 r − δ l , s). To gain further insight into the high SNR regime, we analyzed the asymptotic behavior of (21) and presented it in an easyto-follow form with reasonable accuracy. At high SNR values, the outage probability of the FSO system can be expressed as where G c and G d denote the system's coding gain and diversity order, respectively. For this purpose, we invoke asymptotic expansions of the Fox-H function [60, Theorems 1.7 and 1.11]. Based on (21), the asymptotic behavior of the outage probability for theγ r → ∞ is derived as The diversity gain of the considered system over atmospheric turbulence conditions is inferred as G d = min{ ϕ 2 r s , α r s , β r s }.
B. Average Bit Error Probability
Substituting (19) where (1, s) and (U 1 , U 2 ) = (1 − ϕ 2 r − δ l , s), c and d are modulation specific parameters. Similar to the P out , the average BER can be expressed asymptotically, at high SNR, as
V. SECURITY ANALYSIS
In this section, we analyze the secrecy performance of the MU-MIMO FSO system in the presence of a single eavesdropper E equipped with a P aperture. When T sends its signals to R, they can be intercepted by E, as shown in Fig. 1. It is assumed that E is in the receiving plane of R, and thus, it can also overhear the transmission from T . Therefore, E can combine its signals received from T to obtain an enhanced signal version using equal gain combining (EGC). Typically, EGC is usually able to achieve better combining performance due to its low complexity implementation compared to other combining techniques. Therefore, we consider using EGC at E to maximize its ability to interpret the transmitted message. Thus, the received signal at E is given by where I ite is the irradiation coefficient of the interception link between the ith user of T and the tth interception aperture of E, and v e is an AWGN sample at E. Since E applies the EGC scheme, the combined SNR observed at E can be expressed as [62] (26) For this case, the PDF of the T → E links, which follows a Malaga-M distribution in the form of the Meijer G-function, is given as follows [63] where α e and β e are the new shaping parameters, defined as α e = P α E and β e = P β E , α E and β E are the two parameters related to atmospheric conditions via the T → E links, ϕ e is the ratio of the equivalent beam radius, and the standard deviation in the E-plane,γ e = η 2 x 2 k MP N o . The CDF expression is obtained using (27) where X i 1,k , X i 2,p , X i 3,q , Ω i 1,k , Ω 2,p , and Ω 3,q are given in Table I. The above series representation of the CDF of Malaga-M distribution of (29), contains three power series. Each series contains summation terms with only exponents of X. Therefore, it is easy to compute an integral containing the proposed series representation compared to a complicated function-based representation in (28). By applying the ratio test for each power series in (29), it can be shown that the above series representation contains converging power series with an infinite radius of convergence. Using the Cauchy ratio test to check convergence, the series X k is absolutely convergent if [64,Eq. (10)] as lim k→∞ | X k +1 X k | < 0. The radius of convergence of the first subseries in (29) is given by It can be seen from (30) that the order of k in the denominator is higher than the numerator. So, applying limit k → ∞ in (30) gives 0; hence, r 1 is absolutely convergent [64]. In a similar manner, the series in X i 2,p Ω 2,p and X i 3,q Ω 3,q can be shown to be convergent.
Hence, the CDF of the T → E links with M and P apertures in T and E, respectively, is given by [63] where F γ ie (γ) is given by (29). Hence, the PDF of the T → E links with respect to γ e can be obtained as where f γ ie (γ) is given by (27). Substituting (27) and (29) into (32) and applying the binomial theorem yields the expression for f γ e (γ) shown in the following, where where τ is given in Table I.
A. Intercept Probability Based on TAS Opportunistic Scheme
Intercept probability is an important metric that provides further insight into the secrecy performance of a communication system. It is defined as the probability that the capacity of the Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. main link is less than that of the wiretap channel. Mathematically, the intercept probability is formulated as follows [65] where C r denotes the channel capacity of the legitimate T → R links and C e is the channel capacity of the eavesdropping T → E links where f γ e is the PDF of the wiretap links obtained from (33). The integration of (35) can be calculated by substituting (18) and (33) where (ψ 1 , ψ 2 ) = (1 + ϕ 2 r + δ l , s) , (1, s), To analyze the asymptotic behavior of the intercept probability in the range of high SNR values, we first recall [60, Eq. (1.8.15)] and [60, Eq. (1.8.5)] and further perform some algebraic manipulations as follows
VI. SECURITY-RELIABILITY TRADE-OFF
In this section, we develop an expression of the SRT for the MU-MIMO FSO system that considers the system's predetermined outage threshold. The outage threshold corresponds to the threshold SNR γ th below which detection is very unlikely at the given data rate. An interception occurs when the eavesdropper detects the signal with an SNR above this threshold. Meanwhile, the SRT can be represented as follows [6] P int,th = P r (γ r ≤ γ e , γ e > γ th ) where P r (γ r ≤ γ e ) and P r (γ e > γ th ) represent the intercept probability and the outage probability of the legitimate T → R and T → E links, respectively. Eq. (41) can be rewritten as Thus, by substituting (29) and (38) in (42), we have The asymptotic behavior of the P int,th is derived for theγ r → ∞ as
VII. NUMERICAL RESULTS
Selected simulation results of the MU-MIMO FSO system considering eavesdropping situations are provided and analyzed in this section using the above analytical expressions. We investigate the effects of atmospheric turbulence, generalized pointing errors, the number of users, and the power received from the eavesdropper on the outage probability, intercept probability, and SRT performance of the system under consideration. For simplicity, we assume independent, identically distributed main and wiretap channels. Moreover, the average optical power of the FSO links is assumed to be b o = 0.25, Ω = 0.5, and ρ = 0.95. The graphs were generated in MATLAB using analytical equations and Monte Carlo simulations with 10 6 channel realizations to confirm the accuracy and usefulness of the proposed approximation. Unless stated otherwise, the considered system and channel simulation parameters are listed in Table II. First, results corresponding to the outage performance analysis described in Section IV are shown in Fig. 3 for moderate turbulence as a function of average electrical SNR,γ r , with γ th = 5 dB. In Fig. 3, two normalized jitter values (σ x /a, σ y /a) = {(2, 1), (9,7)} are considered to carefully analyze how T → R links are affected by generalized pointing errors. As seen from Fig. 3, a more significant number of users, K, leads to a lower outage probability, indicating a linear relationship between the diversity order and K. Moreover, the best performance is obtained when the jitter value is small (σ x /a, σ y /a) = (2, 1), since the photodetector is better positioned in this case. Moreover, the analytical curves agree with the simulation curves in the high SNR region, proving the derived formulas' correctness. However, the analytical results in the low SNR region deviate from the simulation results as clearly shown in Fig. 3. This is due to the approximation method in (8), where only the first two terms in the power series expansion of the modified Bessel function are considered to simplify the MIMO Malaga-M distribution. Fig. 4 shows the outage probability as a function of K for selected values of γ th withγ r = 5 dB, (σ x /a, σ y /a) = (9, 7), and moderate atmospheric turbulence. Fig. 4 shows that the values of outage probability decrease linearly as the number of users increases, implying that the order of diversity has a linear relationship with K. Moreover, the results show that a higher number of apertures in T and R leads to significantly improved performance. This is because the diversity order of the system increases as the number of K, M and N increases. Compared to the system performance at a low value of γ th , the negative effect of a higher value of γ th on the system's outage probability is easily seen.
Next, we discuss how the threshold SNR affects the MU-MIMO FSO system's reliability. Fig. 5 shows the system's outage probability as a function of the threshold SNR, γ th , for different atmospheric turbulence conditions. The results show that the increase of γ th and the strength of atmospheric turbulence significantly deteriorates the outage performance for both detection schemes of R. The results also show that the system's performance decreases faster as the number of M and N decreases; in particular, the system achieves worst with K, M, N = 2. Moreover, the heterodyne detection outperforms the IM/DD detection for all values of N , as shown in Fig. 5.
Next, we investigate the transmission distance of the MU-MIMO FSO system by plotting the outage performance as a function of the average distance between T and R,d, for different numbers of K, M, N as shown in Fig. 6. In this case,γ r = 5 dB, (σ x /a, σ y /a) = (9, 7), and strong atmospheric turbulence. The results show that the outage probability increases with the length of the link, regardless of the number of K, M , or N , since the power received by R decreases for the most extended transmission distances. This phenomenon also illustrates how the scintillation index approaches saturation as the length of the link or the strength of the turbulence increases. One possible explanation is that the direct transmission component, which dominates FSO transmission, is affected by atmospheric turbulence when the selected user is far from the transmitter. In addition, the results show that heterodyne detection performs better than the IM/DD scheme at all transmission distances. While the outage probability of the SISO system increases dramatically with increasing link length, the P out of the MU-MIMO FSO system is much less extreme. Fig. 7 shows the outage performance as a function of the average SNR of the MU-MIMO and SISO FSO systems with selected values of γ th . In this case, the atmospheric turbulence is assumed to be strong with (σ x /a, σ y /a) = (9, 7). As can be seen, the P out values considering both γ th values decrease rapidly with increasing SNR. Moreover, the results demonstrate that the P out value is enhanced with a more significant number of K, M, N compared to the SISO case. Additionally, when γ th is 0 dB, the FSO system achieves the lowest P out . The Monte Carlo simulations and the asymptotic results agree well with the exact results and show the accuracy of the obtained expressions at high SNR values, as shown in this figure.
Regarding the average BER analysis, we will highlight the effect of the receiver aperture's size in combination with the atmospheric turbulence conditions on system performance as shown in Fig. 8. In this case, we consider K, M, N = 2 and an ignorant misalignments effect. It is clear from this figure that decreasing turbulence conditions (i.e., moderate turbulence conditions) leads to better BER performance regardless of the receiver aperture size and type of detection scheme in R. However, for a larger aperture's size of the receiver (i.e., a = 10 cm) a lower BER was achieved. This is because the variations in intensity at R result in a variance in received power that depends on the size of the receiver aperture. Increasing the receiver aperture size decreases the power variance and improves the bit error rate. Where the receiver aperture must be large enough to collect sufficient power to reduce scintillation effects at a given range, but also small enough to be a practical size for cost-effectiveness. In addition, it can be noticed that the degradation of the system performance due to strong turbulence is greater when IM/DD detection is employed. For instance, under moderate turbulence and s = 1, the system gained about 6 dB with a = 10 cm in comparison with the case of a = 5 cm when BER =10 −5 . Nevertheless, the analytical curves agree with the simulation curves in the high SNR region, but, the analytical results depart from the simulation results in the low SNR region as clearly shown in this figure.
Next, we investigate the effects of system and channel parameters on the security performance of the MU-MIMO FSO. Fig. 9 presents the security performance represented by the intercept probability as a function of the average SNR in dB for legitimate T → R links withγ e = 10 dB, K, P = 4, and (σ x /a, σ y /a) = (9,7) with different M, N, k numbers and varying atmospheric turbulence conditions. First, we find that the values of the interception probability decrease with increasing M and N at all turbulence strengths. Moreover, the intercept probability decreases linearly with increasing of SNR values, indicating a linear relationship between diversity order and SNR. The results also show that the case of M, N, k = (4, 2, 1) performs best under weak turbulence. A continuous decrease in the strength of turbulence on the FSO links would lead to a significant decrease in the intercept probability values. For example, when M, N, k = (4, 2, 1), the MU-MIMO system achieves a P int of 10 −6 in weak turbulence withγ r = 28 dB, which increases to 1.3 × 10 −4 in the case of the SISO system and the same turbulence condition. The agreement between the analytical and simulation curves at high SNR values proves the accuracy of our analysis. Now we will demonstrate the impact of the number of users and the average SNR of the T → R links on the security performance of the considered system, as shown in Fig. 10. In this case, we set the number of P to 4 and (σ x /a, σ y /a) = (9,7). It can be observed that the performance of P int with a large value ofγ r is superior to that with a low value ofγ r at all turbulence strengths. Moreover, the P int decreases with increasing K. This is because as K increases, the legitimate links between the T and R increase, improving the overall SNR value of the receiver and thus decreasing the P int value. However, as expected, the system performs better in weak and moderate turbulence than in strong turbulence. Fig. 11 illustrates the combined effects of the number of users, number of apertures, and SNR of T → E links on the intercept probability of the investigated system. In this case, a strong turbulence strength, (σ x /a, σ y /a) = (9,7), and P = 4 are considered. The system achieves lower P int with higher K, M , and N . This outcome can be explained by the high power received by R when K, M , and N increase. Whenγ e = 3 dB, the system's performance is worse than the P int atγ e = 0 dB because the T → E link receives more power in this case.
From the previous results in Figs. 3-11, it can be deduced that the simulations and the asymptotic results agree well with the exact results at high SNR values, which proves the accuracy of the expressions derived previously in Sections IV, V, and VI. However, the curves deviate from each other at low SNR regimes due to the approximation of (8). In the following, we will discuss the combined effect of turbulence, generalized nonzero boresight pointing errors, the number of users, and the number of eavesdroppers' apertures on the SRT of the system under consideration in Figs. 12 and 13. Fig. 12 shows the effects of generalized pointing errors on the SRT of the system in the presence of strong turbulence and P = 4. In this case, we setγ r andγ e to 5 dB. The numerical results show that the intercept probability decreases with the deterioration of P out , indicating the close relationship between the two indicators. However, when the value of jitter is low (σ x /a, σ y /a) = (2, 1) and the number of K, M, N is high, the system has better performance because the received power by R is increased, which leads to a decrease in the interception probability. Moreover, with the adopted TAS scheme, the MU-MIMO system performs better than the SISO system. Additionally, a good match between analytical and simulation results can be seen at high values of P out and P int . In Fig. 13, we display the SRT of the system under consideration for different numbers of K, M, N , with P = 4. In this case, we assumeγ r =γ e = 5 dB, with (σ x /a, σ y /a) = (9,7). Fig. 13 shows that the SRT improves under all turbulence conditions when K, M , and N increase from 2 to 4. However, as expected, the intercept and outage probability results under weak turbulence conditions are better than those under moderate and strong turbulence conditions. Additionally, the agreement between the analytical and simulation results at high values of P out and P int shows the accuracy of our SRT analysis.
Further, we have considered the models used in [12], and [42] to compare the results under moderate turbulence condition, γ r =γ e = 5 dB, and (σ x /a, σ y /a) = (9,7). This comparison is shown in Fig. 14. It can be observed that the results obtained for the considered MU-MIMO FSO system is superior to the models used in [12], and [42].
The general observation of the previous results shows that increasing the number of users K resulted in an increase in the number of links between the T and the legitimate receiver R, which improves the overall SNR value of the receiver and thus improves the overall system performance including the outage probability, the interception probability, and the SRT, as demonstrated in (20), (38), and (43).
VIII. CONCLUSION
We examined the performance of the MU-MIMO FSO system using TAS opportunistic scheme, which utilizes the best user selection by the authorized receiver. The outage probability, average BER, intercept probability, and SRT closed-form expressions were developed based on the adopted power series expansion of the Meijer-G function, assuming a generalized Malaga-M distribution and nonzero boresight pointing errors for MIMO FSO links. In addition, an asymptotic closed-form expression for the outage probability, average BER, intercept probability, and SRT was also derived for high SNR values. The proposed analysis sheds substantial light on the secrecy performance of the MU-MIMO FSO system in the presence of eavesdropper attacks using the developed expressions for the interception probability. The main findings of this work show that increasing the number of users and the number of apertures of the transmitter and receiver improves the system performance regardless of channel conditions. Moreover, the simulations and the asymptotic results agree well with the exact results at high SNR values showing the accuracy of the obtained expressions. Finally, the adopted TAS scheme of the MU-MIMO FSO system provided significant improvement in security, reliability, and SRT performance compared to the SISO FSO system. | 2023-03-29T15:13:57.986Z | 2023-06-01T00:00:00.000 | {
"year": 2023,
"sha1": "22fd4c3ea0908efb75045b6a51e32437dcb4146c",
"oa_license": "CCBY",
"oa_url": "https://ieeexplore.ieee.org/ielx7/4563994/4814557/10081357.pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "c0b79b47a53067137b53bc63376e5838828d8ee4",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": []
} |
267967204 | pes2o/s2orc | v3-fos-license | Review on the Role of BRCA Mutations in Genomic Screening and Risk Stratification of Prostate Cancer
(1) Background: Somatic and germline alterations can be commonly found in prostate cancer (PCa) patients. The aim of our present study was to perform a comprehensive review of the current literature in order to examine the impact of BRCA mutations in the context of PCa as well as their significance as genetic biomarkers. (2) Methods: A narrative review of all the available literature was performed. Only “landmark” publications were included. (3) Results: Overall, the number of PCa patients who harbor a BRCA2 mutation range between 1.2% and 3.2%. However, BRCA2 and BRCA1 mutations are responsible for most cases of hereditary PCa, increasing the risk by 3–8.6 times and up to 4 times, respectively. These mutations are correlated with aggressive disease and poor prognosis. Gene testing should be offered to patients with metastatic PCa, those with 2–3 first-degree relatives with PCa, or those aged < 55 and with one close relative with breast (age ≤ 50 years) or invasive ovarian cancer. (4) Conclusions: The individualized assessment of BRCA mutations is an important tool for the risk stratification of PCa patients. It is also a population screening tool which can guide our risk assessment strategies and achieve better results for our patients and their families.
Introduction
Based on epidemiological data, prostate cancer (PCa) has been shown to be the most common of all urologic malignancies and second only to lung cancer as regards cancerrelated mortality among male patients in developed countries [1].The incidence of the disease has been steadily increasing over recent years [2], with the lifetime risk being equal to 12%, while the median age of diagnosis is 66 years according to the SEER database [3].Although the widespread use of PSA as a screening tool has markedly increased the number of patients being diagnosed with occult disease [4], early detection has substantially decreased both cancer-specific mortality and advanced disease on initial diagnosis by 40% and 75%, respectively [5,6].
Commonly accepted risk factors for PCa include age, African race, several genetic factors and a family history of PCa [7].More specifically, PCa is considered to be one of the malignancies with the greatest heritability component (up to 57% genetic contribution) [8], with several studies suggesting that 8-12% of all patients presenting with advanced disease may harbor a germline mutation in a tumor-suppressor gene [9,10].Interestingly, it has been shown that individuals with a family history of PCa have 2.5 times higher probability of being diagnosed with the same disease when compared to the general population [11].Researchers analyzing data form the Swedish Cancer Database concluded that the three malignancies with the highest familial cancer rate are prostate, breast and colorectal cancer (20.2% vs. 13.6% vs. 12.8%) [12].An integrative analysis on patients diagnosed with advanced disease showed that somatic and germline alterations were found in up to 90% of those suffering from metastatic castration-resistant cancer [13].So far, the most wellstudied germline mutations relating to PCa are those of BRCA 1/2, CHEK2, NBM and ATM [10,14,15].
The main scope of the current review is to specifically examine the impact of BRCA mutations in the context of PCa, as well as their significance as a genetic biomarker in the era of efforts towards personalized screening.Emphasis is given to studies discussing prostate cancer risk stratification and early prostate cancer detection.Since our aim was to perform a narrative review, we included only "landmark" publications.
Role of Family History and BRCA Genes in PCa
In 2015, a study conducted by Liss et al. showed that male patients who had a history of PCa in their families were not only more susceptible to developing the same malignancy but they were also at a greater risk of dying from PCa, with both results being statistically significant.The same study highlighted the value of early PSA testing in family members of PCa patients, as it can lead to lower cancer-specific mortality rates and a possible survival benefit [16].Likewise, a study performed by Brandt et al. showed that the hazard ratios for prostate cancer diagnosis and risk of death from prostate cancer in men increased with the number of affected relatives and decreased with increased age [17].Also worth noting is the fact that we should treat familial and hereditary cancer as two distinct clinical entities.The hereditary form can be attributed to identifiable mutations (most commonly mutations regarding BRCA genes), while the familial form covers a larger spectrum (approximately 15-20% of PCa cases), with individuals having a positive family history in common but no identifiable genetic alteration.Nevertheless, the degree to which a positive family history defines the probability of developing cancer is highly variable and is also affected by factors such as age and degree of relation with the affected individuals.According to Carter et al., the proportion of PCa cases that could be attributed to mutated high-risk alleles was found to be approximately equal to 40% for men younger than 55 years, which is much higher than the 9% for individuals older than 85 years [18].Moreover, a meta-analysis conducted by Zeeger et al. showed that men with an affected father had more than 50% lower relative risk of receiving a diagnosis of PCa than men with an affected brother (2.2 vs. 3.4), whereas this risk practically disappeared among second-degree family members [19].Bratt et al. provided a nationwide population study, reporting the association between family history and diagnosis of prostate cancer in different cancer risk categories (any prostate cancer, nonlow-risk prostate cancer and high-risk prostate cancer).According to them, the probabilities of intermediate and high-risk prostate cancer are better for counselling men with a positive family history of prostate cancer [20].Having this in mind, it is easy to understand that in order to achieve a better estimation of the lifetime risk of being diagnosed with PCa, based solely on heritability factors, we should take into consideration the status of close relatives as well as the number of affected individuals in a family pedigree and their age at diagnosis.In this diagnostic algorithm, we should also take into account a family history from the maternal family branch [21].
BRCA 1 and BRCA 2 genes were first discovered almost three decades ago (1994 and 1995, respectively) after carefully examining the biological and genetic background of families with an abnormally high prevalence of breast and ovarian cancer [22,23].These genes are responsible for 30-70% and approximately 90% of all hereditary breast and ovarian cancer cases, respectively.According to several studies, individuals harboring mutations in these genes have a lifetime risk of developing breast cancer of up to 85%, while the risk for ovarian cancer ranges from 20 to 40%.Individuals carrying BRCA1 mutations were also found to be at higher risk of developing other types of cancer, such as pancreatic and cervical.As far as prostate cancer is concerned, the risk was found to be age-dependent, affecting carriers younger than 65 years old [24,25].
In terms of biology, BRCA genes belong to the family of DNA damage repair (DDR) genes.The main role of these genes is to repair several DNA aberrations taking place during the cell cycle, thus providing genomic stability and ensuring an uneventful distribution of genetic material to the daughter cells following mitotic cell division [26].In cases where this system fails, individuals become susceptible to malignancies such as breast, ovarian, prostate and pancreatic cancer.In particular, BRCA 1 mutations were associated with an increased risk of ovarian and breast cancer, whereas BRCA 2 carriers were more susceptible to pancreatic and prostate cancer [27].According to several studies, the percentage of patients with PCa that harbor a BRCA2 mutation ranges between 1.2% and 3.2%, and the percentage is even smaller when it comes to BRCA1 [28,29].Nevertheless, those genes (especially BRCA2) are considered to be responsible for most cases of hereditary PCa, with mutations in BRCA2 and BRCA1 increasing the risk by 3-8.6 times and up to 4 times, respectively, when compared to the general population [24,28,30,31].
So far, several studies have shown that PCa in individuals with mutated BRCA1/2 genes is in general more aggressive and associated with lower overall survival when compared to cases of male patients with normal alleles [32][33][34][35].Back in 2019, Castro et al. conducted a retrospective analysis based on data from 79 BRCA mutation carriers and 1940 non-carriers, all of whom were diagnosed with PCa [33].According to the authors, BRCA 1/2 mutations were found to be directly correlated with a higher risk of locally advanced disease, a higher Gleason score, nodal infiltration and distant metastases at the time of diagnosis.Moreover, they showed that BRCA2 mutations should be considered an independent negative prognostic factor based on both the significantly shorter 5-year cancer-specific survival and metastasis-free survival among mutation carriers.
Just to further examine the strength of the connection between BRCA mutations and PCa risk, Ibrahim et al. carried out a retrospective analysis by dividing a total of 102 men into two cohorts based on their genetic profile (mutation of either BRCA1 or BRCA2) [36].Based on their findings, almost one-third of the under-examination individuals had at least one type of cancer, with prostate cancer being the most common (2 patients in the BRCA1 group and 11 patients in the BRCA2 group), followed by breast, skin and urothelial cancer.The above findings are in accordance with previous studies, supporting the theory that BRCA2 mutation carriers are far more susceptible to malignancies than BRCA1 carriers [25,30,[37][38][39].According to a well-designed case-control study on a special population of males of Ashkenazi origin, BRCA2 mutation carriers were shown to be at greater risk of developing PCa when compared to the age-matched control group [40].Finally, according to the PROREPAIR-B trial, which prospectively examined only patients with metastatic castration-resistant disease, a germline BRCA2 mutation was characterized as an independent negative prognostic factor with a statistically significant lower CSS rate (17.4 vs. 33.2months, p = 0.027).In this particular study, a statistically significant difference was also noticed in terms of treating BRCA2 carriers; namely, the use of ARTA (abiraterone or enzalutamide) was associated with better CSS and PFS in comparison with taxanes [41].
BRCA Testing in Prostate Cancer and Screening of Men with Known Mutations
Nowadays, it has become clear that PSA, which has been established as the major screening tool for PCa detection, has contributed to a small absolute decrease regarding the risk of death from PCa [42], while bringing with it the risk of overdiagnosis and the unnecessary treatment of quite a few individuals who would otherwise have never experienced clinical manifestations of the disease [43].Under these circumstances, we can easily understand that a universal screening plan is related to a high financial burden as well as several unnecessary biopsies and treatment-related morbidities for individuals with indolent disease and a high PSA value [44,45].On the other hand, we cannot overlook the importance of disease detection at an earlier stage, which gives us the opportunity to use a variety of therapeutic interventions with the aim of preventing cancer progression and metastatic disease [46].Having all the above in mind, it is easy to justify the efforts made in recent years towards finding sufficient and objective data that could support the development of individualized screening plans.In the process of planning such strategies, we should take into consideration not only results from recent studies and population stratification tactics but also the need for shared and informed decision making.
In general, when we suspect the presence of an inherited malignancy or syndrome running through a family, it is of utmost importance that the patient is referred for genetic counseling.Unfortunately, in the case of PCa, there has been no uniform consensus so far regarding the exact criteria that should lead a patient to visit a geneticist.
In the context of PCa, without specific consideration of any particular genetic alteration, the American College of Medical Genetics suggests that a thorough genetic evaluation should take place if any of the following stands true [47]: Two or more relatives receiving a PCa diagnosis at an age of 55 or younger (the relatives should be first-degree).
2.
At least three first-degree relatives with PCa, irrespective of age.3.
Gleason grade 8 or higher and at least two individuals in the family pedigree diagnosed with breast, ovarian or pancreatic cancer.
Again, without specific interest for a specific mutation, the Johns Hopkins groups suggests another set of guidelines for familial PCa, which are as follows [48]: Family pedigree with evidence of PCa in three successive generations.
2.
Two relatives diagnosed with PCa at an early age (≤55 years).
3.
At least three first-degree relatives with PCa.
Based on EAU recommendations, all individuals with high-risk or metastatic disease should be offered genetic counseling at least for the possibility that they have mutations in the BRCA1/2, ATM, FANCA or PALB2 genes.For the patients who are diagnosed with lower-risk disease, testing is recommended if any of the following are true [49]: There is a strong family history of PCa.
2.
At least one member of the family is diagnosed with Lynch syndrome.
3.
There are already known germline mutations in the family or at least one family member has pancreatic or breast or ovarian cancer (possible BRCA2 mutation).
On the other hand, the National Comprehensive Cancer Network (NCCN) has published guidelines specifically for testing BRCA1/BRCA2 status, including patients with any of the following characteristics [50]: Diagnosis of PCa (Gleason ≥ 7) and at least two relatives with prostate (Gleason ≥ 7) or breast or pancreatic cancer.
3.
Diagnosis of PCa (Gleason ≥ 7) at any age and at least one close relative with breast (age ≤ 50 years) or invasive ovarian cancer.Serious efforts to reach a consensus and possibly bridge the gap between the recommendations made by different organizations have been made through studies like IMPACT.This study is an ongoing multicenter observational trial on carriers of BRCA1/2 mutations who have already been diagnosed with PCa [51].The main purpose of the study is to determine the significance of an annual PSA testing protocol for BRCA1/2 carriers compared with non-carriers, with the PSA threshold for a biopsy being that of 3 ng/mL.In their interim analysis, researchers concluded that further follow-up is needed for BRCA1 carriers, but as regards the BRCA2 carriers, early and routine screening with PSA is important, as it was demonstrated that prostate biopsy in this specific population is associated with twice the positive predictive value compared to general population studies.An interim analysis of the IMPACT study showed that 77% of BRCA2 carriers were diagnosed with intermediate and high-risk disease [52].A Dutch multidisciplinary expert panel reached a consensus in a meeting regarding the indications and applications of germline and tumor genetic testing in prostate cancer.Interestingly, they agreed that germline and tumor genetic testing should not be performed in the case of nonmetastatic hormone-sensitive prostate cancer when a relevant family history of cancer does not exist.As far as the metastatic disease is concerned, neither of the above-mentioned tests received panel approval for implementation in M1a HSPC, whereas it was agreed that the final therapeutic decision was not affected by the results.In metastatic CRPC, there was a lack of consensus regarding when tumor genetic testing should be performed and who should evaluate it, thus coming to no recommendations [53].Table 1 summarizes the recommendations of different organizations for genetic testing in prostate cancer.At least three first-degree relatives with PCa.
European Association of Urology (EAU) Recommendations.
Genetic counseling at least for the possibility that they have mutations in BRCA1/2, ATM, FANCA or PALB2 genes.
All individuals with high-risk or metastatic disease.For the patients who are diagnosed with lower-risk disease, testing is recommended if [49]: 1.
There is a strong family history of PCa.
2.
At least one member of the family is diagnosed with Lynch syndrome.
3.
There are already known germline mutations in the family or at least one family member has pancreatic or breast or ovarian cancer (possible BRCA2 mutation).
National Comprehensive Cancer Network (NCCN).Specific BRCA1/BRCA2 status.In an effort to establish risk-adapted guidelines for PCa diagnosis, the NCCN recommends that BRCA1/2 carriers should be tested annually starting at the age of 40 years [54].The EAU follows the same directions with the exceptions of recommending intervals based upon the baseline PSA value and not including BRCA1 carriers in their guidelines [55].Finally, the ACS does not take into consideration the germline status and suggests that men with a first-degree relative diagnosed younger than 65 years should start screening at 40 years (45 years if more than one first-degree relative with PCa) [56].A recent review article by Giri et al. summarizes all existing guidelines for germline testing in prostate cancer, in an effort to promote multidisciplinary patient primary care.The authors also shed light on a very intriguing topic, namely the involvement of primary care physicians in the genetic testing process and algorithm.Through proper education, these professionals can offer individualized patient-centered medical advice and refer individuals with a positive family history at high risk for further evaluation and assessment by specialists [57].
Conclusions
Genetic studies during recent decades have undoubtedly added a lot to our current understanding of prostate cancer biology.Based on the current literature, it seems to be inevitable that genetic testing is very soon going to be an integral part of clinical practice not only for individuals already diagnosed with the disease but also in the setting of population screening.Hopefully, more precise knowledge on genetic predispositions for PCa is going to help us further improve our risk assessment strategies and achieve better results for our patients and their families.
Table 1 .
Genetic testing recommendations for prostate cancer according to different organizations. | 2024-02-27T16:23:31.176Z | 2024-02-22T00:00:00.000 | {
"year": 2024,
"sha1": "18e254b69cdf4ee6fb8d2fbb9ca42bb6bfb5aea1",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1718-7729/31/3/86/pdf?version=1708582927",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "e68625be7621d32089d8b78eafb49f6e1edf9600",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
3449123 | pes2o/s2orc | v3-fos-license | Spin--3/2 to spin--1/2 heavy baryons and pseudoscalar mesons transitions in QCD
The strong coupling constants of light pseudoscalar mesons with spin--3/2 and spin--1/2 heavy baryons are calculated in the framework of light cone QCD sum rules. It is shown that each class of transitions among members of the sextet spin--3/2 to sextet spin--1/2 baryons and that of the sextet spin--3/2 to spin--1/2 anti--triplet baryons is described by only one invariant function. We also estimate the widths of kinematically allowed transitions. Our results on decay widths are in good agreement with the existing experimental data, as well as predictions of other nonperturbative approaches.
Introduction
Theoretical and experimental studies of the flavored hadrons are among the most promising areas in particle physics. From theoretical point of view, this can be explained by the fact that the heavy flavored baryons provide a rich laboratory to study predictions of the heavy quark effective theory. On the other hand, these baryons have many weak and strong decay channels and therefore measurement of these channels can give essential information about the quark structure of heavy baryons. During the last decade, highly exciting experimental results have been obtained on the spectroscopy of heavy hadrons. All ground states of heavy hadrons with c quark have been observed [1]. The new states of heavy baryons are also discovered in BaBar, BELLE, CDF and D/ 0 Collaborations. The operation of LHC will open a new window for more detailed investigation of these new baryons [2].
At present, we have experimental information on the strong one-pion decays for the Σ c [3][4][5], Σ * c [4,6] and Ξ * c [7,8] baryons. The strong coupling constants of pseudoscalar mesons with heavy baryons are the main unknown parameters of these transitions. Therefore, a reliable estimation of these strong coupling constants in the framework of QCD receives great interest. At hadronic scale, the strong coupling constant α s (Q 2 ) is large and hence perturbative theory becomes invalid. For this reason, estimation of the coupling constants becomes impossible starting from the fundamental QCD Lagrangian and some nonperturbative methods are needed. Among many nonperturbative approaches, the QCD sum rule [9] is one of the most powerful method in studying the properties of hadrons. The main advantage of this method is that, it is based on fundamental QCD Lagrangian. In the present work, we estimate the strong coupling constants of pseudoscalar mesons in the transitions of spin-3/2 to spin-1/2 heavy baryons within light cone QCD sum rules method (for more about this method, see [10]). Note that the strong coupling constants of pseudoscalar and vector mesons with heavy baryons in the spin-1/2 to spin-1/2 transitions are studied in [11] and [12].
The rest of the paper is organized as follows. In section 2, the light cone sum rules for the coupling constants of pseudoscalar mesons with heavy baryons in spin-3/2 to spin-1/2 transitions are calculated. In section 3, the numerical analysis of the obtained sum rules is performed and a comparison of our results with the predictions of other approaches as well as existing experimental results is made.
2 Light cone QCD sum rules for the pseudoscalar mesons with heavy baryons in spin-3/2 to spin-1/2 transitions In this section, the strong coupling constants of light pseudoscalar mesons with heavy baryons in spin-3/2 to spin-1/2 transitions are calculated. Before making an attempt in estimating these coupling constants, few words about SU(3) f classification of heavy baryons are in order. Heavy baryons with a single heavy quark and two light quarks can be decomposed into two multiplets, namely, sextet 6 F and anti-triplet3 F due to the symmetry property of flavor and color structures of these baryons. This observation leads to the result of total spin J P = (3/2) + or (1/2) + for 6 F and J P = (1/2) + for3 F . In the present work, we consider J = 3/2 states in 6 F and investigate sextet to sextet (S * SP) and sextet to anti-triplet (S * AP) transitions with the participation of light pseudoscalar mesons, where S * , S and A stand for sextet with spin-3/2, sextet with spin-1/2 and anti-triplet spin-1/2 states, respectively. We now pay our attention to the calculation of the strong coupling constants of pseudoscalar mesons with heavy baryons in spin-3/2 to spin-1/2 transitions. To derive the light cone sum rules for S * SP and S * AP transitions we consider the following general correlation function: where η (i) (x) are the interpolating currents of the heavy baryons with spin-1/2 in sextet (i = 1) and anti-triplet (i = 2) representation andη µ is the interpolating current for the sextet J P = 3 + /2 states. The correlation function (1) can be calculated in terms of hadrons (phenomenological part) and in terms of quark-gluon degrees of freedom in deep Euclidean region, i.e., when p 2 → −∞. Equating then these representations of the correlation function using the dispersion relation, we get the sum rules for strong coupling constants of light pseudoscalar mesons with heavy baryons. We proceed by calculating the phenomenological part of the correlation function. The expression for the phenomenological part is obtained by saturating it with the full set of hadrons carrying the same quantum numbers as the corresponding interpolating current. Isolating the contributions of the ground state baryons, one can easily obtain where m 1 and m 2 are the masses of the initial and final heavy baryons, p+q and p represent their four-momentum, respectively, and dots represent contributions coming from higher states and continuum. It follows from Eq. (2) that in obtaining the phenomenological part of the correlation function, the matrix elements, 0 η (i) (x) B ( p) , B(p)P (q) | B(p + q) and B(p + q) |η µ (0)| 0 are needed. These matrix elements are determined as follows: where λ 2 and λ 1 are the residues of spin-1/2 and spin-3/2 heavy baryons, respectively, g is the coupling constant of heavy baryons with pseudoscalar mesons, and u µ is the Rarita-Schwinger spinor. Using Eqs. (3) and (2) and performing summation over spins of spin-1/2 and spin-3/2 baryons, s u(p, s)ū(p, s) = (/ p + m 2 ) , in principle, one can obtain the expression for the phenomenological part of the correlation function. But at this point appear two unpleasant problems: a) The spin-1/2 baryons also contribute to the matrix element 0 |η µ | B(p, 3/2) of spin-3/2 baryons (see also [11]). Indeed, hence, the current η µ couples to both spin-3/2 and spin-1/2 states. Using Eqs. (2), (4) and (5), one can see that the unwanted contributions coming from spin-1/2 states contain structures proportional to γ µ at the far right end or (p + q) µ . b) The second problem is related to the fact that the structures which appear in the phenomenological part of the correlation function are not all independent. Both these problems can be removed by ordering the Dirac matrices in a specific form. In this work, the Dirac matrices are ordered in the form / q/ pγ µ and the coefficient of the structure / qq µ is chosen in order to calculate the aforementioned strong coupling constant, which is free of the spin-1/2 contributions. Using the ordering procedure, we get the following representation for the coefficient of the selected structure in the phenomenological part: In order to obtain sum rules for the coupling constant appearing in Eq. (6), we need to calculate the correlation function also from the QCD side. Before calculating it, we shall first find the relations among the correlation functions corresponding to different transition channels. In more concrete words, we shall find the relations among the coefficient functions of the structure / qq µ for different transition channels. For this purpose we follow an approach whose main ingredients are presented in [14][15][16][17].
In obtaining the relations among the correlation functions describing various spin-3/2 to spin-1/2 heavy baryon transitions, as well as, in obtaining the theoretical part of QCD sum rules, the forms of the interpolating currents are needed. In constructing the interpolating currents, we will use the fact that the interpolating currents for the particles in sextet (anti-triplet) representations should be symmetric (antisymmetric) with respect to the light quarks. Using this fact, the interpolating current for baryons in sextet representation with J = 3/2 can be written as: where A is the normalization factor, a, b and c are the color indices. In Table 1, we present the values of A and light quark content of heavy spin-3/2 baryons. The general form of the interpolating currents for the heavy spin-1/2 sextet and antitriplet baryons can be written as ( for example see [18]): where β is an arbitrary constant and β = −1 corresponds to the Ioffe current and superscripts s and a stand for symmetric and antisymmetric spin-1/2 currents, respectively. The light quark content of the heavy baryons with spin-1/2 in the sextet and anti-triplet representations are given in Table 2. After introducing the explicit expressions for the interpolating currents, we are ready now to obtain the relations among the correlation functions that describe different transitions. It should be noted here that the relations which are presented below are independent of the choice of structures, while the expressions of the correlation functions are all structure dependent.
In order to obtain the relations among the correlation functions responsible for different transitions, we consider the Σ * 0 b → Σ 0 b π 0 and Σ * 0 → Λπ 0 transitions which describe sextet spin-3/2 to sextet spin-1/2 and sextet spin-3/2 to anti-triplet spin-1/2 transitions, respectively. These invariant functions can be written in the general form as where superscripts (1) and (2) correspond to S * SP and S * AP transitions, respectively. The interpolating current for π 0 is written as: It follows from this expression that, The invariant functions Π ) describe the radiation of π 0 from u, d and b quarks, respectively, and they are formally defined in the following way: Remembering the fact that the interpolating currents for sextet spin-3/2 and sextet spin-1/2 baryons are symmetric with respect to the exchange of light quarks, while the interpolating currents for spin-1/2 anti-triplet baryons are antisymmetric, we can write, Using these relations and Eqs. (9) and (11), we get, where i = 1 (i = 2) and upper (lower) sign describes S * SP (S * AP) transition. The invariant function responsible for the channels by noting that the interpolating currents for Ξ * 0 b , Ξ 0 ′ b and Ξ 0 b can be obtained from the one for Σ * 0 b , Σ 0 b and Λ 0 b by making the replacement d → s, and taking into account the fact that g π 0s s = 0. As a result, we get, The invariant functions corresponding to Ξ channels with the help of the replacement u → d, as a result of which we get, Calculation of the coupling constants of the sextet spin-3/2 to sextet and anti-triplet spin-1/2 transitions with other pseudoscalar mesons can be done in a similar way as for the π 0 meson. Note that in our calculations, the mixing between η and η ′ mesons is neglected and the interpolating current of η meson has the following form: that gives, Using this expression let us consider, for example, the Σ * 0 b → Σ 0 b η and Σ * 0 b → Λ 0 b η transitions. Following the same lines of calculations as in the π 0 meson case, we immediately get, The invariant function responsible for the Ξ * 0 b → Ξ 0 ′ b η and Ξ * 0 b → Ξ 0 b η transition can be written as: The relations among invariant functions involving charged pseudoscalar π ± mesons can be obtained from previous results by taking into account the following arguments. For instance, let us consider the Σ * + b → Λ 0 b π + transition. In the Σ * 0 b → Λ 0 b π 0 transition, the u(d) quark from Σ * 0 b and Λ 0 b baryons forms the finalūu(dd) state, and the d(u) and b quarks behave like spectators. In the case of charged π + meson, the d quark from Λ 0 b and u quark from Σ * 0 b form theūd final state, and the remaining d(u) and b quarks are the spectators. For this reason, one can expect that these two matrix elements should be proportional to each other and explicit calculations show that this indeed is the case. Hence, making the replacement u ↔ d in Eq. (21), we get All remaining relations among the invariant functions responsible for the spin-3/2 and spin-1/2 transitions involving pseudoscalar mesons are presented in the Appendix. After establishing the relations among the invariant functions, we now proceed by calculating the invariant functions from QCD side in deep Euclidean region −p 2 → ∞, −(p + q) 2 → ∞, using the operator product expansion (OPE). The main nonperturbative input parameters in the calculation of the theoretical part of the correlation function are the distribution amplitudes (DAs) of the pseudoscalar mesons. These (DAs) of the pseudoscalar mesons are involved in determining the matrix elements of the nonlocal operators between the vacuum and one pseudoscalar meson states, i.e., p(q) |q(x)Γq(0)| 0 and p(q) |q(x)G µν q(0)| 0 , where Γ is any Dirac matrix. The DAs of pseudoscalar mesons up to twist-4 accuracy are given in [19].
In the calculation of the theoretical part of the correlation function, we also need to know the expressions of the light and heavy quark propagators. The light quark propagator, in presence of an external field, is calculated in [20]: where γ E ≃ 0.577 is the Euler constant, and Λ is the scale parameter. In further numerical calculations, we choose it as Λ = (0.5 ÷ 1) GeV (see [21,22]). The heavy quark propagator in an external field has the following form: where S f ree Q (x) is the free heavy quark operator in x-representation, which is given by: where K 1 and K 2 are the modified Bessel function of the second kind.
Using the explicit expressions of the heavy and light quark propagators, as well as, definition of the DAs of the pseudoscalar mesons, the correlation function can be calculated from the QCD side. Choosing the coefficient of the structure / qq µ from both sides of the correlation function and applying double Borel transformations with respect to the variables −p 2 and −(p + q) 2 , in order to suppress the contributions of higher states and continuum, we get the sum rules for the strong coupling constants of pseudoscalar mesons with sextet spin-3/2 and spin-1/2 heavy baryons as: where M 2 1 and M 2 2 are the Borel masses in the initial and final channels, respectively. The masses of initial and final heavy baryons are very close to each other, so that we can choose The residues λ 1 of spin-3/2 and λ 2 of spin-1/2 are calculated in [23]. The explicit expressions for Π (i) 1 are quite lengthy, so as an example, we present only the Π (1) 1 , which is obtained as: where, The Dα i = dαqdα q dα g δ(1 − αq − α q − α g ), q 1 and q 2 are the light quarks, Q is the heavy quark, the subscript P stands for pseudoscalar meson and the functions A , A ⊥ , T , V , V ⊥ , φ σ , φ ′ σ , φ η , φ P , A and B are the DAs with definite twists for the pseudoscalar mesons. To shorten the above expression, we have ignored the light quarks masses as well as terms containing gluon condensates, but we take into account their contribution when doing numerical analysis. The continuum subtraction is performed using results of the work [15].
Numerical analysis
This section is devoted to the numerical analysis of the strong coupling constants of mesons with spin-3/2 and spin-1/2 heavy baryons. The main input parameters for performing numerical analysis are the DAs of the light pseudoscalar mesons, whose expressions are presented in [19]. The other input parameters appearing in the sum rules are, qq = −(0.24 ± 0.001) 3 GeV 3 , m 2 0 = (0.8 ± 0.2) GeV 2 [13], f π = 0.131 GeV , f K = 0.16 GeV and f η = 0.13 GeV . In the sum rules for the strong coupling constants of light pseudoscalar mesons with heavy baryons, there are three auxiliary parameters, namely Borel mass M 2 , continuum threshold s 0 and the arbitrary parameter β in the expressions of the interpolating currents of spin-1/2 baryons. It is clear that, any physical quantity, like the aforementioned strong coupling constants, should be independent of these auxiliary parameters. Therefore, we try to find so called "working regions" of these parameters, where g B * BP is practically independent of them. The upper limit of M 2 can be obtained by demanding that the higher states and continuum contributions contribute less than, say, 50% of the total dispersion integral. The lower bound of M 2 can be determined by requiring that the highest power in 1/M 2 should be less than (20-25)% of the highest power M 2 . These two conditions allow us to fix the following working regions: 15 GeV 2 ≤ M 2 ≤ 30 GeV 2 for the bottom baryons, and 4 GeV 2 ≤ M 2 ≤ 12 GeV 2 for the charmed baryons. As far as continuum threshold is concerned, we choose it in the interval between s 0 = (m B + 0.5) 2 GeV 2 and s 0 = (m B + 0.7) 2 GeV 2 .
As an example, in Figs. 1 and 2, we present the dependence of the coupling constants for the Σ * 0 c → Λ 0 c π 0 and Ξ * + c → Ξ + c π 0 transitions on M 2 , at different fixed values of β and at s 0 = 10.5 GeV 2 . From these figures, we see that the coupling constants for these transitions exhibit good stability when M 2 is varied in the above mentioned "working region". Depicted in Figs. 2 and 4 are the dependences of the same coupling constants on cos θ at several fixed values of s 0 and at M 2 = 8 GeV 2 , where β = tan θ. From these figures, we observe that when cos θ is varied in the domain −0.3 ≤ cos θ ≤ 0.5, the coupling constants show rather stable behavior and they also have very weak dependence on s 0 . Similar analysis for the strong coupling constants of all S * SP and S * AP is performed and the results are presented in Tables 3 and 4, respectively. For completeness, in these Tables we also present the predictions of the Ioffe current (β = −1) for these coupling constants. Here, we would like to remind that, our obtained domain for cos θ lies inside the more wide interval obtained from analysis of mass sum rules for heavy non strange baryons in [18,24].
Note that only few of the presented coupling constants can be measured directly from the analysis of the decays, and the remaining coupling constants can only be measured, indirectly. At present, the decay widths of the Σ * ++ c → Λ + c π + and Σ * 0 c → Λ + c π − are measured, experimentally and also the upper bounds for the Σ * + c → Λ + c π 0 , Ξ * + c → Ξ 0 c π + , Ξ * + c → Ξ + c π 0 , Ξ * 0 c → Ξ + c π − and Ξ * 0 c → Ξ 0 c π 0 are announced (see [4] and [6]). Using the matrix element for the 3/2 → 1/2π transition, i.e., one can easily obtain the following relation for the corresponding decay width: where | q| is the momentum of the π meson. Using the values of the coupling constants from Tables (3) and (4), and also Eq. (30), we can easily predict the values of the corresponding decay widths. Our predictions on these decays, the experimental results, as well as predictions of other approaches on these coupling constants are presented in Table 5. From this Table, we see that our predictions on decay widths for the above-mentioned kinematically allowed transitions are all in good agreement with the existing experimental results and the prediction of other approaches.
In summary, we calculated the strong coupling constants of spin-3/2 to spin-1/2 transitions with the participation of pseudoscalar mesons within LCSR. Our analysis shows that all S * SP and S * AP couplings are described by only one invariant function in each class of transitions. Moreover, we estimated the widths of the kinematically allowed transitions, which match quite good with the existing experimental data, as well as predictions of other approaches.
Appendix A :
In this appendix we present the expressions of the correlation functions in terms of invariant function Π (1) 1 and Π (2) 1 involving π, K and η 1 mesons. | 2011-05-05T17:02:59.000Z | 2010-12-29T00:00:00.000 | {
"year": 2011,
"sha1": "9b2fbf5421aaef2788c89e35d44c3b968acc0c72",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1012.5935",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "9b2fbf5421aaef2788c89e35d44c3b968acc0c72",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
239595640 | pes2o/s2orc | v3-fos-license | Converting waste PET plastics into automobile fuels and antifreeze components
With the aim to solve the serious problem of white plastic pollution, we report herein a low-cost process to quantitatively convert polyethylene terephthalate (PET) into p-xylene (PX) and ethylene glycol (EG) over modified Cu/SiO2 catalyst using methanol as both solvent and hydrogen donor. Kinetic and in-situ Fourier-transform infrared spectroscopy (FTIR) studies demonstrate that the degradation of PET into PX involves tandem PET methanolysis and dimethyl terephthalate (DMT) selective hydro-deoxygenation (HDO) steps with the in-situ produced H2 from methanol decomposition at 210 °C. The overall high activities are attributed to the high Cu+/Cu0 ratio derived from the dense and granular copper silicate precursor, as formed by the induction of proper NaCl addition during the hydrothermal synthesis. This hydrogen-free one-pot approach allows to directly produce gasoline fuels and antifreeze components from waste poly-ester plastic, providing a feasible solution to the plastic problem in islands.
24. SI line 105: ICP calibration method used? 25. Figure S2: Please add the initial HT sample result as well for comparison.
Reviewer #2: Remarks to the Author: PET contributes significantly to plastic waste generation. In this work, Zhao and co-workers developed a new H2 free method using Cu based catalyst to convert PET into xylene and ethylene glycol. The conversion efficiency is high and the reaction pathway has been well studied. The catalyst structureactivity correlation has also been convincingly described. Overall it is a nice piece of study that deserves to be published in Nat. Commun. after proper revision. 1) The stability of the catalyst. I feel this is a major limitation of the work. There is no adequate information on the stability of the Cu catalyst. Detailed characterizations of the spent Cu catalyst may be provided. Further, the reusability of the catalyst may be studied in more detail. If direct reuse is not possible, what is the reason for catalyst deactivation and whether it is possible to recover catalyst activity by certain treatment.
2) The authors may also comment on the applicability of the catalytic system beyond PET. If it is only applicable to PET, then sorting strategies have to be applied.
3) Methanol is used as solvent and hydrogen donor. How much methanol is decomposed during the process? Does the consumption of methanol match that of PET conversion? 4) It would also be good if the authors comment on how to purify products and reuse unreacted methanol. 5) It is interesting to see the example of Phuket Island. Did the authors really use samples collected there, or it is just based on literature report? This needs to be made clear. If the sample from Phuket Island is not used, then it should be removed from Figure 5. Related parts in MS should be modified as well.
Reviewer #3: Remarks to the Author: The manuscript "Converting waste PET plastics into automobile fuel and antifreeze components" by Zhao and coworkers describes a novel methodology for the depolymerization of PET waste using methanol as the solvent and hydrogen source, and a Cu-based catalyst. This new method is very important due to the use of an alcohol as reducing agent and also a non-toxic and earth abundant metal catalyst. The work is well written and the discussion of the results appropriate, involving an extensive study of reaction conditions. Some aspects of the reaction mechanism were also included in this study.
I recommend the publication of this manuscript in nature communications, after major revisions: -Authors should test this method using other alcohols, for example ethanol and isopropanol, as the hydrogen source and verify that PET depolymerization and p-xylene formation also occur. -Authors should also indicate whether the PET alcoholysis reaction can be carried out at temperatures below 210 °C and what yields are obtained.
-To study the applicability of this method, it would be very interesting if the authors tested this method from the second most used polyester, polybutylene terephthalate (PBT), and checked if it is also possible to obtain p-xylene with good yields. We are grateful for the constructive comments from three reviewers, and based on these suggestions, we have modified and improved our manuscript. The detailed responses to the comments of the reviewers are listed in blue together with the text of the original material.
Manuscript ID: NCOMMS-21-40636A
Title: Converting waste PET plastics into automobile fuel and antifreeze components
Reviewer #1:
The article describes hydrogen gas-free conversion of PET plastic to para-xylene and glycol. In itself the work is interesting, but many questions remain regarding the catalytic compounds, rationale, comparison with competing technology, as well as experimental details that I would like the authors to address. I have listed them below.
1. The rationale for converting PET waste into fuel baffles me a bit. In this work PET is converted first to glycol and DMT, after which DMT is converted to PX for fuel applications. This would mean that valuable resources, especially aromatics, will just be burnt rather than reused. Given the limited amount of resources available on the planet one should focus on reusing resources rather than burning them. Please discuss the rationale of the work in light of this in the introduction.
Reply: Thanks for the Reviewer's suggestion. We agree with you that plastic waste is one of the most valuable wastes and can be considered as a potentially cheap source for the production of industrial fuels and chemicals. In land cities, it is feasible to converted plastics into chemical raw materials through effective chemical 2 recycling methods. However, in some islands, especially those with developed tourism, due to the lack of industry on the island, a large amount of abandoned plastic waste can only be disposed by landfill or incineration. Recycled plastics are also generally limited to feasible applications where low-quality materials are collected, resulting in minimal economic incentives for waste recycling, sorting and processing. At this time, if it can be used as a raw material to output gasoline energy and antifreeze components on the island through a simple process, it will be a very practical way.
2. PET depolymerization has recently been reviewed by Barnard, Rubio and Thielemans in Green Chemistry. This work also introduces a methodology to quantifiably compare different processes based on materials use (using Sheldon's factor) and energy consumption. Please compare the efficiency of this work with the existing literature using this (or a similar) methodology.
Reply: Thanks for the Reviewer's suggestion. We tried to use the environmental factor and environmental energy impact in the Barnard, Rubio and Thielemans's work to evaluate the efficiency of several parallel works. Firstly, energy economy coefficient (ε) is proposed to enable objective comparison on the influence of para meters such as temperature, catalyst type, or proportion of starting materials, where t is the reaction time (in minutes), T is the reaction temperature in degrees celsius, and Y is the yield of the main monomer in mass fraction (which containing the aromatic moiety) in eqn (1). Barnard et al. improved the environmental factor (Efactor) in eqn (4) which took the effect of materials input that results in waste generation into consideration. The environmental energy impact factor (ξ) results from the combination of the two factors above as presented in eqn (5). The best processes would tend to present low values of Efactor and ξ factor and high ε values.
Table R1 clearly showed that this work has the highest ε (1.323E-5°C -1 *min -1 ) due to its excellent product yield (100%) and low reaction temperature (210 °C) and time (360 min). High solvent/PET ratio (197.5) resulted in the high Efactor (37.19). But ξ (2811035°C*min) is still the smallest by combining the above two coefficients, and about twice time than other works. We also tried to redouble the PET and catalyst at the same time, and still get 100% PX yield, greatly reduced Efactor as well as ξ. 3 In addition, only this work uses non-noble metal catalysts, and the obtained products are highly selective. 3. PET recycling in the introduction is heavily focused on conversion to fuel, whereas depolymerization into monomers would economically make more sense (less steps and no loss of resources) yet is not covered. Please add this to the introduction.
Reply: Thanks for the Reviewer's suggestion. We have added the depolymerization of PET into monomers in introduction.
Chemical depolymerization methods, mainly include hydrolysis, glycolysis, and ammonolysis [3][4][5][6] , can reverse the chemical composition of plastics and turn into stable monomer molecules again. However, these methods still face limitations of harsh reaction conditions, low product yield, and purification difficulties. Reply: Thanks for the Reviewer's suggestion. We repeated each experiment three times, and the error bars are added in following figures. Reply: Thanks for the Reviewer's suggestion. Due to the small particle size, poor crystallinity and weak intensity, XRD results maybe not suitable for Rietveld refinement. We compared the results with the standard spectrum (PDF#27-0188) in the database [7][8] , and the results are shown in Figure 2a. Ref. The O1s XPS of reduced catalysts was conducted, as shown in Figure R1. The peak at 530.3 eV was observed in the reduced CuNa/SiO2 and Cu/SiO2, ascribed to Cu2O or CuSiO3 as reported.
[9-10] The peak of 532.5 eV was attributed to the O 1s of SiO2 support [10] . However, we cannot distinguish the Cu2O and CuSiO3 species through the O1s XPS analysis.
Ref.
[9] Ding, J. 7. Line 99: How was partial reduction of Cu + to Cu 0 measured? How much is partial reduction? Please quantify.
Reply: Thanks for the Reviewer's suggestion. The compositions of reduced catalysts were calculated from XPS and XAES analysis [12][13] , as shown in Supplementary to the higher difficulty in reduction. The X-ray induced Auger spectra (XAES) Cu LMM were employed to distinguish between the Cu 0 and Cu + and the deconvolution results are listed. The higher Cu + /Cu 0 ratio (1.87) of CuNa/SiO2 confirmed that after the addition of Na + , copper silicate with a low crystallinity and a dense texture was less likely to be reduced to Cu 0 . A higher ratio of Cu + /Cu 0 was indicative of a higher tendency to both methanol dehydrogenation and DMT hydrodeoxygenation.
This part has been modified in the revised version. 8. Line 112: XRD data in Figure 2S do not say anything about surface species, only bulk species. Please provide and describe surface analysis info as additional info.
Supplementary
Reply: Thanks for the Reviewer's suggestion. To investigate the surface information of Cu/SiO2 samples prepared by different methods, we used XPS and Auger Cu 11 LMM analysis to analyze the distributions of copper species on their surfaces ( Supplementary Fig. 3). On the XPS profiles of HT and DPA samples, there are obvious Cu 2+ satellite peaks (940-950 eV) ( Supplementary Fig. 3a), indicating that Cu 2+ was incompletely reduced. However, the same phenomenon does not appear in the XPS of DPU and IM Cu/SiO2 samples. The Cu LMM X-ray induced Auger spectra (XAES) were employed to distinguish between the Cu 0 and Cu + species ( Supplementary Fig. 3b). The results showed that the ratios of Cu + /Cu 0 in DPU, DPA and IM samples were significantly lower than such ratio in the HT sample, which corresponds to a significant decrease in reactivity of methanol dehydrogenation and PET hydrodeoxygenation (Table 1). This part has been revised in the renewed version. 10. Line 167: The way this is written seems like Cu + has a peak at 952.2 eV and Cu 0 at 932.1 eV. Please rewrite and also specify the orbital assignment.
Reply: Thanks for the Reviewer's suggestion. We have already fixed the typographical error and rewrite it as below.
13. Line 210: "poor crystallinity". Please quantify Reply: Thanks for the Reviewer's suggestion. We used the software MDI-Jade to fit the XRD data of samples firstly, and then calculated the crystallinity of each sample.
The crystallinity and R-values (fitting error) are listed below.
14 Supplementary Line 231: "large interface area" How was this determined? Quantification?
Reply: In order to roughly quantify the interface areas of the formed copper silicate with SiO2, we supplemented the measurement of IR spectroscopy in vacuum and determined the remaining silanols groups on SiO2 (Supplementary Fig. 14). The peak at 3740 cm -1 in the SiO2 sample is attributed to Si-OH groups ( Supplementary Fig. 14) [16][17] . With the traditional hydrothermal method, Cu 2+ in the solution combined with the silanol on the SiO2 surface to form copper silicate, which accelerated the layered copper silicate nucleation and growth significantly. Thus only a small amount of Si-OH can still be detected. Upon addition of 5 NaCl, a large amount of Na + occupied the silanol on the surface of the SiO2, Cu 2+ in the solution could only be combined with the remaining silanol on the SiO2 surface to form scattered and isolated copper silicate particles (Figures 2g and 2i), which was attached to the surface of the carrier. Thus, the formed granular copper silicate showed a large interface area with SiO2, since no remaining Si-OH on CuNa/SiO2 was detected by IR spectra in vacuum ( Supplementary Fig. 14). When 15 Na + was introduced, Na + occupied almost all the Si-OH on the surface, and Cu 2+ can only combined with SiO3 2− in the solution to form granular copper silicate and then deposited on the SiO2. After washed with deionized water, some Si-OH groups are exposed, the peak at 3740 cm -1 is reserved. Therefore, upon adding 15 Na + , this type of copper silicate showed 15 better crystallinity (Supplementary Table 6) and small interface areas with SiO2 ( Supplementary Fig. 14).
This part has been modified accordingly. 16. Line 227: "inhibiting nucleation and growth" How was this determined? 16 Reply: Thanks for the Reviewer's suggestion. Transmission electron microscopy (TEM) images intuitively showed the different morphologies of the two copper silicates formed with and without NaCl introduction during the hydrothermal process. Thus, while the dried precursor of Cu/SiO2 showed a layered copper silicate structure (Figure 2g), the dried precursor of CuNa/SiO2 showed a special state of granular particle accumulation (Figure 2i). Therefore, it is inferred that Na + introduction indeed inhibits the nucleation and growth of layered copper silicate precursor. 17. What is special about the 5:1 ratio Na + /Cu 2+ that this would give rise to the optimal Cu + /Cu 0 ratio? Reply: Thanks for the Reviewer's suggestion.
When the Na + /Cu 2+ ratio is 5:1, the obtained granular copper silicate has the smallest specific surface area, the lowest water content and the poorest crystallinity ( Supplementary Fig. 10-13). TGA tests of the CuNa/SiO2 precursor showed that physisorbed water (2.41%) and crystal water (6.75%) upon addition of 5 NaCl was the lowest among all the samples tested ( Supplementary Fig. 11). This also confirmed that the copper silicate structure was densest at this ratio. N2 adsorption-desorption 17 ( Supplementary Fig. 12) revealed that CuNa/SiO2 had the lowest surface area (46.9 m 2 /g) upon addition of 5 NaCl, indicating that the formed structure was the most compact among the samples tested herein.
A large amount of Na + occupied the silanol on the surface of the SiO2 upon addition of 5 NaCl, thereby inhibiting nucleation and growth of layered copper silicate (Figures 2g and 2i). Cu 2+ in the solution could only be combined with the remaining silanol on the SiO2 surface to form scattered and isolated copper silicate particles, and the compact structure had a small surface area and poor crystallinity. This granular copper silicate grown from the surface of SiO2 is more stable and difficult to be reduced, resulting in a higher Cu + /Cu 0 ratio and benefiting the further methanol dehydrogenation and DMT hydrodeoxygenation.
However, when the amount of added NaCl was too high, Na + occupied all the silanol sites on SiO2, resulting in the precipitation of Cu 2+ with SiO3 2− in solution to form copper silicate, which was then deposited on the SiO2 surface. Compared to the catalyst with 5 NaCl introduced during hydrothermal treatment, this type of copper silicate showed better crystallinity (Supplementary Table 6) and was relatively easier to be reduced to Cu/Cu2O·SiO2 with a low ratio of Cu + /Cu 0 ( Supplementary Fig. 9d). In general, the addition of NaCl in the hydrothermal treatment resulted in the formation of granular copper silicate with a lower crystallinity, smaller specific surface area, and denser texture, which leads to form an optimal Cu + /Cu 0 ratio after reduction. Reply: Thanks for the Reviewer's suggestion. A recent survey of beach sediment along the coastline of the Phuket Island showed that PET (mainly containing beverage bottles, plastic films, and microwave packaging) accounted for ca. 33.1% of the overall plastic sediment (Supplementary Fig. 17). Several common PET plastics that are available on the tourist island were chosen to convert, such as Coca-Cola bottles, McDonald's drink caps, disposable lunch boxes, packaging bags and even some polyester clothes ( Supplementary Fig. 18a). After simple treatment with the raw materials by scissors ( Supplementary Fig. 18b), we obtained 100% yield of p-xylene from different sources of PET plastics at the same catalytic conditions.
Considering that lots of plastic wastes on the island sediments landfill are mixed together, in this context, we used CuNa/SiO2 to catalyze the mixtures of PET and another plastic PBT at 210 °C, and the results showed that mixed plastics can be completely converted to p-xylene as well.
Supplementary Fig. 17 Compositions in sediments along the coast of Phuket island. Reply: Thanks for the Reviewer's suggestion. We added all the chemical purities. Reply: Thanks for the Reviewer's suggestion. Inductively coupled plasma atomic emission spectroscopy (ICP-AES): The content of each element in the catalyst sample was determined by a PerkinElmer Optima 8300 inductively coupled plasma atomic emission spectrometer. The test process was as follows, Firstly, the catalyst sample was dissolved in hydrofluoric acid to ensure that it was completely dissolved and in a clear state. Finally, the solution was diluted to a suitable test range. The five standard solutions were prepared to construct the external standard curve. The content of elements in the samples was determined by external standard curve. We have 23 added this part into the revised manuscript.
25. Figure S2: Please add the initial HT sample result as well for comparison.
Reply: Thanks for the Reviewer's suggestion. We have added the initial HT sample result in the revised version. Comments: PET contributes significantly to plastic waste generation. In this work, Zhao and co-workers developed a new H2 free method using Cu based catalyst to convert PET into xylene and ethylene glycol. The conversion efficiency is high and the reaction pathway has been well studied. The catalyst structure-activity correlation has also been convincingly described. Overall it is a nice piece of study that deserves to be published in Nat. Commun. after proper revision.
1) The stability of the catalyst. I feel this is a major limitation of the work. There is no adequate information on the stability of the Cu catalyst. Detailed characterizations of the spent Cu catalyst may be provided. Further, the reusability of the catalyst may be studied in more detail. If direct reuse is not possible, what is the reason for catalyst deactivation and whether it is possible to recover catalyst activity by certain treatment.
Reply: Thanks for the Reviewer's suggestion. After we tested the catalyst for four bathes, the catalyst deactivated obviously (Supplementary Table 9). According to XRD results, a large amount of Cu 0 was reduced and the size of particles was increased ( Supplementary Fig. 16a). In line with XRD results, TEM image showed that the Cu particles size was also increased after recycling tests ( Supplementary Fig. 16b-c). XAES analysis proved that the ratio of Cu + /Cu 0 drastically decreased from 1.80 to 0.57 ( Supplementary Fig. 16f), which hindered the synergistic effect of Cu 0 and Cu + in the catalytic process. In addition, Cu 2+ was partially reduced by the excess hydrogen produced by methanol dehydrogenation, as shown in XPS of Supplementary Fig. 16d-e.
Concerning on the recovery of catalyst activity, we tried to calcinate the used catalyst at 450°C in an air atmosphere for 4 h, and then reduced it at 450 °C in a hydrogen atmosphere for 4 h. The obtained catalyst continued for cycle testing and the results showed that PX yield attained 74.8%, 62.5% in the runs 1 and 2, while such recovered catalyst totally deactivated in the third run, probably due to the difficulty in rebalancing the ratio of Cu 2+ , Cu + and Cu 0 species in copper silicate catalysts. Fig. 16 (a) XRD patterns of used CuNa/SiO2 catalyst and CuNa/SiO2 (reduced); TEM image of (b) used CuNa/SiO2 catalyst and (c) CuNa/SiO2 (reduced);
2) The authors may also comment on the applicability of the catalytic system beyond PET. If it is only applicable to PET, then sorting strategies have to be applied. 3) Methanol is used as solvent and hydrogen donor. How much methanol is decomposed during the process? Does the consumption of methanol match that of PET conversion?
Reply: Thanks for the Reviewer's suggestion. By calculating the consumption of methanol and the production of hydrogen, the experimental results showed that the values basically match. The following is the calculation process： Reactions involved in the PET conversion process: Real methanol consumption (according to liquid phase)： 28 Experimental Method: 30 mL methanol was diluted 10 times in ethyl acetate with 0.1 mL tetrahydronaphthalene as the internal standard before the reaction and determined the response factor (f) by GC-MS using following formula.
The reacted solution was diluted 10 times with ethyl acetate, the residual amount of methanol is calculated by the following formula subsequently. According to the reduction of peak area, the residual methanol content was 97.6%. 4) It would also be good if the authors comment on how to purify products and reuse unreacted methanol.
Reply: Thanks for the Reviewer's suggestion. Both the products and methanol can be separated by simple distillation based on their different boiling points (see Supplementary Table 10 below). Fig. 18a). After simple treatment with the raw materials by scissors ( Supplementary Fig. 18b), we obtained 100% yield of p-xylene from different sources of PET plastics at the same catalytic conditions. Considering that most plastic wastes on the island sediments landfill are mixture together, in this context, we used CuNa/SiO2 to catalyze the mixture of PET and PBT at 210 °C, and the results showed that mixed plastics can be completely converted to p-xylene as well. The manuscript "Converting waste PET plastics into automobile fuel and antifreeze components" by Zhao and coworkers describes a novel methodology for the depolymerization of PET waste using methanol as the solvent and hydrogen source, and a Cu-based catalyst. This new method is very important due to the use of an alcohol as reducing agent and also a non-toxic and earth abundant metal catalyst.
Supplementary
The work is well written and the discussion of the results appropriate, involving an extensive study of reaction conditions. Some aspects of the reaction mechanism were also included in this study. I recommend the publication of this manuscript in nature communications, after major revisions: 1.Authors should test this method using other alcohols, for example ethanol and isopropanol, as the hydrogen source and verify that PET depolymerization and p-xylene formation also occur.
Reply: Thanks for the Reviewer's suggestion. We have supplemented the experiments on the PET conversion in ethanol and isopropanol in Supplementary Table 3.
Experimental data showed that PET can be well alcoholyzed in both ethanol and isopropanol, obtain 80.3% and 73.5% yields of monomers after 0.5 h, respectively.
However, the hydrogen released from ethanol and isopropanol decomposition over CuNa/SiO2 was not sufficient, which attained only 0.7 and 0.8 MPa incremental pressure at ambient temperature, respectively. In comparison, methanol released as high as 3.8 MPa gases at identical conditions. In the further PET alcoholysis and hydrodeoxygenation tests in ethanol and isopropanol, the gained DMT monomers from PET were not hydrogenated and no p-xylene was formed in ethanol or isopropsanol (Supplementary Table 3 3. To study the applicability of this method, it would be very interesting if the authors tested this method from the second most used polyester, polybutylene terephthalate (PBT), and checked if it is also possible to obtain p-xylene with good yields.
Reply: Thanks for the Reviewer's suggestion. We tested PBT conversion in methanol over CuNa/SiO2 at different temperatures in the same catalytic system, and the results are very similar to PET conversion (Supplementary and 72.4%. DMT was not further converted due to the low gas pressures (0.5 and 0 MPa) at these two temperatures. | 2021-10-23T15:08:22.043Z | 2021-10-21T00:00:00.000 | {
"year": 2022,
"sha1": "807f7a039fdd57eee74f4fd963c5777b83392d16",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41467-022-31078-w.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "683e6f2df17818648b5c6a0317391cb709a5a9ba",
"s2fieldsofstudy": [
"Environmental Science",
"Materials Science",
"Engineering",
"Chemistry"
],
"extfieldsofstudy": [
"Medicine"
]
} |
5776686 | pes2o/s2orc | v3-fos-license | Sequencing Overview of Ewing Sarcoma: A Journey across Genomic, Epigenomic and Transcriptomic Landscapes
Ewing sarcoma is an aggressive neoplasm occurring predominantly in adolescent Caucasians. At the genome level, a pathognomonic EWSR1-ETS translocation is present. The resulting fusion protein acts as a molecular driver in the tumor development and interferes, amongst others, with endogenous transcription and splicing. The Ewing sarcoma cell shows a poorly differentiated, stem-cell like phenotype. Consequently, the cellular origin of Ewing sarcoma is still a hot discussed topic. To further characterize Ewing sarcoma and to further elucidate the role of EWSR1-ETS fusion protein multiple genome, epigenome and transcriptome level studies were performed. In this review, the data from these studies were combined into a comprehensive overview. Presently, classical morphological predictive markers are used in the clinic and the therapy is dominantly based on systemic chemotherapy in combination with surgical interventions. Using sequencing, novel predictive markers and candidates for immuno- and targeted therapy were identified which were summarized in this review.
Introduction
Ewing sarcoma (EWS) is a high-grade sarcoma occurring predominantly in the bones of children and young adolescents, in which it is the third most common primary bone sarcoma, following osteosarcoma and chondrosarcoma. In adults, it occurs less frequently, but at this age, soft tissue and organ related involvement is more common [1,2]. At the cellular level, EWS has a poorly differentiated, stem cell-like phenotype with some degree of neurogenic features. These were partly represented by earlier classification as peripheral primitive neuroectodermal tumors (PNET). In the current World Health Organization (WHO) classification, however, PNET and a clinical variant of EWS known as Askin tumor, arising in the chest wall, are all classified as EWS based on the presence of a unifying pathognomonic chromosomal translocation [1]. This translocation forms a chimera gene fusing the EWSR1 gene with a member of the ETS transcription factor family. Of the EWSR1-ETS translocations, EWSR1-FLI1 is the most common with 85% of the cases. Other partners of EWSR1 are ERG (10%), ETV1, ETV4 and FEV [2]. No difference in survival was observed between the different translocation types [2]. There is an increasing body of evidence from tumors with histopathological appearance of EWS without the involvement of EWSR1 and/or ETS. The clinical relevance of this Ewing-like tumor family from classical EWS is yet unknown and is studied [3][4][5]. The incidence of EWS is three per million and around a nine-fold more in Caucasians compared to Africans [6]. A suggested genetic explanation for this is the presence of intronic Alu elements (retrotransposons) located near the breakpoint region. In the African population, an allele which lacks the majority of the Alu elements has been identified with an allele frequency of 8% [7]. Alu elements are potentially more preferred during recombination and their increase could increase the chance of a translocation to occur [8]. The lack of Alu repeats may contribute, but it cannot be the leading mechanism behind the observed difference in tumor incidence. Furthermore, a similar occurrence in Alu distribution was not observed in other EWSR1 translocation positive sarcomas, like clear cell sarcoma [9]. A large genome-wide association study (GWAS) on EWS identified no single-nucleotide polymorphism (SNP) association at the EWSR1 and ETS breakpoints. However, they did find three SNPs; rs9430161 on chromosome 1 upstream TARDBP (Tat activating regulatory DNA-binding protein), rs224278 on chromosome 10 upstream EGR2 (early growth response 2) and rs4924410 at locus 15q15, which were associated with EWS with odds ratio of 2.2, 1.7 and 1.5, respectively. EGR2 is a target of EWSR1-FLI1 and TARDBP was proposed to be structurally and functionally similar to EWSR1 [10]. Further validation is required for the SNP at 15q15, since multiple genes are located in close proximity of it. The SNPs on chromosome 10 and 1 were more frequent present in Caucasians compared to Africans and could thereby be a factor in the differences in incidence of EWS in different racial patient populations [10]. Recently, another possible cause of the epidemiologic difference in the occurrence of EWS has been proposed. The EWSR1-ETS chimeric protein binds to GGAA microsatellites which differ in distribution between Caucasians and Africans. Caucasians have a higher frequency of repeats of 20-30 GGAA elements compared to Africans, which have a higher frequency of repeats longer than 30 elements. In a reporter gene assay, the highest EWSR1-FLI1 expression was observed when the GGAA microsatellite consisted of 20-30 motifs and this was concordant with the EWS target gene expression in relation to GGAA microsatellite length in EWS cell lines [11,12]. This suggests that the expression inducible capability of EWSR1-ETS can be larger in Caucasians compared to Africans.
Relation between EWSR1-ETS and the Cell of Origin of EWS
There is an ongoing debate on the identification of the cell of origin of EWS. Expression of the fusion protein leads to more stem cell-like phenotypes and expressions of neuro-ectodermal markers [13]. In addition, EWSR1 is expressed in many tissues, its function is poorly understood and the EWSR1 gene is involved in translocations in multiple other tumors [14][15][16][17]. Multiple cells of origin have been suggested, such as mesenchymal stem and neural crest cells [13,18,19]. In order to shed some light on this debate, the effect of induced expression of the chimeric protein in non-tumorigenic cells was investigated. It was expected that the translocation had a large impact on cell homeostasis and interfered at multiple levels in endogenous processes. To study the impact of this gene chimera, primary human fibroblasts were transfected with an EWSR1-FLI1 construct and that led to a TP53 dependent growth arrest. This points towards the need of additional (secondary) changes to be able to transform [20]. Likewise, in other studies, which used EWSR1-ETS transfected adult human mesenchymal stem cells (MSCs), an additional mutation was needed for the cells to form tumors; while transformation was possible using unmodified pediatric MSCs [18,21,22]. Animal models containing inducible EWSR1-FLI1 constructs led to phenotypically varying tumors from malignant peripheral nerve sheath tumors to myeloid/erythroid leukemia [23,24]. These observations directed towards a hypothesis that certain epigenetic changes might be needed to result in an EWSR1-ETS driven tumor and that this partly dictates the phenotype of the tumor. The presently hypothesized cells of origin are MSCs and neural crest cells. This is based on their capability to endure expression of EWSR1-ETS gene chimera without additional mutations, and the finding that transient EWSR1-ETS expression leads to a tumor similar to EWS at the level of expressed cellular markers and micro-array expression data [22,25]. Recently, a new mouse model has been created to mimic EWS using specific selected cells of the embryonic superficial zone of the long bones. In these animals, EWS-like tumors developed without any additional gene modifications. This might be a leap forward in creating a mouse model for EWS [19]. To gain further insight into the tumor specific genetic changes multiple massive parallel sequencing studies were performed at the genome, the transcriptome and the epigenome level (Table 1). By combining the results of these studies, researchers may identify landscape marks in the EWS OMIC atlas explaining some of the mechanisms behind the behavior of Ewing sarcoma with the aim to identify new, targeted therapeutic targets These targets can be validated by combining functional studies and testing In addition, this might shed light on the cell of origin and secondary events necessary for tumor formation and changes that are related to a more therapy resistant or more aggressive phenotype.
Genome Map
To identify possible secondary genetic and genomic alterations related to the development of EWS and its biology, several groups performed genome-wide studies such as: whole genome sequencing (WGS), whole exome sequencing (WES) and whole transcriptome sequencing (WTS) [26][27][28]. These three types of studies included WGS of 123 tumor samples in parallel with the normal tissue derived germline controls, WES of 92 tumors of which 26 with paired normal control and 11 cell lines and WTS of 92 tumors and 42 cell lines resulting in data about structural rearrangements and variations, somatic mutations and expression profiles.
For a long time, EWS was known as a genetically stable tumor with rarely occurring additional mutations. Only a few genomic changes such as TP53 mutations or CDKN2A/CDKN2B deletions were observed in a minority of samples in retrospective studies and they were reported to be associated with an inferior outcome in a multivariate analysis [41,42]. The search for secondary mutations that provide a permissive genetic background, and might explain how the EWSR1-ETS chimera protein transforms cells, remained unsuccessful for over twenty years after the initial identification of the EWSR1-FLI1 fusion gene [43]. The goal of the genome sequencing studies was to identify the missing link in this area. Both WGS and WES studies detected only a very low number of somatic mutations (0.65-0.15 per Mb) although different statistics for analysis were used [27,28]. Similarly, the low number of single nucleotide variations (SNV) in EWS has been reported in an earlier study and was, when compared to other tumors, one of the lowest [29]. Possible causes for the low number of SNVs could be related to the pathognomonic gene fusion acting as a direct tumor driver, and to the young age of onset of the tumor with possible fewer gained environmental mutations. Rhabdomyosarcoma (RMS) consists of both fusion gene positive and negative subtypes and the fusion positive subtype contained significant less mutations compared to the fusion negative subtype [44]. The number of mutations detected in fusion positive RMS was similar to EWS. The number of additional mutations correlated with age in both the RMS and EWS, confirming an age related factor [28,44]. Another retrospective study confirmed that the increased number of somatic mutations was in a univariate analysis correlated to shorter survival time [27]. This might partly explain why an increased age is correlated with inferior prognosis in EWS, but it could also be due decreased tolerance to chemotherapy [45,46]. In biopsies, the most common kind of mutation detected was a C to T transition, which was linked to the common event of deamination of methylated cytosines [28]. The number of mutations was, as expected, increased in post-chemotherapy samples and an association between the increased numbers of novel mutations with a poor patient outcome was observed [28]. In theory, these clones might already have been present but remained undetected due to tumor heterogeneity. Alternatively, these mutations were caused by the treatment resulting in a drug resistance phenotype. This would be very interesting for understanding treatment response prediction. Overall, EWS is from a global genomic perspective a relatively stable tumor with low number of somatic mutations, implying a functional mutation recognition and repair mechanism.
Structural and Copy Number Variant Map
All bona fide EWS contained an EWSR1-ETS translocation and these were detected in all tumors and cell lines tested [26][27][28]. In the study by Brohl et al. [26], however, seven cases were identified with cellular phenotype similarly to EWS but without an EWSR1-ETS translocation, supported by the fact that these samples cluster separately based on RNA expression profile. This observation supports the notion of the existence of a Ewing-like tumor with clinical-and histo-morphological appearances similarly to EWS but carrying other, specific translocations such as, BCOR-CCNB3, EWSR1-NFATc2, FUS-NFATc2 and CIC-FOXO4 and CIC-DUX4 [3][4][5]. As these entities are rare, follow-up studies have to show if these groups should be further stratified based on the genes involved or might be lumped as one clinical entity, Ewing-like sarcoma. None of the sequenced EWS samples detected an additional, commonly occurring translocation co-existing with EWSR1-ETS.
Although EWS tumors with a complex karyotype occur in a minority of cases, there are some common chromosomal alterations. These are gain of chromosome 1q, 8, 12 and loss of 9p21 and 16q [47][48][49][50]. Gain of chromosome 1q and chromosome 16q loss were strongly co-associated caused by an unbalanced translocation der(16)t(1;16) [47,[51][52][53]. The frequency of 1q gain was, in various studies, associated with a dismal prognosis and was higher in chemotherapeutic treated tumors [27,28,47,48,54,55]. The responsible factor for this association was investigated by Mackintosh et al. who compared samples with and without 1q gain and 16 loss. At chromosome 1q, they identified increased expression of the gene Cell Division Cycle Protein 2 (CTD2), also known as Denticleless E3 Ubiquitin Protein Ligase Homolog (DTL), as the suspected factor [48]. DTL is, like TP53, involved in DNA damage repair and could therefore have an effect on tumor progression [56]. The chromosome 1q gain is not a EWS specific aberration, as it is one of the most frequently observed secondary changes in many tumor entities and even in cultured embryonic stem cells [57]. The large heterochromatic regions at 1q12 might be responsible for the frequent translocation breakpoint leading to gain of the long arm of chromosome 1 [58][59][60]. As was observed with 1q, gain of chromosome 8 and 12 was present in many other tumors summarized by the progenetix website [61]. According to this website, these chromosome gains might be linked to pluripotency and proliferation. Chromosome 12 gain has also been observed in cultured human embryonic stem cells [62]. The oncogene associated with the increased tumorigenity for chromosome 12 gain is not clear since next to NANOG it contains many genes including known oncogenes CDK4, ERBB3, GLI1 and MDM2. For chromosome 8 gain, the increased expression of the oncogene MYC may be the attributing factor; however, EWS without the gain of chromosome 8 show similarly high expression of MYC [63][64][65][66]. Homozygous loss of 9p21 is, with about 12%, less common in EWS but could have a large impact since a well-known cell cycle regulator CDKN2A/CDKN2B is in this locus. Huang et al. [41] demonstrated in a retrospective study of 60 patients that the loss of CDKN2A/CDKN2B has a negative effect on the overall survival. Recently, Tirode et al. [27] analyzed 300 EWS samples and did not observe a significant difference in overall survival of patients with or without CDKN2A/CDKN2B loss. This underlines the importance of large sample size in studies of EWS, when trying to predict the effect of genomic alternations on prognosis. However, no data on chemotherapeutic response was presented of these patients, which has been reported to be significantly worse in patients with a CDKN2A/CDKN2B loss [27]. Loss of heterozygosity (LOH) is detected in earlier studies in a minority of the patients and was investigated using micro-satellite instability markers and identification markers, but no overlapping chromosomal regions were detected [67,68]. A recent study examined LOH in only six EWS samples by using SNP microarrays and showed some overlapping chromosomal regions with the ones reported earlier [69]. These were 17p and 11p and may be relevant to verify since TP53 is located at the 17p chromosomal region. In several tumors inactivation of TP53 has been reported due to point mutations or, less frequently, homozygous deletion, or deletion in combination with point mutation due to LOH. Intriguingly, in EWS inactivation of TP53 caused by deletion was found as a rarely occurring event in earlier studies and was not even reported in any of recent large genomic landscape studies [26][27][28]41,48,49,70].
In the recent NGS studies, copy number alterations were detected in EWS but no common alteration was found. The copy number and structural alterations may merely represent secondary changes leading to a complex karyotype, which was found to be negatively associated with survival in earlier studies and was confirmed by genome sequencing study of Tirode et al. all retrospectively [27,47,71].
Mutation Map
Although EWS contains few SNVs, their distribution over the genome is quite specific. The most commonly affected genes found in the genomic landscape studies were STAG2 and TP53 with an occurrence in patients of respectively 9%-21.5% and 5.2% to 7% and both were in a retrospective study in a univariate analysis associated with poor prognosis [26,27,72]. The most commonly mutated gene STAG2 was only recently reported for the first time in EWS [73]. The distribution of the mutations is striking, with a quarter of the cases having a mutation at R216X, which is a possible CpG site and might be linked to a STAG2-DNA methylation pattern (see Figure 1A). The mutated STAG2 status correlated only with an increase in structural variants and no other of the tested parameters [27,28]. This observation may be related to the function of STAG2, as it is a subunit of the cohesin complex and involved in chromatin modeling, chromatin cohesion, repair of stalled replication forks and double-strand breaks (DSBs) [74][75][76][77]. STAG2 or other mutations in the cohesion complex were observed also in other tumors, including glioblastoma, myeloid malignancies, colon cancer and bladder cancer [72,73,78,79]. In colon cancer and glioblastoma, cohesin complex mutations were associated, like in EWS, with an increase in structural variants and aneuploidy [73,79]. In contrast, in myeloid malignancies this was not observed and in bladder cancer an inverse association was reported [80,81]. However, in myeloid malignancies, like in EWS, cohesin mutations were associated with poor prognosis [78]. In addition, when one of the cohesin complex genes was mutated in myeloid leukemia cell lines, less cohesin was bound to the chromatin [80]. Since cohesin is a key regulator of the chromatin structure and consequently influences gene expression, a reduction in the cohesin bound to the chromatin could affect the global gene expression [74,80,82]. TP53 is the second most common mutated gene in EWS and is one of the most common mutated genes in all tumors [83]. The frequency of TP53 mutations is slightly lower compared to earlier reports with an average of 10%. The two most frequent detected TP53 mutations were the p.C176F and p.R273X of which p.R273X has been reported earlier [84]. In the International Agency for Research on Cancer (IARC) database, p.R273X is, like in EWS, a hot spot mutation. Yet, the most frequent TP53 mutation in EWS p.C176F is remarkably not listed as a hot spot in the IARC database. In addition, the IARC database hot spot mutation p.R248Q is detected in only one tumor sample and only in one cell line, although it has been reported more frequent in earlier studies. This suggests that more samples are needed for a clear TP53 mutation pattern (see Figure 1B) [26][27][28]83,85]. Mutations of STAG2 and TP53 showed a trend for co-occurrence with a synergistic negative effect on prognosis when both mutations were present. They are both involved in the checkpoint and repair processes, which may be further abrogated when both genes are mutated [27]. A trend for mutual exclusivity of TP53 mutation and the loss of CDKN2A with only a few exceptions were present. Moreover, CDKN2A loss and STAG2 mutation were mutual exclusive [27,28]. This indicates that CDKN2A and STAG2 may be involved in complementary essential processes such as cell cycle and chromatin remodeling. Having a mutation in both genes may be lethal or redundant for EWS tumors [27,28,41,42,86,87]. To correct errors that may be caused by the relatively low numbers of cases analyzed, validation of these data in a bigger study is necessary. Other somatic gene mutations in EWS, described in three large genomic studies, were low and not recurrent. All three studies reported a different process to be most influenced by these somatic mutations. Tirode et al. [27] found mutations in several epigenetic regulators with EZH2 as the most frequent mutated gene (3/112 cases), whereas Crompton et al. [28] reported mutations in other ETS transcription factors, including ERF (3/46 cases). Brohl et al. [26] reported mutations in the DNA repair pathway, in specific, with the deleterious polymorphism K3326X in BRCA2 (4/55 cases) and a mutation in RAD51 (1/55 case). An earlier study identified only four mutations in 75 EWS tumors with a hotspot array of 275 recurrent mutations across 29 genes which were not reported by these large genomic studies [88]. A recent study in chemotherapy-treated EWS tumors observed mutations which had implications for further targeted therapy response, such as KRAS [30].
Genome-wide sequencing of EWS was expected to show a common secondary event that would help to understand and model Ewing sarcoma and its onset. However, no common secondary event was identified. Overall, EWS was found to be a relatively stable tumor with a low frequency of mutations, which were scattered across the genome and acted dominantly on cell cycle processes. This suggests that these mutations occur during tumor progression and may be used as a marker for tumor progression but are not associated with the onset of EWS. Consequently, this may indicate the involvement of other factors in the onset of EWS pointing to disturbances at the epigenetic level as potential candidate.
Epigenome Map
Epigenetic modification involves both histone and DNA modifications such as acetylation or methylation of histone proteins and methylation of CpG islands. The DNA accessibility for transcription factors and polymerases, and thereby transcription, is partly regulated by these modifications. Classical sequencing reactions are not suited for the detection of epigenetic changes, therefore additional treatments have to be applied to detect these modifications. Examples of treatments to detect DNA methylation are MeDIP-seq, methylated DNA immunoprecipitation sequencing and WGBS, whole-genome bisulfite sequencing [89,90]. More complex approaches should be used to detect modifications influencing histone composition, such as ChIP-seq, chromatin immunoprecipitation sequencing; ChIP-exo, chromatin immunoprecipitation-exonuclease, or the detection of DNase-I sensitive sites [31,91,92]. As these approaches are complex reactions and not uniformly applied in different laboratories, comprehensive epigenome mapping of tumors are rarely published, although the ENCODE project, specifically set up for this, has generated a general overview [31,33,91,92]. Many parts of the epigenome in tumors however have been reported, since it is thought to have great therapeutic potential [93][94][95][96]. Recently an epigenome overview has been published by Tomazou et al. [32] covering the epigenome and transcriptome of EWS cell line A673 with inducible EWSR1-FLI1 knockdown construct. Four separate clusters of histone marks were detected with different effects upon knockdown of EWSR1-FLI1 [32]. Furthermore unique EWS open chromatin structures at distant enhancer and super-enhancers sites were detected, suggesting an important role for epigenomic regulation [32]. This might be related to the earlier described binding of the EWSR1-ETS fusion protein to GGAA containing microsatellite elements at enhancer sites and thereby affecting expression of downstream located genes (see Figure 2A) [31,97,98]. However, experimental evidence is lacking here. Binding to GGAA elements is an ETS specific effect and acts specifically on genes which do not contain a TATA box promoter [99]. Examples of such genes are CAV1, NR0B1 and FCGRT. The binding of EWSR1-FLI1 to GGAA microsatellitesmight lead to multimer formation which is needed to attract sufficient number of chromatin remodelers necessary for the sustained expression [31,34,98].
An important attracted chromatin remodeler for this sustained expression is p300 that acetylates histone 3 lysine H3K27 (H3K27ac). Monomeric EWSR1-FLI1 binding to a single GGAA element could not activate transcription and even inhibited gene expression, marked by the H3K9me3 histone modification (see Figure 2B) [32].This might be due to insufficient attraction and binding of p300 since the fusion protein lacks a p300 binding site while wild-type ETS transcription with p300 binding sites could attract p300 and activate transcription [31,100]. In pediatric mesenchymal stem cells, induction of EWSR1-FLI1 led to a histone pattern at the EWSR1-FLI1 bound GGAA microsatellites which was similar to the pattern in EWS cell lines. Inhibition of EWSR1-FLI1 led to a decrease in activation of histone mark H3K27ac, which supports an active role of EWSR1-FLI1 in chromatin remodeling [31,32]. The H3K27 acetylation was especially associated with EWSR1-FLI1 bound enhancers [32]. It has to be noted that the overlap of ChIP-seq detected EWSR1-ETS binding sites was low with only 21% between EWSR1-FLI1 carrying cell lines and 17.2% between EWSR1-FLI1 and EWSR1-ERG carrying cell lines [31,36]. If these are all cell culture related artifacts or are due to accessibility of the DNA is not known. Another chromatin remodeling complex bound by EWSR1-FLI1 is the NuRD complex containing HDAC2 and HDAC3 proteins. These HDACs, when together with CHD4, can be active in the NuRD complex. Consequently, binding of the NuRD complex to EWSR1-FLI1 leads to repression of gene expression [101]. EWSR1-FLI1 regulated repression of expression was reverted by HDAC inhibitors and inhibiting histone demethylase LSD1, another NuRD complex protein. The NuRD complex is involved in many processes, especially in blood vessel development and integrity [32,[102][103][104]. The interaction of EWSR1-ETS with the epigenetic remodelers is further increased by binding of EWSR1-ETS to the promotor of enhancer of zeste homolog 2 (EZH2) and Sirtuin 1 (SIRT1), thereby upregulating this histone methyltransferase and deacetylase [105,106]. The EZH2 mediated effect in the cell was dependent on HDAC activity, demonstrating a cross interaction between two EWSR1-ETS modulated chromatin remodelers [105]. Overall, a complex interaction between EWSR1-ETS, chromatin and chromatin remodelers is needed in Ewing sarcoma to execute its oncogenic effect. As described earlier, transient expression of EWSR1-ETS in cells from different origin resulted in different phenotypes. This might be, in part, attributed to the chromatin state near GGAA microsatellites. An open chromatin structure at the enhancer and super-enhancer sites, as identified by Tomazou et al. [32], may be needed for the transforming effect of a EWSR1-ETS fusion protein in the development of Ewing sarcoma and if a more closed chromatin state was present an EWSR1-ETS translocation would lead to different effects or cell death. [32]. Although this is an attractive and plausible hypothesis, there is no experimental evidence yet to support this notion. The microenvironment, through for example, proliferative signaling, could greatly influence the chromatin state and have an interplay between EWSR1-ETS oncogenic properties. The other way around, tumor cell induced signaling can change the differentiation status of cells allocated in the tumor and distant microenvironment. in a yet unidentified activation complex and p300, which is needed for efficient transcription. The activation complex may bind to H3K4me3 and H3K27ac histone marks which, in turn, may lead to upregulation of the epigenetic modifiers SIRT1 and EZH2; (B) EWSR1-ETS repression complex binds to single GGAA elements and scavenges for p300, but, as it is insufficient to create an activating complex, it may recruit NuRD repression complex which may lead to further repressed expression. In addition, these repression sites are marked with H3K9me3 histone mark.
The type of mutations identified in EWS tumors pointed towards presence of methylated CpG sites, as mentioned in the genome map chapter. DNA methylation in EWS is studied only in a limited number of studies that used various techniques. In a recent relative small retrospective study by Park et al. [107], it was shown that patients with a poor outcome had increased methylation of CpG islands compared to patients with a better outcome, although the total hypermethylated genes was limited with only 10% of the investigated genes [107]. Their observation showed a similar proportion of genes with methylated CpG islands to an earlier study on DNA methylation using a different methylation micro-array [108]. Although the proportions were similar, the majority of the actual detected genes identified were different, having only six genes in common (LYN, EPHA3, ESR1, MAP3K1, NGFR and SOX17) in two studies. Compared to clear cell sarcoma and rhabdoid tumor of the kidney the same low number of hypermethylation of CpG islands was observed, but the number of significant hypomethylated genes was similar [109]. Since this study contained only four Ewing sarcoma samples, a larger study with more samples using the same platform should be performed. Whole genome DNA methylation was also performed in the earlier mentioned epigenome-wide study of Tomazou et al. [32]. Through WGBS, they observed less DNA methylation at actively expressed genes compared to non-expressed genes, suggesting an involvement of DNA methylation in the EWSR1-ETS mediated gene expression effect. However knockdown of EWSR1-ETS did not change the DNA methylation pattern. An alternative method to investigate the DNA methylation would be by using the PACBIO RSII sequencer system (Pacific Biosciences, Menlo Park, CA, USA). This system can detect the methylated CpG sites during the sequencing process and, as it does not need any amplification or chemical modification step, it has no probe bias. An unbiased sequencing approach could help to identify DNA methylation pattern in primary tumors and see if the pattern is the same in EWS tumors compared to cell lines. Since cell lines are used as models for EWS tumors and DNA methylation at whole genome level is only studied in cell lines.
Transcriptome Map
An EWSR1-ETS rearrangement affects gene expression levels, as mentioned above. In addition, it affects the expression of non-coding RNAs and splicing of RNAs by binding to the polymerase II complex protein hsRPB7 and to RNA helicase A (RHA) (see Figure 3) [40,110,111]. The effect of EWSR1-ETS on gene expression levels has been investigated with microarrays and studied in cell and animal models [19,20,22,112]. A meta-analysis of earlier micro-array studies was performed and compared the expression levels of other sarcomas demonstrating a specific EWS signature [113]. Knockdown studies of the most common fusion protein EWSR1-FLI1 revealed that it causes both downregulation and upregulation of numerous genes involved in extracellular and intracellular processes [35,114]. Downregulated genes were involved in extracellular signaling and signaling regulation, including multiple chemokines and interleukins (such as, CXCL8, CCL2 and IL1A) [38,101]. Upregulated genes were involved in neural differentiation, transcription and cell cycle and included membrane proteins [114][115][116][117]. Examples of external validated membrane proteins upregulated by EWSR1-ETS fusion protein are STEAP1, GPR64, CD99, CAV1 and CHM1 [116,[118][119][120][121]. These membrane proteins are interacting with the surrounding tumor microenvironment, thereby contributing to the high vascularization and invasive properties of EWS [116,119,122]. Validation of the EWSR1-ETS upregulated transcription factors NKX2.2, NR0B1, GLI1, BCL11B and E2F3 demonstrated an extensive attribution to the aggressive and stem-like phenotype of EWS [115,[123][124][125][126]. EWSR1-ETS affects gene expression, mainly downregulation, both directly and indirectly. Directly, by binding GGAA microsatellites and indirectly, by interacting with the NuRD co-repressor complex and upregulating above mentioned transcription factors [101,115,125]. Transcription initiation is commonly not regulated by one but multiple transcription factors which interact with each other. By interacting with transcription factors, such as E2F3 and Sp1, EWSR1-ETS enhances its ability to induce gene activation [35,126,127]. Although EWSR1-ETS needs variable different cellular processes for its effect at the transcriptome level, the EWSR1-ETS map was observed to be relatively stable. When comparing the transcriptomes of cell lines with tumors in a principle component analysis, only the first principle component of pathways was significantly different. The principle component consisted of tumor-microenvironment pathways in EWS tumors and metabolic pathways in cell lines [28]. EWSR1-ETS fusion protein acts as an aberrant transcription factor that influences the regulation of mRNA, lncRNA and miRNA expression levels. In addition, by binding to RHA additional transcripts can be bound and this might interfere with the stability of these transcripts. Alterations in epigenetic activity lead to up-and downregulation of a number of transcription factors and thereby interfere indirectly with gene expression. Furthermore, EWSR1-ETS fusion protein binds to the spliceosome and thereby altering splicing processes. By acting on these mediators, multiple cellular pathways are affected. The summarizing gene ontology clusters of the upregulated (green) cellular pathways are cell cycle, membrane proteins, IGF signaling and transcription and the downregulated is extracellular signaling (red). The main processes influenced by these gene ontology clusters are an increase in proliferation, pluripotency, migration and angiogenesis. EWSR1-ETS affects not only the expression of genes but also the expression of non-coding RNAs, including both micro RNAs (miRNAs) and long non-coding RNAs (lncRNAs) [22,38,128]. miRNAs are regulating more than 60% of the human genes by mainly binding at the 3'UTR of the mRNA [129]. Around 10% of the studied miRNAs are significantly affected in EWS, both in down-and upregulation [128,130,131]. Affected pathways are diverse and include important tumorigenic pathways such as IGF signaling, chromatin remodeling, pluripotency and DNA damage repair [128,[131][132][133][134]. An relatively small EWS retrospective patient survival association study on miRNAs identified a survival association with increased miRNA34a expression [135]. miRNA34a is thought not to be influenced by EWSR1-FLI1 itself but its activity is regulated by TP53 and NF-κB and is associated with survival also in a retrospective glioblastoma study [136][137][138]. It regulates expression of proteins involved in growth pathway signaling, apoptosis, chromatin remodeling and genomic stress [136][137][138]. miRNA analysis at whole transcriptome scale might be successful to identify more miRNAs regulated by EWSR1-ETS or which are predictive for therapy.
Long non-coding RNAs are relatively recently discovered as functionally relevant and have functions both in epigenetic and post-translational regulations [139,140]. For example, MALAT1 is a commonly expressed lncRNA, which is involved in angiogenesis and cell cycle progression [141,142]. Brunner et al. [143] studied expression of lncRNAs in a large tumor panel of both sarcomas and carcinomas including EWS. A large number of known and novel lncRNAs differentially expressed in EWS were identified, including ALDH1L1-AS2, DICER1-AS1 and LINC00277 [143]. In later research, LINC00277 (EWSAT1) was the only lncRNA which was significantly overexpressed in EWSR1-FLI1 transfected pediatric MSCs and downregulated in EWS cell lines when treated with EWSR1-FLI1 shRNA [38]. LNC00277 induction on its own affected expression levels of numerous genes which overlapped with EWSR1-FLI1 target genes. Its effect was established partly by interacting with the RNA binding protein HNRNPK. A number of splice variants of LINC00277 were described of which LINC00277-2 was dominantly expressed in EWS. The modus of action of the various splice variants is unknown until now.
After transcription, RNAs are spliced and alternative splicing increases the functional diversity of proteins and noncoding RNAs. Splicing is regulated by multiple protein complexes and by interfering in this regulation many cellular processes can potentially be affected [144]. EWSR1 is involved in one of these protein complexes as scaffold protein [145,146]. EWSR1-ETS, missing the C-terminal part of EWSR1, interferes in the EWSR1 complex mediated splicing and causes the deregulated splicing of EWSR1 complex targeting RNAs [39,[146][147][148]. One of the processes interfered with by EWSR1-ETS binding that was investigated in depth is the binding to RNA helicase A (RHA). RHA has both functions in DNA and RNA unwinding and stabilization [40,149]. Especially the RNA binding of RHA was inhibited by EWSR1-ETS binding and the new EWSR1-ETS RHA complex could bind additional targets, which were enriched for transcripts involved in extracellular signaling processes [40]. The consequences at the cellular level of the splicing interference has been illustrated by the splicing of CCDN1 (cyclin D1), a cell cycle regulator and Vascular Endothelial Growth Factor A (VEGF-A). The normal a-isoform of CCND1 is exported from the nucleus during G1 phase to stop the cell cycle but by EWSR1-ETS interference the relative quantity of the b-isoform is increased in EWS. This isoform is not exported and increases proliferation of EWS cells [150]. VEGF-A splicing can result in both more and less angiogenic isoforms and, by the interference of EWSR1-ETS, the equilibrium between these isoforms is shifted to the more angiogenic isoform VEGFA-165. The effect of this shift is an increase in angiogenesis, which correlates to the highly vascularized histological features of Ewing sarcoma [151]. Despite the number of fundamental studies on the mechanism affected by EWSR1-ETS, limited studies are published on RNA targets. Whole transcriptome sequencing could be used to map the RNA targets of which splicing are affected by EWSR1-ETS but this has not yet been reported. In conclusion transcriptome mapping has shown to be of high value to characterize EWS and identify potential targets and survival markers [37,38,106,135].
Understanding Therapy Sensitivity and Identifying Target Candidates Using the Ewing Sarcoma Sequencing Overview
Before the introduction of chemotherapy, the overall survival of patients with EWS was about 10% using surgery alone. Early observation showed increased radiosensitivity of Ewing sarcoma and therefore radiotherapy as monotherapy was introduced, but the majority of patients still died of metastasis within two years when only radiotherapy was used [152,153]. The introduction of systematic chemotherapy increased the overall survival from ten percent to nowadays sixty to seventy-five percent for a localized tumor at diagnosis [2,154,155]. However, when a patient has recurrent disease, which is the case in thirty percent, or presents a metastatic disease at diagnosis, the overall survival drops to ten to forty percent [45,46]. As these patients are young, longtime curing is the treatment prospective, rather than stabilizing and short-term benefit. This translates in intense treatments but, as a pay-off, these have large consequences for long-term survival of EWS patients [156]. Prognostic markers for survival or treatment sensitivity may help to personalize the treatment [157]. New treatment protocols are needed to increase the patient survival with the least long-term effect. The uncovering of the mechanism of disease specific pathways serves as the basis for the development of targeted drugs to treat patients with the highest efficacy and the least side effects. For this, an EWS OMIC overview, using the results obtained by sequencing from various sources (Table 1), could help to increase this fundamental understanding and could lead to the identification of novel therapeutic candidates for systemic, targeted and immunotherapy.
DNA Damage Response and Repair: Systemic and Targeted Therapy
Chemotherapy is an essential part in the treatment of EWS [154,158]. Over the past decades, the combination of chemotherapeutics, dosage and administration protocol has been adjusted to improve tumor response and reduce toxicity [155,159,160]. The present standard treatment protocol for EWS is based on combination of vincristine, doxorubicin, ifosfamide or cyclophosphamide and etoposide [2,159]. Most of these are DNA damaging drugs. As EWS contains a limited number of secondary mutations, it is likely that these tumors have an intact DNA damage response mechanism. Alkylating and double strand break causing agents consequently activate this mechanism leading to growth arrest and apoptosis of EWS cells. The hypothesis of an intact DNA damage response mechanism correlates with the chemotherapeutic resistance of TP53 mutated EWS tumors, a key gene in this mechanism, since these tumors do not have an intact DNA damage response mechanism [41]. Fusion positive RMS has a limited number of mutations, similarly to EWS, but it is less sensitive to chemotherapy compared to translocation negative RMS [44]. This implies that downstream EWSR1-ETS effects may partly be responsible for the chemotherapy and radiotherapy sensitivity. In accordance, EWSR1-ETS associated DSBs have been identified and radiation induced damage turnover in EWS was reduced compared to osteosarcoma [161,162]. This makes the damage repair pathway a promising candidate to target. Compromising the DNA damage repair pathway via inhibition of poly (ADP ribose) polymerase 1 (PARP1) did indeed lead to inhibited proliferation in EWS cell lines and potentiated the response to temozolomide and irinotecan [19,[161][162][163]. However, in EWS xenografts and patients treated with a PARP1 inhibitor, only the combination with temozolomide or irinotecan was effective [162,164]. In colon cancer xenografts this effect was observed as well [165]. The sensitivity of colon cancer to PARP1 inhibition is hypothesized to be related to a less functional homologous recombination due to cohesin complex aberrancy [75,77,166]. In glioblastoma cells, a correlation between PARP1 sensitivity and the presence or absence of the cohesin complex gene STAG2 was demonstrated [167]. In the EWS PARP1 inhibition studies both STAG2 wild type and mutant cell lines were sensitive to PARP1 inhibition in combination with chemotherapy [163,168]. A specific role for STAG2 in this is therefore unlikely in EWS. Overall, it seems EWSR1-ETS interferes in the DNA damage repair pathway by a yet unexplainable way based on data obtained from genome and transcriptome sequencing studies leading to chemotherapy sensitivity. Identifying the TP53 independent DNA damage response and repair pathway could open novel therapeutic options.
Targeting Chromatin Remodeling; EWSR1-ETS and Its Binding Partners
As mentioned, EWSR1-ETS intervenes in chromatin remodeling in multiple ways and the chromatin state around GGAA microsatellites might be related to the oncogenic capacity of EWSR1-ETS. Hence, chromatin remodeling is a good target. Understanding the action of EWSR1-ETS fusion protein in this could be used to design novel therapeutic agents that either occupies its GGAA microsatellite binding sites, targets the chromatin remodeling or blocks binding of its partners in transcription.
Chemotherapeutic drugs induce DNA damage by binding to the DNA, but the same binding can interfere with the binding of EWSR1-ETS to the DNA. Between these DNA binding chemotherapeutic agents, there is a difference in binding specificity, where Cisplatin and Doxorubicin are suggested to be less specific than Actinomycin D for removal of EWSR1-ETS from the DNA [169]. However, due to the heavy systemic side effects Actinomycin D is no longer used to treat Ewing sarcoma patients in the U.S. [170]. Trabectedin, a toxin from the sea squirt Ecteinascidia turbinate, is believed to be more specific against EWSR1-ETS DNA binding sites. In vitro studies in EWS and myxoid liposarcoma, another fusion gene holding tumor demonstrated a high efficacy and showed interference with the activity of EWSR1-ETS and EWSR1-CHOP fusion protein, respectively [158,171]. In a clinical trial, however, trabectedin alone did not show a significant effect on overall survival in EWS [172].
Targeting chromatin remodelers, for example LSD1 and HDAC2, that attribute to the EWSR1-ETS oncogenic potential has been shown to be effective in vitro and in xenografts [37,106,173,174]. The effect of inhibiting LSD1 was even analyzed by whole transcriptome sequencing in cell lines with EWSR1-FLI1 or EWSR1-ERG translocation to identify the overall effect on gene expression. Inhibition affected numerous genes including well-known target genes like CAV1, NKX2.2. This inhibition study strengthened the role of the NuRD complex in the transcriptome wide effect of EWSR1-ETS [37]. HDAC2 inhibition by Vorinostat had a similar effect on the EWSR1-ETS repressed genes, replicating an earlier HDAC2 inhibition study in EWS, but did not affect genes directly activated by EWSR1-ETS [37,175]. A third potential identified target is the EWSR1-ETS upregulated histone deacetylase SIRT1 which was identified in a EWSR1-FLI1 knock-down screen and inhibition was effective in EWS cell lines in vitro and in xenografts [106]. Furthermore, SIRT1 is regulated by miRNA34a expression, a prognostic factor in EWS, and both are associated with TP53 activity [135,137].
Riggi et al. [31] demonstrated that EWSR1-ETS can function as initiator in chromatin remodeling but needs to recruit other proteins for transcription initiation. Agents that inhibit the interaction between EWSR1-ETS and its binding partners by blocking the binding sites of EWSR1-ETS can be fruitful and a daunting task at the same time due to the disordered structure of EWSR1-ETS [176]. The identification of the small molecule YK-4-279 as inhibitor of the EWSR1-ETS binding to RHA confirmed that this approach is indeed promising and had a broad transcriptomic influence in EWS [40,177,178]. YK-4-279 treatment resulted in the same effect at splicing level as EWSR1-ETS inhibition and inhibited the binding of transcripts by the EWSR1-ETS RHA complex [39,40]. In vivo experiments suggest that combining this agent with other treatments could be especially effective, as shown in the combination with TP53 reactivating agent Nutlin3a in a zebrafish model [179].
Targeting EWSR1-ETS Influenced Extracellular Signaling, Transcriptome Mapping as a Lead
At the transcriptome level, EWSR1-ETS influences various pathways involved in intracellular processes and tumor microenvironmental processes that are needed for EWS development and maintenance. Both these processes are vital according to the processes collectively described as the hallmarks of cancer by Hanahan and Weinberg [180]. Major pathways by EWSR1-ETS affected are involved in extracellular signaling and membrane protein signaling. At a histo-morphological level, this is reflected by a stem-cell like tumor with high vascularization and a clinically observed high metastatic potential. Involved EWSR1-ETS key target pathways responsible for these features might be identified by transcriptome sequencing.
A well-known EWSR1-ETS targeted pathway is the IGF pathway, which is involved in tumor growth, metastasis and angiogenesis [181,182]. EWSR1-ETS increases the IGF1 pathway activity by upregulation of IGF1 expression and downregulation of insulin growth factor binding protein 3 and 5 (IGFBP3, IGFBP5) and various IGF pathway targeting miRNAs [114,117,133]. Targeting this pathway by small molecules or by monoclonal antibodies was shown to be highly effective in cell lines and it inhibited the angiogenesis in xenografts. In clinical trials, IGF1R treatment resulted in partial success due to no-response or quick resistance while a small group of patients remained stable. By studying the long-term responding patients at OMIC levels, we could identify the cause of their tumors sensitivity to anti-IGF therapy, which, in turn, would identify patients for anti-IGF therapy and understand the mechanism of the gained resistance [183][184][185][186][187]. In addition, combination chemotherapy with anti-IGF therapy is being investigated and may be an option. The combination of OSI-906, a dual inhibitor of IGF1R and IR, with trabectedin showed promising preclinical results [188].
The introduction of anti-angiogenic therapy with promises for all cancer types has been taking full media coverage with high initial expectations. Massive efforts to develop novel anti-angiogenic agents have led to several novel targeted therapy approaches. Although anti-angiogenic therapy alone was found to be insufficient, a combination with other treatment modalities may be effective [189]. As Ewing sarcoma is highly vascularized, targeting angiogenesis has been investigated in several in vitro and in vivo studies with success [185,[190][191][192]. This was translated into multiple clinical trials testing anti-angiogenic drugs (NCT00516295, NCT01946529, NCT01492673 and NCT02243605). However, a pilot study in which chemotherapy with or without bevacizumab (a VEGF-A inhibitor) was used did not show a positive effect of bevacizumab [193]. Vascular mimicry has been observed in Ewing sarcoma and may be enhanced under hypoxic conditions that might reverse the anti-angiogenic effect [194][195][196].
As was demonstrated by micro-array studies and verified by transcriptome sequencing, EWSR1-ETS represses extracellular signaling proteins, including chemokines [28,38,113]. Chemokines are involved in all important tumor microenvironmental processes and elucidating the relation between the presence or absence of these chemokines may lead to new candidate targets or reactivating agents which would increase the chemokine expression [197]. The expression levels of the pro-inflammatory chemokines CXCL9 and 10 have been linked to the number of infiltrating T-cells and subsequently with a better overall survival in EWS patients in a relatively small retrospective study in a univariate analysis [198]. Treatment with interferon gamma (IFN-ƴ) enhanced the expression levels of pro-inflammatory chemokines and sensitizes resistant EWS cells in vitro to tumor necrosis factor apoptosis-inducing ligand (TRAIL)-induced apoptosis [198,199]. If these processes are related to each other is unknown. The only chemokine receptor which is highly expressed in EWS and not repressed by EWSR1-ETS is CXCR4 [38,200]. As a key factor in the tumor-microenvironment processes, especially metastasis, it is a very interesting receptor to study in EWS as a potential biomarker and therapeutic candidate [201,202]. Its RNA expression was, like many other tumors including osteosarcoma, correlated with lung metastasis and in vitro membrane CXCR4 positive cell lines migrated towards a CXCL12 gradient [200,203,204]. In contrast, no CXCR4 was detected at protein level in EWS lung metastases with immunohistochemistry but was positive in the chemotherapy-naïve tumor biopsies where it correlated with tumor volume [205]. The cause of the contradiction is unknown up to now and may be attributed to different CXCR4 isoforms or the abundant post-transcriptional modifications of CXCR4 [206][207][208][209][210].
The mentioned pathways are just examples of the many candidate pathways which can be targeted in EWS. These candidate pathways connect intracellular processes with interactions in the microenvironment and can therefore be ideal for combined therapy. OMICs can contribute to identify targets at DNA and RNA level, but since these pathways are highly interconnected with each other additional post-transcriptional studies are needed for target validation and understanding the role of these signaling pathways in EWS. An example of such a post-transcriptional study is a knockdown study [211].
Targeting EWS with Immunotherapy
Immunotherapy is based on the use of two general mechanisms: (1) activating the native immune system (2) priming natural killer (NK) cells or cytotoxic T-cells for antigens specifically overexpressed in the tumor to treat. The performed OMIC studies in EWS can be of value in both cases. Expression of antigen presenting and NK cell ligands in EWS samples can be determined retrospectively and EWS specific antigens can be identified.
Determination of the tumor-associated leukocytes in pediatric tumors showed an increase in macrophages and almost lack of dendritic compare to adult tumors as a common feature the almost lack of dendritic cells [212]. The determination of the presence of intratumoral leukocytes is particularly important since high numbers of CD8+ T-cells have retrospectively been found to be associated with improved survival in an univariate analysis [198]. Indirect attraction of these CD8+ T-cells to EWS may be enhanced by the IFN-ƴ therapy, since it upregulates pro-inflammatory chemokine expression levels [198]. Activation and attraction of the T-cells is presently not tested in a clinical trial but activation of endogenous or donor NK cells has been shown to be effective in EWS and phase I and II clinical trials and are currently open for enrollment (NCT01287104, NCT02100891) [213,214]. The efficacy of recognition may in fact be increased by combining this with the earlier mentioned chromatin remodeler inhibitors. In vitro HDAC inhibitor enhanced the NKGD ligands expression in EWS cell lines, which are essential for NK mediated lysis [215]. For improved long-term NK activation and to overcome tumor mediated downregulation of NKGD, prolonged ex vivo activation or antibody dependent cytolysis is needed [216]. From a preclinical perspective, allografting may be a beneficial adjuvant therapy in combination with either EWSR1-ETS blocking therapy or standard chemotherapy.
The second general method of priming cytotoxic T-cells for tumor specific membrane proteins overexpressed by EWS or unique HLA presented peptides has been investigated. Proposed targets are the tumor specific membrane proteins like PRAME, GPR64 and STEAP1 [217][218][219]. However, ex vivo priming of T-cells for antigens like PRAME and STEAP1 could not yet induce a prolonged antitumor immune response in preclinical studies. T-cells could not interact with the endogenous presented antigens at EWS cell lines or the T-cells which did recognize the presented antigens were classified as exhausted T-cells according to high PD1 expression [220,221]. The EWSR1-ETS upregulated proteins EZH2 and CHM1 were successfully used to prime allo-restricted T-cells but these are proteins expressed in many other tissues and could have serious side effects and lead to non-tumor specific targeting [221]. The ideal antigen to prime T-cells for would be EWSR1-ETS itself. The potential of this hypothesis was tested and a EWSR1-FLI1 specific antigen was identified and verified as an activating antigen for cytotoxic T-cells, but no follow-up study has been presented [222]. A possible cause for the less effective recognition demonstrated in EWS cell lines, and the potential clinical limiting factor, is the loss of major histocompatibility complex (MHC) class I and class II, which are needed for a proper immune response [223]. Genetic engineering of T-cells with chimeric antigen receptors directed to overexpressed surface markers is independent of the MHC class system. These can be designed against proteins and even phosphoglycolipids [224]. The singular tested surface marker with this method is the neural ganglioside GD2 which was expressed in all 10 EWS cell lines tested and 12 of 14 analyzed patients [225]. A follow-up xenograft study was successful showing a reduction in tumor growth and number of tumors but effect on total survival was not significantly different [224]. For further research, antigens should be selected for EWS specific surface markers to prevent non-tumor cells to be targeted, like stem cells which do not express MHC class II complex but do express certain surface markers [226]. For example, the aforementioned GD2 is also expressed in neural crest cells and mesenchymal stem cells [227].
Overall, immunotherapy can be promising in combination with systemic or targeted therapy. Both NK-cell and T-cell related therapies have potential, especially the antibody targeted NK-cells and genetically modified T-cells. Genetically modified T-cell therapy in EWS is just starting and all membrane proteins and potential tumor specific splice variants of membrane proteins could be targets for these T-cells. In addition, the ability to target glycolipid structures opens a complete new set of possibilities, but these cannot be identified by sequencing.
Conclusions
By sequencing EWS at the genome, epigenome and transcriptome level, an atlas can be created which would help to fundamentally understand EWS and help to identify important nodes as therapeutic candidates. The EWSR1-ETS translocation is the characteristic pathognomonic alteration found in all tumors so far. The fusion protein act as a strong transforming oncogene and, in experimental conditions, the transfection of cells with normal cellular backgrounds rather leads to oncogene-induced apoptosis than to transformation. Recent studies, however, showed that stem cells from young individuals with the necessary permissive background did form tumors, pointing towards the importance of epigenetic controlling in cellular/tissue differentiation in providing the necessary niche for the transformation [19,22,25,38]. Therefore, mapping these genomic and functional genomic alterations can lead to identification of the cell of origin, improvements in prediction of clinical outcome, and discovery of novel therapeutic targets. These prospects have led to numerous, independent investigations using various approaches related to OMICs.
As a result, several novel findings and confirmations of earlier observations were collected. For example, some secondary structural alterations can be detected in a subset of the tumors that can identify a patient with unfavorable prognosis. Despite huge efforts to identify secondary mutations that provide a permissive background to transform cells with the pathognomonic EWSR1-ETS translocation, only a limited number of secondary point mutations were detected in EWS. Of these, STAG2 and TP53 were the most frequently mutated genes and mutations in these genes were associated with inferior prognosis. The value of these mutations as prognostic markers has to be validated in a prospective study. Transcriptome sequencing projects excluded the possibility of recurrent co-occurring fusion genes that would be responsible for the transformation to endure the fusion protein. Based on these massive sequencing efforts, it is likely that the EWSR1-ETS fusion can propagate transformation in cells with less differentiated features and the epigenetic landscape of these primitive cells form a permissive niche for oncogenic transformation. EWSR1-ETS is, both at the transcriptome and epigenome level, the most dominant actor both by activation and repression transcription and needs cooperation of binding partner proteins like chromatin remodelers. The identified key pathways in Ewing sarcoma and the EWSR1-ETS chromatin remodeling binding partners include promising candidate targets. This needs to be validated with in functional studies in combination with the epigenome and transcriptome analyses.
The advantage of the genomic stability of EWS is that the endogenous pathways controlling DNA damage recognition and apoptosis are still intact and could potentially be activated when targeted specifically and especially together with agents acting on the basis of the EWSR1-ETS network of epigenomic and transcriptomic changes. For example, blocking the interaction with its binding proteins could be a very efficient combination therapy. By deciphering this network for both targeted therapy as well as immunotherapy, novel key target candidates can be identified. In the future, hopefully these therapies could, together with conventional chemotherapy, improve the outcome of these young patients.
Conflicts of Interest
The authors declare no conflict of interest. | 2016-04-23T08:45:58.166Z | 2015-07-01T00:00:00.000 | {
"year": 2015,
"sha1": "1df3138d7e87f50e9caa848e207e7d2c35dd8294",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/16/7/16176/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "1df3138d7e87f50e9caa848e207e7d2c35dd8294",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
22589258 | pes2o/s2orc | v3-fos-license | Structural flexibility modulates the activity of human glutathione transferase P1-1. Influence of a poor co-substrate on dynamics and kinetics of human glutathione transferase.
Presteady-state and steady-state kinetics of human glutathione transferase P1-1 (EC) have been studied at pH 5.0 by using 7-chloro-4-nitrobenzo-2-oxa-1,3-diazole, a poor co-substrate for this isoenzyme. Steady-state kinetics fits well with the simplest rapid equilibrium random sequential bi-bi mechanism and reveals a strong intrasubunit synergistic modulation between the GSH-binding site (G-site) and the hydrophobic binding site for the co-substrate (H-site); the affinity of the G-site for GSH increases about 30 times at saturating co-substrate and vice versa. Presteady-state experiments and thermodynamic data indicate that the rate-limiting step is a physical event and, probably, a structural transition of the ternary complex. Similar to that observed with 1-chloro-2,4-dinitrobenzene (Ricci, G., Caccuri, A. M., Lo Bello, M., Rosato, N., Mei, G., Nicotra, M., Chiessi, E., Mazzetti, A. P., and Federici, G. (1996) J. Biol. Chem. 271, 16187-16192), this event may be related to the frequency of enzyme motions. The observed low, viscosity-independent kcat value suggests that these motions are slow and diffusion-independent for an increased internal viscosity. In fact, molecular modeling suggests that the hydroxyl group of Tyr-108, which resides in helix 4, may be in hydrogen bonding distance of the oxygen atom of this new substrate, thus yielding a less flexible H-site. This effect might be transmitted to the G-site via helix 4. In addition, a new homotropic behavior exhibited by 7-chloro-4-nitrobenzo-2-oxa-1,3-diazole is found in Cys-47 mutants revealing a structural intersubunit communication between the two H-sites.
) (2) but a "poor" substrate for the Pi and Mu class isoenzymes having k cat NBD-Cl values only 2-4% of those found with CDNB (2). Steady-state experiments performed at pH 5.0 in the presence of NBD-Cl or CDNB indicate that a minimum rapid equilibrium random sequential bi-bi model is sufficient to describe the catalytic mechanism by GST P1-1, similar to that observed with CDNB at pH 6.5 (3). Presteady-state kinetics, thermodynamic data, and viscosity variation experiments suggest that the rate-limiting step with NBD-Cl is, probably, a structural transition of the ternary complex. Interestingly, intra-and intersubunit communications between active sites are triggered by this new co-substrate. These findings have been tentatively explained on the basis of the x-ray crystal structure.
Spectrophotometric Measurements-Steady-state kinetics of GST P1-1 with NBD-Cl as co-substrate were followed spectrophotometrically at 419 nm where the GS-NBD conjugate has its absorption maximum (⑀ 419 nm ϭ 14.5 mM Ϫ1 cm Ϫ1 ) (2). Activity with CDNB was measured at 340 nm where the product absorbs (⑀ 340 nm ϭ 9.6 mM Ϫ1 cm Ϫ1 ) (6). Spectrophotometric measurements were performed with a double-beam UVICON 940 spectrophotometer (Kontron Instruments) equipped with a cuvette holder thermostatted at 25°C. Kinetic experiments were done in 1 ml (final volume) of 0.1 M sodium acetate buffer, pH 5.0, containing 5-10 g of GST P1-1 and variable amounts of substrates. The reaction rates were measured at 0.1-s intervals for a total period of 12 s. Initial rates were determined by linear regression and corrected for the spontaneous reaction. Kinetic data with NBD-Cl were collected by varying NBD-Cl from 10 to 200 M and GSH from 12 to 500 M over a matrix of * This work was partially supported by a grant of Ministero dell'Università e della Ricerca Scientifica e Tecnologica (fund 40%) and by a grant from National Research Council, Progetto Finalizzato ACRO. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Kinetic data with CDNB were obtained at pH 5.0 by varying CDNB and GSH from 50 M to 1 mM over a matrix of 36 substrate concentrations.
Fluorometric Measurements-Binding of GSH, EtS-NBD, and NBD-Cl to the enzyme was followed in 0.1 M sodium acetate buffer, pH 5.0, by the intrinsic fluorescence quenching of the enzyme after ligand addition with the procedure previously described (8). Fluorescence measurements were done with a Perkin-Elmer LS-5 fluorometer with a sample holder at 25°C. Excitation was at 280 nm and the emission was at 340 nm.
Data Analysis-The nomenclature of kinetic parameters was essentially that of Segel (7). The dissociation constant (K) for substrates and the substrate analogue EtS-NBD was obtained by fluorescence measurements by fitting the data to the simplest model by assuming a single and noncooperative binding site per subunit as shown in Equation 1: where Y corresponds to the fractional saturation, K is the dissociation constant, and [L] is the ligand concentration. The dissociation constant values were used for the analysis of kinetic data that were fitted to a rapid equilibrium random sequential bi-bi model, according to Scheme 1. The parameters to be varied in the fitting procedure were only limited to V max and the coupling factor ␣, according to Equation 2: where v is the initial velocity, [A] and [B] are GSH and NBD-Cl (CDNB) concentrations, and K A and K B their respective dissociation constants (without the second substrate) obtained by fluorescence experiments. The dissociation constant by fluorometry for CDNB is uncertain due to its high absorbance at 280 nm and its low affinity for GST; K B CDNB was then left free to vary in the kinetic fits. ␣ is the coupling factor. V max is expressed as [product]⅐s Ϫ1 ⅐[GST] Ϫ1 . Parameters derived from binding data (K A and K B
NBD-Cl
) and from the fit of kinetic data (V max , K B CDNB , and ␣) were used as follows: (i) to calculate the theoretical curves of log K A app versus log [B], i.e. to correlate the variation of the affinity of GST P1-1 for a substrate to the concentration of the second substrate; data were fitted according to Equation 3. The fluorescence data for binding of the unreactive substrate analogue EtS-NBD in the presence of GSH were analyzed by using Equation 3. The theoretical curve is calculated by using values of K A and
K B
EtS-NBD obtained by fluorometry and ␣ by kinetic fit. Binding and kinetic data were analyzed with a software package WIN-MATLAB (Math Works-South Natick, MA).
Stopped Flow Analysis-Stopped-flow measurements were performed on a stopped-flow spectrophotometer consisting of a High-Tech SHU-51 rapid mixing device thermostatted at 25°C equipped with a Jasco J600 spectropolarimeter. Light path of the cell was 0.2 cm. Deadtime of the instrument was Յ5 ms. Presteady-state analysis was performed by rapid mixing of GST P1-1 (about 100 M active sites) in 0.1 M sodium acetate buffer, pH 5.0, containing 4 mM GSH with an identical volume of 0.4 mM NBD-Cl in the same buffer. Blank was done in the same conditions without GST. The increase of absorbance at 419 nm was monitored every 1 ms. Spectrum of the species which accumulates in the dead time (Յ5 ms) was constructed point by point from the absorbance values measured at different wavelengths (from 300 to 600 nm) after rapid mixing 100 M GST active site in 0.1 M sodium acetate buffer, pH 5.0, containing 500 M GSH with an identical volume of 100 M NBD-Cl in the same buffer. At each wavelength, the absorbance value at t ϱ was subtracted from the absorbance value at t°, and this value was then added to the absorbance of the product at the same wavelength (9).
Cooperativity-The cooperativity of C47S and C47S/C101S mutants toward GSH was assayed by varying GSH concentration (from 5 M to 1 mM) at 0.2 mM NBD-Cl constant concentration (in 0.1 M sodium acetate buffer, pH 5.0, and 25°C). Similar experiments at fixed GSH concentration (0.5 mM) and variable NBD-Cl (from 2.5 M to 0.2 mM) were used to check the cooperativity toward NBD-Cl. Kinetic data were analyzed by the KaleidaGraph (version 2.0.2) (Abelbeck Software) computer program and fitted to a rate equation expressing positive cooperativity as previously reported (8).
Viscosity Dependence of Kinetic Parameters-The effect of viscosity on kinetic parameters and on cooperativity was assayed by using 0.1 M sodium acetate buffer, pH 5.0, containing variable glycerol concentrations. Viscosity values () at 25°C were calculated as described (10) and randomly checked with an Ostwald viscometer. Viscosities are reported relative to 0.1 M sodium acetate buffer, pH 5.0.
Molecular Modeling-Modeling was performed on a Silicon Graphics Indigo 2 workstation using the software modeling package INSIGHT II (Biosym Technologies Inc.). Molecular dynamics simulations and energy minimization calculations were performed using DISCOVER (Biosym Technologies Inc.). A model of NBD-Cl was built and energyminimized to tidy up its geometry. The ligand was roughly positioned in the H-site of the human GST P1-1 so it was bound in a productive mode (i.e. the chlorine atom was positioned close to where the thiol group of GSH would be located when bound). Molecular dynamics simulation was performed on the complex, followed by 200 steps of conjugate gradient energy minimization to relieve any residual strain and to optimize any potential electrostatic and van der Waals contacts. This approach was applied to NBD-Cl, CDNB, and their glutathione conjugates.
RESULTS
Steady-state Kinetics-Steady-state kinetics with NBD-Cl at pH 5.0 was analyzed and compared with that obtained with CDNB under the same conditions. Thermodynamic constants obtained by fluorometry (Table I) were used for fitting the SCHEME 1 TABLE I Dissociation constants and kinetic parameters at pH 5.0 and pH 6.5 ␣ is the coupling factor; K A is the dissociation constant for the GST-GSH complex; K B is the dissociation constant for the GST-co-substrate complex. V max is coincident to k cat .
Although k cat NBD-Cl Х k cat CDNB at pH 5.0, NBD-Cl must be considered a poor substrate for GST Pl-1 as this pH value is the optimal one for the NBD-Cl reaction, while k cat CDNB is 76 s Ϫ1 at pH 6.5. b Data from Ref. 12 kinetic data as reported under the "Experimental Procedures." As already found with CDNB as co-substrate at pH 6.5 (3), the rapid equilibrium random sequential bi-bi model (Scheme 1) or the mathematically equivalent steady-state ordered bi-bi mechanism was a minimum model that describes satisfactorily the experimental data for both co-substrates at pH 5.0 (Fig. 1). A diagnostic procedure to distinguish between these two mechanisms is based on product inhibition studies (7). The pattern of inhibition at pH 5.0 by GS-NBD (Fig. 2) is consistent with a rapid equilibrium random mechanism (7). Data presented in the form of a double-reciprocal plot (Fig. 1) provide evidence of the convergence of the lines above the abscissa axis, showing a coupling factor ␣ ϭ 0.036 with NBD-Cl and ␣ ϭ 0.38 with CDNB. These fractional numbers indicate that the affinity for GSH increases in the presence of NBD-Cl about 30 times (and vice versa), but only 3 times in the presence of CDNB (and vice versa) (7). The agreement between fitting procedures and experimental data is evident in Fig. 3 which shows the plots obtained by using Equations 3-5. Furthermore, the results obtained from fluorescence quenching measurements for the binding of a nonreactive analogue of the second substrate (EtS-NBD), in the presence of variable GSH concentrations, indicate that the kinetic synergism between the G-site and H-site must be related to a mutual modulation of the dissociation constants (Fig. 4).
Presteady-state Kinetics-A careful investigation of the presteady-state kinetics of GST P1-1 with NBD-Cl as co-substrate does not confirm our preliminary evidence for an accumulation of GS-NBD product before the attainment of the steady state (12). No evident burst phase is now observed at 419 nm where GS-NBD absorbs strongly (2) (Fig. 5, inset). Some technical problems were responsible for that earlier report; the major problem was an inadequate washing volume for the cell (based on advice by manufacturer). Furthermore, the spectral image of the accumulating species, obtained as described under "Experimental Procedures," is very similar to that of NBD-Cl with a little red shift of about 3 nm (Fig. 5). This spectrum has been assigned to the NBD-Cl complexed enzyme, since a very different spectrum is expected for the Meisenheimer (or ) complex which occurs as an intermediate in similar aromatic substitution reactions (13). On the basis of the present data and of the previously reported k cat independence of the nature of the leaving group (12), we suggest that the ratelimiting step is a physical event, and it must occur between the ternary complex formation and the chemical step, as found for the CDNB/GSH-catalyzed reaction (1). Thermodynamic data also indicate different rate-determining events for the catalyzed and uncatalyzed reactions. In the latter the -complex formation was found to be the rate-determining step (12). Calculations from the Arrhenius plots shown in Fig. 6 give an activation energy for the enzymatic NBD-Cl/GSH reaction of about 20 kJ/mol higher than that for the spontaneous one. This may be a paradoxical result if a coincident rate-limiting step occurs in both reactions.
Cys-47 Mutants Display Cooperativity toward Both GSH and NBD-Cl-When CDNB is the co-substrate, the substitution of Cys-47 by Ala or Ser does not change the k cat value but lowers the affinity of the G-site for GSH (about 9-fold for the C47S mutant and 14-fold for the C47S/C101S double mutant) and triggers a marked positive cooperativity toward GSH (8). An increased flexibility of helix 2, due to the lack of a peculiar electrostatic bond between Cys-47 and Lys-54, may be the basis of this structural interaction between subunits (4, 8). Conversely, no homotropic effect by CDNB is observed (8). A more complex behavior is found with NBD-Cl. As shown in Table II the affinity toward both GSH and NBD-Cl decreases 4 -10-fold in the mutants while k cat values are quite unchanged. The homotropic behavior by GSH is again recovered both for C47S mutant and C47S/C101S double mutant (Table II) with Hill coefficients very similar to that obtained in the presence of CDNB (8). Surprisingly, a positive cooperativity is observed toward NBD-Cl with Hill coefficient values of 1.56 and 1.40 for the single and double mutants, respectively (Table II). This reveals a structural intersubunit communication between the two H-sites of the dimeric enzyme which is absent or not detectable in the wild-type and in Cys-47 mutants with CDNB as co-substrate.
Viscosity Effect on Wild Type-In the preceding paper (1) we demonstrated a viscosity dependence both of k cat and K m GSH values for the GST-catalyzed reaction of CDNB with GSH. This has been interpreted assuming that kinetic parameters depend on some diffusion-controlled motions of the protein. In particular, helix 2 flexibility seems to be involved in determining the G-site affinity, although it is unclear whether it plays a role in the k cat modulation. By using NBD-Cl as co-substrate, k cat , K m GSH , and K m NBD-Cl remain almost unchanged even at a high viscosity value (Table II). These data may be interpreted by means of the x-ray structure of the enzyme (see "Discussion"); moreover, they are a good control to show that the viscosity dependence of k cat and K m GSH previously observed in the presence of CDNB (1) is not due to a nonviscosity solvent effect (14).
Viscosity Effect on C47S and C47S/C101S Mutants-As ob- (Table II). On the contrary the positive cooperativity toward GSH, as expressed by the Hill coefficient, is markedly lowered by increasing the viscosity (Table II). Thus it appears that the flexibility of helix 2, responsible for the GSH cooperativity in these mutants (1,8), is still diffusion-controlled in the presence of NBD-Cl as co-substrate. Conversely, the homotropic behavior of NBD-Cl is viscosity-independent (Table II). This new intersubunit interaction between H-sites should involve protein regions different from helix 2.
DISCUSSION
Steady-state kinetic data at pH 5.0 with NBD-Cl (or CDNB) as co-substrate are consistent with a rapid equilibrium random sequential bi-bi model (Figs. 1-3). The effects of change of the leaving group (12) and of temperature (Fig. 6) on k cat and presteady-state kinetic data (Fig. 5) suggest that the ratelimiting step in the NBD-Cl/GSH system is a physical event that occurs before the -complex formation, i.e. a structural transition of the ternary complex as occurs in the CDNB/GSH system (1). However, with NBD-Cl a strong kinetic and binding synergism between H-and G-site is observed (Figs. 1 and 4 and Table I) which implies the existence of a mutual communication between these binding sites. Only a little positive modulation appears with CDNB at the same pH value ( Fig. 1 and Table I). Unlike CDNB, NBD-Cl gives a very low and viscosityindependent k cat value (Table II). These findings may be tentatively interpreted on the basis of x-ray structure of GST P1-1 (15). A computer modeling shows the rings of NBD-Cl packed between the two aromatic rings of Phe-8 and Tyr-108. Furthermore, the oxygen atom of the 5-membered ring of NBD-Cl (for structure, see Fig. 5) is in hydrogen-bonding distance of the hydroxyl group of Tyr-108 (3.1 Å) which resides in helix 4 (residues 83-109) (Fig. 7). This hydrogen bond, missed in the CDNB complex (distance from the oxygen atoms of the nitro groups Х5 Å), may in part account for the high affinity of GST toward NBD-Cl (␣K B ϭ 3.3 M and K B ϭ 93 M), and it probably causes an increased rigidity of the active site.
There is evidence that helix 4 displays diffusion-controlled motions in absence of ligands (1). If these motions are assumed to be related to the conformational transition of the ternary complex (rate-limiting), the observed low k cat value and its viscosity independence (in the presence of NBD-Cl) are the logical consequence of a less flexible helix 4. In fact, the energy barrier for a rate-limiting structural transition depends on the relative contributions of internal and external friction (14); thus, increased intramolecular constraints could minimize the effect of the external friction due to solvent viscosity. 2 In addi- 2 During the revision of this paper, according to a reviewer's suggestion, we checked the effect of nondenaturing urea concentrations (up to 2 M) on k cat values. With NBD-Cl (in 0.1 M sodium acetate buffer, pH 5.0), k cat increases up to 200% of the basal value at 2 M urea. No change of k cat has been observed with CDNB as co-substrate (in 0.1 M potassium phosphate buffer, pH 6.5). By assuming that nondenaturing urea concentrations only lower the internal friction of a protein, its different effect on k cat values suggests that the internal constraints play a crucial role in determining the structural transition energy barrier only in the presence of NBD-Cl as co-substrate. These findings agree well with the conclusions of this paper. tion, helix 4 provides a number of important residues close to or in contact with the bound GSH such as Asp-94, Glu-97, and Asp-98 (15). Hence, a structural perturbation due to the interaction of NBD-Cl with Tyr-108, localized on the H-site, may be transmitted to the G-site via helix 4 yielding the observed synergistic effect between the H-and G-sites responsible for the increased affinity toward substrates. As concerns K m GSH , its viscosity-independent value (Table II) seems no more related to diffusion-controlled motions of helix 2 that, on the contrary, are relevant in determining K m GSH in the CDNB/GSH system (1). It is possible that the increased rigidity of the active site due to NBD-Cl binding may be transmitted to helix 2. Alternatively, helix 2 motions may be always diffusiondependent, but the K m GSH value is mainly determined by GSH interaction with helix 4 promoted by NBD-Cl. The viscosity effect on the Hill coefficients of Cys-47 mutants (Table II) suggests that helix 2 (whose flexibility modulates this parameter (1)) is still diffusion-controlled even in the presence of NBD-Cl.
Preliminary evidence obtained with a mutated GST P1-1, where Tyr-108 is replaced by Phe, confirms the influence of the hydroxyl group of Tyr-108 on k cat and K m NBD-Cl as these kinetic parameters increase remarkably and become viscosity-dependent in this mutant. 3 Comparison of the crystallographic data and of the catalytic properties toward NBD-Cl in Alpha, Pi, and Mu GST isoenzymes indirectly confirms the above suggestions. Both Pi and Mu class isoenzymes, characterized by low k cat and K m NBD-Cl values (2), have equivalently located hydroxyl groups (Tyr-108 and Tyr-115, respectively) which may form a hydrogen bond with the heterocyclic portion of NBD-Cl. On the contrary, Alpha class GST, which displays higher k cat and K m NBD-Cl values (2), lacks this hydrogen bond as a Val residue replaces the Tyr residue. | 2018-04-03T06:05:16.720Z | 1996-07-05T00:00:00.000 | {
"year": 1996,
"sha1": "219f704c3e1fe898266cfc62633db0ba54aa2d01",
"oa_license": "CCBY",
"oa_url": "http://www.jbc.org/content/271/27/16193.full.pdf",
"oa_status": "HYBRID",
"pdf_src": "Highwire",
"pdf_hash": "eeaf4ec75be2d4983d9755a69d2e2c441a515992",
"s2fieldsofstudy": [
"Biology",
"Chemistry"
],
"extfieldsofstudy": [
"Chemistry",
"Medicine"
]
} |
243480507 | pes2o/s2orc | v3-fos-license | Effects of the Two Doses of Dexmedetomidine on Sedation, Agitation, and Bleeding During Pediatric Adenotonsillectomy
Background Due to the importance of dexmedetomidine and its different dosages, here, we aimed to investigate and compare the effectiveness of the doses of 1 µg/kg and 2 µg/kg of dexmedetomidine in sedation, agitation, and bleeding in pediatrics undergoing adenotonsillectomy. Methods This double-blinded randomized clinical trial was performed on 105 pediatric patients that were candidates for adenotonsillectomy. Then, the patients were divided into three groups receiving dexmedetomidine at a dose of 2 µg/kg, diluted dexmedetomidine at 1 µg/kg, and normal saline. The drugs were administered 15 minutes before operations via the intravenous method. The duration of extubation, mean arterial pressure (MAP), heart rate (HR), and SPO2 in the recovery were recorded. We also collected data regarding patients’ sedation and agitation every 15 minutes. Results Our data showed no significant differences between the groups of patients regarding MAP, HR, and SPO2. However, the mean sedation score was significantly higher in patients receiving dexmedetomidine (2 µg/kg), and this score was lowest in the control group at the time of entrance to the recovery room. The patients that received dexmedetomidine at a dose of 1 µg/kg had the lowest agitation score after 45 minutes of being in the recovery room, and the patients treated with dexmedetomidine at a dose of 2 µg/kg had the lowest agitation score after 60 minutes of being in the recovery compared to other groups of patients. Conclusions The use of the doses of 1 µg/kg and 2 µg/kg of dexmedetomidine was associated with proper sedation and a significant reduction in agitation. The patients also had lower amounts of bleeding. We recommend that anesthesiologists should pay more attention to dexmedetomidine at a dose of 2 µg/kg, especially in pediatric surgical procedures.
Background
Adenotonsillectomy is one of the most common surgeries in children. The two main indications for this surgery are upper airway obstruction and infection. The obstruction may be in the nasopharynx or the oropharynx (1,2). The prevalence of childhood adenotonsillectomy was 2.5 per 1000 person-years based on epidemiologic results (3,4). Perioperative complications of adenotonsillectomy include hemorrhage, respiratory decompensation, velopharyngeal incompetence, and subglottic stenosis (5). Bleeding and unstable hemodynamics are two critical complications of adenotonsillectomy that should be prevented (6).
Because tonsillar hypertrophy is more common in children, especially at the age of 3 to 6 years, and the risk of bleeding during and after surgery is high, maintaining stable hemodynamic stability during surgery is very important (7). Prolonged surgery and hemodynamic instability are associated with more complications in children (8).
Controlled hypotension should be used with caution to minimize the risk of damage to vital organs. Important risks that may arise from using controlled hypotension include the possibility of coronary, cerebral, or renal circulatory failure (9).
Dexmedetomidine is a central alpha 2 adrenergic agonist with both sedative and analgesic effects. Dexmedetomidine is an anxiolytic, sedative, and pain medication and is notable for its ability to provide sedation without risk of respiratory depression (10,11). The use of this drug in children has increased due to its neuroprotective properties. According to the available findings, dexmedetomidine is one of the few drugs that does not cause cognitive impairment after anesthesia in children (12,13).
Various studies have reported the benefits of using dexmedetomidine in surgery, including induction of controlled hypotension and reduced bleeding. With controlled hypotension following the administration of dexmedetomidine, blood pressure and the heart rate (HR) are decreased, and a safe operative procedure can be conducted (14)(15)(16).
Another important aspect of surgical procedures, especially in pediatrics, is the intensity of sedation and agitation during recovery. Increased agitation can seriously influence the pediatric experience of surgical procedures; therefore, proper management and control of agitation via suitable sedation can play a critical role in pediatric anesthesia (17)(18)(19). It has been indicated that dexmedetomidine could provide acceptable sedation and prevent agitation in children undergoing surgeries.
So far, studies have investigated the use of dexmedetomidine in reducing bleeding and agitation during various operations, and different dosages have been confirmed in this regard. However, to the best of our knowledge, no study has compared the effects of two different doses of 1 µg/kg and 2 µg/kg of dexmedetomidine on these issues during adenotonsillectomy among pediatrics. We should also note that the dose of 2 µg/kg of dexmedetomidine is currently used in clinics, and its beneficial effects have been indicated, but due to possible complications of this dosage, including bradycardia, we aimed to investigate the usage of dexmedetomidine at a dose of 1 µg/kg.
Objectives
The aim of the present study was to evaluate and compare these two distinct doses of dexmedetomidine in pediatrics.
Methods
This double-blinded randomized clinical trial was performed in 2020 in Imam Hossein pediatrics hospital affiliated with the Isfahan University of Medical Sciences. The study population consisted of pediatric patients aged 3 -10 years that were candidates for adenotonsillectomy. The study protocol was approved by the research committee of Isfahan University of Medical Sciences, and the Ethics Committee confirmed the research protocol (Ethics code: IR.MUI.MED.REC.1398.729 and the Iranian registry of clinical trials (IRCT) code: IRCT20200325046853N1).
Inclusion Criteria
The inclusion criteria were the age of 3 -10 years, being a candidate for adenotonsillectomy, the American Society of Anesthesiologists (ASA) classification equal to I or II, and signing the written informed consent to participate in this study by the parents. Patients with the following criteria did not enter the study: having a previous medical disease (such as cardio-pulmonary, renal, endocrine, etc.), disorders making the patients vulnerable to bleeding, impaired pre-operative coagulation tests, including prothrombin time (PT), partial thromboplastin time (PTT), and international normalizing ratio (INR), and having metabolic disorders, such as phenylketonuria or galactosemia.
Exclusion Criteria
The exclusion criteria were any complications during the surgery, including abnormal bleeding and hemodynamic instability.
The informed consent was taken from the parents after a full explanation about the methods and drugs that were used in this study, their complications, and benefits.
The sampling method was non-probability sequential, in which all cases who met the inclusion criteria were included in the study until the sample size was completed. A total number of 105 patients entered the study. Demographic data of all cases, including age, gender, and weight were collected. Patients were separated from their parents after pre-medication with intravenous midazolam (0.1 mg/kg) and brought to the operating room. They were then anesthetized with fentanyl (2 µg/kg), sodium thiopental (5 mg/kg), and atracurium (0.5 mg/kg), and intubated with an orotracheal tube with an appropriate size, and connected to a ventilator and anesthetized with a 1% isoflurane retainer for the maintenance of anesthesia.
In order to comply with the double-blind conditions of the study, the doses of 2 µg/kg and 1 µg/kg of diluted dexmedetomidine in a volume of 10 cc were inserted in the syringes. Another syringe was made without dexmedetomidine and contained only 10 cc of normal saline. The specialist labeled the syringes A, B, and C and provided them to the researcher on a daily basis. Thus, until the data collection was completed, the researcher and the data registrar did not know the type of drugs in the syringes. Even the data analyst was unaware of the type of groups. The drugs were administered 15 minutes before the beginning of the operations via the intravenous (IV) method. Then, the patients were divided into three groups using random allocation software. Group A received 1 µg/kg diluted dexmedetomidine prepared by Elixir Pharmaceutical Company in a volume of 10 cc slowly over ten minutes, 15 minutes before surgery by IV method. Group B received 2 µg/kg diluted dexmedetomidine in a volume of 10 cc slowly over ten minutes, and group C received 10 cc normal saline slowly over 10 minutes. Mean arterial pressure (MAP), HR, and SPO 2 were recorded every 15 minutes before induction and during anesthesia; the volume of bleeding during surgery was also recorded on a CC basis. Patients were extubated after surgery and transferred to the recovery room. The duration of extubation (from the time of discontinuation of anesthesia to the time of extubation of patients) was recorded, and in the recovery room, MAP, HR, and SPO 2 were recorded every 15 minutes. The duration of stay in the recovery room based on Modified Aldrete Score (20) was recorded in three groups and compared with each other.
In the recovery room, the patient's degree of agitation was recorded every 15 minutes by the nurse observer of the patient's condition according to the Pediatric Anesthesia Emergence delirium (PAED) criteria in three groups and compared (21). According to this score, the states of inconsolable, restlessness, awareness of surroundings, purposeful actions, and having eye contact were scored from 0 to 4, and higher scores indicated worse conditions. A score equal to or greater than 10 was considered as the emergence of delirium and was treated with propofol at a dose of 1 mg/kg. We also measured the amounts of propofol usage in patients. The Ramsay Sedation Scale (RSS) was recorded in three groups every 15 minutes (22). Based on this score, the score of the patients varies from 1 (awake, anxious, agitated, or restless) to 6 (asleep, no response to light, glabella tap, or loud noise).
Statistical Analysis
The obtained data were entered into the statistical package for social sciences (SPSS) version 24. We used an independent t-test and repeated measures ANOVA to compare data at different time points in different groups. A Pvalue < 0.05 was considered as the significance threshold.
Results
In the present study, 105 pediatrics were entered based on inclusion criteria. The mean age of the pediatrics was 6.39 ± 2.05 years. Also, 57 patients (54.3%) were boys and 48 patients (45.7%) were girls. The mean weight of patients was 21.78 ± 6.54 kg. The age, gender, and weight of the subjects in the three groups showed no significant difference (P > 0.05 for all items) ( Table 1). Table 2 compares the MAP, HR, and SPO 2 values at different time points among three groups. Based on the evaluations, we observed no significant differences between groups of patients regarding the mentioned variables (P > 0.05) ( Table 2). Evaluation of sedation and agitation scores in patients showed significant differences among the groups. Based on our data, the mean sedation score was significantly higher in group B and this score was lower in the control group at the time of entrance to the recovery room (P= 0.031). We observed no differences between the patients regarding the sedation score. Evaluation of agitation score also showed that group A had the lowest agitation score after 45 minutes of being in the recovery room (P= 0.041), and group B had the lowest agitation score after 60 minutes compared to other groups of patients (P= 0.003) ( Table 3). Based on our data, the mean recovery duration was significantly lower in group A compared to other cases (P= 0.007), and the amount of bleeding was significantly lower in group B (P= 0.001). No significant differences were observed between patients regarding extubation duration (P= 0.807) and propofol usage (P= 0.182) ( Table 4).
The mean recovery duration was significantly lower in group A compared to other cases (P= 0.007), and the amounts of bleeding were significantly lower in group B (P= 0.001).
Discussion
During surgical procedures, especially in pediatrics, sustaining stabilized hemodynamics, reducing the duration of recovery, reducing the amounts of bleeding, and proper control of the sedation and agitation during the recovery should be critically managed.
The present study evaluated the use of dexmedetomidine at two different dosages compared to the placebo for the management of these surgical procedure complications. Based on our results, the use of dexmedetomidine had no significant effects on the patient's hemodynamics, but the mean sedation score was significantly higher in group B, and this score was lowest in the group C at the time of entrance to the recovery room showing the effectiveness of dexmedetomidine (2 µg/kg) in providing proper sedation. Assessments of agitation also showed that the administration of dexmedetomidine (1 µg/kg) led to a reduced agitation score after 45 minutes of the recovery, and group B had the lowest agitation score after 60 minutes. We also showed that the mean recovery duration was significantly lower in group A and the amounts of bleeding were significantly lower in group B.
The use of dexmedetomidine in reducing agitation after surgical procedures has been previously evaluated. In the present study, we showed that both 1 µg/kg and 2 µg/kg dosages of dexmedetomidine reduced the agitation score of pediatrics and the patients that received the dose of 2 µg/kg had lower scores within 60 minutes. It has been indicated that a single dose injection of dexmedetomidine during the surgeries could reduce the post-operation agitation of pediatrics.
A study by Guler et al. was conducted on 60 children undergoing adenotonsillectomy.
They injected dexmedetomidine at a dose of 0.5 µg/kg, and the sedation and agitation scores of the pediatrics were assessed.
Anesth Pain Med. 2021; 11(5):e118424. Based on their reports, the agitation score, frequency of complications, such as cough or vomiting, and bleeding volume were significantly lower in patients receiving dexmedetomidine (0.5 µg/kg) (23). Sato et al. showed that the administration of dexmedetomidine at a dose of 0.3 µg/kg could reduce the agitation levels in pediatrics undergoing surgical procedures, but they also mentioned that further studies using high doses are required (24).
The findings of the present study were in line with these reports showing the effectiveness of dexmedetomidine in reducing the post-operative agitation scores; how-ever, the important point is that none of the previous studies have compared two different dosages of dexmedetomidine and there is still doubt regarding the most effective dosage of this drug.
In 2019, Zhang in China showed that dexmedetomidine at a dose of 0.3 µg/kg could be helpful in reducing the agitation and providing suitable sedation in pediatrics (25). These data are consistent with the findings of our study. We found that both doses (1 µg/kg and 2 µg/kg) of dexmedetomidine are effective; however, the dose of 2 µg/kg caused a longer effect on lowering the agitation than The mean sedation score was significantly higher in group B, and this score was lower in the control group at the time of entrance to the recovery room (P= 0.031).
Group A had the lowest agitation score after 45 minutes of the recovery (P= 0.041), and group B had the lowest agitation score after 60 minutes of the recovery compared to other groups of patients (P= 0.003). the dose of 1 µg/kg and dexmedetomidine at 2 µg/kg provided a better sedation score in pediatric patients.
Another finding of the present study was a reduction in recovery duration and bleeding volume in pediatric patients that received dexmedetomidine at the doses of 1 µg/kg and 2 µg/kg, respectively.
These findings could also have clinical importance, especially in surgical procedures that are associated with high bleeding volumes. Phan and Nahata evaluated the clinical uses of dexmedetomidine in pediatric patients. A review on previous studies on dexmedetomidine showed that most studies have compared a single dose of 1 µg/kg or 0.5 µg/kg of dexmedetomidine with the placebo and reported its beneficial effects. They also showed that dexmedetomidine could provide a better surgeon's satisfaction due to reduced bleeding in the surgical field (26). Other studies have also mentioned that the administration of dexmedetomidine could reduce the bleeding volume and also provide better sedation (27)(28)(29). However, these studies have administered a single dose of dexmedetomidine. The main point of our study was the comparison of two different dosages of dexmedetomidine in pediatrics.
Limitation
One of the limitations of the current study was not evaluating the surgeon's satisfaction with the surgical field vision. However, our results showed significant effects of dexmedetomidine in providing proper sedation and reducing agitation in children.
Conclusions
Considering that dexmedetomidine at a dose of 1 µg/kg is associated with decreased agitation and shorter time spent in the recovery room and regarding the complications of dexmedetomidine that could be observed using both dosages, we recommend that anesthesiologists should pay more attention to the beneficial characteristics of dexmedetomidine at a dose of 1 µg/kg, especially in pediatric surgical procedures.
Acknowledgments
We would like to thank Dr. Aryan Rafiee Zadeh, as well as all healthcare professionals of the Al-Zahra hospital, for their cooperation and support.
Footnotes
Authors' Contribution: A. S. conceived and designed the evaluation and drafted the manuscript. H. A. participated in designing the evaluation, performed parts of the statistical analysis, and helped to draft the manuscript. H. S. re-evaluated the clinical data, revised the manuscript, performed the statistical analysis, and revised the manuscript. S. S. collected the clinical data, interpreted them, and revised the manuscript. | 2021-11-05T15:16:08.105Z | 2021-10-01T00:00:00.000 | {
"year": 2021,
"sha1": "6366d5a0fff4a721a3ca3f70dd6fbbdab5855c5a",
"oa_license": "CCBYNC",
"oa_url": "https://sites.kowsarpub.com/aapm/cdn/dl/f70295f2-43fb-11ec-81c6-3ffc602a8db6",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c22ebb56b2153f22e29119a885d7e9c6c49e0186",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
490106 | pes2o/s2orc | v3-fos-license | ER to Golgi transport: Requirement for p115 at a pre-Golgi VTC stage.
The membrane transport factor p115 functions in the secretory pathway of mammalian cells. Using biochemical and morphological approaches, we show that p115 participates in the assembly and maintenance of normal Golgi structure and is required for ER to Golgi traffic at a pre-Golgi stage. Injection of antibodies against p115 into intact WIF-B cells caused Golgi disruption and inhibited Golgi complex reassembly after BFA treatment and wash-out. Addition of anti-p115 antibodies or depletion of p115 from a VSVtsO45 based semi-intact cell transport assay inhibited transport. The inhibition occurred after VSV glycoprotein (VSV-G) exit from the ER but before its delivery to the Golgi complex, and resulted in VSV-G protein accumulating in peripheral vesicular tubular clusters (VTCs). The p115-requiring step of transport followed the rab1-requiring step and preceded the Ca(2+)-requiring step. Unexpectedly, mannosidase I redistributed from the Golgi complex to colocalize with VSV-G protein arrested in pre-Golgi VTCs by p115 depletion. Redistribution of mannosidase I was also observed in cells incubated at 15 degrees C. Our data show that p115 is essential for the translocation of pre-Golgi VTCs from peripheral sites to the Golgi stack. This defines a previously uncharacterized function for p115 at the VTC stage of ER to Golgi traffic.
E
NDOPLASMIC reticulum (ER) to Golgi traffic has been extensively studied and a general framework for the process has been elucidated (for reviews see Bannykh et al., 1998;Hong, 1998;Glick and Malhotra, 1998).Export from the ER defines the first step of the pathway and is mediated by the recruitment of the coat protein (COP) 1 II coats (composed of sec23/24 and sec13/ 31 complexes and Sar1p) and the subsequent budding of COP II-coated vesicles (for review see Barlowe, 1998).The coat proteins direct the sorting and concentration of proteins into the budding profiles, resulting in the preferential recruitment of cargo and exclusion of resident ER proteins (Aridor et al., 1998).COP II vesicles bud from specialized ER export sites morphologically defined as an assembly of centrally located tubules surrounded by budding profiles still attached to the ER (Bannykh et al., 1996;Bannykh and Balch, 1997).Budded COP II vesicles fuse with each other and with already fused vesicles to form the polymorphic tubular structures that have been called ERGIC (ER-Golgi intermediate compartment), intermediate compartment, or vesicular tubular clusters (VTCs) (Saraste and Svensson, 1991;Balch et al., 1994;Tang et al., 1995;Bannykh et al., 1998).VTCs comprise a morphologically and functionally complex compartment integrating the anterograde ER to Golgi complex and the retrograde recycling transport pathways.VTCs progress from the cell periphery towards the microtubule organizing center and the adjoining Golgi complex on microtubules in a dynein/dynactin-mediated process (for review see Lippincott-Schwartz, 1998).During their limited life span, VTCs undergo maturation by the selective recycling and acquisition of specific proteins and the sequential exchange of specific molecules (e.g., COP II coats exchange for COP I coats on VTCs [Pepperkok et al., 1993;Peter et al., 1993;Aridor et al., 1995;Scales et al., 1997]).Current evidence favors the view that COP I vesicles mediate the selective retrieval of proteins from the Golgi back to the VTCs and/or the ER, but their participation in anterograde traffic has not been ruled out (Lowe and Kreis, 1998).VTCs accumulate in the peri-Golgi region and, once there, seem to form a more extensive array of tubules, sometimes called the cis-Golgi network or the CGN.The exact relationship between VTCs and the CGN is unclear and it is possible that VTCs fuse to make a CGN or that the CGN is a large aggregate of individual VTCs.
Fusion of COP II vesicles to form VTCs and the subsequent assembly of VTCs to form cis-Golgi requires the concerted function of numerous molecules, among them the compartment-specific GTPases rab1 (Plutner et al., 1991;Tisdale et al., 1992;Pind et al., 1994) and rab2 (Chavrier et al., 1990;Tisdale et al., 1992;Tisdale and Balch, 1996), the general soluble fusion factors NSF ( N -ethylmaleimide-sensitive fusion factor) and SNAP (soluble NSF attachment protein), and a set of distinct integral membrane proteins v-SNARES and t-SNARES (vesicle-and target-SNAP receptors) believed to confer specificity to fusion (Söllner et al., 1993; for reviews see Rothman, 1994;Hay and Scheller, 1997;Nichols and Pelham, 1998).The molecular mechanism underlying membrane traffic is highly conserved, and in yeast, ER to Golgi transport is also dependent on the function of the compartment-specific GTPase Ypt1p (Segev et al., 1988), the general fusion factors Sec18p (NSF homologue) and Sec17p (SNAP homologue), and the specific interactions of ER-Golgi SNAREs (Bos1p, Bet1p, Sed5p, Sec22p, and Ykt6p) (for reviews see Ferro-Novick and Jahn, 1994;Mc-New et al., 1997).SNAREs mediate membrane interactions at relatively short distances (an assembled SNARE pair has been shown to be only 14 nm; Hanson et al., 1997), and it appears that donor and target membranes interact at longer distances through the selective interactions of specific tethering proteins (Orci et al., 1998; for review see Pfeffer, 1996).In yeast, tethering of COP II vesicles to Golgi membranes (yeast do not have morphologically identifiable VTCs) occurs before SNARE complex formation and is dependent on the function of a peripheral membrane-associated protein, Uso1p (Nakajima et al., 1991;Sapperstein et al., 1996;Cao et al., 1998).
A mammalian homologue of Uso1p, p115, has been characterized and shown to have a multidomain structure composed of an NH 2 -terminal globular head, central coiled-coil region, and a COOH-terminal acidic domain (Barroso et al., 1995;Sapperstein et al., 1995;Yamakawa et al., 1996).Like Uso1p, p115 assembles into parallel dimers.Functionally, p115 has been identified as a requirement in in vivo-reconstituted assays measuring cis-to medial-Golgi transport (Waters et al., 1992), transcytotic traffic (Barroso et al., 1995), regrowth of Golgi cisterna from mitotic Golgi fragments (Rabouille et al., 1995(Rabouille et al., , 1998;;Nakamura et al., 1997), and cisternal stacking (Shorter and Warren, 1999).Consistent with the role of p115 in Golgi events, p115 has been localized predominantly to the Golgi at the light microscope level.However, by electron microscopy, p115 was found preferentially associated with VTCs adjacent to the Golgi stack (the close proximity of VTCs to the Golgi [Ladinsky et al., 1999] makes them difficult to resolve from Golgi elements by immunofluorescence), and to cycle extensively between the Golgi and earlier compartments of the secretory pathway (Nelson et al., 1998).Such localization and cycling behavior is characteristic for proteins involved in ER to Golgi traffic, suggesting that p115, like Uso1p, might participate in this stage of the secretory pathway.
Here, we show that p115 is critical for the dynamic maintenance and the de novo assembly of morphologically normal Golgi complexes, and that p115 is required for ER to Golgi traffic at a pre-Golgi VTC stage, after the rab1requiring step but preceding the Ca 2 ϩ -requiring step.Biochemical and morphological analysis of cargo VSV glycoprotein (VSV-G) protein progression from the ER to the Golgi indicated that p115 function is not essential for events leading to the formation of VTCs, but that p115 is required for the subsequent delivery of VTCs from the peripheral sites, where they form, to the Golgi stack.Characterization of the p115-arrested VTCs showed that the cis-Golgi enzyme mannosidase I (Mann I), but not Mann II, specifically redistributed from the Golgi stack and colocalized with such VTCs.Our results indicate that p115 is required for traffic progression along the secretory pathway at a pre-Golgi VTC step, suggesting that p115 is essential for events leading to VTC delivery to the Golgi stack.
Antibodies
Rabbit polyclonal antibodies against p115 were generated by immunization with purified rat p115 (Barroso et al., 1995).For affinity purification of antibodies, purified GST-p115 fusion protein was run on 7.5% acrylamide gels, and transferred to nitrocellulose.Strips of nitrocellulose containing p115 were incubated with immune serum in PBS, 5% dried milk, and 0.1% Tween 20 for 3 h at room temperature.Bound antibodies were eluted with 0.1 M glycine, pH 3.0, and then neutralized with 1/10 vol 1 M phosphate buffer, pH 7.4.These anti-p115 antibodies were either used in transport assays or cross-linked to protein A-Sepharose 4 FF (Pharmacia Biotech, Inc.) with dimethyl pimelimidate (Pierce Chemical Co.) according to the manufacturer's protocol, and subsequently used for p115 immunodepletion experiments.Mouse polyclonal antibodies against GM130 were generated by immunization with inclusion bodies from bacteria expressing full-length GM130 (Nelson et al., 1998).Anti-giantin monoclonal G1/133 (Linstedt and Hauri, 1993) and anti-ERGIC-53 monoclonal G1/93 (Schweizer et al., 1988)
Generation of GST-p115 Construct
Full-length p115 was cloned into the BamHI-NotI restriction sites of the pGEX-6P-2 GST vector (Pharmacia Biotech, Inc.).p115 was obtained by PCR using as a template a pBluescript-II-KS plasmid (Stratagene) encoding rat p115 cDNA previously described (Barroso et al., 1995).GST-p115 fusion protein expression and purification were performed according to the manufacturer's protocol.
Purification of Cytosolic p115
p115 was affinity-purified from rat liver cytosol using affinity-purified polyclonal antibodies raised against p115 cross-linked to beads as previously described (Barroso et al., 1995).The p115 preparation was judged homogenous by Coomassie blue staining.
Microinjection
Ascites raised against p115 (3A10) were concentrated by centrifugation in a Microcon 30 (Pierce Chemical Co.) and dialyzed into MI buffer (48 mM K 2 HPO 4 , 140 mM NaH 2 PO 4 , 4.5 mM KH 2 PO 4 , pH 7.2).Denaturation was by boiling to 100 Њ C for 3 min.Before injection, solutions were centrifuged at 15,800 g for 20 min, and then mixed with 1/10 vol of Texas red-dextran (TR-dextran, 10,000 D, 50 mg/ml) and 10 ϫ MI buffer.Injections into the cytoplasm of WIF-B cells were performed with an Eppendorf transjector 5246 and micromanipulator 5171 attached to a Zeiss Axiovert 100 microscope.The pressure was 50 hectoPascal for 0.1-0.2s using an Eppendorf femtotip needle.After injection, cells were incubated with Ham's F12 medium (Gibco Laboratories) at 37 Њ C for 2-3 h, fixed, and processed for immunofluorescence.In some experiments, 2.5 g/ml BFA was used.Fluorescent images were collected with the MCID analysis system (Zeiss Axiovert 100 microscope) or with a Zeiss confocal microscope (LSM410).Digitized images were cropped, assembled, and labeled in Adobe Photoshop.
Semi-intact Cell ER-Golgi Transport Assay
The ER to Golgi transport assay was performed as described previously (Beckers et al., 1987;Schwaninger et al., 1992b).In brief, normal rat kidney (NRK) cells, or LEC-1 (a CHO-derived cell line deficient in NAGT-1 and analogous to CHO15B) cells were grown on 10-cm petri dishes to 80-90% of confluence and infected with the temperature-sensitive strain of the vesicular stomatitis virus, VSVtsO45 at 32 Њ C for 3-4 h (Bergmann, 1989).The cells were pulse-labeled with 35 S-trans label (200 mCi/ml; ICN) at the restrictive temperature (42 Њ C) for 10 min, chased with complete medium for 5 min, and perforated by hypotonic swelling and scraping to make semi-intact cells.A transport reaction was performed in a final total volume of 40 l in a buffer that contained 25 mM Hepes-KOH, pH 7.2, 75 mM potassium acetate, 2.5 mM magnesium acetate, 5 mM EGTA, 1.8 mM CaCl 2 , 1 mM N -acetylglucosamine, ATP regeneration system (1 mM ATP, 5 mM creatine phosphate, and 0.2 IU rabbit muscle creatine phosphokinase), 5 l rat liver cytosol, 5 l of semi-intact cells in 50 mM Hepes-KOH, pH 7.2, and 90 mM potassium acetate.Transport was initiated by transfer of cells to 32 Њ C.After 90 min of incubation, cells were pelleted, resuspended in appropriated buffer, and digested with endoglycosidase H (endo-H) or endoglycosidase D (endo-D) as described previously (Schwaninger et al., 1992b).The samples were analyzed on 8% SDS-PAGE and fluorography.The transport was quantified using GS-700 imaging densitometer (Bio-Rad Laboratories).For antibody inhibition of transport assay, affinity-purified p115 antibodies were added into the complete assay cocktail and incubated on ice for 30 min before use.Staging experiments were performed as previously described (Peter et al., 1998).
Immunodepletion of Rat Liver Cytosol
Rat liver cytosol was prepared as previously described (Dascher et al., 1994).Immunopurified anti-p115 antibodies cross-linked to protein A-Sepharose were incubated (2 h at 4 Њ C) with rat liver cytosol, and the level of p115 depletion was tested by SDS-PAGE and immunoblotting.
Morphological Analysis of VSVtsO45 G Protein Transport
This analysis was performed as described previously (Plutner et al., 1992).
In brief, NRK cells plated on coverslips were infected with VSVtsO45 at 32 Њ C for 30 min, followed by an incubation at 42 Њ C for 3 h, and then shifted to ice and permeabilized with digitonin (20 mg/ml).Coverslips were incubated in various transport cocktails at 32 Њ C for 90 min.For antibody inhibition of the transport assay, affinity-purified p115 antibodies were added into the complete transport mix and incubated on ice for 30 min followed by an incubation at 32 Њ C for 90 min.Transport was determined by transferring coverslips to ice and fixing them in 3% formaldehyde/PBS for 10 min.The coverslips were processed for double label immunofluorescence.
Immunofluorescence Microscopy
Cells grown on glass coverslips were washed three times in PBS and fixed in 3% paraformaldehyde in PBS for 10 min at room temperature.Paraformaldehyde was quenched with 10 mM ammonium chloride, and cells were permeabilized with PBS, 0.1% Triton X-100 for 7 min at room temperature.The coverslips were washed (three times, 2 min per wash) with PBS, and blocked in PBS, 0.4% fish skin gelatin, 0.2% Tween 20 for 5 min, followed by blocking in PBS, 2.5% goat serum, 0.2% Tween for 5 min.Cells were incubated with primary antibody diluted in PBS, 0.4% fish skin gelatin, and 0.2% Tween 20 for 45 min at 37 Њ C. Coverslips were washed (five times, 5 min per wash) with PBS and 0.2% Tween 20.Sec-ondary antibodies coupled to FITC or rhodamine were diluted in 2.5% goat serum and incubated on coverslips for 30 min at 37 Њ C. Coverslips were washed with PBS and 0.2% Tween 20 as above, and mounted on slides in 9:1 glycerol/PBS with 0.1% q-phenylenediamine.Fluorescence patterns were visualized with an Olympus IX70 epifluorescence microscope.Optical sections were captured with a CCD high resolution camera equipped with a camera/computer interface.Images were analyzed with a power Mac 9500/132 computer using IPLab Spectrum software (Scanalytics Inc.).
Antibodies against p115 Disrupt Golgi Structure and Prevent Reassembly of Golgi Stacks In Vivo
The requirement for p115 has been previously studied only in in vitro-reconstituted assays (Waters et al., 1992;Barroso et al., 1995;Rabouille et al., 1995Rabouille et al., , 1998)).To examine the function of p115 in vivo, monoclonal anti-p115 an- tibodies were microinjected into WIF-B cells (Shanks et al., 1994), and their effect on Golgi structure was monitored by the distribution of Mann II.As shown in Fig. 1, injection of anti-p115 antibodies had an effect on Golgi structure.A field of cells is shown by phase-contrast (A), and the injected cells are identified by their content of coinjected TR-dextran (B).Within 2 h after injection, the Golgi complexes in all injected cells were disassembled, and Mann II was present in relatively large punctate structures dispersed throughout the cells (C).The structures appeared superficially similar to those observed in cells transfected with mutant forms of p115 lacking portions of the globular head or the coiled-coil tail (Nelson et al., 1998).Injection of control antibodies against the bile canalicular plasma membrane protein 5 Ј -nucleotidase did not perturb Golgi structure (D-F).ER to Golgi traffic can be disrupted by brefeldin A (BFA), which inhibits a guanine nucleotide exchange factor for ADP ribosylation factor; thus, preventing assembly of COP I coats, and ultimately blocking normal traffic between ER and the Golgi (Donaldson et al., 1992b;Helms and Rothman, 1992).This results in COP I-independent tubulation of the Golgi and the relocation of Golgi proteins into the ER (Misumi et al., 1986;Fujiwara et al., 1988;Lippincott-Schwartz et al., 1989).BFA-induced membrane redistribution has given valuable insights into components and mechanisms involved in regulation of the secretory pathway (Klausner et al., 1992), and we combined BFA treatments with microinjection of anti-p115 antibodies to characterize the effect of anti-p115 antibodies on ER-Golgi traffic.
The redistribution of Mann II from the Golgi into the ER in control WIF-B cells or in cells injected with anti-p115 antibodies was examined first.As shown in Fig. 2, Mann II was detected in a normal perinuclear Golgi pattern in uninjected cells or in injected cells immediately after the antibody injection and before BFA treatment (A).The injected cells were identified by the presence of anti-p115 IgG (B).When an analogous field of cells was treated with BFA for 30 min, Mann II relocated from the Golgi to the ER in uninjected and injected cells (C).Only a single time point was analyzed, and it remains possible that the kinetics of Mann II redistribution may have varied in injected and uninjected cells.Some Golgi staining was still evident in uninjected and injected cells after the BFA treatment (C, arrowheads).After 30 min, the majority of anti-p115 IgG was associated with membranes, presumably representing p115 localization (D), as shown previously (Nelson et al., 1998).The results suggest that anti-p115 antibodies do not block the Golgi to ER redistribution of Mann II induced by BFA treatment.
To examine if the antibodies inhibit the movement of Mann II from the ER, WIF-B cells were first treated with BFA for 30 min to relocate Mann II from the Golgi to the ER, and then injected with anti-p115 antibodies.The temporal sequence of Golgi recovery after BFA wash-out was followed.As shown in Fig. 3 A, punctate structures dispersed throughout the cell were detected after a 10-min BFA wash-out in both, uninjected and injected cells.The size and distribution of the elements were comparable, suggesting that the antibodies do not block early stages of BFA recovery.A difference was seen after 30 min of BFA wash-out, when defined Golgi structures could be seen in uninjected cells (C, arrows), while significantly smaller, dispersed structures were present in the injected cells (C, arrowheads).The effect was also evident after 120 min of BFA wash-out, when compact Golgi structures were seen in uninjected cells (E, arrows), whereas dispersed punctate structures persisted in the injected cells (E, arrowheads).In some injected cells, a more Golgi-like Mann II pattern was seen (E, asterisk), but the structure appeared less compact and organized.These results suggest that anti-p115 antibodies prevent the reassembly of normal Golgi complexes.
p115 Is Associated with Functional VTCs Moving VSV-G Protein from the ER to the Golgi
The inhibitory effects of anti-p115 antibodies on the maintenance and reassembly of normal Golgi structure, coupled with the previous finding that p115 is present on VTCs and cycles between the Golgi and earlier secretory compartments (Nelson et al., 1998), suggested a potential role for p115 in ER to Golgi transport.To analyze if p115 is present on functional VTCs transporting cargo from the ER to the Golgi, we used a temperature-sensitive strain of the vesicular stomatitis virus (VSVtsO45) as a transport marker (for review see Bergmann, 1989).The viral VSV-G protein fails to exit the ER at 42 Њ C, the nonpermissive temperature, but after shifting the cells to the permissive temperature of 32 Њ C, a wave of VSV-G protein enters the secretory pathway, and its movement from the ER to the Golgi can be monitored morphologically (Pepperkok et al., 1993;Balch et al., 1994).NRK cells infected with the virus and cultured at the nonpermissive temperature for 3 h contain VSV-G protein in the ER, whereas p115 is predominantly detected in the Golgi region (Fig. 4, A-C).When infected cells were subsequently shifted from 42 to 15 Њ C and incubated at 15 Њ C for 3 h, VSV-G protein movement to the Golgi was arrested in peripheral VTCs that also contained p115 (Fig. 4, D-F).We have shown previously that p115 colocalizes with the VTC marker, ERGIC-53 (Schweizer et al., 1988) when VTCs are preferentially accumulated during low (15 Њ C) temperature incubation (Saraste and Svensson, 1991;Nelson et al., 1998).Bonafide Golgi proteins such as galactosyl-transferase or Mann II do not redistribute to peripheral VTCs after low temperature treatment (Lippincott-Schwartz et al., 1990;Nelson et al., 1998; see Fig. 10 E).When infected cells were shifted from 42 to 32 Њ C for 1 h, VSV-G protein and p115 colocalized in the Golgi (Fig. 4, G-I).These results indicate that p115 is a component of VTCs that move VSV-G protein from the ER to the Golgi, and raise the possibility that p115 could be involved in VTC dynamics.
p115 Is Required for ER to Golgi Transport
Although p115 has been identified as a cis-to medial-Golgi transport factor (Waters et al., 1992), its yeast homologue, Uso1p, has been shown to act in ER to Golgi traffic (Nakajima et al., 1991;Cao et al., 1998).To examine directly p115 participation in ER to Golgi traffic, we used a previously developed semi-intact cell transport assay (Beckers et al., 1987).NRK cells were infected with VSVtsO45 and radiolabeled at 42 Њ C. Cells were permeabi-lized to remove endogenous cytosol, supplemented with exogenous transport cocktails, and shifted to 32 Њ C to initiate VSV-G protein transport.Delivery of VSV-G protein to the Golgi was assessed by its carbohydrate processing, as defined by endo-H resistance.VSV-G protein oligosaccharide chains are processed during transport by the sequential actions of enzymes localized throughout the Golgi stack.The processing involves the sequential function of Mann I and N -acetylglucosamine transferase I (NAGT-1), both considered cis-Golgi enzymes (Schwaninger et al., 1992b), and of Mann II, localized in the medial/ trans Golgi stack (Velasco et al., 1993).After processing by Mann I, VSV-G acquires endo-D sensitivity, whereas subsequent processing by NAGT-1 and Mann II confers endo-H resistance.As shown in Fig. 5 A, when complete transport cocktail was added to permeabilized cells, ف 50% of VSV-G protein was processed to an endo-H-resistant form (top band, lane 2), and this is set as 100% processing.The percent processing is analogous to the level of processing reported previously (Tisdale and Balch, 1996;Subramanian et al., 1996).In contrast, when transport was analyzed with an ATP-depleting system, VSV-G protein remained sensitive to endo-H (bottom band, lane 1), and this is set as 0% processing.The addition of increasing amounts of affinity-purified anti-p115 antibodies (from 0.1 to 0.8 g) led to a dose-dependent inhibition of VSV-G protein processing to the endo-H-resistant form (lanes 3-6).To provide quantitative data on VSV-G processing, analogous data from repeated experiments ( n ϭ 3) were evaluated by densitometry, and the average of relative percent is presented in the accompanying bar graph.The relative processing was reduced by 15% in the presence of 0.1 g of antibody with Ͼ 80% inhibition when 0.4 g of anti-p115 antibodies were Figure 3. Antibodies against p115 block Golgi complex reassembly during BFA wash-out.WIF-B cells were treated with BFA for 30 min, and then injected with mAbs against p115 mixed with TR-dextran.Cells were fixed after 10 min (A and B), 30 min (C and D), or 120 min (E and F) of BFA wash-out, and processed for immunofluorescence using antibodies against Mann II (A, C, and E).Injected cells were identified by their content of TR-dextran (B, D, and F), and are traced in outline in A, C, and E. After 10 min of BFA wash-out, Mann II relocation to punctate structures is indistinguishable in injected and uninjected cells.After 30 and 120 min of BFA wash-out, Mann II appears in morphologically normal Golgi structures in uninjected cells, but remains in punctate structures in injected cells.Arrowheads mark Golgi in injected cells, arrows point to Golgi in uninjected cells.Asterisk denotes a more compact Golgi complex in injected cells.added.When preimmune antibodies were added to the transport assay, normal processing of VSV-G protein was observed (data not shown).
To ensure that the inhibitory effects of the anti-p115 antibodies were due to an interaction with p115, the antibodies were preincubated either with GST-p115 fusion protein or GST transferred to nitrocellulose strips.The nonbound fractions were tested for inhibitory activity in the semi-intact transport assay.As shown in Fig. 5 B, lane 3, preincubation of the antibody with GST-p115 nitrocellulose strips efficiently neutralized its inhibitory effect on traffic, and processing of VSV-G protein to the endo-H-resistant form was comparable to the control situation with complete cytosol (lane 2).In contrast, inhibition was still apparent with antibodies incubated with GST strips (lane 4), and processing of VSV-G protein to the endo-H-resistant form was comparable to that in the absence of ATP (lane 1).These results suggest that anti-p115 antibodies block ER to Golgi transport through a specific interaction with p115.
To extend these findings, VSV-G protein transport in the presence of limiting amounts of p115 was analyzed.In the semi-intact cell transport assay, exogenous rat liver cytosol must be added to provide cytosolic and peripheral membrane proteins released during permeabilization (Beckers et al., 1987).p115 is peripherally associated with membranes, and during cell disruption, is largely released into the cytosol (Waters et al., 1992;Nelson et al., 1998).Rat liver cytosol contains high amounts of p115 ( ف 0.5 g/mg).To examine if such exogenously added p115 is required for transport, p115 was removed from the cytosol by immunodepletion.The extent of immunodepletion was analyzed by immunoblotting (Fig. 5 C, p115 panel, lanes 2 and 3 compared with lane 4).The immunoblot was evaluated by densitometry, and the relative amount of p115 present is shown in the accompanying bar graph.Up to 60% of cytosolic p115 was removed in lane 4. When such p115-depleted cytosol was used in the transport assay, VSV-G protein processing to the endo-H-resistant form was inhibited by ف 90%.Immunodepletion of cytosol with nonspecific antibodies (affinity-purified rabbit anti-goat IgG, lane 3) or preimmune serum (data not shown) had no effect on VSV-G protein transport.Reactions containing complete transport cocktail (lane 2) or cocktail containing an ATP-depleting system (lane 1) were analyzed as positive and negative controls and defined as 100 and 0% VSV-G protein processing, respectively.
To examine if addition of purified p115 could rescue transport, p115-depleted cytosol (analogous to that in lane 4) was supplemented with increasing amounts of purified p115.p115 was purified from rat liver cytosol on affinitypurified anti-p115 antibodies cross-linked to a protein A-Sepharose column.As shown in Fig. 5 D, the material eluted from the anti-p115 IgG column (lane 1) contains a A-C) or incubated at 15ЊC for 3 h (D-F) or at 32ЊC for 1 h (G-I) before fixation.Cell were processed for double label immunofluorescence using antibodies against p115 (A, D, and G) and antibodies against VSV-G protein (B, E, and H).At 42ЊC, VSV-G protein is present in the ER (B), whereas p115 is predominantly localized to the Golgi (A).p115 and VSV-G protein colocalize in peripheral VTCs after 15ЊC incubation (D-F), and in the Golgi after 32ЊC incubation (G-I).Bar 10 m.
band of ف 110 kD, whereas no such band was eluted from a control preimmune IgG column (lane 2).To ensure that the 110-kD band was p115, the 110-kD region was excised from a gel, subjected to tryptic digestion and the peptides were separated and characterized by MALDI mass spectrometry combined with sequence database searching.The peptide mass map of 10 major peptides (Fig. 5 D) match with p115, and no other rat protein in that molecular mass range was found in the search.Analogous purified p115 was added to the p115 depleted cytosol.As shown in Fig. 5 C, p115 panel, lanes 5-7, and bar graph, the supplemented cytosol contained 110, ف 140, and ف 160% of p115 in untreated cytosol, respectively.When such supplemented cytosols were used in the transport assay, VSV-G protein was processed to the endo-H-resistant form with ف 40, ف 50, and ف 100% processing efficiency, respectively (VSV-G panel, lanes 5-7, and bar graph).Taken together, these results indicate that p115 is essential for VSV-G protein delivery from the ER to the Golgi.
p115 Is Required for ER to Golgi Transport at a Step after the rab1 Requirement and before the Ca 2 ϩ Requirement
Previous work suggested that p115 functions to tether donor and acceptor membranes before membrane fusion (Barroso et al., 1995;Sönnischen et al., 1998), but did not place the p115 requirement relative to requirements for other known transport factors.Studies on Uso1p suggested that its function is regulated by Ypt1p (Sapperstein Figure 5. p115 is essential for ER to Golgi transport.ER to Golgi transport was performed in semi-intact NRK cells.Transport is measured as the percentage of VSV-G protein processed from the endo-H-sensitive (S) to the endo-H-resistant (R) form.(A) Transport reactions contained complete transport cocktail (lanes 2-6), or cocktail with ATP-depleting system (lane 1).Reactions were supplemented with increasing amounts of affinity-purified anti-p115 antibodies (lanes 3-6).Transport of VSV-G protein was proportionally inhibited in the presence of antibodies against p115.Analogous gels (n ϭ 3) were quantitated by densitometry and the averages are presented in the bar graph.Transport in lane 1 is set as 0% and in lane 2 as 100%.(B) Transport reactions contained complete transport cocktail (lane 2), cocktail with ATP-depleting system (lane 1), or cocktail supplemented with anti-p115 antibodies preincubated with GST-p115 (lane 3), or GST (lane 4).Preincubation of anti-p115 antibodies with GST-p115 neutralized their inhibitory effect on transport.(C) Transport reactions contained complete transport cocktail (lanes 2-7), or cocktail with ATP-depleting system (lane 1).In lane 3, transport cocktail contained cytosol preincubated with control IgGs.In lane 4, transport cocktail contained cytosol preincubated with anti-p115 antibodies.Reduced level of p115 is visible in lane 4 and leads to inhibition of VSV-G protein transport.Increasing amounts of purified p115 were added to the cytosol shown in lane 4. Addition of purified p115 overcame the inhibitory effect of p115 removal and supported VSV-G protein transport (lanes 5-7).Analogous gels ( n ϭ 3) were quantitated by densitometry, and the averages are presented in the bar graph.Transport in lane 1 is set as 0% and in lane 2 as 100%.An aliquot of each transport reaction was probed by immunoblotting with anti-p115 antibodies and the immunoblot is shown in panel p115.The same amount of p115 was used in lane 1 as in lane 2 and only reaction in lane 2 was analyzed.(D) Rat liver cytosol was incubated with anti-p115 antibodies or control IgGs cross-linked to protein A-Sepharose.Bound material was eluted and analyzed by SDS-PAGE.An -011فkD band was visible after Coomassie blue staining in material eluted from anti-p115 column (lane 1) but not in control eluate (lane 2).The -011فkD band was excised, digested with trypsin, and the resulting peptides analyzed by MALDI mass spectrometry.The peptide mass map of the 10 most abundant peptides (marked by asterisks) matched the sequence of p115.et al., 1996;Cao et al., 1998), and Ypt1p and its mammalian homologue rab1 are essential for ER to Golgi transport (Plutner et al., 1990(Plutner et al., , 1991;;Schwaninger et al., 1992b;Ferro-Novick and Jahn, 1994;Nuoffer et al., 1994).To order the sequence of rab1-and p115-requiring transport steps in mammalian cells, staging experiments were performed.In the first stage, p115-depleted cytosol was added to the semi-intact cell transport assay, and the cells were incubated at the permissive temperature for 60 min to allow VSV-G protein to accumulate in the p115-arrested compartment.The cells were collected, and in the second stage, incubated with either p115-depleted cytosol, complete cytosol, rab1-depleted cytosol, or complete cytosol supplemented with anti-rab1 antibodies.Rab1-depleted cytosol contained Ͻ 5% of the rab1 found in untreated cytosol (data not shown), and 3 g of anti-rab antibodies were used, an amount similar to that shown previously to be inhibitory in the semi-intact cell transport assay (Plut-ner et al., 1991).As shown in Fig. 6 A (gel and bar graph, lane 1), when p115-depleted cytosol was used in both stages of the transport assay, Ͻ 10% of VSV-G protein was processed to the endo-H-resistant form.This represents the background of the assay and is set as 0% relative processing.When complete cytosol (lane 2) was added in the second stage, ف 35% of VSV-G protein was transported to the Golgi and acquired endo-H resistance (set as 100% relative processing), the same extent as when completed cytosol was used in both stages (data not shown).Significantly, when rab1-depleted cytosol (lane 3), or complete cytosol containing anti-rab1 antibodies (lane 4) were used, comparable ف 90% level of relative processing was observed, indicating that progression of VSV-G protein from the p115-depletion blocked compartment to the Golgi does not require rab1.
In agreement with previous results showing partial inhibition after rab1 depletion (Peter et al., 1998), when stages I and II were performed with rab1-depleted cytosol, VSV-G transport was blocked by ف 60% (lane 5).When complete cytosol was added in the second stage (lane 6), VSV-G protein processing was Ͼ 80% of that in lane 2. In contrast, when p115-depleted cytosol was added in the second stage (lane 7), VSV-G protein processing was inhibited by ف 60%, the same extent as when rab1-depleted cytosol was used in both stages (lane 5).This suggests that movement of VSV-G protein from the rab1-depletion blocked compartment to the Golgi requires p115.Together, the data indicate that rab1 is required before the p115-requiring step of VSV-G protein transport.
Using the same ER to Golgi transport assay, it has been shown that Ca 2 ϩ is required at a last transport step before membrane fusion (Beckers and Balch, 1989;Aridor et al., 1995).To determine if addition of anti-p115 antibodies or p115 depletion inhibits ER to Golgi transport at a stage before or after the Ca 2 ϩ requirement, staging experiments were performed.In the first stage, 10 mM EGTA was added to the semi-intact cell transport assay, and the cells were incubated at the permissive temperature for 60 min to allow VSV-G protein to accumulate in the EGTAarrested compartment.The cells were collected, and in the second stage, incubated with either cytosol containing 10 mM EGTA, complete cytosol, p115-depleted cytosol, or complete cytosol supplemented with anti-p115 antibodies.As shown in Fig. 6 B (gel and bar graph, lane 1), when EGTA was added to both stages of the transport assay, ف 3% of VSV-G protein acquired endo-H resistance.This represents the background of the assay and is set as 0% relative transport.When complete cytosol (lane 2) was added in the second stage, ف 35% of VSV-G protein was processed, and this represents 100% relative processing.The ف 35% processing is analogous to the percent processing observed when complete cytosol is added to both stages of transport (data not shown).Significantly, when p115-depleted cytosol (lane 3), or complete cytosol containing p115 antibodies (lane 4) was added in the second stage, ف 100% relative processing was observed, suggesting that progression of VSV-G protein from the Ca 2 ϩ -arrested compartment to the Golgi does not require p115.
This was confirmed by reverse staging.As shown in Fig. 6 B (lanes 5 and 6), addition of p115-depleted cytosol to both stages of the transport assay resulted in %51ف of Figure 6.p115 acts after the rab1 requirement and before the Ca 2ϩ requirement.Two-stage ER to Golgi transport was performed in semi-intact NRK cells.In stage I, cells were incubated for 60 min at 32ЊC.Cells were then pelleted and resuspended in fresh transport cocktail and incubated in stage II for 45 min at 32ЊC.Transport is measured as the percentage of VSV-G protein processed from the endo-H-sensitive (S) to the endo-H-resistant (R) form.(A) VSV-G protein is minimally processed when p115depleted cytosol is used in both stages (lane 1), and this is set as 0% relative processing.VSV-G protein is processed when complete cytosol (lane 2) is used, and this is set as 100% relative processing.VSV-G protein is processed when rab1-depleted cytosol (lane 3) or cytosol containing anti-rab1 antibodies (lane 4) is added during the second stage.VSV-G protein is not processed when rab1-depleted cytosol is used in both stages (lane 5), but is processed when complete cytosol is used in the second stage (lane 6).VSV-G protein is not processed when p115-depleted cytosol is added during the second stage (lane 7).The staging indicates that p115 acts after the rab1-requiring step of transport.(B) VSV-G protein is minimally processed when EGTA is added to both stages (lane 1), and this is set as 0% relative processing.VSV-G protein is processed when complete transport cocktail is used in both stages (lane 2), and this is set as 100% relative processing.VSV-G protein is processed when p115-depleted cytosol (lane 3) or anti-p115 antibodies (lane 4) are added during the second stage.VSV-G protein is not processed when p115depleted cytosol is added to both stages (lanes 5 and 6) or when EGTA is added to the second stage (lane 8), but is processed when complete cytosol is added to the second stage (lane 7).The staging indicates that p115 precedes the Ca 2ϩ -requiring step of transport.
VSV-G protein acquiring endo-H resistance %04ف( relative processing).(In these experiments, p115 depletion was incomplete [data not shown] and accounts for the incomplete inhibition of VSV-G protein transport.)When complete cytosol was added in the second stage (lane 7), %03ف of VSV-G protein was processed %09ف( relative processing).When cytosol containing EGTA was added in the second stage, %5ف of the VSV-G protein was transported to the Golgi and acquired endo-H resistance %01ف( relative processing) (lane 8).This suggests that movement of VSV-G protein from the p115-depletion blocked compartment to the Golgi stack requires Ca 2ϩ .Together, the results indicate that p115 is required before the Ca 2ϩ -requiring stage of VSV-G protein transport.
Morphological examination of VSV-G protein localization in the presence of EGTA suggests that Ca 2ϩ is required for VTC delivery to the Golgi stack (Aridor et al., 1995).Our finding that p115 is required at a stage of transport before the Ca 2ϩ requirement suggested that p115 might also be required at that stage.
p115 Is Required for a VTC Step before the Delivery of VSV-G Protein to the Golgi Stack
To define where anti-p115 antibodies and p115 depletion blocked transport of VSV-G protein, morphological transport assays were performed.Addition of anti-p115 antibodies 01ف( ng IgG/l transport assay, an amount analogous to that shown to inhibit the biochemical assay in Fig. 5 A, lane 5) to the morphological transport assay prevented VSV-G protein from moving to the Golgi (Fig. 7, A-C).A representative experiment (from Ͼ10 analyses) is presented.VSV-G protein was not detected in the ER, indicating that anti-p115 antibodies had no effect on VSV-G protein exit from the ER and its delivery to post-ER transport intermediates, but prevented delivery of such intermediates to the Golgi.This resulted in accumulation of VSV-G protein in peripheral VTCs, most of which were labeled with anti-p115 antibodies (C, arrowheads).Anti-p115 antibodies were also detected in the Golgi (C, arrows), perhaps because of incomplete removal of p115 Figure 7. Antibodies against p115 block ER to Golgi transport of VSV-G protein at a pre-Golgi stage.NRK cells were infected with VSVtsO45 for 3 h at 42ЊC.Cells were permeabilized and supplemented with complete transport cocktail supplemented either with anti-p115 antibodies (A-C) or control antibodies (D-F).After transport at 32ЊC for 90 min, cells were processed by double label immunofluorescence using anti-p115 (A and D) and anti-VSV-G protein (B and E) antibodies.Addition of anti-p115 antibodies to the transport assay has no effect on VSV-G protein exit from the ER, but prevents VSV-G protein transport to the Golgi (A-C) and causes accumulation of VSV-G protein in scattered VTCs.Arrowheads point to peripheral structures containing VSV-G protein and anti-p115 antibodies.Arrows indicate Golgi elements labeled with anti-p115 antibodies but lacking VSV-G protein.Addition of control antibodies to the transport assay had no effect on VSV-G protein transport, and VSV-G protein was efficiently delivered to the Golgi (D-F).Bar 10 m. from membranes during permeabilization.Alternatively, anti-p115 antibodies might act by trapping p115 on Golgi membranes through the formation of inactive complexes.Addition of equivalent amounts of preimmune antibodies to the transport assay had no effect on VSV-G protein transport, and VSV-G protein was efficiently delivered to the Golgi, where it colocalized with p115 (D-F).The same pattern was observed when monoclonal anti-Mann II antibodies were added to the reaction (data not shown).
When p115-depleted cytosol (analogous to that unable to support the biochemical assay in Fig. 5 C, lane 4) was used in the morphological transport assay, VSV-G protein was also not transported to the Golgi, as evidenced by its lack of colocalization with Mann II (Fig. 8, A-C).Two representative cells from two different experiments are shown in these panels.VSV-G protein localized predominantly to peripheral VTCs, indistinguishable from those seen in the presence of anti-p115 antibodies (compare to Fig. 7 B).VSV-G protein was almost exclusively in peripheral VTCs, and was not present to any significant extent in the ER, confirming that p115 is not involved in ER exit and delivery of cargo to more distal transport intermediates.VSV-G protein was efficiently transported to the Golgi when complete cytosol was used (Fig. 8, D-F).Taken together, these results indicate that p115 function is required after the formation of post-ER VTCs but before their delivery to the Golgi stack.This defines a novel function for p115, and identifies the first step of membrane transport in which p115 participates.
Differential Processing of Arrested VSV-G Protein
Since addition of anti-p115 antibodies or p115 depletion inhibits VSV-G protein delivery to the Golgi and causes its accumulation in peripheral VTCs, we expected that the arrested VSV-G protein will not be processed by any of the Golgi glycosyl-modifying enzymes.In agreement, our data show that the arrested VSV-G protein remains endo-H sensitive, indicating a lack of processing by NAGT-1 and Mann II.To examine whether the arrested VSV-G protein is processed by the cis-Golgi enzyme Mann I, endo-D resistance/sensitivity was analyzed in LEC-1 and NRK cells.
Mutant LEC-1 cells that lack NAGT-1 were used first because in these cells oligosaccharide modifications stop after Mann I processing (Stanley et al., 1975).VSV-G protein is endo-D resistant while in the ER and becomes endo-D sensitive after being transported and processed by Mann I.When infected cells were permeabilized and used in the semi-intact cell transport assay in the presence of an NRK cells were infected with VSVtsO45 for 3 h at 42ЊC.Cells were permeabilized and supplemented with transport cocktails containing p115-depleted cytosol (A-C) or complete cytosol (D-F).After transport at 32ЊC for 90 min, cells were processed by double label immunofluorescence using anti-VSV-G protein (A and D) and anti-Mann II (B and E) antibodies.Depletion of p115 from the transport assay had no effect on VSV-G protein exit from the ER, but prevented VSV-G protein transport to the Golgi (A-C) and caused accumulation of VSV-G protein in peripheral VTCs lacking Mann II.In the presence of complete cytosol, VSV-G protein was efficiently delivered to the Golgi (D-F).Bar 10 m.ATP-depleting system, VSV-G protein was endo-D resistant (Fig. 9 A, ϪATP).When complete transport cocktail was used, a proportion )%06ف( of VSV-G protein became endo-D sensitive, and this is taken as the standard to which other reactions are compared (Fig. 9 A, ϩATP).Unexpectedly, VSV-G protein was processed to the endo-D-sensitive form when anti-p115 antibodies were added to the transport assay (Fig. 9 A, anti-p115) in an amount analogous to that found to block acquisition of endo-H resistance in NRK cells (Fig. 5 A, lane 6).Addition of control antibodies had no effect on VSV-G protein processing (Fig. 9 A, control).Similarly, when p115-depleted cytosol was used in the transport assay, VSV-G protein was processed to the endo-D-sensitive form (Fig. 9 B, gel and bar graph, lanes 5 and 6).Even a significant depletion of p115 (Ͼ80% of p115 was depleted in lanes 5 and 6 as compared with lane 2) had no effect on the amount of VSV-G protein sensitive to endo-D.The level of p115 %02ف( of control) that supported the processing to the endo-D-sensitive form in LEC-1 cells was unable to support processing to the endo-H-resistant form in NRK cells (compare Fig. 9 B, lane 6, to Fig. 5 C, lane 4).The level of endo-D processing in reactions containing cytosol immunodepleted with anti-p115 antibodies was similar to that when control (Fig. 9 B, lane 3) or preimmune (Fig. 9 B, lane 4) antibodies were used for immunodepletion.
To test VSV-G protein processing by Mann I while in pre-Golgi transport intermediates, we arrested VSV-G protein transport by incubating the LEC-1 cells at 15ЊC (Rowe et al., 1998).As shown in Fig. 9 C, VSV-G protein is resistant to endo-D when cells are incubated at 42ЊC (lanes 1 and 4) and becomes endo-D sensitive when incubated at 32ЊC (lanes 2 and 5).When cells were incubated at 15ЊC, %07ف of VSV-G protein was endo-D sensitive (lanes 3 and 6), suggesting that the VSV-G protein was processed by Mann I.
Analogous experiments were performed in NRK cells, but both endo-H and endo-D resistance/sensitivity of VSV-G protein arrested by various treatments were tested.These digestions were done in parallel since only endo-D sensitivity of an endo-H-sensitive form of VSV-G will define whether VSV-G protein is processed by Mann I (A) VSV-G protein is endo-D resistant when transport cocktail lacking ATP is used (ϪATP), and this is set as 0% relative processing.A proportion of VSV-G protein is endo-D sensitive when complete transport cocktail is used (ϩATP), and this is set as 100% relative processing.A similar proportion of VSV-G protein is endo-D sensitive when complete transport cocktails containing control antibodies (control) or anti-p115 antibodies (anti-p115) are used.(B) VSV-G protein is endo-D resistant when transport cocktail lacking ATP is used (lane 1), and this is set as 0% relative processing.A proportion of VSV-G protein is endo-D sensitive when complete transport cocktail is used (lane 2), and this is set as 100% relative processing.A similar proportion of VSV-G protein is endo-D sensitive when complete transport cocktails containing cytosol immunodepleted with control antibodies (lane 3), preimmune antibodies (lane 4), or anti-p115 antibodies (lanes 5 and 6) are used.Analogous gels (n ϭ 3) were quantitated by densitometry and averages are presented as a bar graph.An aliquot of each transport reaction was probed by immunoblotting with anti-p115 antibodies and the immunoblot is shown in panel p115.The relative amounts of p115 in each transport reaction are presented in the bar graph.The amount of p115 in lane 1 is set as 100%.(C) LEC-1 cells were infected with VSVtsO45 for 3 h at 42ЊC, and either analyzed directly or incubated at 32ЊC for additional 1 h or at 15ЊC for additional 3 h before analysis.Cells were collected and analyzed directly or after endo-D digestion.Processing of VSV-G protein from the endo-D-resistant (R) to the endo-D-sensitive (S) form is shown.A single VSV-G band is seen in untreated samples regardless of incubation temperature (lanes 4-6).VSV-G is resistant to endo-D when retained in the ER during the 42ЊC incubation (lane 1), but becomes endo-D sensitive when transported to the Golgi during a 32ЊC incubation (lane 2) or when arrested in peripheral VTCs during a 15ЊC incubation (lane 3).(D, lanes 1-3 and 5-7) ER to Golgi transport was performed in semi-intact NRK cells.After transport, the sensitivity/resistance of VSV-G protein to endo-H and endo-D was analyzed in parallel.VSV-G protein is endo-H sensitive (lane 1) and endo-D resistant (lane 5) when transport cocktail lacking ATP is used.VSV-G protein is endo-H resistant (lane 2) and endo-D resistant (lane 6) when complete transport cocktail is used.VSV-G protein is endo-H sensitive (lane 3) and endo-D resistant (lane 7) when complete transport cocktails containing anti-p115 antibodies are used.(D, lanes 4 and 8) NRK cells were infected with VSVtsO45 (3 h at 32ЊC), pulsed-labeled at the restrictive temperature (10 min at 42ЊC) and chased in complete medium for 3 h at 15ЊC.Cells were collected and digested with either endo-H (lane 4) or endo-D (lane 8).VSV-G protein is endo-H sensitive and endo-D resistant when arrested in peripheral VTCs at 15ЊC.but not NAGT-1 and Mann II.As shown in Fig. 9 D, when transport was performed in the presence of an ATPdepleted transport cocktail, VSV-G protein is endo-H sensitive (lane 1) and endo-D resistant (lane 5).When transport was performed in the presence of a complete transport cocktail, VSV-G becomes endo-H resistant (lane 2) and is endo-D resistant (lane 6).When affinity-purified anti-p115 antibodies were added to the complete transport assay, VSV-G protein is endo-H sensitive (lane 3) and endo-D resistant (lane 7).These results differ from the results of analogous experiments performed in LEC-1 cells (Fig. 9 A), in which addition of affinity-purified anti-p115 antibodies leads to VSV-G protein that is endo-D sensitive.
In an attempt to clarify this difference, we analyzed the endo-H and endo-D sensitivity of VSV-G protein arrested in peripheral VTCs when NRK cells are incubated at 15ЊC.As shown in Fig. 9 D, lanes 4 and 8, VSV-G protein was endo-H sensitive and endo-D resistant.This result differs from that in LEC-1 cells, where an endo-D-sensitive form was observed after 15ЊC incubation.Together, the data show that endo-D-sensitive VSV-G protein is seen in LEC-1 cells in the presence of anti-p115 antibodies or when cells are incubated at 15ЊC, whereas the same treatments of NRK cells result in a VSV-G protein that is endo-D resistant.A likely explanation for this difference is that the lack of p115 requirement in LEC-1 cells is due to incomplete inactivation/removal of endogenous p115 from those cells as opposed to NRK cells.Alternatively, since we observe a difference in VSV-G protein endo-D sensitivity in intact LEC-1 and NRK incubated at 15ЊC, oligosaccharide processing of VSV-G protein might be different in the two cell lines.Currently, we have no explanation for the different results in LEC-1 and NRK cells, and cannot define whether p115 is required for transport of VSV-G protein into a Mann I-containing compartment.
Mann I Redistributes from the Golgi When ER to Golgi Transport Is Blocked by 15ЊC Incubation or p115 Depletion
The differences in the biochemical results in LEC-1 and NRK cells prompted us to examine Mann I localization in traffic-arrested NRK cells since all previous morphological analyses (Figs. 4,7,and 8) have been done in NRK cells.Previous EM immunoperoxidase studies indicated that in NRK cells, Mann II is localized in the medial-cisternae, whereas Mann I is localized in the trans-cisternae and the TGN, and at the light immunofluorescence level, both Mann II and Mann I are concentrated in the Golgi region (Velasco et al., 1993).In agreement, Mann I and II colocalized in the Golgi region in cells incubated at 37ЊC, although Mann I was also present in a more diffuse non-Golgi pattern (Fig. 10, A-C).Significantly, Mann I redistributed from the Golgi to peripheral punctate structures when cells were incubated at 15ЊC, whereas Mann II did not show such redistribution (Fig. 10, D-F).To characterize these pre-Golgi elements, VSVtsO45-infected cells were shifted from 40 to 15ЊC for 3 h, and then processed to localize Mann I and VSV-G protein.As shown in Fig. 10, G-I, Mann I redistributed from the Golgi to peripheral structures, some of which (arrowheads, I) contained VSV-G protein.In contrast, Mann II did not show redistribution (Fig. 10, J-L).
To determine whether Mann I can redistribute to pre-Golgi VTCs arrested by p115 depletion, semi-intact cells incubated with transport assays supplemented with complete cytosol or with p115-depleted cytosol were processed for double label immunofluorescence to localize VSV-G protein and Mann I.As shown in Fig. 10, M-O, when transport was performed with complete cytosol, Mann I was detected in a typical perinuclear Golgi structure, where it colocalized with VSV-G protein.In contrast, when transport was performed with p115-depleted cytosol, Mann I was found colocalizing with peripheral VTCs (Fig. 10, P-R, arrowheads in R point to a VTC-containing VSV-G protein and Mann I).As already shown in Fig. 8, Mann II did not show such relocation.Together, these results indicate that Mann I can cycle to pre-Golgi compartments during cargo transit from the ER to the Golgi.Furthermore, our data show that p115 is not required for events before and including the recycling of Mann I from the Golgi to peripheral compartments, but is essential for the subsequent events leading to delivery of cargo and Mann I to the Golgi.
Requirement for p115 in Maintaining Golgi Structure
Functional p115 appears to be required for normal Golgi morphology in intact cells since injection of anti-p115 antibodies into WIF-B cells resulted in the appearance of Golgi fragments scattered throughout the injected cells.This antibody effect appears specific since microinjection of anti-5Ј-nucleotidase antibodies does not lead to changes in Golgi morphology.Normal Golgi architecture varies in different cells, but seems to be dependent on the balance of incoming and outgoing membrane traffic through the structure.Perturbation of the anterograde or retrograde traffic by various treatments of cells leads to Golgi disruption.Specifically, changes in Golgi morphology were observed in cells overexpressing ADP-ribosylation factor (Dascher and Balch, 1994), injected with rab1a mutant (Wilson et al., 1994), or anti--COP antibodies (Pepperkok et al., 1993).Similarly, treatment of cells with BFA (Misumi et al., 1986;Fujiwara et al., 1988;Lippincott-Schwartz et al., 1989), nocodazole (Storrie and Yang, 1998), or okadaic acid (Pryde et al., 1998) has a dramatic effect on Golgi integrity.The exact mechanism of the disruptive action of anti-p115 antibodies on Golgi structure is currently unknown, but it is possible that the antibodies perturb transport pathways between the ER and the Golgi.Specifically, the antibodies could inhibit p115 activity in vivo by blocking p115 interactions with GM130 (Nakamura et al., 1997;Nelson et al., 1998), giantin (Sönnischen et al., 1998), or another, as yet unidentified, protein.Alternatively, anti-p115 antibodies might act by trapping p115 on the membranes through the formation of inactive complexes.
Presence of anti-p115 antibodies in cells did not prevent the redistribution of Mann II to the ER after BFA treatment, but the anti-p115 antibodies prevented the reassem-bly of Golgi complexes during subsequent BFA wash-out.The requirement for functional p115 manifested late, after exit of Golgi enzymes from the ER and delivery to peripherally scattered punctate structures.Significantly, these structures did not move towards the MTOC and did not coalesce into a centrally located Golgi complex.The effect of anti-p115 antibodies was similar to that of anti-EAGE (anti--COP) antibodies, which also interfere with the proper reassembly of compact Golgi complexes in a juxtanuclear region during a BFA wash-out (Scheel et al., 1997).Although our findings place the functional requirement for p115 in ER to Golgi traffic between the VTCs and the Golgi, it remains to be determined which step of the anterograde or retrograde transport is blocked by anti-p115 antibodies in vivo.
Requirement for p115 in ER to Golgi Transport
The previous finding that p115 is abundant on VTCs (Nelson et al., 1998) was extended in this study to show that Figure 10.Mann I relocates from the Golgi to arrested pre-Golgi VTCs.NRK cells grown at 37ЊC (A-C) or incubated for 3 h at 15ЊC (D-F) were processed for double label immunofluorescence using anti-Mann I (A and D) and anti-Mann II (B and E) antibodies.In cells grown at 37ЊC, Mann I colocalized with Mann II in the Golgi region (C), but after 15ЊC incubation, Mann I relocated to peripheral punctate structures and did not colocalize with the Golgi localized Mann II (F).NRK cells infected with VSVtsO45 were incubated for 2 h at 42ЊC and for an additional 3 h at 15ЊC (G-L).Cells were processed for double label immunofluorescence using anti-Mann I (G) and anti-VSV-G protein (I) antibodies, or anti-Mann II (J) and anti-VSV-G protein antibodies (K).Mann I is present in dispersed punctate structures, some of which contain VSV-G protein (I, arrowheads).Mann II remains within the Golgi and does not relocate to peripheral structures containing VSV-G protein (L).NRK cells infected with VSVtsO45 were incubated for 3 h at 42ЊC, permeabilized, and supplemented with transport cocktails containing complete cytosol (M-O) or p115-depleted cytosol (P-R).After transport at 32ЊC for 90 min, cells were processed for double label immunofluorescence using anti-Mann I (M and P) and anti-VSV-G protein (N and Q) antibodies.In reactions containing complete cytosol, VSV-G protein is delivered to the Golgi where it colocalizes with Mann I (O).In reactions containing p115-depleted cytosol, Mann I relocates to pre-Golgi VTCs containing arrested VSV-G protein (R, arrowheads).Bar, 10 m.
p115-containing VTCs are functional and carry cargo VSV-G protein from the ER to the Golgi.The possible involvement of p115 in this part of the secretory pathway was examined using previously established ER to Golgi transport assays that measure biochemically or morphologically the movement of VSV-G protein from the ER to the Golgi in semi-intact cells.Our biochemical results show that anti-p115 antibodies or p115 depletion specifically block VSV-G protein transport before acquisition of endo-H resistance, believed to represent the delivery of VSV-G protein to the cis/medial-cisternae of the Golgi stack.Unexpectedly, morphological examination showed that in the presence of anti-p115 antibodies or p115 depletion, VSV-G protein was not delivered to the Golgi stack, but instead was arrested in pre-Golgi peripheral VTCs.These findings indicate that p115 is not required for VSV-G proteins to efficiently exit the ER and be delivered to morphologically normal peripheral VTCs, but is essential for the subsequent delivery of such VTCs to the Golgi stack.
What is the role of p115 in ER to Golgi traffic?Because p115 is predominantly associated with VTCs and p115 depletion arrests transport in VTCs, we favor the hypothesis that p115 functions while on VTCs.A possible role for p115 could be to mediate delivery of essential components from the Golgi to VTCs, by tethering Golgi-derived recycling vesicles to VTCs before fusion.This hypothesis is suggested by the finding that a subpopulation of Golgiderived COP I vesicles (albeit produced under the nonphysiological conditions of GTP␥S treatment) is enriched in giantin, a protein known to bind p115 (Sönnischen et al., 1998).Giantin is an extended coiled-coil transmembrane protein and the interaction between p115 on VTCs and giantin on COP I vesicles could theoretically span the distance of Ͼ400 nm, raising the possibility that p115 on VTCs might act to catch and tether giantin-containing Golgi-derived COP I vesicles.Fusion of such vesicles would deliver components essential for VTC maturation and subsequent delivery to the Golgi stack.
Alternatively, p115 might function to mediate VTC motility on microtubules.This is suggested by the fact that the p115-induced block in ER to Golgi traffic is biochemically and morphologically similar to that caused by 15ЊC incubation, when dynein/dynactin-mediated movement of VTCs on microtubules is inhibited (Presley et al., 1997).Furthermore, VSV-G protein is arrested in peripheral VTCs in cells overexpressing the dynamitin subunit of the dynactin complex (Presley et al., 1997), and their pattern is indistinguishable from VTCs arrested by anti-p115 antibody addition or p115 depletion.Overexpression of dynamitin disrupts the dynactin complex, separating the Arp-1 filament from the dynein-binding arm of p150 Glued (Echeverri et al., 1996;Burkhardt et al., 1997) and prevents microtubule-mediated motility.Interestingly, a link between microtubule movement and rab proteins has been established by the finding that the Golgi-localized rab6 interacts with rabkinesin, a Golgi-associated kinesin-like motor protein that binds to microtubules and is likely to mediate membrane movement along microtubules (Echard et al., 1998).Whether rab1 might also have motor proteins (perhaps of the dynein/dynactin family) as one of its effectors remains to be determined.However, it is interesting that p115 structure is similar to that of kinesin (a dimer with globular heads, a central coiled-coil tail, and a tail domain), although p115 does not contain an ATP-binding/ hydrolysis domain and does not interact with microtubules (data not shown).
It is equally possible that p115 is required to recruit or activate other uncharacterized factors that will ultimately allow VTC transport and fusion.We do not consider it likely that p115 function in VTC dynamics involves interaction with GM130, since GM130 is not detected on VTCs (Nelson et al., 1998) or Golgi-derived COP I vesicles (Sönnischen et al., 1998), and appears restricted to the Golgi stack.It is more probable that p115-GM130 interactions occur after the VTCs are delivered to the Golgi stack.
Functional p115 was shown to be required in ER to Golgi traffic after the rab1-dependent step of transport.Rab1 has been shown previously to be required after COP II vesicle budding and after the assembly of peripheral VTCs (Peter et al., 1998).In semi-intact cells supplemented with the rab1a (N124I) mutant that is defective for guanine nucleotide binding, VSV-G protein has been shown at the ultrastructural level to accumulate in tubular structures with associated coated vesicular profiles (Pind et al., 1994), but the morphology of the p115-arrested transport intermediates remains to be analyzed.The behavior of p115 and rab1 in mammalian cells is analogous to that of Uso1p and Ypt1 in yeast, since Ypt1p appears to act upstream of Uso1p, and might be indirectly involved in recruiting Uso1p to the membranes (Cao et al., 1998).Whether rab1 performs a similar function for p115 is uncertain since significant differences in the dynamics of transport intermediates (e.g., yeast do not appear to have extensive VTCs) are evident.
p115 was found to act before the Ca 2ϩ -requiring step, which is considered to be the last stage of transport before fusion (Aridor et al., 1995).SNARE pairing precedes Ca 2ϩ requirement (Chen et al., 1999), and our data are consistent with the reported requirement for Uso1p before SNARE complex formation (Sapperstein et al., 1996;Barlowe, 1997;Cao et al., 1998).
Cycling of Golgi Enzymes to Pre-Golgi Compartments
This report documents the first instance of a morphologically observed relocation of Mann I from the Golgi stack to peripheral pre-Golgi structures.The relocation is apparent in cells in which ER to Golgi transport is blocked by 15ЊC incubation or p115 depletion, suggesting that Mann I (like other Golgi proteins) cycles between the Golgi and the ER.The mechanism of Golgi protein recycling is largely undefined.Recycling might occur directly from the Golgi to the ER, since direct delivery via tubular structures is observed during BFA-induced relocation of Golgi proteins, and at least one Golgi glycosylating enzyme, GalT, can be detected in similar tubules under physiological conditions (Sciaky et al., 1997).The BFAinduced tubular relocation of Golgi proteins is COP I-independent (Donaldson et al., 1991).Alternatively, recycling might occur from the Golgi to VTCs.This pathway would be consistent with the biochemical data showing that VSV-G protein arrested in pre-Golgi transport intermediates can be processed by NAGT-1, which recycles via small vesicles from the Golgi to the pre-Golgi intermedi-ates (Love et al., 1998).The recent results showing that NAGT-1 binds coatomer in vitro (Dominguez et al., 1998), and NAGT-1, as well Mann II and Gal-T, bud from the Golgi in COP I-coated vesicles (Lanoix et al., 1999) suggest that COP I-coated vesicles might transport Golgiderived glycosyl transferases to pre-Golgi transport intermediates.Based on the structural similarity of Mann I to other glycosyltransferases (a type II transmembrane protein), it is possible that it also recycles via Golgi-derived COP I-coated vesicles.
Previous work has defined the kinetics with which Golgi resident proteins cycle between the Golgi and the ER.Several Golgi proteins such as KDEL-R, TGN38 (Cole et al., 1998), and gp27 (Fullekrug et al., 1999), and glycosylating enzymes such as N-acetylgalactosaminyltransferase-2 (GalNAc-T2) and GalT (Storrie et al., 1998) have been shown to cycle with half lives between 20 min (e.g., the KDEL-R; Cole et al., 1998) and2-3 h (e.g., gp27 [Fullekrug et al., 1999]; GalNAc-T2 and GalT [Storrie et al., 1998]).Based on our observation that the majority of Mann I relocates from the Golgi and appears in blocked VTCs during a 90-min incubation at 37ЊC (the time course of the semi-intact transport assay), it appears that Mann I cycles with relatively fast kinetics.Although the semiintact cell system used in our studies is not designed to analyze recycling pathways, the results showing relocation of Mann I, but not of Mann II to peripheral VTCs suggests that Mann II cycles with significantly slower kinetics.
The recycling kinetics of Mann I seem more similar to those of the ERGIC proteins, KDEL-R and ERGIC53, which cycle rapidly at 37ЊC.These proteins also relocate to peripheral VTCs during longer (usually 3ف h) incubations at 15ЊC.However, Mann I does not appear to be a bonafide component of the ERGIC since EM analysis showed it to be localized throughout the Golgi stack, with highest concentration in the trans-cisternae and the TGN, rather than the cis-face of the stack (Velasco et al., 1993).This suggests that rapid cycling is not restricted to residents of the ERGIC, and that proteins with steady state distribution in more distal Golgi compartments including the TGN, also contain signals for rapid cycling.Whether the recycling mechanisms are the same and involve COP I vesicles, or are distinct remains to be elucidated.
Our morphological results have direct bearing on the controversy regarding the mode of cargo transport through the Golgi.The finding that Mann I efficiently relocates from the Golgi stack to colocalize with peripheral cargocontaining VTCs is less consistent with the anterograde vesicular traffic model than with the maturation model for exocytic traffic.Our results are not consistent with the anterograde vesicular traffic model, in which glycosyl-transferases remain relatively stationary in distinct Golgi cisternae while the cargo is shuttled from proximal to distal cisterna in small vesicles.The observed relocation supports the maturation model, in which cargo progresses through the secretory pathway in a compartment that is remodeled by the sequential recycling of processing enzymes from later Golgi compartments.
Figure 1 .
Figure 1.Antibodies against p115 cause Golgi complex disassembly.mAbs against p115 mixed with TR-dextran (A-C), or mAbs against 5Ј-nucleotidase mixed with TR-dextran (D-F) were microinjected into the cytoplasm of WIF-B cells.Phase images of cells are shown (A and D).Injected cells are identified by TR-dextran (B and E) and are traced in outline in C and F. Cells were fixed 2 h after injection and processed for immunofluorescence using antibodies against Mann II (C and F).Cells injected with anti-p115 antibodies, but not uninjected cells or cells injected with control antibodies show disassembly of Golgi complexes.Bar, 10 m.
Figure 2 .
Figure 2. Antibodies against p115 do not block BFA-induced retrograde Golgi to ER traffic.Mouse mAbs against p115 were microinjected into the cytoplasm of WIF-B cells.Cells were fixed immediately after injection (A and B) or after a 30-min BFA treatment (C and D).Cell were processed for double label immunofluorescence using antibodies against Mann II (A and C) or against mouse IgG (B and D).Relocation of Mann II from the Golgi to the ER appears indistinguishable in injected and uninjected cells.Golgi remnants were detected in some cells after the 30-min BFA treatment (C, arrowheads).
Figure 4 .
Figure 4. p115 is associated with VTCs moving cargo VSV-G protein from the ER to the Golgi.NRK cells were infected with VSVtsO45 at 42ЊC for 3 h.The cells were either fixed (A-C) or incubated at 15ЊC for 3 h (D-F) or at 32ЊC for 1 h (G-I) before fixation.Cell were processed for double label immunofluorescence using antibodies against p115 (A, D, and G) and antibodies against VSV-G protein (B, E, and H).At 42ЊC, VSV-G protein is present in the ER (B), whereas p115 is predominantly localized to the Golgi (A).p115 and VSV-G protein colocalize in peripheral VTCs after 15ЊC incubation (D-F), and in the Golgi after 32ЊC incubation (G-I).Bar 10 m.
Figure 8 .
Figure8.p115 depletion blocks ER to Golgi transport of VSV-G protein at a pre-Golgi stage.NRK cells were infected with VSVtsO45 for 3 h at 42ЊC.Cells were permeabilized and supplemented with transport cocktails containing p115-depleted cytosol (A-C) or complete cytosol (D-F).After transport at 32ЊC for 90 min, cells were processed by double label immunofluorescence using anti-VSV-G protein (A and D) and anti-Mann II (B and E) antibodies.Depletion of p115 from the transport assay had no effect on VSV-G protein exit from the ER, but prevented VSV-G protein transport to the Golgi (A-C) and caused accumulation of VSV-G protein in peripheral VTCs lacking Mann II.In the presence of complete cytosol, VSV-G protein was efficiently delivered to the Golgi (D-F).Bar 10 m.
Figure 9 .
Figure 9. VSV-G protein is differentially endo-D resistant/sensitive in LEC-1 and NRK cells.(A and B) ER to Golgi transport was performed in semi-intact LEC-1 cells.After transport, the sensitivity/resistance of VSV-G protein to endo-D was analyzed.(A)VSV-G protein is endo-D resistant when transport cocktail lacking ATP is used (ϪATP), and this is set as 0% relative processing.A proportion of VSV-G protein is endo-D sensitive when complete transport cocktail is used (ϩATP), and this is set as 100% relative processing.A similar proportion of VSV-G protein is endo-D sensitive when complete transport cocktails containing control antibodies (control) or anti-p115 antibodies (anti-p115) are used.(B) VSV-G protein is endo-D resistant when transport cocktail lacking ATP is used (lane 1), and this is set as 0% relative processing.A proportion of VSV-G protein is endo-D sensitive when complete transport cocktail is used (lane 2), and this is set as 100% relative processing.A similar proportion of VSV-G protein is endo-D sensitive when complete transport cocktails containing cytosol immunodepleted with control antibodies (lane 3), preimmune antibodies (lane 4), or anti-p115 antibodies (lanes 5 and 6) are used.Analogous gels (n ϭ 3) were quantitated by densitometry and averages are presented as a bar graph.An aliquot of each transport reaction was probed by immunoblotting with anti-p115 antibodies and the immunoblot is shown in panel p115.The relative amounts of p115 in each transport reaction are presented in the bar graph.The amount of p115 in lane 1 is set as 100%.(C) LEC-1 cells were infected with VSVtsO45 for 3 h at 42ЊC, and either analyzed directly or incubated at 32ЊC for additional 1 h or at 15ЊC for additional 3 h before analysis.Cells were collected and analyzed directly or after endo-D digestion.Processing of VSV-G protein from the endo-D-resistant (R) to the endo-D-sensitive (S) form is shown.A single VSV-G band is seen in untreated samples regardless of incubation temperature (lanes 4-6).VSV-G is resistant to endo-D | 2014-10-01T00:00:00.000Z | 1999-12-13T00:00:00.000 | {
"year": 1999,
"sha1": "1b7f21b563997b329f63eb6402c9d1a56a75cafe",
"oa_license": "CCBYNCSA",
"oa_url": "http://jcb.rupress.org/content/147/6/1205.full.pdf",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "1b7f21b563997b329f63eb6402c9d1a56a75cafe",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology"
]
} |
18898253 | pes2o/s2orc | v3-fos-license | Global Patterns in the Implementation of Payments for Environmental Services
Assessing global tendencies and impacts of conditional payments for environmental services (PES) programs is challenging because of their heterogeneity, and scarcity of comparative studies. This meta-study systematizes 55 PES schemes worldwide in a quantitative database. Using categorical principal component analysis to highlight clustering patterns, we reconfirm frequently hypothesized differences between public and private PES schemes, but also identify diverging patterns between commercial and non-commercial private PES vis-à-vis their service focus, area size, and market orientation. When do these PES schemes likely achieve significant environmental additionality? Using binary logistical regression, we find additionality to be positively influenced by three theoretically recommended PES ‘best design’ features: spatial targeting, payment differentiation, and strong conditionality, alongside some contextual controls (activity paid for and implementation time elapsed). Our results thus stress the preeminence of customized design over operational characteristics when assessing what determines the outcomes of PES implementation.
Introduction
Payments for environmental services (PES) have become an increasingly popular tool for environmental management, supplementing policy tools that were previously widely focused on command-and-control measures. Dozens of programs to reward provision of environmental services (ES) are currently being implemented at multiple geographical scales around the world See e.g. www.watershedconnect.com, www.ecosystemarketplace.com and http://www.oas.org/ dsd/PES/Database.htm#). PES feature direct, conditional contracts to achieve a negotiated environmental outcome between a provider and a user of environmental services [1]. The underlying rationale is that if the service user's gain from pro-environmental action is sufficiently large, there may be a good case for compensating the service-providing landowners for choosing a profit-wise second-best, but environmentally more benign resource use. This type of "Coasean deal" implies using a positive economic incentive to enhance a positive (or avoid a look at the interplay between PES implementation characteristics, contextual and institutional factors (including the role of the public and private sectors), and evidence of environmental additionality. There has been extensive debate whether PES should be broadly defined, or follow a narrower, theoretically-based definition [2,5,8,10]. In the following, we choose a narrower delimitation, focused on conditionality as sine qua non PES criterion [4]. While this choice evidently caps our sample size, it allows us to look at interventions that are truly comparable. We refer to schemes fitting all five criteria as "canonical PES", and include in our sample only schemes that deviate moderately from this ideal setup (see below). We also only include cases with one or more pre-existing academic assessments in the literature, aiming thus at assessing only PES schemes with scientifically validated sources of information.
By adopting a globally scoped systematic literature review, including all terrestrial environmental services, we reach a sample of 55 PES schemes. Compared to [34] and [35], we extend the set of design-oriented (e.g. type of payments, use of baselines, monitoring) and context-specific variables (e.g. size, region, services transacted, actors involved) in our statistical analysis. Compared to [36], our twice as large sample enhances the statistical analysis, and treats new research questions. We thus believe our sample is as broad as the current state of PES affairs permits, and yet generically similar to be able to compare initiatives.
This article is structured as follows. In Section 2, we describe in detail our sample and methods. In Section 3 (Results), we follow a sequential approach, going from open-ended exploratory methods to an analysis with an assumed causal direction. We start with (a) descriptive statistics, then turn to (b) a categorical principal component analysis of PES variables, and lastly (c) analyse econometrically how PES design likely affects environmental additionality. Finally, we discuss the scope of our findings, and their implications for future research and policies (Section 4).
Sample and Methods
While hundreds of PES schemes are reported loosely upon in the literature, most contributions do not provide sufficient in-depth information to be useful for quantitative meta-analysis. We selected our case studies based on different methodological guidelines for meta-analysis, derived from clinical and social sciences [37][38][39]. We carried out a systematic literature research drawing on various pre-existing PES databases, qualitative PES reviews, journal storage (JSTOR) and other web-based (Google Scholar) sources (Table 1). From the resulting identified records, we filtered out cases where payments to ES providers (i) could not be confirmed to have occurred at least once (never mind whether the program was still ongoing); (ii) did not conform well to our conditionality-focused definition for PES; or (iii) did not in the PES case study deliver sufficient descriptors to meet our minimum data standards.
This narrowed our sample to 90 literature references referring to a total of 55 PES cases worldwide (counting until mid-2014 when our statistical analysis was begun), of which 47 are ongoing (See S1 Table for meta-analysis references and Fig 1 for the PRISMA selection diagram). Fig 2 shows the geographic and thematic focus of the selected PES cases. They feature payments for the protection and restoration of watersheds (Water) (with 22 cases, the most frequently targeted ES), for biodiversity conservation (Biodiversity) (10 cases), for climate change mitigation through carbon sequestration or avoided deforestation (Carbon) (8 cases), and for multiple services from agriculturally dominated systems (Multiple-Agriculture) (12 cases). Region-wise, countries with emerging economies dominate, especially Latin America with 23 cases. Industrialised countries have less PES cases, though some are huge government schemes that outsize small-scale initiatives by orders of magnitude. We did not weight our observation in terms of scheme size, but treated each case equally. Hence, our analysis comes to put relatively more emphasis on a series of smaller-scale PES schemes that predominate in non-OECD countries.
We built a database of in total 50 basic variables, a dozen of which we employ analytically. The main categories are: 1. implementation modes (broad PES scheme descriptors, land-ES link, spatial extent); 2. program design (monitoring, sanctioning, baselines, ES targeting, differentiation of payments); 3. program institutional arrangements (ES buyer and provider types, source of payments, presence of intermediaries, transaction setting); 4. program funding (degree of user-financing, length of the contract, type and level of payment); 5. scheme adherence to our PES "canonical" standards (indexed degree of fit with PES theory); and 6. evidence of environmental additionality (unit measured, method, level).
Generally, we combined categorical (no hierarchy between levels), ordinal (hierarchical between levels), and continuous variables (cardinal ranking) in our analysis. S1 Database describes all variables in detail, but a few points are worth flagging. We registered which actors participated in different implementation stages, and classified them between public, private commercial (enterprises) and private non-commercial (NGOs, foundations, grassroots organizations) sectors (iii). Under funding sources (iv), we describe which sectors assumed which payments (cash and in-kind). Costs were divided into upfront (e.g. information, design, capacity building and negotiation costs) versus recurrent management costs (e.g. administrative, implementation, monitoring costs), and payments proper. Schemes totally or mainly funded by the public sector were classified as public-sector schemes; correspondingly for the two other sectors.
Fitness to a "canonical PES" (v) is a composite indicator derived from the sum of the criteria that refer to the compliance or not with the five PES definitional criteria mentioned above, plus the degree of transfer directness (adapted from [8]). This composite indicator allows us to capture the degree of implementation closeness to PES theory, using an ordinal indicator with scores ranging from 6 to 17 (S2 Table). The variable "economic setting" differentiates between degrees of competition on both provider and user sides (e.g. monopsony, oligopsony, club, market) [35]. Descriptive statistics of the main variables are shown in Table 2.
Finally, environmental additionality (vi) was clearly the most challenging variable, but also of great interest [40]. One previous meta-analysis used expert scores [35], another implementer self-assessment [34] of the degree of environmental success. Potentially we could have focused exclusively on studies using so-called rigorous impact assessment methods. Due to the scarcity of those so far in assessing PES [41], a recent systematic PES literature review applying that filter ended up with just 11 studies, six of which alone on Costa Rica's PSA program [42]. Paradoxically, even among those six "rigorous" studies on Costa Rica, non-trivial differences exist in their bottom-line additionality, including because of variability in the methods used (e.g. matching techniques and controlling for sources of impact heterogeneity). Even when researchers have not yet produced the perfect data sets for quantitative impact evaluation, practitioners and policy makers still need to make real-world choices about how to design their interventions, taking stock of the current state of affairs in the best possible way. In this study, we have thus opted for including both quantitative and qualitative documentation on environmental additionality, and classified it according to the precision of the assessment methods, resulting in five levels (S3 and S4 Tables). Consequently, we opted for the least pretentious outcome assessment: a binary variable to differentiate between, on the one hand, likely zero or low and on the other, likely considerable additionality in ES delivery and/or a targeted land-use proxy, such as forest cover. In addition, we performed various sensitivity analyses on our additionality assessments. From the total of 55 PES cases, for four simply no attempt had been made to gauge environmental additionality, thus restraining the number of cases for the additionality analysis to 51.
Descriptive statistics
Our first interest is whether, as hypothesized in the PES literature, substantial differences exist between the characteristics of public vs. privately funded PES schemes [2,17]. Geographically, sector funding exhibits three patterns: In Latin America, about one tenth of cases are private non-commercial initiatives (9%), one fourth is private commercial, while two thirds (65%) are publicly funded cases. Europe, North America and Asia present a slightly higher frequency of publicly funded PES (70%). In Africa, privately funded schemes clearly predominate (85%). Half of these private schemes are run by the private commercial sector, especially for eco-tourism and wildlife. Do sectors alternate their lead of PES schemes over time? For one fourth of our cases, this is the case. The sector initiating the PES scheme (carrying upfront costs) also continues covering the recurrent management costs in 73% of cases. Where shifts occur, most frequently the public sector covers the initial costs, transferring the project afterwards to the private commercial sector to manage it. For instance, the Scolel Té project in Southern Mexico had initial funding from the UK Department for International Development (DFID) in partnership with the Mexican university El Colegio de la Frontera Sur (Ecosur) and the University of Edinburgh to kick-start this forest carbon scheme, while currently it is managed by the private company Ambio [43]. Such sequential transfer of leadership between private and public environment sectors has also been observed in other sectors [22].
How do per-hectare payments vary across the sample? Our oldest PES scheme, the Conservation Reserve Program (USA), started in 1985. Hence, we adjusted payments for inflation, converting to real US$ from 2003 -the median starting year of assessed PES cases (Source: http://www.oecd-ilibrary.org/statistics). Mean payments and size of PES schemes are both significantly different across ES targeted (ANOVA F = 9.5 p = 0.00 for payments, ANOVA F = 3.6 p = 0.02, for area). Watershed PES schemes are the smallest in size (Fig 3A), but record the highest annual per-hectare payments (3B), compared to other ES. Carbon, biodiversity and multi-functional agriculture PES schemes score similar in size, although carbon PES show large variability. Multifunctional agricultural PES schemes vary the least in payment amounts, and account for large areas. The lowest payments are for biodiversity PES, but with large variance. Unsurprisingly, public schemes are on average larger in size than private ones, especially vis-à-vis non-commercials, but also pay more per hectare (3C, 3D). As expected, payments vary much less in public than in private schemes (see S5 Table for details) Variable patterns and clusters PES schemes typically exhibit differences across services and implementing sectors (public vs. private), but do characteristics cluster in more systematic ways, as hypothesized in the PES literature [4,20]? In this subsection, we use multi-dimensional categorical principal component (CatPCA) and cluster analyses to explore underlying patterns. The analyses were conducted in a matrix of 55 PES (cases) and eight relevant categorical variables (attributes), including those synthetizing PES categories as defined by [35], and the composite index describing the "canonical PES" fitness. CatPCA reduces the multidimensional space represented by this matrix into a given number (usually two or three) of orthogonal (independent) dimensions. Each dimension is defined through a combination of the variables, and represents in decreasing order a given percentage of the total variability in the matrix. The data in the matrix have a strong internal consistency and reliability (Cronbach's Alpha = 0.87 -S6 Table). Fig 4 shows the first two dimensions of the analysis, representing a 52.1% of the total variability of our 55 cases per 8 variables matrix.
We additionally performed a cluster analysis (distance measure: Sorensen; group linkage method: group average) resulting in three distinct groupings of cases and attributes. The drawn border lines in Fig 4 delimit the PES schemes belonging to each of the three clusters. The combined CatPCA and cluster show which schemes have closest coordinates, and thus are more similar, the degree of concentration of PES schemes within groups, and the distance between groups. Variables with extreme values in one or both of the two axes define the cluster characteristics, since each axis is a vector composed by the weights of all the variables. On the contrary, variables close to the origin of the quadrant lines are variables that are present in two or more groups, and therefore do not explain the differences between groups. We identify the following three groupings: • Group 1 (G1) has typical "agri-environmental public PES" characteristics, being defined by the association of: farmland ecosystems, multi-functional agricultural ES, public sector carrying upfront, management and payment costs, monopsony setting, low-to-medium scores for the fitness to a "canonical PES", source of payments is not ES users; • Group 2 (G2) could be called "NGO-led biodiversity PES", as it is defined by the association of: semi-arid ecosystems, biodiversity as main ES target, non-profit sector carrying upfront, management and payment costs, oligopsony economic setting, high score for the fitness to a "canonical PES"; • Group 3 (G3) we labeled "Private commercial carbon and water PES", being defined by the association of: PES schemes targeting carbon-related ES, private sector carrying upfront, management and payment costs, high to very-high scores with regard to canonical PES index, source of payments is ES users.
While the PES literature previously had recognized important differences between public and private PES schemes [9,17], our analysis here thus also points to important differences within the private group between commercial and NGO-led PES schemes, e.g. with respect to ES focus and financing sources: a threefold sectoral categorization is thus empirically more justified. Conversely, it is also interesting to observe that forest ecosystems are placed very close to the intersection of both axis, suggesting that conserving forested ecosystems through PES is being approached from a combination of agri-environmental, carbon and biodiversity schemes. c) PES design and the degree of additionality In almost any PES literature review, a key reader question will be "what schemes worked?", i.e. which combination of ES objectives, contextual preconditions, and design elements has led to successful outcomes? As explained above, in answering this question we are handicapped by an extreme scarcity of rigorous PES impact evaluations, and by a sample size that moderately limits our degrees of freedom in statistical analysis. In response, we attempt in this section to answer econometrically a more limited sub-question: What key PES design factors may predict whether a PES scheme achieves at least some environmental additionality? An emerging literature on the principles of environmentally effective PES design points to particularly three critical factors of high potential [44][45][46][47][48]: 1. Spatial targeting of contracts-vis-à-vis hot-spots of high ES intensity, and high threat (leverage of change), respectively: by using pre-identified spatial filters to give explicitly higher focus to areas of potentially high ES gains (e.g. biodiversity hotspots) and high leverage (e.g. current deforestation hotspots), the chances for making a measurable environmental difference increase; 2. Differentiated payments-vis-à-vis variable provision costs across ES providers: whenever ES providers are heterogeneous in their profit opportunities (due to different asset holdings and technologies, market access, preferences, etc.), then offering them different payments levels (or even variable contract types) leads to greater cost efficiency than paying everybody the same (per hectare, family), thus potentially stretching the scheme's environmental gains; 3. Conditionality degree-the implementer's combined efforts to monitor and sanction incompliance: PES schemes that are perceived by ES providers to be ill-monitored and -enforced will often eventually lead to widespread non-compliance, as cashing in on PES while following 'business as usual' becomes a profitable cheating strategy. The hypothesis here is that PES schemes which go serious about implementing this quintessential PES feature will also tend to perform better with respect to their environmental outcomes.
For spatial targeting (i), we distinguished three progressive levels: a) no targeting, b) either ES density or threat targeted; c) both density and threat targeted. Differentiated payments (ii) were classified binarily, according to whether or not more than one single payment level (predominantly per-hectare, but also per-household payments) was used. Conditionality (iii) we constructed as the product of indices for documented monitoring and sanctioning efforts, respectively.
In addition, we included three control variables in the estimation. First, the public vs. private commercial and non-commercial funding distinction proved its importance in the preceding analysis. Second, baselines and additionality assessments tend to differ substantially whether PES are "asset-building" (i.e. paying for added environmental value, such as tree planting) or "activity-restricting", (i.e. made for avoiding projected damage, such as reducing deforestation) [40]. Finally, the time elapsed since project start could also be stage-setting, e.g. in terms of learning-by-doing effects on environmental effectiveness being allowed to kick in as the implementation process proceeds. Table 3 shows the result of the binary logistical regression testing the predictive level of the above described critical design variables on environmental additionality for the 51 cases for which it was possible to obtain an estimation of additionality. Overall, the model has a robust fit (Cox & Snell R 2 = 0.52; Nagelkerke R 2 = 0.79) correctly predicting 97% of additionality and 83% of no additionality cases (overall predictive accuracy of 94%); the Hosmer-Lemeshow test to assess the null hypothesis that the model prediction does not fit perfectly with the observed data was not significant (H-L p = 0.55), confirming the goodness of fit of the model. Multicollinearity checks are also robust indicating no collinearity among the variables included in the model (Tolerance>0.2; VIF<3; S7 and S8 Tables; VIF: Variance Inflation Factor).
Starting with the control variables (a) in Model I, we see that it makes the expected difference for assessed additionality whether the PES scheme is asset-building versus preventively conserving nature by avoiding projected pressures. This expresses probably that it is much easier to verify the achievement of an additional outcome when an asset has been added (e.g. planting a tree) than when a projected damage has been avoided (e.g. leaving a threatened tree standing) (dummy significant at 5% level). Interestingly, the two sectoral control variables do not come out as significant, in spite of having been flagged as important determinants of clusters in the above. We attribute this to the inclusion of design variables: much of what we see as sector differences may hide different PES design principles. Finally, the implementation time variable is significant (5% level), but with an unexpected negative sign, indicating that older schemes tend to be less additional than more recent ones. Since most of the schemes are still ongoing, implementation time thus widely comes to reflect the calendar year when the scheme was implemented. While there is thus little evidence of learning-by-doing impacts within each PES scheme, implementers may indeed learn more from each other (e.g. through various PES networks), so that newer PES schemes are able to avoid some features that have not worked well for other implementers in the past, in terms of reaching their environmental objectives. Turning now to the three PES design variables, they were all estimated with the expected positive sign (i.e. increasing additionality), though with slightly different levels of significance: payment diversification and the degree of conditionality had a strong effect (significant at 5% level), while spatial targeting narrowly misses that threshold (significant at 10% level). This is an important finding, which reconfirms the recommendations for implementation that are being put forward in the more theoretically orientated PES literature.
Our results could potentially have been affected by the precision level of the environmental additionality estimations. To control for the sensitivity of the model to this factor, we run the previous model including the created ordinal variable describing the precision of additionality measures (Model II in Table 3). The resulting model is robust (Cox & Snell R 2 = 0.55; Nagelkerke R 2 = 0.83; overall predictive accuracy of 96%; no multicollinearity) and confirms the predictive significance of the three critical design factors: Diversification of payments, spatial targeting and conditionality.
Discussion and Conclusion
Meta-analyses can help us recognizing global empirical patterns of PES. As the PES literature flourishes and new case studies are continuously being added, we have the opportunity to gain more knowledge about commonalities in implementation and outcomes. In this article, we took advantage of the expansion of PES case studies in recent years to construct a global database. We used a fairly narrow and explicit definition of PES, in order to identify comparable schemes. The cases selected show a balanced number of schemes by geographical region, ecosystem type, ecosystem service and economic sector involved, showing that the systematic review procedure succeeded in limiting the risk of selection bias. For instance, the limited number of African PES schemes included in our study reflects the actual slower up-take of PES in this continent. We included quantitative and qualitative data from what we see functionally as "genuine" PES schemes, using a narrow PES definition that ensures comparability. We also filtered out cases without the availability of sufficient reliable documentation. We then conducted some exploratory analysis of the emerging patterns of implementation, going from purely descriptive bivariate statistics to principal component analysis of emerging variable clusters, and finally to test some theoretically sustained hypotheses regarding the association of alleged key PES design variables with positive environmental additionality.
From our descriptive analysis, we noted significant differences between publicly and privately financed PES schemes (e.g. public schemes feature larger in area and more costly schemes), but also between the different ES that these schemes target. Public sector PES participation is high in Europe and Asia (with a tradition for public-sector environmental management), yet very low in Sub-Saharan Africa, where public sector institutions have lower capacity to organize PES schemes. Latin America, the prime region of PES implementation, displays a large variety of arrangements.
Next, we identified through categorical principal component analysis three dominating clusters, which we label "agri-environmental public PES", "NGO-led biodiversity PES", and "private commercial carbon PES". Each of these interacts with a set of variables (ecosystem, target ES, lead actors, etc.). While we thus reconfirm the hypothesis from the pre-existing PES literature that public and private PES differ significantly, our results also point to important differences between the two types of private PES schemes: the typical commercial PES (private for-profit company) and the non-commercial (non-governmental not-for-profit organizations) exhibit distinct patterns. Still, also some permeability between the three groups remains, showing that PES implementation is also the result of a hybridization and cooperation between public and private sectors.
Finally, we also attempted to scrutinize which PES design factors influenced environmental outcomes, as measured by a binarily defined proxy of environmental additionality. We confirmed the significance of spatial targeting (for ES density and threat) and, to a statistically somewhat lesser extent, of payment differentiation (as opposed to uniform payments) and the degree of conditionality (monitoring and sanctioning efforts applied). Payments for building environmental assets were also more likely to be additional than payments for avoiding damages. Interestingly, the public vs. private sector distinction, applied here as a control variable, no longer came out as significant. This could imply that the main differences in outcomes between public and private PES manifest themselves through divergences in technical PES design principles, rather than being sui generis differences. Our additionality findings are robust to the sensitivity analysis with respect to the precision in measuring environmental additionality: stronger precision does not condition the validity of the model, and the key design variables remain significant.
As so often in complex interactions between social and biophysical systems, could there potentially be problems of endogeneity in the relations we tested for in our binary logistical regression analysis? For instance, additionality targets may certainly from the outset be lower in public PES schemes, which tend to have more side-objectives. However, this should be controlled for by our sector variables. In principle, we believe that environmental additionality as a response variable constitutes a true 'bottom line' of the interplay between PES contexts and the design of PES interventions. We are thus confident that endogeneity does not constitute a major problem, if any. Yet, as a precautionary measure and sensitivity check, we also ran the model at three different probability cut-off levels (0.25, 0.50 and 0.75), in order to capture the influence of the ex-ante probability of the environmental additionality in the model. We find no changes in model descriptors although the highest accuracy is obtained with the equalprobability assumption of 0.50 (S9 Table). This reinforces our conviction that our current model specification is robust.
For future research, we believe our global-comparative analysis could eventually be improved by including more cases, as they become available, and by more consolidated information about rigorously evaluated key PES outcomes, in both environmental and socioeconomic terms. However, the current scarcity of rigorous impact evaluations applies not only to PES, but basically to any conservation tool other than protected areas [32,41]. While our additionality measure and some of the design proxies are admittedly still rough approximations, our results can be seen as a first set of pointers, to be tested subsequently in more sophisticated ways with more and better data.
Nevertheless, our analyses give an interesting indication that, after controlling for various contextual factors, the application of the best PES design principles may add value to the environmental outcomes. This represents a call for greater efforts of using state-of-the-art principles to make PES design more sophisticated, e.g. in terms of monitoring, targeting and differentiation. As a recommendation, this will probably not be favored across the board by all PES implementers. Yet, to put it conversely, our results indicate that there may be ample efficiency costs attached to the over-simplification of policies and interventions in environmental incentive programs such as PES.
Supporting Information S1 Table. ANOVA and t-test means differences for payments amounts and size of PES programs, by sector and targeted ES. | 2018-04-03T02:13:49.651Z | 2016-03-03T00:00:00.000 | {
"year": 2016,
"sha1": "394aa76565cdcd856a7ec7172a520e2044445034",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0149847&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "9d8c7283478350495125290a93cf2c205904dfd2",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Business",
"Medicine"
]
} |
243881840 | pes2o/s2orc | v3-fos-license | A Descriptive Analysis on the Impact of COVID-19 Lockdowns on Road Traffic Incidents in Sydney, Australia
COVID-19 has had tremendous effects worldwide, resulting in large-scale death and upheaval. An abundance of studies have shown that traffic patterns have changed worldwide as working from home has become dominant, with many facilities, restaurants and retail services being closed due to the lockdown orders. With regards to road safety, there have been several studies on the reduction in fatalities and crash frequencies and increase in crash severity during the lockdown period. However, no scientific evidence has been reported on the impact of COVID-19 lockdowns on traffic incident duration, a key metric for crash management. It is also unclear from the existing literature whether the impacts on traffic incidents are consistent across multiple lockdowns. This paper analyses the impact of two different COVID-19 lockdowns in Sydney, Australia, on traffic incident duration and frequency. During the first (31 March–28 April 2020) and second (26 June–31 August 2021) lockdowns, the number of incidents fell by 50% and 60%, respectively, in comparison to the same periods in 2018 and 2019. The proportion of incidents involving towing increased significantly during both lockdowns. The mean duration of crashes increased by 16% during the first lockdown, but the change was less significant during the subsequent lockdown. Crashes involving diversions, emergency services and towing saw an increase in the mean duration by 67%, 16%, and 47%, respectively, during the first lockdown. However, this was not reflected in the 2021 data, with only major crashes seeing a significant increase, i.e., by 58%. There was also a noticeable shift in the location of incidents, with more incidents recorded in suburban areas, away from the central business area. Our findings suggest drastic changes in incident characteristics, and these changes should be considered by policymakers in promoting a safer and more sustainable transportation network in the future.
Introduction
SARS-CoV-2, also known as the 2019 novel coronavirus (COVID- 19), has had tremendous effects worldwide, resulting in large-scale death and upheaval. At the beginning of March 2020, the World Health Organization (WHO) declared it a pandemic, the first one in over 100 years since the 1918 influenza pandemic. As of September 2021, COVID-19 has claimed over 4 million lives, infected over 200 million people and is still an ongoing worldwide crisis with over 600,000 daily cases worldwide. COVID-19 has meant that governments have taken measures and introduced unprecedented policies for our time, which have had a widespread impact on the way we live our day-to-day lives. These policies have been aimed at reducing travel and mobility in large cities to combat community transmission of COVID-19. Governments have imposed varying degrees of limitations on mobility, internationally and locally. These policies and orders by governments, also known as "lockdown" orders, restrict people from leaving their homes except for essential reasons, and office workers and students have moved to remote work or learning. An abundance of studies have shown that traffic patterns have changed worldwide as working from home has become dominant, with many facilities, restaurants and retail services being closed due to these orders. Urban traffic around the world, without any exception, has observed a drastic reduction in traffic congestion. For example, cities such as Milan, Paris, Madrid and Manchester saw a reduction in congestion by 85%, 84%, 83%, and 76%, respectively, during the lockdown period compared with the baseline scenario [1]. The reduced traffic movement has resulted in a reduction in fuel consumption [2], road crashes and fatalities [3][4][5][6] and air pollution [7,8], and an increase in travel time reliability [9]. On the other hand, lockdowns have also been associated with more frequent harsh acceleration and braking events, mobile phone usage [3], and increased crash severity [4].
Road traffic safety can be characterised by different metrics such as crash frequency, fatalities, rates, severity, and duration [10,11]. There has been much emphasis on the reduction in fatalities and crash frequencies during the lockdown period because of the relative ease in obtaining such data. There are a few studies on crash severity and rate [12], but to the best of the authors' knowledge, there is no study on the impact of lockdown on traffic incident duration. On the one hand, decreased congestion and crash frequency could help the emergency services to reach crash locations faster, thus reducing crash duration. On the other hand, increased crash severity could increase the difficulty in clearing the incident. Therefore, assessing the net impact of lockdowns on crash duration is of interest.
Furthermore, the existing studies on the impact of COVID-19 have only focused on crashes. Other traffic incidents, such as breakdowns, may not result in as much emotional, social, and economic losses as crashes but still significantly reduce roadway capacities and operating conditions. Although vehicles' quality and safety performance have been improved over the years, these incidents cannot be avoided entirely and continue to be a critical problem for road users, authorities, and roadside assistance companies [13,14].
Finally, it is unclear from the existing literature whether the impacts on traffic incidents were consistent across multiple lockdowns. In this context, the current paper aims to quantify the impacts of multiple COVID-19 lockdowns on road safety, specifically on incident duration and frequency in Sydney, Australia. The rest of the paper is organised as follows. First, a review of a few studies on the impacts of COVID-19 on crash characteristics is provided. This is followed by a description of the data and methodology used in this paper. Finally, the results are described and followed by the conclusions and potential future directions.
Literature Review
Current studies show consistency in regard to the fact that traffic-related incidents have decreased significantly due to stay-at-home orders, which limit mobility [3,15]. Saladié et al., (2020) highlight the impact of strict lockdowns in Tarragona province in Spain, with the reduction in crashes (74.3%) being higher than the decrease in mobility (62.9%) during that period [15]. Another study from Spain reported a 67% reduction in the number of crashes in the city of Santander [16]. A study from Greece observed a 62% reduction in crashes, a 68% reduction in fatalities and a 48% drop in serious injuries [17]. However, another study from Greece reported a 57% reduction in traffic but only a 42% reduction in crashes [18]. Muley et al. (2021) reported a 30% drop in traffic movement and a 37% reduction in crash frequency in Qatar [19]. A study conducted in Greece and Saudi Arabia showed a 6% to 11% increment in speeding violations and up to 12% increase in harsh braking events during March and April 2020 [3]. In California, researchers estimated that the crash frequency dropped by 60% [20]. In another study, researchers observed a 65% reduction in collisions during the lockdown in Toronto and a 59% reduction during the "re-opening" period [21].
A study conducted across the USA reported that minor crashes decreased by up to 23%, but the most severe crashes went up by 18% [22]. In Missouri, researchers found a significant reduction in crashes leading to minor or no injuries, but not in crashes resulting in severe or fatal injuries [6]. In the state of Ohio, USA, Li et al. (2021) observed a reduction of up to 55% in crashes, 47% in injuries, 34% in severe or fatal injuries, and 44% in traffic volume [23].
Some studies observed varying impacts of lockdowns on road safety among different sub-groups of the population. For example, Lin et al. (2021) noticed spatial heterogeneity in the reduction in crashes in the cities of Los Angeles and New York. They observed that the proportion of crashes increased for "Hispanic" and "Male" groups, and the crash hot spots shifted from higher-income areas to lower-income areas [24]. A study in West Virginia evaluated the crashes involving injury across different age groups and gender and observed 44% fewer injuries overall [25]. Rapoport et al. (2021) noticed a reduction in injuries by 64.7% among older drivers (aged 80 years and above) and 22.9% among middle-aged drivers (35−54 years) in Ontario, Canada. However, they did not find any significant changes among drivers of other age groups [26].
A few researchers also studied the impact of lockdowns on crash rate (i.e., crash frequency divided by traffic volume), a standardised measure of relative safety on roadway segments that is more easily interpretable than crash frequency [11,27]. For example, Doucette et al., (2020) found that single-vehicle crash rates increased 2.29 times, and single-vehicle fatal crash rates increased 4.10 times during the initial weeks of the stay-athome periods in Connecticut, USA [12]. However, in a separate study using an extended dataset, researchers found that crash rates returned to average during the post-stay-at-home order [28].
As evident from the above review, there has been much emphasis on evaluating the impacts of COVID-19 lockdowns on crash frequency and fatalities. There are also a few studies on crash rate and severity. It is interesting to note that most studies report a crash frequency reduction in the range of 50-60%, apart from a few exceptions. Furthermore, most research has been done in European and North American countries.
Research studies on various traffic incident characteristics are still being conducted in different areas of the world at varying degrees of lockdown. At the time of writing, there is no single study on the impact of COVID-19 on crash duration and other types of incidents such as vehicle breakdowns. Since many countries have implemented multiple lockdowns, the consistency of their impact on road safety has not yet been fully evaluated.
Area of Study
The study area for the analysis is Sydney, the capital of the state of New South Wales (NSW) in Australia. Sydney has a population of more than 5 million and is the most populous city in entire Oceania. Following outbreaks of COVID-19, the first lockdown period for NSW began on 31 March 2020. The extent of this lockdown order restricted residents of Greater Sydney from travelling except for essential purposes such as grocery shopping, work and education (if a person is unable to work or learn at their residence), compassionate grounds, or any other essential services. This order was loosened beginning on the 28 April 2020, where up to 2 visitors were allowed. On the 10 May 2020, cafés and restaurants were allowed to seat up to 10 people, and outdoor gatherings of up to 10 people were also allowed. Figure 1 highlights the onset of COVID-19 cases in NSW, which necessitated restrictions to stop the community transmission of COVID-19 [29].
More recently, from the last week of June 2021, a more restrictive lockdown began to be implemented with the spread of the Delta variant of COVID-19, which has been 97% more transmissible than the original COVID-19 strain [30]. It brought record infection rates in NSW, and consequently, stricter mobility restrictions were imposed. These included the same restrictions as in the first lockdown in 2020 and various other measures over two months from the 26 June 2021 to the current time of writing (early September). Some examples of restrictions that were implemented at various stages of this lockdown include: • Mobility: Individuals were limited to a travel zone of 10 km from their residence. It was subsequentially decreased to 5 km. • People from "high-risk" Local Government Areas (LGAs) were restricted from leaving their LGA unless they were authorised workers. This was constantly revised and added upon as new areas became COVID-19 hotspots. • Public transport: Changed to Sunday timetables. • Work: Construction work was paused for 2 weeks (19 July 2021 to 30 July 2021).
Initially, the restrictions were limited to the city of Sydney. However, on the 14 August 2021, statewide restrictions were made to contain the spread of COVID-19. more transmissible than the original COVID-19 strain [30]. It brought record infection rates in NSW, and consequently, stricter mobility restrictions were imposed. These included the same restrictions as in the first lockdown in 2020 and various other measures over two months from the 26 June 2021 to the current time of writing (early September). Some examples of restrictions that were implemented at various stages of this lockdown include: • Mobility: Individuals were limited to a travel zone of 10 km from their residence. It was subsequentially decreased to 5 km. • People from "high-risk" Local Government Areas (LGAs) were restricted from leaving their LGA unless they were authorised workers. This was constantly revised and added upon as new areas became COVID-19 hotspots. • Public transport: Changed to Sunday timetables.
• Work: Construction work was paused for 2 weeks (19 July 2021 to 30 July 2021). Initially, the restrictions were limited to the city of Sydney. However, on the 14 August 2021, statewide restrictions were made to contain the spread of COVID-19.
Crash Fatality Data
The data on crash fatalities were obtained from the Australian Road Deaths Database (ARDD) [31], which contains all known fatalities recorded in Australia. As of the time of writing, the data on fatalities were correct to August 2021.
The following information was available for each fatality record: • Time and Date • Location: State, Local Government Area (LGA) • Heavy Vehicle/Bus/Truck Involvement • Road information: Speed limit and type of road • Fatality information: Road user, gender and age
Traffic Incident Data
The data on traffic incidents were obtained from the Transport for New South Wales (TfNSW) live traffic website from 1 January 2018 to 31 August 2021, a total of more than
Crash Fatality Data
The data on crash fatalities were obtained from the Australian Road Deaths Database (ARDD) [31], which contains all known fatalities recorded in Australia. As of the time of writing, the data on fatalities were correct to August 2021.
The following information was available for each fatality record: •
Traffic Incident Data
The data on traffic incidents were obtained from the Transport for New South Wales (TfNSW) live traffic website from 1 January 2018 to 31 August 2021, a total of more than 3.5 years of records [32]. The website is managed by the NSW Transport Management Centre (TMC). It is integral in providing monitoring, communication, and traffic management systems to respond to and clear traffic incidents and provide accurate and up-to-date information for travellers on the NSW road network.
The following information is available for each traffic record:
Methodology
Traffic congestion was reduced by close to 40% in metropolitan Sydney during the first lockdown and by up to 75% during the second lockdown, according to the TomTom traffic index [33]. We evaluated whether there was a similar effect on the number of fatalities and other crash characteristics, such as crash frequency, duration and severity. As we had no access to severity indicators such as injuries, proxies for severity were extracted from the dataset. These proxies include the involvement of heavy vehicles, towing, emergency services, and traffic diversions, which are often correlated with more severe crashes [34][35][36][37]. Incident duration is also correlated with severity, as more severe incidents take longer to clear [36,38]. A descriptive analysis of the impact of COVID-19 on fatalities, frequency and incident duration is provided, comparing April 2020 with the same period during April 2018 and 2019 and by comparing the data from the second lockdown with the corresponding period in 2018 and 2019.
Like Doucette (2021) and Saladié (2020), we analysed the incident ratios for each incident category and tested for statistical significance across these time periods using a Chisquare test. Furthermore, we assessed the impact of COVID-19 lockdown orders on incident characteristics and the proportion of incidents involving heavy vehicles (trucks, buses, and trailers), incidents involving towing, fatal crashes and incidents labelled as major by Transport for NSW's Traffic Management Centre. The mean incident duration for each period was also analysed for different categories of incidents such as crashes and breakdowns, with an independent two-way t-test. All statistical tests were conducted with a statistical significance value of 0.05.
Methodology
Traffic congestion was reduced by close to 40% in metropolitan Sydney during the first lockdown and by up to 75% during the second lockdown, according to the TomTom traffic index [33]. We evaluated whether there was a similar effect on the number of fatalities and other crash characteristics, such as crash frequency, duration and severity. As we had no access to severity indicators such as injuries, proxies for severity were extracted from the dataset. These proxies include the involvement of heavy vehicles, towing, emergency services, and traffic diversions, which are often correlated with more severe crashes [34][35][36][37]. Incident duration is also correlated with severity, as more severe incidents take longer to clear [36,38]. A descriptive analysis of the impact of COVID-19 on fatalities, frequency and incident duration is provided, comparing April 2020 with the same period during April 2018 and 2019 and by comparing the data from the second lockdown with the corresponding period in 2018 and 2019.
Like Doucette (2021) and Saladié (2020), we analysed the incident ratios for each incident category and tested for statistical significance across these time periods using a Chi-square test. Furthermore, we assessed the impact of COVID-19 lockdown orders on incident characteristics and the proportion of incidents involving heavy vehicles (trucks, buses, and trailers), incidents involving towing, fatal crashes and incidents labelled as major by Transport for NSW's Traffic Management Centre. The mean incident duration for each period was also analysed for different categories of incidents such as crashes and breakdowns, with an independent two-way t-test. All statistical tests were conducted with a statistical significance value of 0.05.
Fatalities
A comparison of the fatalities reported for April 2020 compared to the same period in 2019 and 2018 (Table 1) reveals a 30% decrease on average. While the sample size is small, this 30% decrease in fatalities compared to a 50% decrease in crash frequency during the same period (refer to Section 5.2) highlights that the impact on fatalities was not the same as the number of crashes. Similarly, fatalities were reduced by 53% during the second lockdown, compared to the 64% reduction in overall crashes. As can be seen in Table 1, the sample size is small and therefore, statistical tests could not be conducted to test whether the changes are significant or not. However, we can see that there has been an increase in the proportion of fatalities that have occurred on highways, and this is represented through the shift in traffic and incidents from Sydney's Central Business District (CBD) to highway roads, as seen in Figures 2 and 3. We also see that the proportion of fatalities involving a heavy vehicle (bus, truck or articulated vehicle) has increased. Furthermore, the proportion of fatalities involving a single vehicle has increased, a finding consistent with a past study [12]. However, the proportion of fatalities involving pedestrians and multiple vehicles has been reduced.
Incident Frequency
A comparison of incident frequency highlights a significant decrease in crashes and vehicle breakdowns during lockdown periods. Figure 4a,b showcase the overall trends in the frequency of crashes and breakdowns with important dates (during and close to the first lockdown) relating to COVID-19 highlighted. There were significant decreases following the spread of COVID-19 throughout the community as national attention began to focus on the first recorded death on the 3 March 2020. Following this, the NSW government began to enact policies and orders aimed at stopping the spread. A strict lockdown commenced on the 31 March, and the significant decline in traffic crashes continued to a nearly 60% decrease compared to peak traffic incidents in 2019. Similarly, vehicle breakdowns dropped by 50%. A comparison of the same period in 2019 and 2018 is provided to highlight these drops. Figure 5a,b show a similar decrease in incident frequency when the second major lockdown was announced in Sydney in late June 2021. However, we see that traffic crashes decreased by more than 70% at the peak due to the stricter restrictions in place. Similarly, vehicle breakdowns dropped by 58%. Table 2 outlines the changes in incident frequency and characteristics and the changes in incident makeup. For each incident involving a certain characteristic, the absolute numbers and proportion relative to the total incidents recorded are presented. The incident rate ratio (IRR) of the incident types during the first and second lockdowns compared to the corresponding periods in 2019 and 2018 also highlight changes for certain indicators of severity. A total of 598 reported incidents that were labelled as breakdowns or crashes occurred during the first lockdown. The proportion of breakdowns and crashes did not change significantly, with IRR values of 1.02 and 0.98, respectively. Figure 5a,b show a similar decrease in incident frequency when the second major lockdown was announced in Sydney in late June 2021. However, we see that traffic crashes decreased by more than 70% at the peak due to the stricter restrictions in place. Similarly, vehicle breakdowns dropped by 58%. Table 2 outlines the changes in incident frequency and characteristics and the changes in incident makeup. For each incident involving a certain characteristic, the absolute numbers and proportion relative to the total incidents recorded are presented. The The proportion of crashes resulting in fatalities increased during this period, with an IRR of 1.4; however, this was not statistically significant (χ 2 = 0.60). The proportion of crashes causing diversions increased significantly, with an IRR of 1.78, which was statistically significant (χ 2 = 3.85, p < 0.01). Diversions required for incidents were associated with an increase in incident duration by up to 60% [36], which could represent a subsequent increase in severity or more directly cause traffic delays. Crashes involving bicycles saw an increase in proportion of incidents by 1.48; however, this was not statistically significant.
The presence of heavy vehicles such as buses, trucks, and articulated vehicles saw a proportionally similar uptick, with the IRR of heavy vehicle crashes and breakdowns increasing to 1.27 (χ 2 = 2.91, p < 0.1) and 1.06 (χ 2 = 0.46), respectively. This would be expected as trucks are a critical part of the NSW economy, transporting goods and various materials to and from Sydney. Despite a lockdown, any effects on heavy vehicle mobility would not be as significant as the impacts on light vehicles, given that lockdown restrictions did not impact their movement. Instead, Freight Australia reported that freight vehicles saw increased activity in a critical effort to keep supply chains open and ensure that communities were adequately supplied [39]. With most heavy vehicles travelling on highspeed highways, this could further increase the severity and fatalities from incidents [40]. The proportion of crashes that required towing also occurred at an increased rate, with an IRR of 1.60 (χ 2 = 4.57, p < 0.05), with the result being statistically significant. Once again, crashes requiring towing have been shown to increase the average incident duration, and subsequently, they could be a proxy indicator for an increase in severity [36]. Table 3 outlines the same changes for the second lockdown in Sydney. A total of 1208 reported incidents that were labelled as breakdowns or crashes occurred in this lockdown period. With different measures undertaken during this period, we also saw different incident characteristics. Most notably, the composition of incidents labelled as crashes saw a statistically significant decrease from a proportion of 0.51 in 2019/2018 to 0.42 (χ 2 = 29.94, p < 0.01). Furthermore, both crashes and breakdowns involving towing saw a statistically significant increase in their incident rate (χ 2 = 33.46, χ 2 = 9.61, p < 0.01). Similarly, crashes involving bicycles saw a significant uptick in incident rate; however, unlike 2020 this was statistically significant. Chi-square test statistic for 1 degree of freedom. χ 2 > 3.841 for p < 0.05. The number in red (and bold) indicates that the IRR has increased during the lockdown period and it is statistically significant. The number in green (and bold) indicates that the IRR has decreased during the lockdown period and it is statistically significant.
However, unlike the 2020 lockdown, there was no statistically significant change in the incident rates for crashes or breakdowns involving heavy vehicles. We could attribute this to the different rules and restrictions put in place as COVID-19 cases began escalate rapidly. Some of the most notable restrictions that would have contributed to this decrease were the 2-week pause on construction work from 19 July to 30 July 2021, and the capacity reduction on construction sites to 50% maximum capacity [41]. A change in public transit to weekend timetables from the 23 August would consequently reduce bus movement [42]. There have also been reports of agitation, protests and delays in the road freight industry due to regular changes in border rules and COVID-19 testing requirements [43,44]. Figure 6a,b showcase the seven-day rolling mean duration for crashes and breakdowns recorded in Sydney, along with key dates for COVID-related events in NSW. It is seen that the first lockdown induced a significant change in the mean incident duration for crashes compared to the same period in previous years when there was no lockdown. However, after the lockdown orders are loosened towards the end of the April, mean incident duration can be seen at its lowest point in three years. On the other hand, no discernible trend can be seen in the breakdown duration during the first lockdown. For the 2021 lockdown, we also see an increase in the duration of crashes for a certain period, most notably from the last week of July to the first week of August 2021 ( Figure 7a). The trend for the breakdown duration is also mostly similar, apart from a decrease in the second half of August (Figure 7b). As seen in Table 4, the mean crash duration increased by 16% during the first lockdown (p-value = 0.03). The breakdown duration decreased by 3%; however, it was statistically insignificant. The direction of the changes in duration was consistent in the second lockdown; however, they are statistically significant. An opportunity for analysis could be to explore further the impact of lockdown compliance with the changes in incident As seen in Table 4, the mean crash duration increased by 16% during the first lockdown (p-value = 0.03). The breakdown duration decreased by 3%; however, it was statistically insignificant. The direction of the changes in duration was consistent in the second lockdown; however, they are statistically significant. An opportunity for analysis could be to explore further the impact of lockdown compliance with the changes in incident duration as Sydney saw increased anti-lockdown sentiment, with protests being prominent during the second lockdown [44,45]. The number in red (and bold) indicates that the mean incident duration has increased during the lockdown period and it is statistically significant.
Incident Duration
Crashes involving diversions, emergency services and towing saw an increase in the mean duration from 2018/2019 to 2020. However, this was not reflected in the 2021 data, with only the major crashes seeing a significant increase. As these incident characteristics often reflect more severe incidents, this suggests that incidents became more severe during the first lockdown as individuals began to adjust their driving behaviour. Finally, none of the breakdown categories saw a significant change in duration during both of the lockdowns. Only the breakdowns involved in towing were reduced in duration during the second lockdown, with the p-value (0.08) close to being significant.
Conclusions
The impact of COVID-19 has been felt throughout the world. As a public crisis that has had devastating ramifications, there is significant evidence suggesting a shift in mobility patterns, and subsequently, a possible shift in driving behaviour and safety characteristics.
In the case of Sydney, there have been positive outcomes with a clear decrease in total crash and vehicle breakdown frequency during the lockdown periods. The decrease in crashes to the tune of 50-60% is consistent with most of the existing studies in Spain, Greece, Saudi Arabia and the USA [15][16][17][18][19]. However, characteristics that indicate more severe incidents (such as incidents labelled as major, fatal and/or involving towing) have increased proportionately, as the decreases have not been as significant as the overall incident rate. This observation is in line with studies done in Missouri and Connecticut in the USA.
With regard to crash duration, we observed a 16% increase during the first lockdown but no significant change during the subsequent lockdown. One can hypothesise that incident duration maybe shorter during the lockdowns as the attending groups no longer have to sit through peak hour traffic because of the overall reduction in congestion and traffic volumes. The alternative hypothesis is that because of the increased incident severity, it is challenging for the attending agencies to clear the incident quickly. The latter hypothesis was likely more pronounced during the first lockdown, while the second lockdown had a mix of both, thus leading to an insignificant change in overall duration.
There are various opportunities for further research. As COVID-19 begins to settle in, studying the effects of mobility changes and permanent shifts in travel patterns and incident characteristics, if any, would be appropriate. It would be interesting to analyse the spatial and temporal shifts in traffic incident hotspots and duration and their relationship with the socio-economic characteristics.
A key limitation within our study is the data source that was used. The data provided a wide range of information that is used for traveller information through NSW's live traffic website. However, there is a lack of information regarding more specific incident characteristics such as the exact emergency services that attended, the speed at which the incidents took place, traffic flow conditions, etc. Furthermore, Sydney's first lockdown was relatively short compared to other countries with multiple stages of mobility restrictions and health orders. Compiling a longer and more comprehensive dataset involving lockdowns with similar restrictions in different areas such as the wider Australia region would be advantageous, as the impact of different policies and various measures can be studied in relation to their impact on mobility and road traffic incidents. This is due to the evolving nature of mobility restrictions seen in Sydney, particularly with new information and restrictions on mobility often changing every week.
As such, COVID-19 reveals both challenges and opportunities for policy makers to design initiatives that aim to reduce future traffic demands, while also considering the impacts and ramifications on incident characteristics and ways to minimise this. We hope COVID-19 provides insight into what a future mobility network focused on sustainable and public transit solutions would be able to achieve for road safety.
Data Availability Statement:
The data presented in this study are available on request from the corresponding author. | 2021-11-10T16:15:56.959Z | 2021-11-01T00:00:00.000 | {
"year": 2021,
"sha1": "092c6df0a0e618fdb7e445adbdb45298e80b46d0",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1660-4601/18/21/11701/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "cd56d66e9a005d790c127b18086bb60ccf348489",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Medicine"
]
} |
15020046 | pes2o/s2orc | v3-fos-license | Measurement of the Cross Section Asymmetry of the Reaction gp-->pi0p in the Resonance Energy Region Eg = 0.5 - 1.1 GeV
The cross section asymmetry Sigma has been measured for the photoproduction of pi0-mesons off protons, using polarized photons in the energy range Eg = 0.5 - 1.1 GeV. The CM angular coverage is Theta = 85 - 125 deg with energy and angle steps of 25 MeV and 5 deg, respectively. The obtained Sigma data, which cover the second and third resonance regions, are compared with existing experimental data and recent phenomenological analyses. The influence of these measurements on such analyses is also considered.
I. INTRODUCTION
Single-pion photoproduction has been used extensively to explore the electromagnetic properties of nucleon resonances, and most determinations of the Nγ resonance couplings [1] have been obtained through multipole analyses of this reaction. The study of N * properties, in general, has enjoyed a resurgence, driven by the growth of new facilities worldwide. Many precise new measurements have focused on the photoproduction of pions and other pseudoscalar mesons, in the hope that they would reveal states not seen previously in the photoproduction and elastic scattering of pions. Associated studies have suggested both new states and couplings to existing states which contradict those found in analyses of the full pion production database.
These results demonstrate the influence of measurements sensitive to new quantities and suggest that older analyses may have been based on insufficient or flawed data sets. A survey of existing data in the 1 GeV region shows most polarization measurements to have only one or two angular points at a given beam energy, often with rather large uncertainties. This allows a great deal of freedom in multipole analyses, and is clearly insufficient to pick out more than the strongest resonance signals.
One quantity which has consistently influenced these new analyses is the beam polarization quantity Σ. Examples include the E2/M1 ratio [2], where new measurements of the differential cross section and Σ changed this ratio by nearly a factor of two. Measurements of Σ in η photoproduction were shown [3,4] to be sensitive to the A 3/2 /A 1/2 ratio of photo-decay amplitudes for the D 13 (1520), implying a ratio very different from the pion photoproduction value. The behavior of Σ in kaon photoproduction has also proved [5] to be crucial in determining the character of a bump seen in precise new total cross section measurements. In π + n photoproduction, results for Σ from GRAAL have been used to argue for a change in the A 1/2 amplitude for the S 11 (1650) [6].
In this experiment, we have performed systematic measurements of the cross section asymmetry Σ for the reaction γp → π • p [7] simultaneously with measurements of Compton scattering γp → γp [7,8], over the energy range E γ = 0.5 -1.1 GeV and at pion CM angles θ * π = 85 • − 125 • . The experimental data (158 data points) from π • -meson photoproduction constitute the first systematic and high statistics measurement of Σ for this reaction channel over the present kinematic region. Preliminary results of the measurements have been presented in [9].
In Sections 2 and 3, we describe the methods employed in this experiment, with special emphasis on the control of systematic errors. In Section 4, we describe how these measurements, and the recent measurements from GRAAL [25], have effected the GW multipole analyses. Comparisons to the Mainz fits [10] are also made. Finally, in Section 5, we summarize our results.
II. EXPERIMENTAL SETUP
The experiment was carried out on the linearly polarized photon beam of the 4.5 GeV Yerevan Synchrotron. The same setup has been used for the parallel study of two reactions: Compton scattering and photoproduction of π • -meson off protons. The experimental setup is shown in Fig. 1.
The linearly polarized photon beam is generated by coherent bremsstrahlung (CB) of 3.0-3.5 GeV electrons on the internal 100 µm diamond crystal target. The beam has 2-3 msec long pulses with a 50 Hz repetition rate and an intensity of about 5 × 10 9 eq. photons/sec. The beam, shaped by a system of collimators and sweeping magnets, has a 10×10 mm size at the position (H 2 ) of the liquid hydrogen target (H 2 ) (9 cm in length). It is transported in a vacuum pipe to the Wilson-type quantameter (Q). The energy spectrum is measured and controlled by a 30-channel pair spectrometer PS-30 (energy resolution 1-2%) [11]. The whole range of bremsstrahlung energies are covered by 5 scans with respective current settings on the PSM analyzing magnet. The full energy spectra of bremsstrahlung on amorphous and crystal targets, measured by the pair spectrometer PS-30, are presented in Fig. 2.
The recoil protons are detected by the magnetic spectrometer (MS) consisting of a doublet of quadrupole lenses (L 1 , L 2 ), a bending magnet (BM), a telescope of four scintillation trigger counters (S 1 − S 4 ) and a system of coordinate detectors, including a scintillation hodoscope (H p ) and seven multiwire proportional chambers (MWPC xy ), allowing the reconstruction of momentum, azimuthal and polar angles of the outgoing proton. Time-of-flight measurements were carried out between S 1 and S 4 counters on a flight base of 9 m for particle identification (p, π). The MS covered a solid angle of ∆Ω ≈ 3.5 msr, and its angular and momentum resolutions were σ θ ≈ 0.3 • , σ φ ≈ 0.2 • and σ P /P ≈ 1%, respectively. The incident photon energy and the CM scattering angle were reconstructed on average with an accuracy of σ Eγ /E γ ≈ 1% and σ θ * ≈ 0.6 • , with corresponding acceptances of ≈ 18% and ≈ 7.5 • , respectively.
The π • decay photons were detected by theČerenkov spectrometer (Č s ) consisting of a veto counter (V), a lead converter (Pb), scintillation hodoscope (H xy ) for x and y coordinate analysis, trigger counter (T), and a lead glassČerenkov counter of full absorption (Č). The energy resolution of theČerenkov counter could be parameterized as σ(E)/E = 0.08/ √ E without converter and σ(E)/E = 0.1/ √ E with converter (E being the photon energy in GeV).
The kinematics of the analyzed process was completely determined by defining the kinematic parameters of the proton in the magnetic spectrometer. For identification of the process γp → π • p, the protons in the magnetic spectrometer were detected in coincidence with the photon detection branch. The coordinate detectors of theČerenkov spectrometer were in addition used for correlation analysis between two detecting branches to identify the low rate yield of the Compton scattering process [7,8].
An independent experimental study was carried out to check the reconstruction accuracy in the magnetic spectrometer (MS) [7]. Monte Carlo calculations, intended to estimate the influence of different possible factors on the reconstruction precision, indicate that inaccuracies in setting the positions, angles and fields of magnetic elements, poor quality of track reconstruction in MWPC xy , additional multiple scattering in materials not taken into account, influence equally the reconstruction accuracy of both the interaction vertex and the scattering angles in MS. Therefore, an experimental study of the vertex position reconstruction accuracy was defined as an effective test of the characteristics of the MS, and was easily carried out by means of a point-like target. Good agreement between the experimental results for vertex reconstruction and Monte Carlo calculations (Fig. 3) provides reliable support for the performance of our proton detection branch within designed precisions.
Finally, the possibility to treat Compton scattering data [7,8], despite significant difficulties caused by the physical background process of π • photoproduction with approximately two orders of magnitude higher cross section, and with practically the same kinematic parameters, also points to a reliable operation of the detection system within designed accuracies.
III. EXPERIMENT AND DATA ANALYSIS
During the measurements, several factors were considered to minimize the influence of systematic uncertainties. A particular emphasis has been placed on maintaining the stability of the coherent peak position. In the case of its deviation, the event registration was paused, until respective correction procedure was made. Such checks of the coherent peak were carried out every 40-50 sec. In Fig. 4, the measured peak region is presented, superimposed on the full coherent bremsstrahlung spectrum, for the cases with correct and distorted peak position. A possible influence on the asymmetry results from instabilities in subsystems was compensated by periodical alternating the orientation of the photon beam polarization during the measurement of the same kinematic point. Also, during data acquisition the trigger rates, efficiencies of the coordinate detectors, distributions of kinematic variables were controlled. Hardware and software of all these control procedures was realized through the CAMAC standard. Fiber optic lines were used to manage the current of the pair spectrometer magnet (PSM) and the electronics of the MS trigger, placed immediately on the MS.
During the off-line analysis, prior to the reconstruction step, the time-of-flight measurements between S 1 and S 4 were analyzed, and the events with a proton candidate in MS were selected (Fig. 5). A soft cut was made, mainly at low energies, on the energy response of thě Cerenkov detector, to reduce the contribution of accidental coincidences and multiparticle events resulting from the energetic part of the coherent bremsstrahlung spectrum [7]. Then, the selected events were sent for reconstruction of the kinematic parameters of proton and the kinematics of the event.
Within further selection, the interaction point in the target was reconstructed and events falling out of the target limits by more than one standard deviation are cut.
The yield of the γp → π • p reaction was defined from the number of selected events in the time spectrum of start-stop measurements between two detection branches (Fig. 6), by fitting and subtracting the respective number of background events in the peak region.
Several corrections were applied to compensate for the contribution of various factors. The results from the analysis of Compton scattering process were used to correct for the influence of that source. Further corrections were applied, compensating for the rate reductions due to the dead time of data acquisition and efficiency of different subdetectors. The maximum trigger rates was about 5 event/sec, and the dead time did not exceed 10%. Another insignificant contribution, resulting from the empty target cell, was measured separately and amounted to ∼1%.
The background pγ− rate is mainly dominated with double pion production processes generated by the high energy tail of CB spectrum. This background contribution has been determined in additional measurements, such as: • pγ−rate dependence on the coherent peak energy of the CB spectrum at fixed kinematics of the setup, allowing the significant variation in the relative yield of single and multiple pion production processes.
The data obtained were then compared with results of Monte Carlo calculations. The contribution of background processes was estimated to be less than 5%.
Calculations of the polarization of the CB spectra were realized by a solution of integral equations for intensity and polarization of the coherent spectra [12]. In the energy range of the measurements, the photon beam, polarization ranged from 50% to 70%.
The cross section asymmetry Σ of the reaction γp → π • p is defined as: where P γ is the polarization of photon beam and C ⊥, are the normalized yields for the photons polarized perpendicular and parallel to the reaction plane. The systematic errors of the measurements were dominated mainly by the accuracy of the linear polarization calculations. The uncertainties in σP γ /P γ have been evaluated to be within ≈ 2.3% in full energy range and are included in the errors of the experimental data. The possible shift in the reconstructed photon energies was calculated from the uncertainty in the MS optics and did not exceed 0.4%. Finally, overall systematic uncertainty for each energy set did not exceed 3%.
It is also important to note, that most kinematic bins were formed from at least two independent data samples collected at different kinematic settings of the detector. The final asymmetry was calculated from the corresponding results in such overlapping bins. On the other hand, comparison of the results on such equivalent bins, produced from different event samples, with different acceptance conditions and polarized incident photons from different parts of the CB spectrum, provided a good check for systematic uncertainties in the obtained results.
The most systematic measurements of Σ for γp → π • p, over the resonance region, exist for the angle θ * π = 90 • (Fig. 7a). From Fig. 7a, one can see, as a whole, a good agreement between the values of prior measurements and our present results. Near 900 MeV, our data disagree with data from MIT [14] and some early SLAC [13] measurements, but are in good agreement with later measurements from another SLAC group [15]. Our points are also consistent with the results of this group for the angle θ * π = 110 • (Fig. 8b). From Fig. 9, it is seen that our results for the angular distributions at E γ = 700, 750, and 800 MeV are in agreement with the existing data as well.
In Figs. 7 − 9, the experimental data are also compared with the predictions of phenomenological analyses from the Mainz [18] and GW [19] groups. While both analyses reproduce the qualitative behavior of Σ, a shape discrepancy is noticeable in the GW fit. The main difference between multipole analyses and experiment is observed above E γ = 700 MeV where there have been few previous measurements. Inclusion of our data in the GW fit results in an improved description, but shape differences remain. The Mainz fit appears to have a shape more consistent with the data. (It should be noted that some of the older fits, in particular the fit of Ref. [21], predict a shape consistent with the Mainz result.) It is interesting to compare these results to those recently published by the GRAAL collaboration [6]. In that work, a similar comparison was made for Σ measurements in the reaction γp → π + n. There a deviation from GW predictions was noted in the 800 − 1000 MeV range, at backward angles. It was suggested that this discrepancy could be removed by a change in the N(1650) photo-decay amplitude (A 1/2 ). We have examined the GRAAL angular distribution at 950 MeV (W CM = 1630 MeV) and agree that a change (reduction) in the E 1/2 0+ multipole can account for the shape difference. Here too, for neutral pion production and energies near 900 MeV, a reduction of this multipole in FA00 by about 20% (in modulus) results in an improved description. A reduction of the same amount is found in comparing the energy-dependent and single-energy solutions. The effect is displayed in Fig. 10. Note that this modification to the GW fit, while improving the agreement with data, actually worsens the agreement with the Mainz prediction at the most backward angles.
V. CONCLUSION
The obtained data on cross section asymmetry (Σ) are in good agreement with existing experimental data and in at least qualitative agreement with phenomenological predictions [18,19,21]. We have found evidence in support of the GRAAL claim that some discrepancies with previous GW analyses could be linked to the E 1/2 0+ multipole and possibly the N(1650) resonance contribution. It will be useful to have a Mainz fit including these data in order to isolate differences with the GW fits. A set of fits employing identical data sets would also be useful in determining whether differences are due to the chosen formalism or data constraints. Work along this line is in progress [22].
In summary, results of present experiment on Σ asymmetry have significantly improved the existing data base in the second and third resonance regions. Forthcoming measurements of cross section and polarization observables in the photoproduction of mesons (JLab [20,23], GRAAL [25]), may lead to more successful determinations of the underlying scattering amplitudes and a more precise determination of resonance photocouplings. | 2014-10-01T00:00:00.000Z | 2000-11-08T00:00:00.000 | {
"year": 2000,
"sha1": "7d5973a140063468f9df006522446135a160f3fb",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/nucl-ex/0011006",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "28cb0d508723caa3e19c0c4587360a2deae228fa",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
119295939 | pes2o/s2orc | v3-fos-license | Running-Mass Inflation Model and Primordial Black Holes
We revisit the question whether the running-mass inflation model allows the formation of Primordial Black Holes (PBHs) that are sufficiently long-lived to serve as candidates for Dark Matter. We incorporate recent cosmological data, including the WMAP 7-year results. Moreover, we include"the running of the running"of the spectral index of the power spectrum, as well as the renormalization group"running of the running"of the inflaton mass term. Our analysis indicates that formation of sufficiently heavy, and hence long-lived, PBHs still remains possible in this scenario. As a by-product, we show that the additional term in the inflaton potential still does not allow significant negative running of the spectral index.
Introduction
to determine the spectral index, its running, and the running of the running. In section 4 we perform a numerical analysis where we compute the spectral index at PBH scales exactly, by numerically solving the equation of motion of the inflaton field. Finally, we conclude in section 5.
Formation of Primordial Black Holes
PBHs may have formed during the very early universe, and if so can have observational implications at the present epoch, e.g. from effects of their Hawking evaporation [1] for masses < ∼ 10 15 g, or by contributing to the present "cold" dark matter density if they are more massive than 10 15 g [3]. (These PBHs would certainly be massive enough to be dynamically "cold" 1 ).
The traditional treatment of PBH formation is based on the Press-Schechter formalism [10] used widely in large-scale structure studies. Here the density field is smoothed on a scale R(M). In the case at hand, R(M) is given by the mass enclosed inside radius R when R crossed the horizon. The probability of PBH formation is then estimated by simply integrating the probability distribution P (δ; R) over the range of perturbations δ which allow PBH formation: δ th < δ < δ cut , where the upper limit arises since very large perturbations would correspond to separate closed 'baby' universes [3,11]. We will show that in practice P (δ; R) is such a rapidly decreasing function of δ above δ th that the upper cutoff is not important. The threshold density is taken as δ th > w, where w = p/ρ is the equation of state parameter describing the epoch during which PBH formation is supposed to have occurred [3]. Here we take w = 1/3, characteristic for the radiation dominated epoch which should have started soon after the end of inflaton. However the correct value of the threshold δ th is quite uncertain. Niemeyer and Jedamzik [12] carried out numerical simulations of the collapse of the isolated regions and found the threshold for PBH formation to be 0.7. We will show that PBHs abundance is sensitive to the value of δ th .
The fraction of the energy density of the Universe in which the density fluctuation exceeds the threshold for PBH formation when smoothed on scale R(M), δ(M) > δ th , which will hence end up in PBHs with mass ≥ γM, 2 is given as in Press-Schechter theory by 3 f (≥ M) = 2γ ∞ δ th P (δ; M(R))dδ . (2.1) Here P (δ; M(R)) is the probability distribution function (PDF) of the linear density field δ smoothed on a scale R, and γ is the fraction of the total energy within a sphere of radius R that ends up inside the PBH. 1 It has been argued [9] that BH evaporation might leave a stable remnant with mass of order of the Planck mass, which could form CDM; we will not pursue this possibility here. 2 Throughout we assume for simplicity that the PBH mass is a fixed fraction γ of the horizon mass corresponding to the smoothing scale. This is not strictly true. In general the mass of PBHs is expected to depend on the amplitude, size and shape of the perturbations [12,13]. 3 We follow ref. [14] in including a factor of two on the right-hand side; this factor is not very important for delineating the inflationary model parameters allowing significant PBH formation. Moreover, we set δ cut to infinity.
For Gaussian fluctuations, the probability distribution of the smoothed density field is given by 4 .
(2.2)
This PDF is thus uniquely determined by the variance σ δ (R) of δ, which is given by In order to compute the variance, we therefore have to know the power spectrum of δ, P δ (k) ≡ k 3 /(2π 2 ) |δ k | 2 , as well as the volume-normalized Fourier transform of the window function used to smooth δ, W (kR). It is not obvious what the correct smoothing function W (kR) is; a top-hat function has often been used in the past, but we prefer to use a Gaussian window function 5 , Finally, on comoving hypersurfaces there is a simple relation between the density perturbation δ and the curvature perturbation R c [16]: The density and curvature perturbation power spectra are therefore related by The mass fraction of the Universe that will collapse into PBHs can now be computed by inserting eqs.(2.6) and (2.4) into eq.(2.3) to determine the variance as function of R. This has to be used in eq.(2.2), which finally has to be inserted into eq.(2.1). Since we assume a Gaussian P (δ) in eq.(2.2), the integral in eq.(2.1) simply gives an error function.
In order to complete this calculation one needs to relate the mass M to the comoving smoothing scale R. The number density of PBHs formed during the reheating phase just after the end of inflation will be greatly diluted by the reheating itself. We therefore only consider PBHs which form during the radiation dominated era after reheating is (more or less) complete. The initial PBHs mass M PBH is related to the particle horizon mass M by 6 when the scale enters the horizon, R = (aH) −1 . Here the coefficient γ, which already appeared in eq.(2.1), depends on the details of gravitational collapse. A simple analytical calculation suggest that γ ≃ w 3/2 ≃ 0.2 during the radiation era [3]. During radiation domination aH ∝ a −1 , and expansion at constant entropy gives ρ ∝ g −1/3 * a −4 [17] (where g * is the number of relativistic degrees of freedom, and we have approximated the temperature and entropy degrees of freedom as equal). This implies that where the subscript "eq" refers to quantities evaluated at matter-radiation equality. In the early Universe, the effective relativistic degree of freedom g * is expected to be of order 100, while g * ,eq = 3.36 and k eq = 0.07Ω m h 2 Mpc −1 (Ω m h 2 = 0.1334 [8]). The horizon mass at matter-radiation equality is given by where a −1 eq = (1 + z eq ) = 3146 and (assuming three species of massless neutrinos) Ω rad,0 h 2 = 4.17 × 10 −5 . Then it is straightforward to show that Note that M PBH ∝ R 2 , not ∝ R 3 as one might naively have expected. Recall that M PBH is related to the horizon mass at the time when the comoving scale R again crossed into the horizon. Larger scales re-enter later, when the energy density was lower; this weakens the dependence of M PBH on R. Moreover, the lightest black holes to form are those corresponding to a comoving scale that re-enters the horizon immediately after inflation. 7 The Gaussian window function in eq.(2.3) strongly suppresses contributions with k > 1/R. At the same time, the factor k 4 in eq.(2.6) suppresses contributions to the integral in eq.(2.3) from small k. As a result, this integral is dominated by a limited range of k−values near 1/R. Over this limited range one can to good approximation assume a power-law primordial power spectrum with fixed power n S 8 P Rc (k) = P Rc (k R )(k/k R ) n S (R)−1 , with k R = 1/R. With this ansatz, the variance of the primordial density field at horizon crossing is given by for n S (R) > −3. The power P Rc is known accurately at CMB scales; for example, P Rc (k 0 ) = (2.43 ± 0.11) × 10 −9 at the "COBE scale" k 0 = 0.002 Mpc −1 [8]. In order to relate this to the scales relevant for PBH formation, we parameterize the power spectrum as (2.12) It is important to distinguish between n S (R) and n(R) at this point. n S (R) describes the slope of the power spectrum at scales k ∼ k R = 1/R, whereas n(R) fixes the normalization of the spectrum at k R ≫ k 0 . The two powers are identical if the spectral index is strictly constant, i.e. if neither n S nor n depend on R. However, in this case CMB data imply [8] that n = n S is close to unity. Eqs.(2.11) and (2.12) then give a very small variance, leading to essentially no PBH formation. Significant PBH formation can therefore only occur in scenarios with running spectral index. We parameterize the scale dependence of n as [19]: recall that we are interested in R ≪ 1/k 0 , i.e. ln(k 0 R) < 0. The parameters α S and β S denote the running of the effective spectral index n S and the running of the running, respectively: (2.14) Eq.(2.13) illustrates the difference between n(R) and n S (R). The latter has an expansion similar to eq.(2.13), but with the usual Taylor-expansion coefficients, 1 in front of α S and 1/2 in front of β S . One therefore has Setting n S (k 0 ) = 1 for simplicity, eq.(2.15) implies n S (R) = 2n(R) − 1 for β S = 0, and n S (R) = 3n(R) − 2 for α S = 0. We will compute the variance σ(R), and hence the PBH fraction f , for these two relations; they represent extreme cases if neither α S nor β S is negative. The result of this calculation is shown in figure 1. Here we have fixed γ = 0.2, and show results for two choices of the threshold δ th and three choices of n(R). We see that scenarios where n(R) = 1.3 (or smaller) are safe in the SM, because there is no model-independent limit on f for M PBH < 10 10 g [4]. As noted earlier, PBHs contributing to Dark Matter today must have M PBH > ∼ 10 15 g; at this mass, they saturate the DM relic density if f ≃ 5 ×10 −19 . 9 Figure 1 shows that this requires n(R) ≃ 1.37 (1.41) for δ th = 0.3 (0.7). The dependence on n S is much milder. Figure 1 also illustrates a serious problem that all scenarios that aim to explain the required CDM density in terms of post-inflationary PBH formation face. We just saw that this can happen only if the spectral index n increases significantly between the scales probed by the CMB and other cosmological observations and the scale R ≃ 10 −9 pc relevant for the formation of 10 15 g PBHs. However, n must then decrease rapidly when going to slightly smaller length scales, since otherwise one would overproduce lighter PBHs. For example, successful Big Bang Nucleosynthesis requires [4] f (10 13 g) ≤ 2 × 10 −20 , about 12 orders of magnitude below that predicted by keeping n(R) fixed at the value required for having 10 15 g PBHs as CDM candidates. We will come back to this point at the end of our paper.
Another problem is that the expansion of n(R) in eq.(2.13) will generally only be accurate if | ln(k 0 R)| is not too large. This is the case for cosmological observations, which probe scales > ∼ 1 Mpc. The expansion becomes questionable for the scales probed by PBH formation. For example, fixing k 0 = 0.002 Mpc −1 , eq.(2.10) gives | ln(k 0 R)| = 41.1 for M PBH = 10 15 g. Fortunately within the framework of a given inflationary scenario this second problem can be solved by computing n(R) and n S (R) exactly, rather than using the expansion (2.13).
Observational bounds on n S and α S can e.g. be found in ref. [8]. By requiring that these parameters remain within the 2σ ranges for all scales k between 10 −4 Mpc −1 and 10 Mpc −1 , we find β S < ∼ 0.03. Here we used the error bars derived from the analysis of WMAP7 data plus data on baryonic acoustic oscillations (BAO), ignoring possible tensor modes (as appropriate for small-field models), and including independent measurements of the Hubble constant as a prior (i.e. we used the "WMAP7+BAO+H 0 " data set). Here we use the pivot scale k 0 = 0.0155 Mpc −1 where n S (k 0 ) and α S (k 0 ) are uncorrelated. On the other hand, eq.(2.13) shows that for α S = 0 we only need β S (k 0 ) ≃ 0.0016 in order to generate sufficiently large density perturbations to allow formation of 10 15 g PBHs. Even if we set α S (k 0 ) equal to its central value, α S (k 0 ) = −0.022, we only need β S (k 0 ) ≃ 0.0032. Including the running of the running of the spectral index thus easily allows to accommodate PBH formation in scenarios that reproduce all current cosmological observations at large scales.
Of course, this kind of model-independent analysis does not show whether simple, reasonably well-motivated inflationary models exist that can generate a sufficiently large β S . One scenario that quite naturally accommodates strong scale dependence of the spectral index is the running-mass inflation model [5,6]. We now apply our results to this model.
Running-Mass Inflation Model
This model, proposed by Stewart [5,6], exploits the observation that in field theory the parameters of the Lagrangian are scale dependent. This is true in particular for the mass of the scalar inflaton, which can thus be considered to be a "running" parameter. 10 The running of the mass parameter can be exploited to solve the "η−problem" of inflation in supergravity [20]. This problem arises because the vacuum energy driving inflation also breaks supersymmetry. In "generic" supergravity models the vacuum energy therefore gives a large (gravity-mediated) contribution to the inflaton mass, yielding |η| ∼ 1. However, this argument applies to the scale where SUSY breaking is felt, which should be close to the (reduced) Planck scale M P = 2.4×10 18 GeV. In the running mass model, renormalization group (RG) running of the inflaton mass reduces the inflaton mass, and hence |η|, at scales where inflation actually happens. There are four types of model, depending on the sign of the squared inflaton mass at the Planck scale, and on whether or not that sign change between M P and the scales characteristic for inflation [21].
The simplest running-mass model is based on the inflationary potential where φ is a real scalar; in supersymmetry it could be the real or imaginary part of the scalar component of a chiral superfield. The natural size of |m 2 φ (M P )| in supergravity is of order V 0 /M 2 P . Even for this large value of m 2 φ , which gives |η| ∼ O(1), the potential will be dominated by the constant term for φ 2 ≪ M 2 P ; as mentioned above, running is supposed to reduce |m 2 φ | even more at lower φ 2 .
The potential (3.1) would lead to eternal inflation. One possibility to end inflation is to implement the idea of hybrid inflation [22]. To that end, one introduces a real scalar "waterfall field" ψ, and adds to the potential the terms 11 Here λ and κ are real couplings, and the coefficient of the ψ 2 term has been chosen such that ψ remains frozen at the origin. Once φ 2 falls below this critical value, ψ quickly approaches its final vacuum expectation value, given by ψ 2 = 24V 0 κ , while φ quickly goes to zero, thereby "shutting off" inflation. However, in this paper we focus on the inflationary period itself; the evolution of perturbations at length scales that left the horizon a few e-folds before the end of inflation should not be affected by the details of how inflation is brought to an end. 12 During inflation, the potential is thus simply given by eq.(3.1). Here m 2 φ (φ) is obtained by integrating an RG equation of the form where β m is the β−function of the inflaton mass parameter. If m 2 φ is a pure SUSY-breaking term, to one loop β m can be schematically written as [5,24] where the first term arises from the gauge interaction with coupling α and the second term from the Yukawa interaction λ Y . C and D are positive numbers of order one, which depend on the representations of the fields coupling to φ, m is a gaugino mass parameter, while m 2 s is the scalar SUSY breaking mass-squared of the scalar particles interacting with the inflaton via Yukawa interaction λ Y .
For successful inflation the running of m 2 φ must be sufficiently strong to generate a local extremum of the potential V φ for some nonvanishing field value, which we call φ * . The inflaton potential will obviously be flat near φ * , so that inflation usually occurs at field values not very far from φ * . We therefore expand m 2 φ (φ) around φ = φ * . The potential we work with thus reads: is given by the scale dependence of the parameters appearing in eq. (3.4). In contrast to earlier analyses of this model [21,24,25], we include the ln 2 (φ/φ * ) term in the potential. This is a two-loop correction, but it can be computed by "iterating" the one-loop correction. 13 Since the coefficient g of this term is of fourth order in couplings, one will naturally expect |g| ≪ |c|. However, this need not be true if |c| "happens" to be suppressed by a cancellation in eq.(3.4). Including the second correction to m 2 φ (φ) seems natural given that we also expanded the running of the spectral index to second (quadratic) order.
Recall that we had defined φ * to be a local extremum of V φ , i.e. V ′ (φ * ) = 0. This implies c; this relation is not affected by the two-loop correction ∝ g. 12 It has recently been pointed out that the waterfall phase might contribute to PBH formation [23]. 13 There are also "genuine" two-loop corrections, which can not be obtained from a one-loop calculation, but they only affect the term linear in ln(φ/φ * ). They are thus formally included in our coefficient c.
In order to calculate the spectral parameters n S , α S and β S defined in eqs.(2.14), we need the first four slow-roll parameters, defined as [16]: 14 where primes denote derivatives with respect to φ. All these parameters are in general scaledependent, i.e. they have to be evaluated at the value of φ that the inflaton field had when the scale k crossed out of the horizon. The spectral parameters are related to these slow-roll parameters by [16]: Slow-roll inflation requires |ǫ|, |η| ≪ 1. Combining eqs.(3.5) and (3.6) we see that we need V 0 ≫ cφ 2 L, gφ 2 L 2 , where we have introduced the short-hand notation In other words, the inflaton potential has to be dominated by the constant term, as noted earlier.
Eqs.(3.6) then imply two strong inequalities between (combinations of) slow-roll parameters: |ǫ| ≪ |η| ; |ǫη| ≪ |ξ 2 | . (3.9) The first relation means that n S − 1 is essentially determined by η. Similarly, both relations together imply that α S is basically fixed by ξ 2 , while only the last two terms in the expression for β S are relevant; these two terms are generically of similar order of magnitude. Finally, the factors of V appearing in the denominators of eqs.(3.6) can be replaced by V 0 . The spectral parameters are thus given by: The powers on ξ 2 and σ 3 are purely by convention; in particular, ξ 2 could be negative.
where L has been defined in eq.(3.8). Clearly the spectral index is not scale-invariant unless c and g are very close to zero. Note that V 0 appears in eqs. (3.10) only in the dimensionless combination cM 2 P /V 0 , while φ only appears via L, i.e. only the ratio φ/φ * appears in these equations.
In contrast, the absolute normalization of the power spectrum is given by (in slow-roll approximation) This normalization is usually quoted at the "COBE scale" k 0 = 0.002 Mpc −1 , where P R = 2.43 × 10 −9 [8]. Applying our potential (3.5) to eq.(3.11), replacing V by V 0 in the numerator, we see that P Rc not only depends on cM 2 P /V 0 and L, but also on the ratio V 0 /(M 2 P φ 2 ). We can thus always find parameters that give the correct normalization of the power spectrum, for all possible combinations of the spectral parameters.
We want to find out whether the potential (3.5) can accommodate sufficient running of n S to allow PBH formation. There are strong observational constraints on n S and α S . It is therefore preferable to use these physical quantities directly as inputs, rather than the model parameters cM 2 P /V 0 , L and g/c. To this end we rewrite the first eq.(3.10) as: .
(3.12)
Inserting this into the second eq.(3.10) gives: (3.13) We thus see that the running of the spectral index is "generically" of order (n S − 1) 2 ; similarly, the running of the running can easily be seen to be ∝ (n S − 1) 3 . This is true in nearly all inflationary scenarios that have a scale-dependent spectral index. Eq.(3.13) can be solved for g/c. Bringing the denominator to the left-hand side leads to a quadratic equation, which has two solutions. They can be written as: 14) where we have introduced the quantity Since g and c are real quantities, eq.(3.14) only makes sense if the argument of the square root is non-negative. Note that the coefficients multiplying r S and r 2 S inside the square root are both non-negative. This means that the model can in principle accommodate any non-negative value of r S . However, small negative values of r S cannot be realized. It is easy to see that the constraint on r S is weakest for L = 0. The argument of the square root is then positive if 2r S + r 2 S > 0, which implies either r S > 0 or r S < −2. Recalling the definition (3.15) we are thus led to the conclusion this bound should hold on all scales, as long as the potential is described by eq.(3.5). Note that it is identical to the bound found in ref. [25], i.e. it is not affected by adding the term ∝ L 2 to the inflaton potential. This is somewhat disappointing, since recent data indicate that α S is negative at CMB scales. Even the generalized version of the running mass model therefore cannot reproduce the current 1 σ range of α S . However, at the 2 σ level significantly positive α S values are still allowed. Let us therefore continue with our analysis, and search for combinations of parameters within the current 2 σ range that might lead to significant PBH formation. Using eqs.(3.12) and (3.14) we can use n S (k 0 ) − 1, α S (k 0 ) and L 0 ≡ ln(φ 0 /φ * ) as input parameters in the last eq.(3.10) to evaluate β S (k 0 ). This can then be inserted into eq.(2.13) to see how large the density perturbations at potential PBH scales are.
This numerical analysis is most easily performed at the "pivot scale", where the errors on n S and α S are essentially uncorrelated; at k−values above (below) this scale, n S and α S are correlated (anticorrelated) [26]. The pivot scale for the "WMAP7+BAO+H 0 " data set we are using is k 0 ≡ k pivot = 0.0155 Mpc −1 [8]. 15 At this scale, observations give [8]: n S (k 0 ) = 0.964 ± 0.013 ; α S (k 0 ) = −0.022 ± 0.020 . (3.17) We saw above that requiring the correct normalization of the power spectrum at CMB scales does not impose any constraint on the spectral parameters. However, the model also has to satisfy several consistency conditions. To begin with, it should provide a sufficient amount of inflation. In the slow-roll approximation, the number of e-folds of inflation following from the potential (3.5) is given by: where we have introduced the dimensionless quantitỹ recall that it can be traded for n S (k 0 ) − 1 using eq. This can be inverted to give where we have introduced The problem is that the denominator in eq.(3.21) vanishes for some finite value of k. This defines an extremal value of ∆N: A negative value of ∆N ex is generally not problematic, since only a few e-folds of inflation have to have occurred before our pivot scale k 0 crossed out of the horizon. However, a small positive value of ∆N ex would imply insufficient amount of inflation after the scale k 0 crossed the horizon. In our numerical work we therefore exclude scenarios with −5 < ∆N ex < 50, i.e. we (rather conservatively) demand that at least 50 e-folds of inflation can occur after k 0 crossed out of the horizon. A second consistency condition we impose is that |L| should not become too large. Specifically, we require |L(k)| < 20 for all scales between k 0 and the PBH scale. For (much) larger values of |L| our potential (3.5) may no longer be appropriate, i.e. higher powers of L may need to be included.
Note that eq.(3.21) allows to compute the effective spectral index n S (k) exactly: .
(3.24)
This in turn allows an exact (numerical) calculation of the spectral index n(k): (3.25) In our numerical scans of parameter space we noticed that frequently the exact value for n(k) at PBH scales differs significantly from the values predicted by the expansion of eq.(2.13); similar statements apply to n S (k). This is not very surprising, given that | ln(k 0 R)| = 39.1 for our value of the pivot scale and M PBH = 10 15 g. In fact, we noticed that even if α S (k 0 ) and β S (k 0 ) are both positive, n S (k) may not grow monotonically with increasing k. In some cases n S (k) computed according to eq.(3.24) even becomes quite large at values of k some 5 or 10 e-folds below the PBH scale. This is problematic, since our calculation is based on the slow-roll approximation, which no longer works if n S − 1 becomes too large. We therefore demanded n S (k) < 2 for all scales up to the PBH scale; the first eq.(3.7) shows that this corresponds to η < 0.5. This last requirement turns out to be the most constraining one when looking for combinations of parameters that give large n(k PBH ).
Numerical Results
We are now ready to present some numerical results. We begin in figure 2, which shows a scatter plot of the spectral parameters α S (k 0 ) and β S (k 0 ), which has been obtained by randomly choosing model parametersc [defined in eq.(3.19)], g/c and L 0 in the ranges 16 |c| ≤ 1, |g|M 2 P /V 0 ≤ 1, |L 0 | ≤ 20. We require that n S (k 0 ) and α S (k 0 ) lie within their 2 σ ranges, and impose the consistency conditions discussed above. The plot shows a very strong correlation between β S and α S : if the latter is negative or small, the former is also small in magnitude. Moreover, there are few points at large α S , and even there most allowed combinations of parameters lead to very small β S . The accumulation of points at small α S can be understood from our earlier result (3.13), which showed that α S is naturally of order (n S − 1) 2 < 0.004 within 2 σ. Moreover, β S is naturally of order α 3/2 S . On the other hand, for α S values close to the upper end of the current 2 σ range, we do find some scenarios where β S is sufficiently large to allow the formation of 10 15 g PBHs.
We also explored the correlation between n S (k 0 ) and α S (k 0 ) (not shown). Here the only notable feature is the lower bound (3.16) on α S (k 0 ); values of α S (k 0 ) up to (and well beyond) its observational upper bound can be realized in this model for any value of n S (k 0 ) within the presently allowed range. Similarly, we do not find any correlation between n S (k 0 ) and β S (k 0 ). This lack of correlation can be explained through the denominator in eq. (3.13), which also appears (to the third power) in the expression for β S once eq.(3.12) has been used to trade c for n S − 1: this denominator can be made small through a cancellation, allowing sizable α S even if n S is very close to 1. Since the same denominator appears (albeit with different power) in the expressions for α S and β S , it does not destroy the correlation between these two quantities discussed in the previous paragraph. Figure 2 indicates that large values of n at the PBH scale can be achieved only if α S at the CMB scale is positive and not too small. In figure 3 we therefore explore the dependence of the potential parameters, and of the spectral parameters at the PBH scale on L 0 , for n S (k 0 ) = 0.964, α S (k 0 ) = 0.012. Note that varying L 0 also changes the parameters c and g (or g/c), see eqs. (3.12) and (3.14). The latter in general has two solutions; however, for most values of L 0 , only one of them leads to sufficient inflation while keeping |L| < 20; if both solutions are allowed, we take the one giving a larger spectral index at the PBH scale, taken to be 1.5 × 10 15 Mpc −1 corresponding to M PBH = 10 15 g. Note that n S and n at the PBH scale are calculated exactly, using eqs.(3.7) and (3.25). We find that the expansion (2.13) is frequently very unreliable, e.g. giving the wrong sign for n − 1 at the PBH scale for L 0 < −1. Figure 3 shows thatc is usually well below 1, as expected from the fact that n S − 1 ∝c, see the first eq.(3.10). Moreover, in most of the parameter space eq.(3.14) implies |g| < |c|; recall that this is also expected, since g is a two-loop term. We find |g| > |c| only if |c| is small. In particular, the poles in g/(2c) shown in figure 3 occur only where c vanishes; note that the spectral parameters remain smooth across these "poles".
There are a couple of real discontinuities in figure 3, where the curves switch between the two solutions of eq.(3.14). The first occurs at L 0 ≃ −0.756. For smaller values of L 0 , the solution giving the smaller |g/c| violates our slow-roll condition |n S − 1| < 1 at scales close to the PBH scale. For larger L 0 this condition is satisfied. Just above the discontinuity, where n S is close to 2 at the PBH scale, we find the largest spectral index at the PBH scale, which is close to 0.47. Recall from figure 1 that this will generate sufficiently large density perturbations to allow the formation of PBHs with M PBH = 10 15 g. However, the formation of PBHs with this mass is possible only for a narrow range of L 0 , roughly −0.756 ≤ L 0 ≤ −0.739.
At L 0 = −0.31, c goes through zero, giving a pole in g/c as discussed above. Then, at L 0 = −0.214, the second discontinuity occurs. Here the curves switch between the two solutions of eq.(3.14) simply because the second solution gives a larger spectral index at the PBH scale. Right at the discontinuity both solutions give the same spectral index, i.e. the curve depicting n PBH remains continuous; however,c jumps from about 0.132 to −0.339. The effective spectral index n S at the PBH scale also shows a small discontinuity. Recall from our discussion of eq.(2.15) that n S will generally be larger than n at the PBH scale, but the difference between the two depends on the model parameters.
For very small values of |L 0 |,c becomes very large; this region of parameter space is therefore somewhat pathological. For sizably positive L 0 , n at the PBH scale increases slowly with increasing L 0 , while |c| and |g/c| both decrease. However, the spectral index at the PBH scale remains below the critical value for the formation of long-lived PBHs.
Note that L always maintains its sign during inflation, since L = 0 corresponds to a stationary point of the potential, which the (classical) inflaton trajectory cannot cross. For most of the parameter space shown in figure 3, |L| decreases during inflation. If L 0 < 0 decreasing |L| corresponds to V ′ (φ 0 ) < 0, i.e. the inflaton rolls towards a minimum of the potential at φ = φ * . For L 0 > 0 we instead have V ′ (φ 0 ) > 0, i.e. the inflaton rolls away from a maximum of the potential.
In fact, this latter situation also describes the branch of figure 3 giving the largest spectral index at the PBH scale; since here L 0 < 0, |L| increases during inflation on this branch. This is illustrated in figures 4, which show the (rescaled) inflaton potential as well as the effective spectral index as function of either the inflaton field (left frame) or of the scale k (right frame). Note that all quantities shown here are dimensionless, and are determined uniquely by the dimensionless parametersc defined in eq. (3.19), g/c and L 0 . This leaves two dimensionful quantities undetermined, e.g. V 0 and φ * ; one combination of these quantities can be fixed via the normalization of the CMB power spectrum, leaving one parameter undetermined (and irrelevant for our discussion).
The left frame shows that the (inverse) scale k first increases quickly as φ rolls down from , and of the effective spectral index n S (dashed, red), as a function of the inflaton field φ/φ 0 (left frame) or of the ratio of scales k/k 0 (right frame). In the left frame the dot-dashed (blue) curve shows k/k 0 , whereas in the right frame it depicts φ/φ 0 ; this curve in both cases refers to the scale to the right. We tookc = −0.1711, g/c = 0.09648, L 0 = −0.756; these parameters maximize the spectral index at the PBH scale for n S (k 0 ) = 0.964, α S (k 0 ) = 0.012 (see figure 3).
its initial value φ 0 . This means that φ initially moves rather slowly, as can also be seen in the right frame. Since α S > 0, the effective spectral index increases with increasing k. The right frame shows that this evolution is quite nonlinear, although for k/k 0 < ∼ 10 10 , n S (k) is to good approximation a parabolic function of ln(k/k 0 ). However, for even smaller scales, i.e. larger k, the rate of growth of n S decreases again, such that n S − 1 reaches a value very close to 1 at the scale k = 1.6 × 10 16 k 0 relevant for the formation of PBHs with M PBH = 10 15 g. Recall that we only allow solutions where n S (k) < 2 for the entire range of k considered; figures 4 therefore illustrate our earlier statement that this constraint limits the size of the spectral index at PBH scales.
The left frame of figure 4 shows that the inflaton potential as written becomes unbounded from below for φ → +∞. This can be cured by introducing a quartic (or higher) term in the inflaton potential; the coefficient of this term should be chosen sufficiently small not to affect the discussion at the values of φ of interest to us. Note also that this pathology of our inflaton potential is not visible in the right frame, since assuming φ = φ 0 < φ * at k = k 0 , the inflaton field can never have been larger than φ * : as noted above, it cannot have moved across the maximum of the potential. Figure 2 indicated a strong dependence of the maximal spectral index at PBH scales on α S (k 0 ). This is confirmed by figure 5, which shows the maximal possible n(k PBH ) consistent with our constraints as function of α S (k 0 ), as well as the corresponding values of the parameters c, g/c and L 0 . We saw in the discussion of eq.(3.14) that α S < 0 is only allowed for a narrow range of L 0 . In this very constrained corner of parameter space, n(k PBH ) remains less than 1, although the effective spectral index n S (k PBH ) can exceed 1 for α S (k 0 ) > −1.7 × 10 −4 ; recall that for the given choice n S (k 0 ) = 0.964, solutions only exist if α S (k 0 ) > −3.24 × 10 −4 , see eq. (3.16).
For α S (k 0 ) ≤ 1.5×10 −4 the optimal set of parameters lies well inside the region of parameter space delineated by our constraints. n S (k PBH ) therefore grows very fast with increasing α S (k 0 ). -L 0 n PBH -1 Figure 5: The solid (black) curve shows the maximal spectral index at the PBH scale k PBH = 1.6 × 10 19 k 0 that is consistent with the constraints we impose, as function of α S (k 0 ). The other curves show the corresponding model parameters:c (double-dotted, green), multiplied with −5 for ease of presentation; −g/(2c) (dotted, blue); and −L 0 (dot-dashed, red). n S (k 0 ) is fixed at 0.964, but the bound on n PBH is almost independent of this choice.
For these very small values of α S (k 0 ), the largest spectral index at PBH scales is always found for g = −2c, which implies that the second derivative of the inflaton potential also vanishes at φ = φ * , i.e. φ * corresponds to a saddle point, rather than an extremum, of the potential.
For slightly larger values of α S (k 0 ) the choice g = −2c allows less than 50 e-folds of inflation after k 0 exited the horizon. Requiring at least 50 e-folds of inflation therefore leads to a kink in the curve for n(k PBH ). The optimal allowed parameter set now has considerable smaller |g/c|, but larger |c|.
The curve for the maximal n(k PBH ) shows a second kink at α S (k 0 ) = 0.001. To the right of this point the most important constraint is our requirement that |n S − 1| < 1 at all scales up to k PBH , as discussed in connection with figures 3 and 4. Note that for α S (k 0 ) > 0.014, the optimal parameter choice leads to n S reaching its maximum at some intermediate k close to, but smaller than, k PBH . This leads to a further flattening of the increase of n(k PBH ).
We nevertheless see that for values of α S (k 0 ) close to the upper end of the 2 σ range specified in (3.17) the spectral index at the scale relevant for the formation of 10 15 g PBHs can be well above the minimum for PBH formation found in section 2. Figure 1 formation of considerably heavier PBHs might be possible in running mass inflation. However, larger PBH masses correspond to smaller k PBH , see eq.(2.10). This in turn allows for less running of the spectral index. In order to check whether even heavier PBHs might be formed during the slow-roll phase of running mass inflation, one therefore has to re-optimize the parameters for different choices of k PBH .
The result of such an analysis is shown in figure 6. We see that in the given model, the formation of PBHs that are sufficiently massive, and hence long-lived, to be CDM candidates could only have been triggered during the slow-roll phase of inflation if α S (k 0 ) is more than one standard deviation above its central value. If α S (k 0 ) is at the upper end of its present 2 σ range, PBHs with mass up to 5 × 10 16 g could have formed as result of density perturbations created at the end of slow-roll inflation. These results are again insensitive to the value of n S (k 0 ).
Summary and Conclusions
In this paper we have investigated the formation of primordial black holes in the radiationdominated era just after inflation. We have focused on density perturbations originating from the slow-roll phase of inflation. In section 2 we reviewed the Press-Schechter type formalism for PBH formation. We found that the formation of PBHs with mass larger than 10 15 g, which could form (part of) the cold Dark Matter in the Universe, the spectral index at scale k PBH should be at least 1.37, even for the lower value of 1/3 for the threshold δ th . This value is higher that the value 1.25 found in [27] because here we have assumed that the mass of the collapsed region to form PBH is only 20% of the entire energy density inside the particle horizon.
This spectral index is much above the value measured in the CMB. PBH formation therefore requires significant positive running of the spectral index when k is increased. In contrast, current observations favor a negative first derivative of the spectral index. For the central value n S (k 0 ) = 0.964, where k 0 is the CMB pivot scale, the first derivative α S (k 0 ) would need to exceed 0.020 if it alone were responsible for the required increase of the spectral index; this is outside the current 2 σ range for this quantity. However, the second derivative (the "running of the running") is currently only very weakly constrained. We showed in a model-independent analysis that this easily allows values of n(k PBH ) large enough for PBH formation, even if the first derivative of the spectral index is negative at CMB scales.
In section 3 we applied this formalism to the "running-mass" model of inflation, which had previously been shown to allow sufficiently large positive running of the spectral index to yield PBH formation. In contrast to previous analyses, we included a term quadratic in the logarithm of the field, i.e. we included the "running of the running" of the inflaton mass along with the "running of the running" of the spectral index. We showed that this model can accommodate a sizably positive second derivative of the spectral index at PBH scales. However, this is only possible if the first derivative is also positive and sizable. In fact, like most inflationary scenarios with a smooth potential [28] the model does not permit large negative running of the spectral index at CMB scales. Current data prefere negative running of the spectral index at CMB scales, but are consistent with sufficiently large positive running at the 2 σ level. Moreover, we saw that a quadratic (in ln k) extrapolation of the spectral index to PBH scales is not reliable, and therefore computed the spectral index exactly. Imposing several consistency conditions, we found that density perturbations that are sufficiently large to trigger PBH formation only occur for a very narrow region of parameter space. Among other things, the signs of the parameters of the inflaton potential must be chosen such that the potential has a local maximum, and the initial value of the field must be slightly (by typically less than one e-fold) below this maximum.
We emphasize that obtaining sufficiently large density perturbations at small scales is only a necessary condition for successful PBH formation. One major challenge of this model is that for parameters allowing PBH formation the spectral index keep increasing at yet smaller scales. Parameters that lead to the formation of many PBHs with mass around 10 15 g, which could form the Dark Matter in the universe, would predict the over-production of unstable PBHs, in conflict with data e.g. from the non-observation of black hole evaporation and (for yet smaller masses) Big Bang Nucleosynthesis. This problem seems quite generic for this mechanism. One way to solve it might be to abruptly cut off inflation just after the scales relevant for the formation of the desired PBHs leave the horizon, which could e.g. be achieved by triggering the waterfall field in hybrid inflation. However, this is somewhat in conflict with the use of the slow-roll formalism of structure formation, which is usually assumed to require a few more e-folds of inflation after the scales of interest left the horizon.Moreover, a sharpe end of inflation means that the visible universe only inflated by about 45 e-folds; this solves the horizlatness problems only if the reheat temperature was rather low [17].
We can therefore not state with confidence that formation of PBHs as Dark Matter candidates is possible in running-mass inflation; all we can say is that certain necessary conditions can be satisfied. Even that is possible only for a very limited range of parameters. This also means that constraints from the over-production of PBHs only rule out a small fraction of the otherwise allowed parameter space of this model. | 2011-04-08T08:04:25.000Z | 2011-02-11T00:00:00.000 | {
"year": 2011,
"sha1": "ea4a0c50ce0be62cfd3a43c642e89b69059b5421",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1102.2340",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "ea4a0c50ce0be62cfd3a43c642e89b69059b5421",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
59466754 | pes2o/s2orc | v3-fos-license | The Glycoprotein Growth Factor Progranulin Promotes Carcinogenesis and has Potential Value in Anti-cancer Therapy
Progranulin (PGRN) is a secreted glycoprotein growth factor with tumorigenic roles in a variety of tumors including, among others, breast, ovarian, prostate, bladder, and liver cancer. In some patients, for example with breast, ovarian or liver cancers, high PGRN expression in tumors correlated with a worse outcome. Studies using cell lines and animal models provide evidence that PGRN promotes tumor cell proliferation, migration and survival, and induces drug resistance. Increasing or decreasing PGRN production enhances or inhibits respectively the growth of PGRN-sensitive tumors in vivo. PGRN activity is associated with p44/42 mitogen-activated protein kinase as well as phosphatidylinositol 3-kinases signaling pathways. In addition, PGRN may stimulate the formation of the tumor stroma. As an extracellular regulator of tumorgenesis, PGRN is a potential therapeutic target and biomarker of prognosis in the treatment of various cancers.
Introduction
Growth factors, their receptors and downstream signaling proteins play a cardinal role in carcinogenesis [1]. In addition, many cancers depend upon hormones such as estrogen or androgens to support their growth. The importance of these extracellular signaling systems as targets in the development of anti-cancer drugs has long been recognized. Despite this progress, cancer continues to be one of the major causes of morbidity and mortality and many types of cancer remain plagued by a high loss of life [2]. The search for additional biological targets to expand the arsenal of potential anti-cancer weapons remains a priority. Here we will explore the hypothesis that progranulin (PGRN) is an extracellular regulator of tumor progression that presents novel opportunities as a therapeutic and prognostic target in the treatment of a variety of different cancers. The evidence will be reviewed (i): that PGRN is often over-produced in cancerous tissue and that this over-production correlates with disease progression, (ii): that experimentally increasing or decreasing the production or bioavailability of PGRN by cancer cells makes them respectively more or less tumorigenic in vivo, (iii): that PGRN stimulates mitosis, blocks apoptosis, including apoptosis due to anti-cancer drugs, and promotes invasion and that this occurs through signaling pathways that are well known to have oncogenic potential, (iv): that PGRN may have additional tumor promoting roles, for example by eliciting tumor stroma production.
Progranulin [3] is also known as Granulin-Epithelin Precursor [4], Proepithelin [5], PC-cell-derived growth factor (PCDGF) [6], Acrogranin [7] and Glycoprotein, 88kDa (GP88). Edman sequencing of a protein called epithelial transforming growth factor (TGFe) revealed a sequence almost identical to an internal portion of PGRN, suggesting that TGFe may be a biologically active fragment of PGRN [8]. PGRN is a secretable glycoprotein [9] consisting of tandem repeats of a 12 cysteine module called the granulin or epithelin domain [10][11][12]5,7]. The individual granulin/epithelin modules can be isolated from tissue extracts and urine as individual peptides of approximately 6 kDa [10,11,13], but whether these are biologically active in their own right or simple breakdown products is not fully resolved (See Figure 1). In this article the term PGRN (progranulin) will be used to mean the full length protein and is synonymous with Granulin/Epithelin Precursor, Proepithelin, PC-Cell-derived growth factor, acrogranin and epithelial Transforming Growth Factor. The 6 kDa peptidic fragments that correspond to individual granulin/epithelin modules will be called grn/ epi peptides. Following NCBI usage the gene will be referred to as GRN.
PGRN is a multifunctional protein that has been implicated in early embryonic development [14,15], bone development [16], inflammation [17][18][19] and wound repair [20]. Mutational loss of a single copy of the human GRN gene results in a form of early onset dementia, called frontotemporal lobar dementia that is characterized by neuronal atrophy in the frontal and temporal cerebral lobes [21,22]. Many of these proposed biological roles for PGRN are characterized by cell proliferation, improved survival in the face of an apoptotic challenge, and migration through an extracellular matrix. As will be discussed below it is through these activities that PGRN is thought to promote cancer growth and development.
progression free survival.
Breast cancer: Immunohistochemical analysis of paraffinembedded human breast cancer samples demonstrated that PGRN was overexpressed in 80% [23] and 79% [24] of invasive ductal carcinoma and half of invasive lobular carcinoma, whereas almost no detectable PGRN expression was observed in normal mammary epithelium and benign tumors. PGRN expression significantly correlated with the histological grade of invasive and ductal carcinomas in situ [23], and with the Ki-67 index of proliferation in all invasive carcinomas [23]. P53 was more commonly expressed in those invasive carcinomas with high PGRN expression [23,24]. There was no significant correlation between PGRN expression and the human epidermal growth factor receptor (HER) family member HER-2. PGRN expression was independent of estrogen and progesterone receptors (ER/PR) status in one study [23], whereas Song et al. [24] reported that PGRN positive staining was more common in ER/PR negative samples than that in ER/PR positive tumors. Elkabets et al. [25] reported that high level PGRN staining was significantly associated with breast tumor size, histological and molecular subtypes. High PGRN expression correlated negatively with the luminal A subtype which are generally low grade tumors that express ER [26], but was positively correlated with triple negative breast cancer (tumors with negative expression of ER, PR, and HER-2 [27]) and basal-like breast cancer subtypes [25]. These cancers are often difficult to treat, suggesting that a PGRN-based therapy could significantly supplement existing breast cancer treatment. PGRNpositive expression in cancer specimens correlated with significantly worse overall survival compared to those without PGRN-staining [25]. Recent work demonstrates that PGRN levels are elevated in the serum of patients with breast cancer [28]. In contrast to these studies however, comparing gene expression profiles between 10 tumors from node-positive patients who survived for 5 years after surgery and 10 tumors from those patients who died of breast cancer within 5 years, Asaka et al. [29] concluded that PGRN was significantly downregulated among the patients who died or show worse outcomes and proposed that downregulated expression of PGRN is a prognostic indicator for subdividing node-positive patients into finer groups with "good prognosis".
Ovarian cancer: Ovarian cancer is the deadliest gynecologic malignancy and around 60% of women with it will die from the cancer [30]. In one immunohistochemistry analysis on human primary ovarian carcinoma, effusions, and tissues of metastatic lesions, PGRN protein expression was observed in 95% of ovarian solid tumor specimens, and the expression was detected in all tissue compartments of the ovarian carcinoma, including carcinoma cells, reactive stromal cells, and tumor-associated endothelial cells [31]. In a statistical analysis of GRN gene expression in laser microdissected ovarian tumor epithelium, GRN expression was observed only in invasive ovarian cancer epithelium and was absent in tumors with low malignant potential (LMP) [32]. Among LMP lesions, PGRN protein expression was, however, occasionally detected in the stromal cells. Miyanishi et al. [33] also reported that immunohistochemical staining of PGRN in the epithelial lesion of ovarian cancer is significantly stronger than that in LMP. PGRN has been investigated as a potential prognostic marker for ovarian cancer. A significantly worse overall survival for patients with high PGRN mRNA expression in ovarian cancers was demonstrated [34]. High PGRN protein expression in ovarian effusion tumor cells correlated to a better overall survival, while elevated PGRN expression in the tumor stromal cells correlated with worse overall survival [31]. Han and et al. [35] studied the relationship between the expression of several serum prediction biomarkers for epithelial ovarian cancer among patients in complete clinical remission. Using receiver operating characteristic and area under the curve analyses to define optimal cut-off points, PGRN levels were significantly associated with worse progression-free survival and overall survival. Elevated PGRN levels at 3 months of clinical remission predicted progression at 18 months.
Uterine cancer: PGRN protein staining was observed in the cytoplasm of tumor cells and stromal cells in endometrial cancer samples [36]. In endometrial cancer, PGRN protein expression was not associated with poor overall survival or known biomarkers of prognosis, including stage and grade. Approximately two thirds of the cancers co-expressed PGRN and ER. Elevated PGRN levels were observed in uterine leiomyosarcomas by immunohistochemical analysis, and the PGRN protein expression positively correlated with histological grades of tumors [37]. Those findings suggested that PGRN might be used to become a specific diagnostic marker for uterus leiomyosarcomas.
Prostate cancer: High levels of PGRN immunostaining were observed in prostate cancer tumor cells both among prostatic intraepithelial neoplasia (PIN) specimens and invasive cancer specimens [38,39]. PGRN was absent or showed low level staining in prostatic glands of normal tissues. However, PGRN expression level has no significant correlation with pathological stage, Gleason score, the status of lymph node metastasis, extraprostatic extension, perineural invasion, surgical margins, and vascular invasion. The elevated levels of PGRN staining in PIN suggest that elevated PGRN production may occur in the earlier stages of prostate cancer development.
Bladder cancer : Monami et al. [40] found that PGRN expression was detectable in both bladder cancer cells and normal bladder urothelium, and suggested that PGRN may play important roles in The upper panel shows that progranulin is a glycoprotein that is composed of seven and a half repeats of the granulin/ epithelin module designated A to G. P represents paragranulin, a six cysteine half module. The conserved amino acids of the granulin/epithelin module are shown (where C is cysteine, G is glycine, T is threonine, D is aspartic acid, P is proline and H is histidine. Dots represent amino acid residues that are not highly conserved). The lower panel shows that progranulin can be converted to peptide fragments each of about 6kDa that correspond to the granulin/epithelin modules.
carcinogenesis as well as regulation of normal physiological activity. Significantly higher immunostaining of PGRN was observed in invasive bladder tumors as compared with normal bladder tissues [41]. Lovat et al. [41] analyzed PGRN mRNA expression in bladder cancer using microarray database and found that overexpression of PGRN was observed in primary bladder cancers. PGRN mRNA expression levels was higher in high-grade bladder cancer than that of low-grade bladder cancer, and overexpression of PGRN were observed in patients who died after 5 years of follow-up compared with those alive after 5 years of follow-up treatment [41,42]. PGRN levels in voided urine samples from bladder cancer patients revealed that PGRN level was significantly higher in patients with malignant lesions compared to healthy individuals [43].
Kidney cancer: Elevated PGRN levels were detected in renal cell carcinoma samples using Western and immunohistochemical analysis [44]. PGRN expression correlated positively with the histological grades of tumors.
Liver cancer: Overexpression of PGRN was observed in a majority of hepatocellular carcinoma (HCC) patients using immunohistochemical methods on liver tumor specimens [45,46]. Patients with high PGRN content in clinical specimens showed characteristics of aggressive HCC, such as large size of tumor and venous infiltration [46]. PGRN protein is commonly observed in the cytoplasm of HCC tumor cells [46,47]. Cheung et al. [46,48] also reported that strong expression level of PGRN is significantly associated with early recurrence after curative resection and suggested that overexpression of PGRN is a predictive biomarker of poor outcome for HCC.
Esophageal cancer: PGRN protein is highly expressed in both cytoplasm and nuclei of esophageal squamous cell carcinoma cells [49]. Positive staining was also observed in stroma, including sporadic interstitial cells, esophageal glands, and vascular endothelial cells. In well differentiated tumors, positive staining of PGRN is uneven, whereas uniformly strongly positive staining was found in poorly differentiated tumors. PGRN positive staining was almost un-detectable in normal esophageal endothelium. PGRN positive staining was correlated with tumor infiltration depth and TNM classification.
Gastric cancer: Serological identification of tumour antigens by recombinant expression cloning identified GRN gene as overexpressed in gastric cancer [50]. PGRN was highly expressed in gastric cancer specimens with 2.3-2.9 folds of difference compared with the adjacent non-cancerous tissues. Using immunohistochemical analysis, PGRN protein expression was almost absent in normal gastric tissues, however 88% of cells were PGRN-positive in gastric cancer biopsy specimens [51]. This may not be specific for neoplasia since increased mucosal PGRN staining was also observed in gastritis [51].
Laryngeal cancer: Using immunohistochemical staining on samples from primary laryngeal cancer and adult laryngeal papilloma as well as laryngeal leukoplakia, PGRN protein expression was found to be significantly higher in the cytoplasm of laryngeal squamous cell carcinomas (LSCC) than those of the other lesions [52]. Overexpression of PGRN mRNA levels was also observed in LSCC tissues.
Lung cancer: PGRN was investigated as a new candidate lung cancer biomarker by one large cohort study but was not found informative as a lung cancer biomarker [53].
Brain cancer: PGRN is consistently overexpressed in human glioblastomas tumors compared with normal brain samples [54]. Using microarrays, Martert et al. [55] investigated gene expression profiles between primary human glioblastoma multiforme (GBM) tumors and normal brain tissues, and found that PGRN is upregulated in GBM tumors. PGRN expression was detected in 36.7% of meningioma tumor samples using reverse transcription-polymerase chain reaction (RT-PCR) [56]. The average size (51.5±5.9 cm 3 ) of tumors with PGRN expression was significantly larger than the mean meningioma volume (24.9±2.8 cm 3 ) of PGRN-negative tumors. The mean area of peritumoral brain edema which is associated with malignant development, was significantly larger in PGRN positive tumors than that in tumors with absence of PGRN.
Myeloma and leukemia: PGRN mRNA and protein expression were observed in multiple myeloma (MM) cell lines, and the immunohistochemical analysis showed PGRN positive staining in bone marrow smears from MM patients but no detectable PGRN staining in bone marrow biopsy samples from patients in remission [57]. Larramendy and et al. [58] found that PGRN expression was significantly down-regulated in acute myeloid leukemia as assessed by cDNA microarray analysis.
Infectious agents, progranulin and cancer:
In south-east Asia, chronic infection with the liver fluke Opisthorchis viverrini (O. viverrini) is a major risk factor for bile ducts cancer or cholangiocarcinoma. Screening genes from the transcriptome [59] or the excretory/secretory proteome [60] of O. viverrini that are associated with cancer in humans identified a parasite homologue of human PGRN. PGRN was secreted by the liver flukes and released into bile ducts to stimulate proliferation of biliary epithelial cells to promote host cells to turnover with the potential to increase susceptibility to ductal carcinogenesis [59,60]. Patients infected by Helicobacter pylori (H. pylori) have a higher than average risk of developing gastric cancer [61]. H. pylori infection was found to induce overexpression of PGRN at both mRNA and protein levels in gastric epithelial cells [51] which may contribute to development of gastric cancers.
PGRN promotes tumorigensis in experimental models
Given that PGRN is often over-expressed in cancers, and that PGRN levels are associated with poor outcome, does PGRN have a functional role in carcinogenesis? Over production of PGRN frequently confers a more aggressive phenotype on cancer cells in vitro, as assessed by parameters such as anchorage-independent growth, improved survival or invasion assays (for examples, see refs [34,36,38,40,45,46,57,[62][63][64][65][66]). Increasing the production of PGRN in poorly tumorigenic cells often results in a more malignant phenotype following transplantation into mice. The over-production of PGRN in the estrogen-dependent breast cancer cell line MCF7, for example, results in increased tumorigenicity, estrogen-independence and resistance to the estrogen receptor blocking drug Tamoxifen [63,67]. Similarly, PGRN over-production confers greater tumorigenicity upon liver [46], ovarian [68] and endometrial cancer cells [36] when grown as xenografts in mice. SW13 cells, which derive from an adrenal carcinoma, are highly sensitive to the proliferative effects of PGRN [3,65]. The parental SW13 cells are poorly tumorogenic in vivo, however by over-producing PGRN they become highly tumorigenic in mice [3]. SW13 cells have defects in the p53 [69] and Rb [70] tumor-suppressor systems, and exhibit little detectable expression of cyclin-dependent kinase inhibitors such as p21 or p16 ink4a [69]. Presumably the loss of function of tumor suppressors such as p53 in SW13 cells removes the brakes on the cell cycle, while PGRN provides the acceleration that drives them towards a more malignant state.
Primary cultures of human cells are more difficult to transform by direct gene transfer than rodent cells [71]. Specifically, there is a requirement for genes that prevent senescence (immortalizing genes) to cooperate with genes that activate the RAS-mediated cell signaling pathway and thereby stimulate proliferation [72][73][74]. In particular, RAS-mediated activation of guanine nucleotide exchange factors for the Ral small GTPases appears critical in the transformation of human cells [75]. Very few genes can be successfully substituted for mutant RAS in these primary cell transformation assays. Examples inhuman ovarian surface epithelial (OSE) cells include the oncogenic epidermal growth factor receptor-family member HER-2 [76]. Importantly, PGRN is also able to substitute for oncogenic RAS in the transformation of immortalized human OSE cells [33] as well as in human uterine smooth muscles cells [37]. These primary cells [33,37], that were first immortalized by expression of telomerase (hTERT) and SV40 large-T antigen (SV40) and then forced to express the GRN gene at high levels, were strongly tumorigenic when transplanted into athymic mice. Immortalized cells expressing hTERT and SV40, but that did not over-express the GRN gene, were not tumorigenic [37,33]. Similarly the expression of the GRN gene on its own did not transform the primary cells. The expression of SV40 impedes inhibitory control mechanisms on the cell cycle exerted by the p53 and Rb systems, as well as inhibiting the protein phosphatase 2A [77] and prevents mitogenmediated senescence, while increased expression of the GRN gene appears to provide the thrust that drives the immortalized, but noncancerous cells towards a tumorigenic phenotype. The ability of PGRN to replace oncogenic RAS in transforming immortalized cells suggests that it is a highly oncogenic protein. This interpretation, that PGRN can transform immortalized (pre-cancerous) cells is supported in other systems including the non-transformed but immortal renal epithelial cell line, MDCK, which becomes highly anchorage-independent when engineered to produce elevated levels of PGRN [3].
Just as the over-expression of PGRN promotes a more proliferative and tumorigenic phenotype, the reduction of PGRN mRNA levels may reduce proliferation in vitro and tumor growth in vivo. In tissue culture, proliferation and anchorage-independence can be inhibited by targeting PGRN mRNA in breast cancer [78], SW13 adrenal carcinomal cells [3], MDCK kidney epithelial cells [3], laryngeal cancer cells [52], prostate cancer cells [38], ovarian cancer cells [32,79] and liver cancer cells [80]. PC cells, which are a highly tumorigenic murine teratoma-derived cell line, secrete a growth factor, called PC cell-derived growth factor that is identical in structure to murine PGRN [6]. Targeting PGRN mRNA levels in PC cells abolished their in vivo tumorigenicity [81]. Lowering the mRNA levels of PGRN in other tumorigenic cell lines, including the breast cancer line MDA-MB-468 [78], the laryngeal carcinomal cell line Hep-2 [52], and liver cancer cells [80] also results in profound inhibition of tumor growth in mice. Monoclonal antibodies against PGRN are effective at inhibiting tumor growth of transplanted liver cancer cells in nude mice, and work by targeting both the growth of the tumor cells directly, but also by suppressing tumor angiogenesis [45]. Taken together, these results suggest that progranulin is both sufficient to stimulate a more malignant phenotype in poorly tumorigenic or immortalized cells; that it is necessary for tumor growth in some aggressive cancers, and that it has promise as a molecular target in the development of novel anti-cancer therapies.
PGRN and the tumor stroma
It is well established that the non-transformed mesenchymal cells that surround the cancer and form the tumor stroma have a crucial role in the growth and metastasis of many cancer types [82]. In most healthy tissues, mesenchymal cells such as fibroblasts or endothelial cells express the GRN gene at low or negligible levels [83]. However, in the fibroblasts of the tumor stroma and tumor capillary endothelial cells, GRN gene expression may be abundant. This is well documented for ovarian cancers, where approximately half of the tumors examined displayed PGRN-staining in stromal fibroblasts and in two thirds of the tumors capillaries were PGRN positive [31]. This has important consequences since, as noted above the presence of ovarian stromal PGRN correlated with a poorer outcome [31]. In breast cancer cells PGRN stimulated increased production of the angiogenic proteins vascular endothelial growth factor (VEGF) [84,67] and angiopoietin [67]. Esophageal squamous cell carcinomas exhibited a positive correlation between PGRN and VEGF levels, with the levels of both proteins correlating with microvessel density [49]. Blockade of PGRN by monoclonal antibodies in HCC xenografts resulted in a decrease in tumor microvessel density, which was attributed to reduced production of VEGF [45]. PGRN may also have a direct effect on angiogenesis that is independent of VEGF production since it stimulated the proliferation and migration of endothelial cells in culture [20].
McAllister and colleagues have recently proposed a novel mechanism for tumor stroma formation where an initial robust breast cancer mass, the instigator, stimulates the formation of reactive tumor stroma in a second poorly-growing or indolent tumor located at a distant anatomical site [85,25]. The instigating tumors secrete the prometastatic protein osteopontin that activates the migration of a population of Sca1 + cKithematopoietic stem cells from the bone marrow to the quiescent tumor [85]. Upon taking up residence in the indolent tumor the Sca1 + cKitbone marrow cells secrete PGRN, which in turn activates the growth of a fibroblastic stroma around the quiescent cancer cells [25]. The cancer cells then proliferate under the influence of the recently formed stroma to create new tumor masses [25]. Many of the details of how PGRN contributes to the formation and activity of the stroma remain unclear, but in principle the PGRNsecreting Sca1 + cKitbone marrow cells might initiate the formation of the tumor stroma, while at a later stage intrinsic PGRN production by stromal fibroblasts may supplement or take over the role of the bone marrow cells in promoting and maintaining the growth of the tumor stroma.
PGRN, cell survival and drug resistance
PGRN is a putative survival factor for normal and cancer cells in vitro. This may contribute to the overall growth of the PGRN-sensitive tumors, and may complicate cancer therapy since PGRN appears to confer increased resistance to several classes of anti-cancer drugs. PGRN prevents anoikis in cancer cells [65], a form of apoptosis that occurs when cells detach from their basement membrane [86]. Immunoneutralization of PGRN in ovarian cancer cells results in enhanced apoptosis as assessed by increased caspase-3 activation and poly(ADP-ribose) polymerase cleavage [79]. PGRN inhibits metabolicstress apoptosis in non-transformed fibroblasts [87] and cultured neurons [88,89]. Recent work in C. elegans and with macrophage from Grn knockout mice revealed that in the absence of PGRN the rate of phagocytosis of apoptotic cells increased [90], suggesting that PGRN may inhibit apoptotic clearance of injured or diseased cells.
Endocrine therapy is a mainstay in the treatment of estrogen receptor (ER)-positive breast cancer, with selective estrogen receptor modulators such as tamoxifen [91] playing an essential role in the treatment of ER-positive breast cancer. Over-production of PGRN in estrogen receptor-positive MCF-7 cells induced tamoxifen-resistance [63,67]. The activation of mitogen-activated protein kinase (MAPK) signaling pathways in breast cancer cells by the over-production of growth factors or growth factor receptors promotes proliferation without requiring ER-mediated growth signaling [92]. This confers resistance to antiestrogens [93] thereby rendering antiestrogen therapy in effective [92]. In this regard, PGRN markedly increased MAPK activity in breast cancer cell lines [63]. Moreover, PGRN treatment in MCF-7 cells inhibited the tamoxifen-induced down-regulation of bcl-2 [94]. PGRN may in addition interfere with aromatase therapy since PGRN inhibited the efficacy of the aromatase inhibitor letrozole in a human breast cancer cell line [95]. PGRN expression level was significantly increased in letrozole resistant tumor cells compared with letrozole sensitive cancer cells [95]. Other hormone-based therapies may also be sensitive to PGRN production. For example, PGRN decreased apoptosis induced by the synthetic glucocorticoid dexamethasone (which is used in the therapeutic regimen of multiple myeoloma) in a dexamethasone-sensitive multiple myeloma cell line [96].
Ovarian cancer cells that constitutively over-produce PGRN were resistant to the platinum containing drug cisplatin [68] (which is used to treat ovarian cancer [97]). However, ovarian cancer stromal cells that were sampled after chemotherapy with platinum drugs tended to show reduced staining for PGRN compared to comparable tissues inspected before treatment [31] suggesting a level of complexity in the drug-PGRN relationship. The recurrence of liver cancer and chemoresistance are huge obstacles to provide curative treatment for patients [48,98,99]. In hepatic cancer stem cells [48], inhibiting the expression of GRN gene sensitized the HCC cells to chemotherapy [48], whereas elevated production of PGRN led to resistance to cisplatin and doxorubicin. In chemoresistant HCC cells, PGRN levels were positively associated with expression of the adenosine triphosphatedependent binding cassette ABCB5 which is a drug transporter in liver cancer cells [48]. The interaction between PGRN and ABCB5 may provide a potential mechanism by which PGRN confers drug resistance. Chemotherapeutic resistance in non-small cell lung cancer (NSCLC) increased with the expression of PGRN [100]. Hu and et al. [101,102] studied the correlation between GRN expression in serum or tumor samples and chemotherapeutic sensitivity in NSCLC and found significant higher expression of PGRN in chemoresistant patients than in chemosensitive patients. They concluded that PGRN expression was significantly associated with response of chemotherapy and PGRN might become a biomarker to evaluate chemotherapeutic sensitivity and predict medical prognosis among NSCLC patients.
The mechanism of action of PGRN
A functional receptor for PGRN has not been identified, although chemical cross-linking experiments show that PGRN [103], its constituent grn/epi peptides [104] and TGFe bind in a specific fashion to cell surface proteins [105]. Scatchard analyses indicated that both for PGRN [103] or grn/epi peptides [104] at least two classes of PGRN binding site exist, one of low affinity but high abundance together with other sites of low abundance but high affinity. PGRN binds with sortilin on cell surfaces which is likely to be part of a protein turnover mechanism [106]. PGRN also binds to and inhibits the TNF-receptors [106,107] and associates with the Toll-like receptor-9 [108]. These interactions are important in the regulation of inflammation. Other binding partners for PGRN include non-receptor extracellular matrix proteins such as perlecan [109] and cartilage oligomeric matrix protein [110]. The binding of PGRN to these extracellular matrix proteins modifies the biological action of PGRN either blunting or enhancing the combined proliferative effects of the PGRN-matrix protein pair [110].
PGRN stimulates the MAPK and phosphatidylinositol 3-kinase (PI3K) pathways in cancer cells, as well as in non-transformed fibroblasts and neurons [20,34,38,40,45,57,62,63,65,66,111,112]. Both the MAPK and PI3K signaling systems are essential for PGRN mediated cell division, survival and invasion [65]. Interestingly different cells may show different signaling responses to PGRN, for example, bladder cancer cells show clear activation of MAPK in response to PGRN but little or no PI3K response [38]. Significantly, the bladder cells respond well to the motility and invasive activity of PGRN but not to its proliferative actions [38]. PGRN signaling may interact with integrin signaling pathways through focal adhesion kinase (FAK) [20,65]. In bladder cells PGRN-stimulated the formation of intracellular complexes of MAPK, paxillin and FAK [40] thereby linking the ERK and FAK signaling machinery. FAK is essential for growth factor and integrin-regulated cellular motility (reviewed in [113]). In addition to promoting motility-related signal transduction events, PGRN stimulates the expression of matrix metalloproteinases (MMPs) in cancer cells, including MMP2, MMP9, MMP13 and MMP17 [65,84,114] which probably contributes to the migratory properties of cells stimulated by PGRN. Downstream effects of PGRN signaling include expression of proteins of the cell cycle such as cyclin D1 [62,63,114], and cyclin B [62], the phosphorylation of other signaling intermediates such as glycogen synthase kinase beta [111,112], and the activation of transcriptional regulators such as nuclear factor kappa-B in myeloma cells [96] and JunB in chondrocytes [16].
The activation of the MAPK and PI3K signaling systems are characteristic of all the classic growth factors, however differences between the signaling properties of PGRN and those of conventional growth factors have been postulated. Non-transformed fibroblasts require stimulation by two independent growth factors in order to complete the cell cycle under serum free conditions [115]. One of these signals is provided by an insulin-like growth factor (IGF), while the other signal may be provided by one of a number of growth factors including members of the platelet-derived growth factor (PDGF), fibroblast growth factor (FGF) or epidermal growth factor (EGF) families. Genetically deleting the receptor for IGF-1 prevents the mitotic activity of the IGFs on murine fibroblasts, but also inhibits all other growth conventional factors from stimulating proliferation [116]. PGRN, however, retains the ability to promote cell division in IGF-I receptor negative fibroblasts under serum free conditions, indeed it is the only extracellular protein known to do so [62,117]. It is thought that rather than any ability of PGRN to stimulate distinct signaling pathways it has this property because it stimulates a more prolonged activation of proteins in the MAPK and PI3K pathways than do classic growth factors such as PDGF or EGF [62]. Unlike their wild-type counterparts, IGF-I receptor negative fibroblasts cells (Rcells) are not transformed by most oncogenes [116], although they are transformed by constitutively active mutant G-protein G-alpha 13 [118], and become sensitive to the transformative activity of SV40 T as they age, possibly due to enhanced expression of the EGF-receptor HER-3 [119]. Whether PGRN signaling also interacts with G-alpha 13 or HER-3 is unknown. The ability for PGRN to promote proliferation in cells such as the Rcells, that are otherwise refractory to growth factor stimulation or the transformative effects of oncogenes, is worrisome in that it provides a putative pathway through which cells could escape the therapeutic effects of anti-cancer drugs that work by targeting growth factors and their receptors. A summary of PGRN's role and mechanism in cancer is given in Figure 2.
Although the focus in this article is on PGRN as a mitogen, a cell survival factor and a promoter of invasion, it may have many other biological effects, some of which could contribute to tumor growth. PGRN is anti-inflammatory [17][18][19], which might in principle effect the host's immune response to developing tumors. Further, PGRN may regulate protein turnover since the over-production of PGRN by HeLa cells, or treatment of the cells with PGRN-conditioned medium, stimulates the formation of more and larger lysosomes [120]. Consistent with these observations, treatments that promote lysosome formation, such as sucrose or the expression of the transcription factor EB, stimulate GRN gene expression [120]. PGRN is reported to interact with intracellular proteins such as cyclin-T [121][122][123] which may modulate transcription. It remains uncertain if these interactions contribute to carcinogenesis, but they clearly provide additional possibilities through which the expression of the GRN gene could modulate cell function.
Regulation of GRN gene expression
At present there is no evidence that the GRN gene is amplified during tumorigenesis, suggesting instead that the elevated expression of the GRN gene in cancer cells results from changes in its regulation rather than an increased gene copy number. Mechanisms for the regulation of the GRN gene include mRNA stability [124], microRNAs [125][126][127][128], the action of RNA binding proteins [129], signal transduction by MAPK signaling [79], nuclear hormone receptors [130], and by physiological parameters such as hypoxia [87,131].
GRN gene expression is stimulated by nuclear receptor hormones such as retinoic acid [124] and, in breast cancer cells [130], endometrial cancer cells [36] and the hypothalamus [132] by estrogen. Importantly, the increase in GRN expression that follows exposure to estrogen in breast cancer cells may set in motion a PGRN autocrine growth stimulatory loop [133] since the mitogenic activity of estrogen on MCF-7 breast cancer cells in culture is inhibited by the immunoneutralization of PGRN [63], while enhanced GRN expression in breast cancer cells enables them to form tumors independently of estrogen [67].
Mitogens other than estrogen may also stimulate GRN gene expression. The proliferative actions of endothelin and lysophospahtidic acid (LPA) on ovarian cells were blocked when PGRN was immunoneutralized, suggesting that PGRN acts as an autocrine signal for these mitogens [79]. Both endothelin and LPA promote the expression of the GRN gene [79] through the activation of MAPK by a protein kinase A, calcium and cyclicAMP-dependent mechanism [79]. The up-regulation of GRN gene expression by MAPK may be widespread since it is also reported in neuroblastoma cell lines [131] and in gastric mucosa [51] in response respectively to hypoxia or the presence of the gastritis causing bacterium H. pylori.
Differentiation agents such as retinoic acid and dimethylsulfoxide increase GRN mRNA expression in myelogenous leukemias [124]. This revealed for the first time the importance and complexity of PGRN mRNA stability in GRN gene regulation. In a progranulocytic cell line, for example, differentiation agents promoted faster GRN mRNA turnover, resulting in a rapid but transient increase in PGRN mRNA following stimulation. In contrast, in a promonocytic cell line identical treatments slowed the rate of PGRN mRNA turnover [124] leading to a slower and more prolonged elevation of the GRN mRNA level. Genetic studies on GRN in neurodegenerative disease proved that microRNAs (miRs), in particular miR-659 [125] and miR-107 [127] are critical negative regulators of PGRN mRNA levels. This extends to cancer cells, where miRs belonging to the miR-15/107 gene group were found to suppress PGRN mRNA levels in prostate cancer cells [128]. This may be functionally significant since in leukemic and prostate cancers low miR-15/107 correlates with high GRN expression [128]. GRN gene expression may be negatively regulated by the p53 tumor suppressor system. Restoring a functional p53 to the malignant glioma cell line LN-Z308 which has lost both p53 allelles, results in an decreased secretion of PGRN [134]. Against this however, in HCCs higher GRN expression correlated with higher levels of wild-type but not mutated p53 [47] suggesting a complex relationship between GRN gene expression and P53. Additional control of PGRN mRNA expression through RNAbinding proteins has recently been established. The dying neurons of patients with GRN gene mutations accumulate intracellular aggregates of a cleaved and ubiquitinated DNA-and-RNA binding protein called the TAR DNA Binding protein or TDP-43 [21,22]. The functional relationship between PGRN and TDP-43 in neurodegenerative disease is not well understood, but in normal brain tissue TDP-43 binds to and decreases PGRN mRNA levels [129]. Whether disruptions in the binding of TDP-43 (or similar RNA-binding proteins) to PGRN mRNA have a role in carcinogenesis is unknown.
The proteolysis of secreted PGRN provides an additional control over PGRN levels. PGRN is digested by matrix metalloproteinases (MMPs) including MMP-9 and MMP-14 [135,136] as well as ADAMTS-7 (a member of the "A Disintegrin And Metalloproteinase with Thrombospondin Motifs" gene family) [137]. During inflammation neutrophil-derived enzymes such as elastase and proteinase-3 digest PGRN down to its constituent 6 kDa grn/epi peptides, some of which have biological activity in, for example, the regulation of inflammation [17,138]. This proteolysis is prevented by the secretory leukocyte protease inhibitor (SLPI) which, in addition to inhibiting elastase enzyme activity, physically binds to PGRN and protects it against enzymatic cleavage [17]. Intriguingly there is strong evidence that SLPI and PGRN act in concert to promote ovarian tumor cell mitosis and survival [139,140]. Other proteins in addition to SLPI For details see ref [3]. may protect PGRN from enzymatic degradation, including extracellular matrix proteins such as cartilage oligomeric matrix protein [141], and serum proteins such as high density lipoprotein [142]. The turnover of PGRN at the cell surface is controlled by Sortilin [106,107]. Sortilin was originally identified as a regulator of lysosomal enzyme trafficking [143] and is critical in controlling extracellular PGRN levels in neuronal cells [106,107,144]. Whether it has comparable functions in cancer cells is at present unknown. Taken together, the levels and activity of PGRN in vivo are likely to depend first on the intra-and extracellular factors that regulate PGRN mRNA levels, followed by complex extracellular interactions between PGRN, proteolytic enzymes, and stabilizing proteins such as SLPI that inhibit the proteolysis of PGRN.
Conclusion
PGRN levels are often highly elevated in tumors at many anatomical sites compared to the equivalent normal tissue. PGRN may therefore have potential as a biomarker for disease outcome. PGRN acts as a mitogen, a cell survival factor and a promoter of invasion for a variety of cancer cells. PGRN may have other biological effects that contribute to tumor growth, including the formation of the tumor stroma, the induction of drug resistance, and anti-inflammatory actions. Given the frequency with which PGRN expression occurs in cancers and its tumorigenic biological actions, there is a strong likelihood that studies of the cellular and molecular mechanisms of PGRN action in tumor formation will supply innovative strategies for the development of novel anti-cancer therapies. | 2019-03-31T13:45:03.976Z | 2012-01-01T00:00:00.000 | {
"year": 2012,
"sha1": "f456442e788ab3e80fd9c52d9f1109c3415b4f70",
"oa_license": "CCBY",
"oa_url": "https://www.omicsonline.org/the-glycoprotein-growth-factor-progranulin-promotes-carcinogenesis-and-has-potential-value-in-anti-cancer-therapy-2157-2518.S2-001.pdf",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "2615b047659fe9b76a8023645eef32c3444db48f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Biology"
]
} |
55197423 | pes2o/s2orc | v3-fos-license | An unusual cause of haemorrhage in the subthalamic nucleus of luys: A post-mortem study with 7.0 tesla magnetic resonance imaging correlates
The present case report is the first one of an isolated haemorrhage in the subthalamic nucleus of a patient with an isolated central nervous system relapse of systemic non-Hodgkin lymphoma. The neuropathological findings and 7.0-tesla magnetic resonance imaging correlates are presented.
INTRODUCTION
A cerebrovascular accident, especially a haemorrhage is the most common lesion affecting the subthalamic nucleus of Luys. Less usual causes are a metastasis, a demyelinising process, a head injury, a tuberculoma or as a complication of stereotactic interventions. [1][2][3] We present the case history and the neuropathological findings in a patient with as incidental finding at post-mortem examination a haemorrhage in the right subthalamic nucleus of Luys of unusual origin. Also the post-mortem 7.0-tesla magnetic resonance imaging (MRI) are described.
CASE HISTORY
A 61-year-old man was admitted in January 2009 for fever of unknown origin. Three hepatic lesions together with abdominal adenopathies were observed on computed tomography. Blood analysis revealed signs of inflammation. The liver tests were normal. No improvement of the fever was obtained with antibiotics and steroids. Liver biopsies, performed in February 2009, revealed lymphomas possibly of the Hodgkin type. Remission was obtained by chemotherapy in August 2009. In October 2010 the patient developed generalized seizures. The examination did not show focal neurological signs. On MRI of the brain, however, three infiltrating lesions were observed: one in the right cerebellar hemisphere, one in the right parietal subcortical region and one in the upper brain stem. No haemorrhages were detected at that time. Two stereotactic cerebral biopsies were not able to reveal any tumour invasion. The patient deceased five months later.
MATERIAL AND METHODS
Post-mortem examination was obtained by written informed consent from the nearest family. The brain tissue samples were first used for diagnosis and afterward integrated in the Lille Neuro-Bank, dependant from the Lille University and co-federated by the "Centre des Resources Biologiques", acting as institutional review board.
In addition to the gross examination of the brain, samples for microscopically examination were stained with haematoxylin eosin, luxol fast blue and Perls Prussian blue.
Three coronal sections of a cerebral hemisphere were submitted to T2 and T2* MRI: a frontal one at the level of the head of the caudate nucleus, a central one at the level of the mammillary body and one at the level of the parietal and occipital lobes. In addition one horizontal section of brainstem and cerebellar hemisphere was also examined. A 7.0-tesla MRI Bruker BioSpin SA with an issuer-receiver cylinder coil of 72 mm inner diameter (Ettlingen, Germany) was used, according to a previous described method. [4]
RESULTS
The general autopsy showed bronchopneumonia and many necrotic and old haemorrhagic lesions in the liver. Not tumour cells were detected Gross examination of the brain showed a haemorrhage restricted to the right subthalamic nucleus of Luys. No other lesions were observed on naked eye examination (see Figure 1A). On T2 and T2* weighted MRI images a diffuse hypo-intensity of the basal ganglia was observed in addition to the subthalamic bleed (see Figure 1B-1C)). On microscopic examination siderophages and residual haemosiderin deposits were observed in the subthalamic nucleus of Luys (see Figure 2A). Polymorphic infiltrates of large lymphoid cells with many mitoses were observed in the right basal ganglia and subcortical structures, also invading the upper brainstem. There was also tumoural infiltration of the blood vessel walls. Although some large cells resembled to Sternberg cells the tumour was considered as a non-Hodgkin lymphoma (see Figure 2B).
DISCUSSION
The overall incidence of intracranial haemorrhage among adult patients with haematological malignancy is 2.8%. It is more common among patients with central nervous system involvement by a lymphoma with a cortical predilection site of 83%. [5] However, isolated central nervous relapse involving the brain parenchyma is a rare complication of systemic non-Hodgkin lymphoma. [6] The present case report is the first one of a unique haemorrhage in the subthalamic nucleus related to isolated central nervous system relapse of systemic non-Hodgkin lymphoma. | 2019-03-16T13:13:23.497Z | 2017-03-21T00:00:00.000 | {
"year": 2017,
"sha1": "8b234d03f19b34efbbfac6494cda59cfb72314f7",
"oa_license": null,
"oa_url": "http://www.sciedu.ca/journal/index.php/ijdi/article/download/10959/6905",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "da26a6bf2dedb82cc19cdcf84c39cb4c9beb3d17",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
11689683 | pes2o/s2orc | v3-fos-license | Classical and quantum geometrodynamics of 2d vacuum dilatonic black holes
We perform a canonical analysis of the system of 2d vacuum dilatonic black holes. Our basic variables are closely tied to the spacetime geometry and we do not make the field redefinitions which have been made by other authors. We present a careful discssion of asymptotics in this canonical formalism. Canonical transformations are made to variables which (on shell) have a clear spacetime significance. We are able to deduce the location of the horizon on the spatial slice (on shell) from the vanishing of a combination of canonical data. The constraints dramatically simplify in terms of the new canonical variables and quantization is easy. The physical interpretation of the variable conjugate to the ADM mass is clarified. This work closely parallels that done by Kucha{\v r} for the vacuum Schwarzschild black holes and is a starting point for a similar analysis, now in progress, for the case of a massless scalar field conformally coupled to a 2d dilatonic black hole.
I Introduction
In this paper we clarify aspects of the canonical description of vacuum 1+1 dimensional dilatonic black holes. Our analysis closely mirrors that of [1], in which vacuum Schwarzschild black holes were studied. The motivation for this work stems from our interest in the quantization of 4-d systems corresponding to spherically symmetric collapse. More specifically, we would like to study quantum aspects of spherically symmetric collapse of a massless scalar field in general relativity in 4 spacetime dimensions. This is a difficult task because the classical field equations are not exactly solvable even though the system is effectively 2 dimensional. However, the CGHS [2] model of 2d dilaton gravity with conformally coupled scalar fields is classically exactly solvable and we hope that the model can be quantized nonperturbatively. Since the 4-d system of interest is effectively 2-d we hope to gain insights into nonperturbative quantization of the 4-d case from a study of the nonperturbative quantization of the CGHS model (Note that most quantization efforts for the case in which matter is present, with a few exceptions such as the work of [3,4], have been perturbative in character).
The first step in such a study is to present a clear analysis of the classical and quantum theory of vacuum dilatonic black holes. Bearing in mind our motivations, we would like to cast the analysis in a framework which emphasizes the similarities of this system with the vacuum Schwarzschild black holes. Both the classical and quantum theory of the latter has been clearly analysed in [1] in a canonical framework. In this paper we show that the 2d vacuum dilatonic black holes can be handled using exactly the same approach, which worked for the Schwarzschild case in [1].
In this work, we perform a canonical transformation to new canonical pairs of variables. One of these pairs consist of the mass of the spacetime and the spatial rate of change of the Killing time. An additional canonical transformation results in the Killing time itself, being a canonical variable. The vanishing of the constraints are shown to be equivalent to, modulo some subtelities, the vanishing of two of the new canonical momenta. The true degrees of freedom are parametrised by a canonical pair, one of which is the mass of the spacetime. It's conjugate variable has a clear spacetime interpretation. Quantization of this description is almost trivial because of the simplified form of the constraints. In particular, in contrast to [5], we can find a Hilbert space representation (with an appropriate measure) for the physical operators of the theory.
This work, as we see it, has the following merits. First, it clarifies the physical interpretation of the observables of the theory (note that the observables were also given a physical interpretation in [5]; we go a little further than [5] in that we discuss parametrisation of the times at infinity). Second, the initial choice of variables used in this work is closely related to the spacetime geometry of the black hole and the whole treatment possesses a very close similarity to that in [1] which deals with the vacuum Schwarzschild case. Third, it can be viewed as a prerequisite for a similar treatment of the more interesting conformally coupled matter case.
Along the way, we give a self consistent treatment of asymptotics in the canonical framework (note the analysis in [6] is inconsistent due to a subtle technical reasonthis will be pointed out in Section 4).
The layout of the paper is as follows. In section 2, we review the global structure and the spacetime solution in the conformal gauge, of the vacuum dilatonic black holes. In section 3, we perform the canonical analysis of the action, in analogy to the ADM analysis done in 3+1 dimensions. Our canonical variables are closely tied to the geometry of the black hole spacetime and we do not make the field redefinitions of [3].
In section 4, we give a careful treatment of asymptotics in the canonical framework and identify the total energy of the system with the generator of time translations at spatial infinity. In the rest of the paper we closely mimic the treatment in [1].
The idea in [1] was to use as canonical variables, quantities which were physically significant. One hoped that the constraints of the theory simplified when written in terms of these quantities. The latter were identified by comparing the spacetime line element written in geometrically preferred coordinates with the ADM line element.
Thus, in section 5, we express the mass and spatial rate of change of Killing time in terms of canonical data. In section 6 we use these quantities as canonical variables and see that the constraints simplify when written in terms of these variables. In section 7, we bring the 'times at infinity' into the canonical framework exactly as in [1]. Finally, in section 8, we give a brief description of the quantum theory based on the classical decsription of section7. In section 9 we describe the classical theory by using light cone coordinates as canonical variables and quantize this description.
We have not attempted to review the enormous amount of pertinent literature and we refer the reader to review articles such as [7].
II The spacetime solution
Disregarding boundary terms, the action we deal with is that of [7]: where R is the scalar curvature of the 2 metric g ab , φ is the dilaton field and K is the The solution to the field equations in the conformal gauge is [7] 1 The ranges of (U, V ) are such that 2M K − K 2 UV ≥ 0, with the curvature singularity along the curves 2M K − K 2 UV = 0. Note that (U, V ) here are like the null Kruskal coordinates (U s , V s ) for Schwarzschild.
In the latter, the curvature singularity is at U s V s = 1. Here K provides an extra scale, which allows for the definition of dimensional Kruskal like coordinates, in contrast with the dimensionless Kruskal coordinates for Schwarzschild.
The global structure of the vacuum dilatonic black hole is identical to that of the fully extended Schwarzschild solution. We label the different parts of the spacetime as follows: Region I with U < 0, V > 0 (the right static region), Region II with U > 0, V > 0 (the future dynamical region), Region III with U > 0, V < 0 (the left static region) and Region IV with U < 0, V < 0 ( the past dynamical region). The horizons are at U = 0 and V = 0. We can define (T, ρ) coordinates (the analog of the Killing time and Regge-Wheeler tortoise coordinates for Schwarzschild) in regions I and III as follows: In each of the regions I and III, the line element and the coordinate ranges are and the horizons are at ρ → −∞ with spatial infinity at ρ → ∞. The metric is manifestly asymptotically flat at left and right spatial infinities in these coordinates.
III The canonical form of the hypersurface action
We apply the standard ADM formalism of general relativity in 3+1 dimensions to the 2-d dilatonic black hole spacetimes. The spacetime is foliated by a 1 parameter family of slices 'Σ', the slices being labelled by a time parameter 't', tǫR. Each t =constant slice, Σ, has the topology of R and is coordinatized by a parameter 'r'. Further, each slice is spacelike and extends from left spatial infinity to right spatial infinity, but is otherwise arbitrary (in particular, the slices are not restricted to pass through the bifurcation point (U, V ) = (0, 0) ).
The range of r is −∞ < r < ∞ with left and right spatial infinities being approached as r → −∞ and r → ∞ respectively. The ADM form of the spacetime line element is Here, Λ 2 (dr) 2 is the induced spatial metric on Σ and N and N r are the usual lapse and shift parameters. We substitute this form of the spacetime metric into the action S D and obtain, modulo boundary terms where dots and primes denote derivatives with respect to t and r respectively and we have defined R := e −φ to facilitate comparision with [1]. We shall supplement the above hypersurface action with the appropriate boundary terms in section 4.
The next step in the canonical analysis is to identify the momenta conjugate to the variables R, Λ (N, N a will turn out to be Lagrange multipliers). The momenta are The canonical form of the hypersurface action is where H and H r are the Hamiltonian and diffeomorphism constraints of the theory. N, N r are Lagrangian multipiers. The symplectic structure can be read off from the action and the only non vanishing Poisson brackets are It is straightforward to check that with these Poisson brackets the constraints are first class. It is also easy to see that H r integrated against N r generates spatial diffeomorphisms of the canonical data.
IV Asymptotics
A satisfactory treatment of asymptotics in the context of a Hamiltonian formalism must achieve the following: (i) The choice of asymptotic conditions on the fields must be such that canonical data corresponding to classical solutions of interest are admitted.
(ii)Under evolution or under flows generated by the constraints of the theory, the canonical data should remain in the phase space. For example, the smoothness of the functional derivatives of the generating functions for such flows must be of the same type as the data themselves. Also, the boundary conditions on the data must be preserved by such flows. The latter requirement may be violated in subtle ways and the formalism must be carefully checked to see that this does not happen.
In what follows we display our choice of boundary conditions on the phase space variables. To impose (ii), above, we need to deal with appropriate functionals on the phase space. These are obtained by integrating the local expressions for H and H r against the multipliers N and N r to obtain C(N) := ∞ −∞ drNH and C(N r ) := ∞ −∞ drN r H r . Imposition of (ii) gives boundary conditions on the N and N r . Next, we identify the generator of unit time translations at spatial infinity and show that it is M on shell( M is the parameter in the spacetime metric). Finally, we point out why the analysis in [6] is incomplete.
Since we use the same techniques as in the analysis for 3+1 canonical gravity, we shall not discuss them in detail.
IV.1 Boundary conditions on phase space variables
At right spatial infinity (r → ∞) we impose: Here α + and β + are arbitary (in general time dependent) paramters independent of r .
At left spatial infinity (r → −∞) we impose exactly the same conditions except that r is replaced by −r and the 'right' parameters α + and β + are replaced by the 'left' parameters denoted by α − and β − . Note that, on solution,
IV.2 The diffeomorphism constraint functional
We require that (a)C(N r ) exists. From the boundary conditions (16)-(21), above Thus, for C(N r ) to be finite, the asymptotic behaviour of N r should be The orbits generated by C(N r ) should lie within the phase space. Consideṙ But for R(t, x) to satisfy the boundary condition (16), asymptoticallẏ This is stronger than (23). For N r satisfying (25), it can be verified that C(N r ) generates orbits which preserve the boundary conditions on the canonical variables and is functionally differentiable without the addition of a surface term. Since asymptotic translations of r do not preserve the the boundary conditions on R, there are no generators of such translations in the theory (if there were, we would identify them with spatial momenta).
IV.3 The Hamiltonian constraint functional and the mass
We require that C(N) exists. Using the boundary conditions on the canonical variables, we find that H ∼ e −2K|r| near the spatial infinities. For each of the momentum dependent terms in H this is obvious. The remaining terms, individually, do not have this behaviour; but taken together, conspire to fall off as e −2K|r| . So, for C(N) to exist, it is possible to admit where N + , N − are independent of r.
Next, we require that C(N) be functionally differentiable. The variation of C(N) is of the form: Here Ξ R , Ξ P R , Ξ Λ , Ξ P Λ are local expressions involving the canonical variables and the lapse and their derivatives. The term which obstructs functional differentiability is δC(N) surface and is given by where δC(N) surface vanishes if N → 0 as |r| → ∞ and C(N) is functionally differentiable for such N.
To complete the analysis for the Hamiltonian constraint functional, we must check that the boundary conditions are preserved by the flows generated by C(N). We havė For the boundary conditions to be satisfied, we must requireṖ Λ ∼ O(1) asymptotically. From the equation forṖ Λ and the boundary conditions on the canonical variables, it can be verified that this requirement implies This fixes N ∼ e −2K|r| asymptotically. It can be checked that for such N , C(N) exists, is functionally differentiable and generates flows which preserve the boundary conditions.
Thus, just as in canonical gravity in 3+1 dimensions, the constraints C(N), C(N r ) generate motions which are trivial at infinity. We have seen that the spatial momentum vanishes because constant spatial translations at infinity do not preserve the asymptotic conditions. We now turn our attention to constant time translations at the spatial infinities and identify the generators of such motions with the ADM masses.
Hence, we must impose (26) and (27). (32) still holds since we want the boundary conditions on P Λ to be preserved under evolution. Thus where N + , N − are independent of r. C(N) is not functionally differentiable due to the presence of δC(N) surface . This term, for N with the above behaviour is To restore functional differentiability, we add an appropriate term to C(N) whose variation cancels δC(N) surface . We call the resultant expression, (which is non zero on the constraint surface) the true Hamiltonian H T . It is given by H T generates time translations which are constant at right and left spatial infinities and we can identify and The analysis in [6] is incomplete for the following reason. In [6] the boundary conditions are a little stronger than ours -in particular, the work there assumes α = β (we have suppressed the + and − subscripts). However the true Hamiltonian does not necessarily preserve α = β. This can be seen by looking at the evolution of R and Λ:Ṙ Our boundary conditions, on the other hand, suffer no such deficiency.
V Reconstruction of mass and Killing time rate from canonical data
We would like to construct, out of the canonical data, expressions which on shell, give the mass, M and spatial rate of change of Killing time, T ′ , along Σ. We guess the correct expressions, just as in [1], by comparing the ADM form of the line element (8) with that corresponding to the spacetime solution (7). So we parmeterize T = T (t, r) and R = R(t, r) and put them into (8). Comparision with (7) gives: Note that R 2 = 2M K + e (2Kρ) on solution (47) (To understand the reasons for our choice of signs for the square root in the expression for N see [1]). Substituting this in the expression for P Λ (10) we get an expression for the Killing time rate Finally, making use of equation (43) for Λ 2 , we get an expression for the mass Thus we have an expression for the mass as a function of the canonical data and we can substitute this expression for M in the above expression for T ′ to get the Killing time rate also as function of the canonical data. Note the similarity of the expressions with those in [1]. Finally, from the boundary conditions on the canonical data (16)-(21), we can infer the asymptotic behaviour of our expression for the mass.
We obtain This prompts the definition Since neither M(x) nor P M (x) contain P R , they commute under Poisson brackets with R. However (M, P M ; R, P R ) are not a canonical chart and we need to replace P R with an appropriate variable to have a canonical chart on phase space. We use the same trick as in [1] to guess the new momentum conjugate to R (which we shall refer to as Π R ). We expect Π R to be a density of weight one and require the diffeomorphism constraint to generate diffeomorphisms on the new canonical variables. This prompts the definition Long and straightforward calculations show that Thus, we make a canonical transformation from (Λ, P Λ ; R, P R ) to (M, P M ; R, Π R ).
We can equally well express the old variables in terms of the new.
Here we have defined
VI.2 The constraints
Before writing the constraints in terms of the new canonical variables, it is instructive to show that M(x) is a constant of motion. From (49) and (13,14) it is easy to check Thus, as expected, the mass function doesn't change over the slice. It is also easy to check that as well as (59), (60) and (58) show that M(x) is indeed a constant of motion.
We now proceed to write the constraints in terms of the new canonical variables.
From the expressions for Π R ,(53), and M ′ ,(58), it can be shown that Thus, the vanishing of H, H r is equivalent to the vanishing of M ′ , Π R modulo the vanishing of f . Again, arguing the same way as in [1], we can replace, as constraints, the former with the latter everywhere on Σ (in [1] the argument for replacing the old constraints with the new even when f =0 is essentially one of continuity). We can express the old set of constraints in terms of the new ones (61,62) in matrix notation i.e.: where A is a 2 × 2 matrix whose field dependent coefficients can be read of from (61,62) above. It is curious that even though the individual elements of A maybe ill defined when f = 0, its determinant is independent of f . In fact and is non vanishing as long as R, Λ = 0. Since Λ 2 is the spatial metric, Λ is non zero and R ≥ 0 by virtue of it's definition in terms of the dilaton field, vanishing on shell only at the singularity. So, if Σ is away from the singularity, the behaviour of DetA supports the replacement of the old constraints with the new.
VI.3 The action
Written in terms of the new canonical variables, the hypersurface action is Replacing the old constraints with the new ones gives where N M and N R are new Lagrange multipliers related to the old ones by From (40) , the boundary action S ∂Σ is given by where (from (37,38) and (50)) The total action is just We now a raise a subtle issue (whose analog has gone unnoticed in
VII The parametrized action
Since the form of the action (and even the notation, with the exception of the symbol for the momentum conjugate to R) is identical to that in [1], one can simply follow the discussion of sections VII to IX of that paper. We will be extremely succint in
VIII Quantum theory
We briefly review Dirac quantization of the classical description which follows from the parameterized action (for the quantization following from the unparameterized action we refer the reader to [1]).
Since R(r), m ≥ 0, it is better to make a point transformation on the classical phase space before we quantize. We define and replace the Π R = 0 constraint by Π ξ = 0. To pass to quantum theory, we choose a coordinate representation. Thus Ψ = Ψ(x; T, ξ] (Ψ denotes the wavefunction). The quantum constraints areP So the nontrivial operators in the theory are constructed fromx andp x = ∂ ∂x and the Hilbert space consists of square integrable functions of x on the real line.
IX.1 Classical theory
We now pass to a description in terms of canonical variables which correspond to (on shell) the null 'Kruskal' coordinates U, V of section 2. Recall, from the spacetime solution, that We start with the description in terms of the parametrized action S[m, p, M, P M , R, Π R , N M , N R ].
We define new canonical variables Motivated by (83) above we make a further canonical transformation (withm,p unchanged) to new variables (U, P U , V, P V ): Away from U = 0 or V = 0 (which correspond to the horizon on shell), we can replace the constraints Π R = 0 = P T by If we impose that P U and P V are continuous functions of r on the constraint surface, they must vanish even on the horizons. Then the parametrized action becomes
IX.2 Quantum theory
We describe the results of a Dirac quantization of the theory described by the action above. The fact that m ≥ 0, is handled by usingx := lnm and px :=mp (see section The Hilbert space consists of square integrable functions ofx and we have a quantum theory identical to that in section 8.
X Conclusions
In this work, we have formulated the canonical description of vacuum 2-d black holes in terms of variables which have a clear spacetime interpretation. In doing this we have followed the receipes of [1] which dealt with vacuum Schwarzschild black holes. As in [1], the classical description simplifies to such an extent that quantization becomes very easy.
When we replace the original classical constraints with new ones, care has to be taken where the horizon intersects the spatial slice. In particular, the new constraint functions are assumed to be continuous on the spatial slice and this forces them to vanish even at the horizon.
Because the classical variables used have a clear physical meaning, it becomes easier to interpret the corresponding quantum operators. We have also constructed, explicitly, a classical description (and quantized it) using the light cone coordinates as canonical variables. In [1] the analogs of these objects were the light cone Kruskal coordinates and using them explicitly was a little involved because the Regge -Wheeler tortoise coordinate plays a key role in the transformation from curvature coordinates to Kruskal coordinates.
We [8] are trying to apply methods similar to those used in this work and in the Schwarzschild case, to the 2d black holes with conformally coupled matter.
Ultimately, we hope to learn useful lessons from this work and [8], and efforts of other workers like [3,4] which will help in tackling the system of spherically symmetric (4-d) gravity with a (spherically symmetric) scalar field. | 2014-10-01T00:00:00.000Z | 1995-08-16T00:00:00.000 | {
"year": 1995,
"sha1": "9a64d229332a3bcc2ac96c35b8da5aa79c6acbdd",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/gr-qc/9508039",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "4a895b4b7f234b66d7d179f5a917b27393c13c82",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Medicine"
]
} |
229447127 | pes2o/s2orc | v3-fos-license | Psychosocial Obstacles to Hepatitis C Treatment Initiation Among Patients in Care: A Hitch in the Cascade of Cure
There are limited data examining the relationship between psychosocial factors and receipt of direct‐acting antiviral (DAA) treatment among patients with hepatitis C in large health care organizations in the United States. We therefore sought to determine whether such factors were associated with DAA initiation. We analyzed data from an extensive psychological, behavioral, and social survey (that incorporated several health‐related quality of life assessments) coupled with clinical data from electronic health records of patients with hepatitis C enrolled at four health care organizations during 2017‐2018. Of 2,681 patients invited, 1,051 (39.2%) responded to the survey; of 894 respondents eligible for analysis, 690 (77.2%) initiated DAAs. Mean follow‐up among respondents was 9.2 years. Compared with DAA recipients, nonrecipients had significantly poorer standardized scores for depression, anxiety, and life‐related stressors as well as poorer scores related to physical and mental function. Lower odds of DAA initiation in multivariable analysis (adjusted by age, race, sex, study site, payment provider, cirrhosis status, comorbidity status, and duration of follow‐up) included Black race (adjusted odds ratio [aOR], 0.59 vs. White race), perceived difficulty getting medical care in the preceding year (aOR, 0.48 vs. no difficulty), recent injection drug use (aOR, 0.11 vs. none), alcohol use disorder (aOR, 0.58 vs. no alcohol use disorder), severe depression (aOR, 0.42 vs. no depression), recent homelessness (aOR, 0.36 vs. no homelessness), and recent incarceration (aOR, 0.34 vs. no incarceration). Conclusion: In addition to racial differences, compared with respondents who initiated DAAs, those who did not were more likely to have several psychological, behavioral, and social impairments. Psychosocial barriers to DAA initiation among patients in care should also be addressed to reduce hepatitis C‐related morbidity and mortality.
Psychosocial Obstacles to Hepatitis C Treatment Initiation Among Patients in
P ublic health prevention and control of hepatitis C entails identification of infected persons through universal testing, followed by linkage to clinical services and ensuring that persons linked to care receive and complete effective treatment. (1)(2)(3)(4)(5) On an individual level, eradication of hepatitis C virus (HCV) infection eliminates the possibility of HCV transmission to others and reduces morbidity and mortality from HCVattributable hepatic and extrahepatic disease among persons who no longer engage in transmission-associated behaviors. (4) Despite radical improvement in the treatment of HCV infection since the release of secondgeneration direct-acting antiviral (DAA) medications, uptake of these drugs in many U.S. general health care systems has been low. (6)(7)(8) Investigators have identified a variety of sociodemographic and clinical factors associated with reduced likelihood of treatment, including younger age, non-White race, Medicaid coverage, lower annual income, lesser degrees of liver fibrosis, and ongoing or recent drug/alcohol use or mental health disorders. (6)(7)(8)(9)(10)(11) However, even in settings in which individuals infected with HCV have largely unrestricted access to DAAs, uptake can be suboptimal and nonsustained. (9,12,13) Under these conditions, studies have identified patient factors related to reduced uptake, such as skepticism about treatment effectiveness and tolerability, limited engagement and negative experiences with providers and health care systems, a lack of perceived urgency for treatment, and competing situational priorities and demands. (12)(13)(14)(15) Few studies have directly examined, using validated psychometric instruments, the psychosocial impediments to treatment initiation among identified patients with hepatitis C who receive integrated clinical care in large U.S. health care organizations. A recent study that classified reasons for DAA noninitiation based on clinical record review found that psychosocial issues were the principal barriers to hepatitis C treatment in an urban academic medical practice. (16) In an earlier study in the pre-DAA era, we found a high degree of physical and psychological impairment among patients in our hepatitis C cohort based on responses to a survey that incorporated several quality of life measures as well as information on employment status, drug/ alcohol/tobacco use, recent psychological stressors, and levels of social support. (17) To examine whether these factors were associated with receipt of treatment in the DAA era, we repeated this survey during 2017-2018. Our objective in this analysis was to compare the psychological, behavioral, and social characteristics of patients with diagnosed hepatitis C who initiated DAAs with those who did not initiate DAAs.
stuDy population
We analyzed data collected from adults with chronic HCV infection in the Chronic Hepatitis Cohort Study (CHeCS), an observational study of patients who receive integrated health care services at four sites: Geisinger Health System in Danville, PA; Henry Ford Health System in Detroit, MI; Kaiser Permanente Northwest in Portland, OR; and Kaiser Permanente Hawaii in Honolulu, HI. The criteria for cohort inclusion and analytic methods involved in its derivation have been described in detail. (18,19) The cohort was created based on analysis of electronic health records and administrative data (supplemented with individual chart review) of approximately 2.7 million patients aged ≥18 years who had at least one clinical service visit (i.e., outpatient or inpatient, emergency department, or laboratory test) from January 1, 2006, to December 31, 2013. Patients who met a combination of laboratory-based (i.e., positive HCV RNA) and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)-based criteria identifying them as having chronic HCV infection were included. (18) Among these patients, prospective follow-up data were available through December 31, 2018. The study protocol was reviewed by an Institutional Review Board and approved by the Office for Human Research Protections at each participating study site. The CHeCS investigation follows the guidelines of the U.S. Department of Health and Human Services regarding the protection of human subjects. The CHeCS study protocol was approved and is renewed annually by the institutional review board at each participating site.
DeRiVation oF tHe suRVeyeD CoHoRt
To determine the patient population available for survey invitation, we excluded from the CHeCS hepatitis C cohort those who had died, those who had achieved sustained virologic response (SVR) from any hepatitis C treatment before 2014, and those without contact information. All patients prescribed all-oral second-generation DAA regimens after January 1, 2014, and before survey invitation and had not undergone liver transplant were invited from March through November 2017 to participate. A sample of CHeCS DAAuntreated chronic hepatitis C intended control patients with evidence of continuing care in the previous 5 years (approximately 1:1 untreated to treated at the time of invitation), matched by sex and 5-year birth year range at each site, were also invited to participate in the survey. Patients were sent an invitation letter by U.S. mail at three of the study sites and through the medical record portal at one site (Portland, OR). The letter explained the survey and provided a unique access code for online completion and offered a $25 incentive for participation. If no response was received after 6 weeks, telephone recruiters made up to five attempts to offer the survey as an in-person interview and to encourage online completion of the survey if an interview was declined.
Data ColleCteD By eleCtRoniC HealtH ReCoRDs anD suRVey
We collected demographic information from electronic health records, including age, sex, race/ethnicity, mean annual household income (by census tract geocode), and study site. Clinical data collected included HCV genotype, cirrhosis status, Charlson comorbidity score, and duration of CHeCS follow-up. Cirrhosis was defined according to any of the following criteria: a) fibrosis-4 score >5.88 (20) ; b) liver biopsy equivalent to Metavir F4 or transient elastography results >12.5; or c) ICD-9-CM/ICD-10-CM and Current Procedural Terminology codes consistent with cirrhosis or hepatic decompensation (Supporting Material Appendix S1). (21) Charlson comorbidity scores were calculated from diagnosis codes (excluding liver-related comorbidities) and were used to categorize patients into scores of 0, 1, or ≥2, where a score of 0 indicated no listed comorbidities, 1 indicated a single comorbidity, and a score of 2 or higher indicated multiple comorbidities. (22) Receipt of and start dates for DAA regimens were confirmed based on individual chart review data for all invited participants.
From the survey, we collected information on access to health care, smoking history, drug and alcohol use, employment and work productivity, the presence of affective and anxiety disorders, and life events and social support. Access to health care assessment included questions about difficulty in getting medical appointments, the time and mode of travel required to get to appointments, and the use of the Department of Veterans Affairs health system for hepatitis C care. The survey also incorporated several health-related quality of life instruments, including the Alcohol Use Disorders Identification Test (AUDIT-C), (23) the Work Productivity and Activity Impairment Questionnaire (WPAI), (24) the Patient Health Questionnaire 8 questions (PHQ8) instrument for depression, (25) the Generalized Anxiety Disorder-7 (GAD-7) scale, (26) a validated stressful life events scale adapted from Holmes and Rahe stress scale, (27)(28) Short Form 8 (SF-8) physical and mental components, (29,30) and an abbreviated Medical Outcomes Study Social Support Survey. (31,32) Details and scoring methods of these instruments can be found in Supporting Material Appendix S2.
CompaRisons oF suRVey ResponDents VeRsus nonResponDents anD oF ResponDents WHo initiateD VeRsus DiD not initiate Daas
Among CHeCS patients with hepatitis C invited to participate in the survey, we compared survey respondents with nonrespondents according to data derived from electronic health records alone, which included demographics, study site, and clinical characteristics (HCV genotype, cirrhosis status, Charlson comorbidity score, receipt of DAAs, and duration of CHeCS follow-up). Among patients who responded to the survey, we compared characteristics of those who initiated DAA treatment before their survey response date with those who had not initiated DAAs before their survey response. We compared characteristics among respondents with respect to the aforementioned data from electronic health records as well as surveyderived data elements that pertained to access to health care, smoking history, drug and alcohol use, employment and work productivity, the presence of affective and anxiety disorders, and life events and support. Among respondents who did not initiate DAAs, we examined responses to survey questions that addressed access to care issues, including the specialty care referral experience and self-reported reasons for not starting treatment.
statistiCal analysis
Analyses were conducted using SAS 9.4 (Cary, NC). For univariable analysis, the two-sided chi-square test and t test were used to compare differences for categorical and continuous variables, respectively; we considered P < 0.05 statistically significant. To examine factors associated with DAA initiation among survey respondents, we also conducted multivariable logistic regression analysis, controlling (selected a priori) for age, race, sex, study site, insurance status, cirrhosis status, Charlson comorbidity score, and duration of follow-up.
stuDy population anD DeRiVation oF suRVeyeD CoHoRt
Of 20,349 patients in the chronic hepatitis C cohort, 2,361 had achieved SVR before 2014 and 4,103 had died or had no contact information, leaving 13,885 eligible for the survey. Of these, 2,681 were invited during March through November 2017 to participate; invitees included all 1,408 patients who had been prescribed all-oral second-generation DAA regimens in CHeCS on or after January 1, 2014, and before survey invitation and 1,273 intended control patients (i.e., who had not received DAA before survey invitation), matched by age, sex, and study site. With the passage of time between survey invitation and response, however, 410 (32.2%) of the intended control patients had received DAAs before their date of response to the survey, leaving in effect a final survey-invited sample of 1,818 patients treated with DAA and 863 patients not treated with DAA (n = 2,681).
CompaRison oF suRVey ResponDents WitH nonResponDents
Of these 2,681 patients invited to participate, 1,051 (39.2%) responded to the survey. Compared with survey nonrespondents, respondents were more likely to be non-Hispanic White, aged 51-70 years, and have annual income ≥$30,000, Medicaid or Medicare plus supplemental insurance coverage, and a Charlson comorbidity score ≥1 (Table 1).
CompaRison oF suRVey ResponDents WHo initiateD VeRsus DiD not initiate Daas
Among the 1,051 patients who responded to the survey, we excluded 126 respondents with characteristics that could affect the receipt of DAAs: previous clinical trial participation (n = 30), hepatitis B virus coinfection (n = 16), human immunodeficiency virus (HIV) coinfection (n = 27), and liver transplant between survey selection and completion date (n = 62). We excluded an additional 31 patients who reported DAA receipt on the survey but for whom we could not confirm a DAA prescription or fill order from electronic health records. With respect to analysis of characteristics associated with receipt of DAAs, the final cohort comprised 894 survey respondents, of whom 690 (77.2%) initiated and 204 (22.8%) did not initiate DAAs. Among these 894 respondents, the mean follow-up in the CHeCS was 9.2 years; patients who did not initiate DAAs had significantly more follow-up than those who did (11.2 vs. 8.6 years, P < 0.001).
In the univariable analysis, compared with respondents who received DAAs, those not receiving DAAs more likely were aged ≤40 and >70 years (marginally, P = 0.049), had Medicaid and Medicare (i.e., standard Medicare without a Medicare Advantage plan or a supplemental Medigap plan [e.g., Part E, F]) coverage, had annual income <$50,000, were affiliated with the Pennsylvania study site, and had (Table 3). There were no differences according to sex, HCV genotype, Charlson comorbidity score, travel time necessary to access care, history of military service, employment status, and a noncurrent history of smoking or drug use.
In the logistic regression analysis adjusted for age, race, sex, study site, insurance status, cirrhosis status, Charlson comorbidity score, and duration of follow-up, characteristics associated with lower odds of receiving DAAs included non-Hispanic Black race (adjusted odds ratio [aOR], 0.59 compared with non-Hispanic White race), affiliation with the Pennsylvania site (aOR, 0.59 compared to the Michigan site), difficulty getting medical care in the preceding year (aOR, 0.48 compared with no difficulty), injection drug use in the preceding 6 months (aOR, 0.11 compared with no recent injection), positive AUDIT-C score (aOR, 0.58 compared with negative score), PHQ8 score consistent with severe depression (aOR, 0.42 compared with no depression), homelessness in the preceding year (aOR, 0.36 compared with not homeless), and incarceration in the preceding year (aOR, 0.34 compared with no incarceration) ( Table 2). Patients with compensated cirrhosis had greater odds of receiving DAAs (aOR, 1.77 compared with no cirrhosis).
Among respondents who did not initiate DAAs, we examined responses to survey questions that addressed access to care issues, including the specialty care referral experience and self-reported reasons for not starting treatment. Among those who had not received specialty care, 34.5% said they were never referred, 11.9% did not know whether or why they had not been referred, 10.7% had more pressing medical issues, 8.3% reported they "did not feel sick," 6.0% were unable to pay for additional care visits, 4.8% lacked transportation, and 15.5% had some "other reason." Among respondents referred but not treated, 19.5% reported that the provider said they were "not sick enough," 13.3% were not sure why they were not treated, 11.6% were denied insurance coverage, 5.2% reported other medical conditions, 4.7% reported cost, 3.3% reported alcohol or drug use, 2.8% reported more urgent personal issues, 1.4% were waiting for better treatment options, 0.9% did not want to start DAAs, 0% said they did not start because of a history of nonadherence, 22.1% said it was for "other reasons." (Note: percentages do not add to 100 because blank responses were not included.)
Discussion
To determine patient characteristics associated with receipt of DAAs, we examined responses to an extensive psychosocial survey coupled with electronic health records of approximately 900 patients with chronic hepatitis C enrolled in four large health care organizations in the United States. These patients had a mean follow-up of approximately 9 years. Survey-derived data in univariable analysis revealed that those who did not initiate DAAs were more likely to report adverse behavioral, psychological, and social conditions compared with patients who received treatment. These included perceived difficulties accessing health care providers and dependence on others to get to health care appointments as well as recent injection drug use, alcohol use disorder, current smoking, higher depression and anxiety scores, homelessness, and legal difficulties and incarceration. Compared with patients who received DAAs, those who did not reported significantly higher levels of stress, more hepatitis C-attributed impairment of daily activities, lower levels of physical and mental function, and lesser degrees of social support. Findings were similar in the multivariable model; lower odds of DAA initiation were associated with perceived difficulties getting medical care, severe depression, alcohol use disorder, and recent injection drug use, homelessness, or incarceration. Unlike the univariable analysis results, non-Hispanic Black race was also associated with lower odds of DAA initiation in the multivariable model (as we found in an earlier uptake analysis of our cohort (8) ). Efforts to eliminate hepatitis C hinge foremost on the identification of infected persons and enabling their access to clinical care. However, considerable barriers to DAA initiation, which is a critical stage in the cascade of care, may remain even when a patient with identified hepatitis C is "in care." The steps needed to initiate DAAs may require a persistence and commitment that exceeds the capacity of persons with other more urgent and acute demands and priorities or of those afflicted with comorbid illness. For example, psychiatric conditions, such as severe depression, may impair one's ability to engage the medical system and pursue the often rigorous process of gaining payer approval for DAA treatment. Additional limitations involving social support, transportation to appointments, or concurrent problems related to employment, housing, and legal entanglements, may further complicate the pursuit of treatment. Multiple clinic visits for diagnostic assessment and drug testing, appointments with social workers and patient navigators, and numerous phone calls may be required to complete the preauthorization process; any of these might be impracticable or insurmountable for persons with ongoing psychosocial impairments.
Studies have examined interventions to alleviate barriers to various components of the hepatitis C care cascade. Measures to improve treatment initiation, the focus of this analysis, have included patient education and outreach, colocalization of services, nonspecialist hepatitis C treatment education and care delivery, use of telemedicine, patient navigation programs, and cost management approaches to help defray out-of-pocket expenses. (33)(34)(35)(36) In recent years, government-affiliated health care systems, such as in the U.S. Department of Veterans Affairs, Cherokee Nation, and Alaskan Native Tribal Health Consortium, have demonstrated remarkable improvements in DAA access and uptake, illustrating the potential advantages of unified health delivery systems with relatively homogeneous patient populations. (37)(38)(39)(40) In the private sector, specialty clinics embedded within large health care organizations also have demonstrated the capacity to improve DAA access. During 2014 through 2017, cumulative DAA uptake among Kaiser Permanente Northern California patients with HIV/HCV coinfection was 70%. (41) These patients received health care planning support from case managers and were prioritized for hepatitis C treatment, which was coordinated within each medical facility by a lead infectious disease clinician and a system-wide hepatitis C task force comprised of clinicians, researchers, and community-based advocates. Improving DAA uptake may be more challenging among a less unified hepatitis C population (i.e., not otherwise united by a shared clinical condition, such as HIV coinfection) in private sector health care organizations. For example, during the same time period, we found that approximately 33% of all CHeCS patients with active HCV infection initiated DAAs. (42) However, at the Kaiser Permanente Hawaii study site, nearly 45% initiated treatment. In 2003, this site established a dedicated hepatitis C clinic and began taking a proactive approach to hepatitis C management using a framework to prompt primary care providers to consider specialty care referral for assessment and treatment of patients infected with HCV at the time of diagnosis. (43) In contrast, patients with hepatitis C in more diffuse care networks, particularly those serving nonurban populations, may have challenges in accessing specialty care and DAAs. (44) This may in part explain why patients at the Pennsylvania study site, a network serving a sizable nonurban population, had lower odds of initiating DAAs than those at the other three sites.
Expanding the pool of health care professionals who can provide DAAs can also improve treatment uptake. Data demonstrate that hepatitis C treatment can be effectively delivered by primary care physicians, nurse practitioners, clinical pharmacy specialists, physician assistants, and registered nurses without compromising treatment efficacy or safety. (45,46) Accordingly, the American Association for the Study of Liver Diseases Hepatitis C Guidance Panel recently published simplified treatment algorithms for treatment-naive adults (without cirrhosis or with compensated cirrhosis). (2) These algorithms are designed to be used by any health care provider knowledgeable about hepatitis C, including those without extensive experience who have access to a specialist, and cover guidance on pretreatment assessment, on-treatment monitoring, assessment of response, and posttreatment management.
Although our surveyed population eligible for analysis consisted of approximately 900 respondents with long-term follow-up, our results may not be generalizable to the entire CHeCS hepatitis C cohort (as only 40% of invited patients responded), to other geographic settings, or to cohorts with different characteristics. Indeed, our surveyed population could be viewed as a unique subset of the complete CHeCS hepatitis C cohort as uptake within the surveyed cohort was approximately 77% compared to 33% for the overall hepatitis C cohort during 2014-2017. (42) Our analysis was limited, therefore, in that it was (unintentionally) heavily weighted with survey respondents who received DAAs, which may have hampered our ability to rely on survey responses to understand barriers to treatment in the overall survey-eligible cohort of nearly 14,000 patients. However, it is remarkable that the presence of psychosocial impairments was significantly more frequent among respondents who did not initiate DAAs compared with those who did, given that only 23% of respondents did not initiate DAAs. It is possible that these impairment differences were even more pronounced among the 60% of patients who did not respond to the survey. Also, given that only one sixth of our respondents were Medicaid recipients, it could be reasoned that treatment populations consisting of mostly patients with Medicaid might demonstrate even more pervasive degrees of psychosocial impairment.
Nonetheless, for some variables, such as recent injection drug use, the number of respondents who reported recent use was low (n = 11), so differences between treated and untreated respondents, although statistically significant, may have been underpowered to make definitive assessments. Our study was also limited by the absence of provider perspectives about barriers to DAA initiation, which reduced our capacity to explicate fully the associations between psychosocial impediments and initiation of treatment. For example, it is unknown whether patients who were severely depressed were less likely to seek treatment in the first place, if they were depressed because they had sought treatment but were denied it, or if they were offered treatment but declined to follow through with the process. We did, however, examine specific access to care issues among respondents who did not initiate DAA treatment. Almost half of these patients were not referred or were unsure whether they had been referred to specialty care. Of those who had been referred but not treated, approximately one third reported that either they were told they were "not sick enough" for treatment or were not told at all why they were not offered DAAs. Only small proportions of nontreated respondents explicitly noted an inability to pay for referral visits and medications, that they had been denied insurance coverage for them, or reported having more pressing medical or situational concerns. However, it was probable that some patients were not referred or offered treatment because of provider awareness of the futile nature of preauthorization constraints or because of "other reasons" not acknowledged by respondents; therefore, the low frequency of survey-reported financial and insurance barriers or of unacknowledged issues regarding nonadherence or substance use/mental health problems likely underestimated the true effect of these factors. For example, respondents with Medicaid and Medicare (without supplemental coverage) were less likely to receive DAAs in the univariable analysis. In Michigan, the location of our principal study site, Medicaid coverage for DAAs was delayed and fibrosis restrictions remained in place long thereafter. (47) A similar situation existed in Pennsylvania where another study site was located. The situation is less clear with our Medicare respondents, although there have been reports of difficulties with receipt of DAAs among Medicare recipients, particularly those lacking Part D coverage or with Part D coverage subject to high copays, or with Medicare/Medicaid dual coverage subject to state-specific Medicaid drug coverage rules. (48) Unfortunately, not all our study sites (each in a different state) collected information on Part D coverage, so we were not able to discern whether or to what degree this might be an issue.
Understanding such complex issues might be difficult to unravel with respect to causation, yet our identification of several adverse behavioral, psychological, and social qualities associated with noninitiation of DAAs among these patients in care suggests the presence of additional treatment barriers to be addressed by clinicians and programs dedicated to reducing the morbidity and mortality burden of hepatitis C and ultimately to its elimination as a public health threat. | 2020-12-03T09:01:33.533Z | 2020-11-29T00:00:00.000 | {
"year": 2020,
"sha1": "2433735f76117c70a0b0403d3487c5b7b82c2686",
"oa_license": "CCBYNCND",
"oa_url": "https://aasldpubs.onlinelibrary.wiley.com/doi/pdfdirect/10.1002/hep4.1632",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7c93ba43091b94ebad467bfe95d5c23d319677b6",
"s2fieldsofstudy": [
"Medicine",
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
149960430 | pes2o/s2orc | v3-fos-license | MANDATORY DATA BREACH NOTIFICATION AND HACKING THE SMART HOME : A LEGAL RESPONSE TO CYBERSECURITY ?
This paper will investigate whether the Australian legal and regulatory framework sufficiently addresses cybersecurity concerns particular to the smart home. Specifically, the paper will analyse the extent to which the introduction of the data breach notification scheme in Australia will apply to smart home device manufacturers regulated by the federal Privacy Act 1988 (Cth) regarding device breaches. By examining Australian Privacy Principle 11 and the introduction of mandatory data breach notification, the paper aims to determine whether the Australian privacy model of Principles-Based Regulation is capable of providing a market-based solution to cybersecurity concerns in the smart home.
I INTRODUCTION
The law has traditionally recognised the home as a private and passive space, wherein there is a reasonable expectation of privacy. 1 Cory J in the Canadian case R v Silveira stated that 'there is no place on Earth where persons can have a greater expectation of privacy than within their dwelling house.' 2 The smart home marks a new frontier in the digital disruption caused by the emergence of the Internet of Things landscape.While the smart home promises users unparalleled freedom and flexibility, there is a risk that the devices converging to create the smart environment may have poor in-built security measures, making infiltration attractive to hackers. 3 The dangers are especially prevalent in a world where smart home devices are becoming progressively interconnected in the amount of data that is transferred and stored between them. 4The smart home, therefore, presents a primary example of the challenges facing an increasingly digitised society.Given the steady increase in demand for smart home device products, and the relative concerns which have been raised as to their level of security and consumer privacy protection capabilities, the problems presented by the insecurity of data and its systems arise as significant concerns in the smart home environment. 5art IV of the paper analyses the frameworks introduced in Part III, and their potential application to smart home devices.It is stated that Australian Privacy Principle 11 is unlikely to apply in the smart home environment, and so the introduction of the mandatory data breach notification scheme is analysed to determine whether the scheme may be of potential relief for consumers of smart home devices.Further tensions within the data breach notification scheme and APP 11 are identified, and inconsistencies highlighted.It is argued that the increasing interconnectivity of data in smart home devices does not readily fit within the traditional conception of privacy law frameworks.
Part V concludes the paper by noting that the impact that mandatory notification laws will have on smart home devices remains unknown.Nevertheless, the application of smart home device data breaches to a Principles-Based Regulation approach does not provide a clear market-based solution to the joint rise in the smart home market and the increasing sharing of data between internet connected devices and platforms.
II CYBERSECURITY AND THE SMART HOME A Background
The introduction of the smart phone offered users instantaneous access to information and 'all-in-one functionality', which slowly gained the trust and dependency of its consumers. 7The smart home finds its natural evolution from this trust and dependency, enlivening the vision to entrench devices and the These may include unsecured factory default settings and passwords, which can leave consumers vulnerable to hackers easily installing malware to gain access to entire home networks. 6Australian Cyber Security Centre, '2017 Threat Report' (Report, Australian Government Australian Cyber Security Centre, 2017). 7 'ultimate digital assistant' into the home. 8The smart home forms part of the broader Internet of Things ('IoT') landscape.While there is a myriad of potential definitions of IoT,9 in essence it refers to the: Interconnection of sensing and actuating devices providing the ability to share information across platforms through a unified framework, developing a common operating picture for enabling innovative applications.This is achieved by seamless ubiquitous sensing, data analytics and information representation with cloud computing as the unifying framework. 10e smart home is concentrated on 'smart connectivity of objects with existing networks and contextaware computation using network resources.' 11The smart home signals a shift in the increase of invisible infrastructure, where technology is no longer monolithic but now has a malleable duality, capable of constant change. 12It is a point of intense contact between the user and the device. 13Through the implementation of smart home devices, the ultimate vision is to create 'ambient computing' in the home where 'smart devices disappear into the background, consumers only [having] to consider the tasks they want performed, and no longer have to consider which device … will be capable of performing that task.' 14 In order to achieve this, various computational nodes ('smart home devices') are connected in the home.Hidden within these smart home devices are advanced technological processes of collection, storage and use. 15For example, the smart television, with its abilities of content sharing and web browsing, is widely considered a major step towards the convergence of computing and entertainment. 16Two issues arise in the context of the smart home: the collection, and the consolidation of, information. 1
Collection and Consolidation of Information
Smart home devices can be generally categorised into four segments: safety, health, energy and entertainment orientated devices.17Manufacturers and software providers of smart home devices, provided they meet the annual turnover requirement of over three million Australian dollars per financial year,18 may be regulated by the Privacy Act 1988 (Cth) ('Privacy Act') in relation to obligations regarding online privacy and data protection. 19The manner in which data is collected is generally unobtrusive, in furtherance of the ultimate vision of ambient computing, and consumers of these devices may not necessarily understand the breadth of data collection potentially occurring in a single smart home network. 20Data is collected and consolidated through a multiplicity of devices to provide the user with 'familiarity', storing consumer preferences such as light brightness in a smart light bulb, or temperature settings in a smart thermostat. 21Interconnectivity between smart home devices, such as a lightbulb and thermostat, facilitate the objective of ambient computing by creating a network of sensors that detect external elements such as light, temperature and motion. 22The devices then collect, send and receive data autonomously between each other for ultimate control and monitoring by a smart home user. 23A consequence of modern data flow is that the data collected and sent from a smart home device will invariably be stored and received through a multitude of international servers.Though beyond the scope of exploration in this paper, considerable difficulties are presented by the regulation of such transnational data transfer, collection and storage. 24These challenges are many and varied and are of particular prevalence from the perspective of international organisations in consideration of the varying standards for compliance across numerous jurisdictions of operation in which data may be shared. 25e unique and varying nature of data collected thus increases an individual's digital trail, and goes 'much closer to knowing and understanding the unique complexities and individual features of human beings' than may be expected. 26For instance, the Google Home 'may combine personal information from one service with information, including personal information, from other Google services.' 27The device may combine user data from various sources such as Gmail, Google Drive and the user's web history. 28The device also offers third party application integration, allowing aggregation of data from platforms such as Uber, Spotify and FitBit. 29 order for a smart home to provide ever-present assistance and functionality to users, connected devices must establish a presence in the home alongside other internet connected devices and facilitate the transfer of data between them to perform tasks. 30 devices in a smart home network by relaying data through 'transmissions' which are secured through 'protocols', typically through Wi-Fi in a home gateway router. 31However, where information is stored and data is capable of being accessed through multiple and potentially unlimited numbers of devices, invariably issues of mixed ownership arise due to the diversity of entities dealing with data on multiple devices and managing the increasing interconnectivity between them.Richard Mason, an information management scholar, foreshadowed in the 1980s that eventually the 'increased collection, handling and distribution of information will pose serious threats to the privacy, accuracy and accessibility of personal information.' 32The handling of such data stored in a smart home thus raises questions in relation to legal responses to potential hacks, and obligations on entities to provide cybersecurity protocols. 33
B Cybersecurity Threats to the Smart Home
There are three cybersecurity threats which are of particular relevance to the smart home.These are data and identity theft, device hijacking, and ransomware.The Australia Cyber Security Centre ('ACSC') 2017 Threat Report ('the Report') discusses the prevalence of these risks to cybersecurity more generally. 34The Report emphasises the increasing sophistication in attacks by cyber criminals, but notes that many networks are compromised using 'publicly known vulnerabilities' which have known mitigations. 35In the context of the smart home, the infrastructure of the devices comprising the home environment may expose the network to shared vulnerabilities. 36This may arise either from poor cybersecurity protocols in a particular device, or result from outdated software nearing the end of its product life-cycle. 37Nevertheless, hackers may target these vulnerabilities and infiltrate a smart home network through physical proximity to the home, or remote activation and access of the sensors in the smart devices. 38
Data and Identity Theft
The potential for identity theft and crime in internet connected devices is not necessarily particular to the smart home. 39Data and identity crimes have an estimated annual economic impact of over two billion dollars in Australia; four to five per cent of Australians are victims of identity crime resulting in financial loss annually. 40udies have shown, however, that there is a link between the increase in the amount of personal information stored in a network and the incentive for hackers to breach that network and commit data and identity theft crimes. 41Although these cyber-attacks can be widespread, smart homes present a more susceptible and attractive target for hackers due to their complex interconnected nature.This is because the sheer volume of data stored in even a small number of connected smart home devices provides more opportunity and incentive for hackers to extract personal information than would be possible from 'less rich data sets'. 42Where multiple devices are connected on a single smart home network, the network becomes increasingly vulnerable to hacking due to a larger 'attack surface'. 43ccess to one device may provide a hacker with a gateway into all of the smart home devices connected on that network and the data that they store. 44e most common method of data and identity theft in the smart home is through credential-harvesting malware, where hackers bypass security protocols through social engineering and 'credential phishing'. 45The granular data collected in the smart home through a multiplicity of devices compiles to form a unique digital profile of the user.The digital profile is capable of detailing both the consumer's behaviour, such as viewing habits on a smart television or energy consumption on a smart meter, 46 and providing essential information which may be used for document forgery, such as in passports or drivers' licences. 47Hackers may either use stolen data personally or sell it on dark web marketplaces for use in financial crime or identity theft. 48The nature of this personal information also appeals to stalkers, who by accessing the data may gain knowledge of a potential target's home and their lifestyle patterns, and may make inferences based on physical proximity. 49Device Hijacking The purpose of smart home devices is to automate processes and simplify tasks. 50The hyperconnectivity of devices in a smart home environment necessarily entails high levels of communication and data transfer between different smart devices over a range of protocols and technologies. 51These protocols contain differing levels of security, and could thus allow a 'weak link' to be identified by a hacker for targeting, allowing them to gain access to the whole smart home network. 52For example, the Philips Hue lightbulb has been criticised for its poor security, as the bulb does not encrypt data before it is transferred to another device. 53This may allow a hacker to send commands to override and infiltrate the second device merely by gaining access to the lightbulb. 54A similar situation may also arise for smart devices with outdated software, which increases the device's susceptibility to a security breach. 55Studies have shown that many smart home devices are configured with identical or substantially similar software and firmware, which increases the potential for a hacker to exploit common vulnerabilities in a range of devices connected on a single smart home network. 56ima facie, individual data from a single smart device such as a lightbulb may not necessarily provide access to a wide range of data on consumer behaviour. 57However, preferences stored in devices like smart lightbulbs may indicate whether or not a consumer is presently at the home by sending a 'current status' update. 58This would provide the hijacker with 'a source of close, granular and intimate data on the activities and behaviour' of the smart home's inhabitants.' 59Further, once a device is hijacked, a 'man-in-the-middle' attack can be made between smart home devices as a result of the 'weak link' in the smart home environment. 60'Man-in-the-middle' attacks involve the hijacker making independent connections with various devices and relaying communications between them. 61milarly to cases of data and identity theft, unique data from multiple devices can be obtained via device hijacking, which allows hijackers to gain contextual knowledge about the individuals and inhabitants of a smart home. 62Pieced together, the inferences made based on learned behaviour have the potential to 'paint a near complete and accurate digital portrait of users.' 63 From utilising this method, a hacker in physical proximity to an infiltrated smart home may remotely access the compromised devices and use this to create a physical attack on the inhabitants.Smart thermostats may be used to increase heating system temperatures and cause pipes to burst by altering user inputs, 64 or surveillance cameras may be remotely turned on to view activities of inhabitants inside the home. 65Ransomware Ransomware is a method used by financially-motivated hackers to extort funds from victims by blocking access to, or controlling, user data. 66The method is a persistent and prevalent threat both in Australia and worldwide, with an 'increasing frequency and variation of campaigns' being reported. 67hen this method is applied to a smart home environment, manipulation of data in the devices may be pushed to extremes in the pursuit of revenue generation. 68For example, distributed denial-of-service ('DDoS') attacks may be made to shut down a home network or tamper with devices. 69Hackers may then demand a ransom through an internet connected printer to restore access. 70Alternatively, a hacker who has gained control of a smart home network may orchestrate a physical attack through a smart device and deny inhabitants access to security devices like smart locks or garage openers. 71Smart televisions are also vulnerable to malicious malware.Malware 'Revoyem' redirects users on smart televisions, through its web browsing facilities, to child-pornographic-themed pages, and demands payment to 'clean' the system. 72I LEGAL RESPONSES TO THE SMART HOME Cybersecurity in Australia is not directly regulated by a single governing piece of legislation.Rather, there exists a patchwork of different laws, regulations and guidelines which regulate conduct and place obligations on 'entities' subject to the Privacy Act. 73Non-compliance with those obligations render an entity liable to punishment and enforcement under the civil penalty framework imposed by the Privacy Act. 74This part of the paper will examine and discuss the current privacy law framework in Australia in relation to potential forms of relief that may be sought by an affected smart home device user following a hack.Specific emphasis will be placed on the rationale of Australia's Principles-Based Regulation framework, reasonable steps to protect personal information under Australian Privacy Principle 11, and the introduction of the mandatory data breach notification scheme.
A Principles-Based Regulation
The privacy regime adopted in the Australian model is based on the 1980 Organisation for Economic Co-operation and Development's (OECD) Guidelines on the Protection of Privacy and Transborder Flows of Personal Data ('OECD Guidelines'). 75The OECD Guidelines, and hence Australia's regulatory framework of legal obligations in information security, are modelled on Principles-Based Regulation ('PBR').Australia's adoption of PBR is widely accepted. 76Of particular interest to this paper is the OECD Guidelines' advancement of the 'security safeguards principle', which states that 'personal data should be protected by reasonable security safeguards.' 77 PBR distinguishes the regulator from the regulated.In Australia, these most commonly amount to the Office of the Australian Information Commissioner ('OAIC') and the entities subject to the authority of the Privacy Act. 78Julia Black, a leading scholar in PBR, has explained the theory as effectively involving a shift in responsibility from the regulator to the regulatee. 79The delegation of regulatory function is described as a conscious and deliberate intention by the regulator to influence the regulatee's internal systems of management and control. 80The delegation of control inherent in this theory is consistent with 'meta-regulation'. 81PBR requires and assumes a high level of trust and cooperation on the part of the regulatee to be competent and responsible, maintaining 'regulatory conversation' with the regulator. 82It reinforces the notion of the 'self-observing, responsible organisation.' 83 prevent the market being dis-incentivised, the regulatee, assumedly understanding its own environmental context, self-regulates.It is assumed that these entities, through corporate culture, will maintain a level of corporate social responsibility to consumers, particularly in the form of cybersecurity. 84The Australian model has been referred to as 'light touch regulation' by its national government, as maximum flexibility is maintained in allowing entities freedom to meet principle-based statutory outcomes by developing innovative forms of compliance. 85PBR can be contrasted to a hierarchical rule-based regime, where 'bright line' and specific rules are adopted. 86PBR is argued to provide an advantage over the hierarchical approach by identifying broad principles which encourage compliance with the spirit rather than the letter of the law. 87The model attempts to prevent the stifling of progress, particularly at the design level, by not burdening entities with obligations to incorporate specific security features to strengthen the protection and integrity of a particular device. 88wever, the PBR regime has been criticised for allowing regulators to act retrospectively, increasing the level of uncertainty of consumers and regulatees as to their standing regarding current conduct and measures, and reducing predictability of regulatory responses to future disputes.It is argued that PBR provides inadequate protection to consumers by creating a corporate culture of adhering to the very 'minimum level' of compliance, hence failing to afford certainty and predictability to consumers. 89ey to the successful implementation of PBR, therefore, is the manner in which it is implemented and the institutional context which surrounds it.Without this context, PBR's 'light touch' regulation may lead to a market consensus of risk-taking in the pursuit of profit over product safety, 90 and the use of ineffective compliance systems based on internal organisational control. 91Australian Privacy Principle 11 The Australian Privacy Principles (APP) were introduced to the Privacy Act under the Privacy Amendment (Enhancing Privacy Protection) Act 2012, 92 and commenced operation in 2014. 93The APPs replaced the now-repealed National Privacy Principles and Information Privacy Principles. 94UT Law Review -Vol 18, No 2 | 277 They are designed as a broad 'technology-neutral approach' for application to current and future technologies, and reflect PBR by acting as 'high-level principles' to guide data management practices of entities regulated under Privacy Act. 95APP 11 does not mandate specific security obligations on entities. 96Each entity ultimately takes the onus and responsibility of determining how to comply with the APPs in the context of their specific circumstances and the data management practices in which they employ. 97 the context of obligations for cybersecurity in the smart home, APP 11 is of most relevance as it relates to security of personal information. 98APP 11.1 states that if an APP entity 'holds personal information' it must take 'such steps as are reasonable in the circumstances' to protect the information from: 'misuse, interference and loss', as well as 'unauthorised access, modification or disclosure.' 99 'Personal information' is defined under section 6(1) of the Privacy Act as 'information or an opinion about an identified individual, or an individual who is reasonably identifiable.' 100 An entity 'holds' personal information if the information complies with the definition of 'personal information' under section 6(1) and the entity 'has [physical or electronic] possession or control of a record that contains the personal information.' 101
Such Steps as are Reasonable in the Circumstances
The Explanatory Memorandum to the Privacy Amendment (Enhancing Privacy Protection) Bill 2012 states that 'reasonable steps in the circumstances' is an objective assessment, but that 'objectively reasonable steps' depend on the 'specific circumstances of each case.' 102 It is dependent on the relevant risks within an entity and their particular devices. 103For example, it would be unreasonable to implement high cybersecurity protocols in a device that has low privacy risks where the costs of taking such steps are high. 104This reflects the underlying reasoning of PBR as the regulated entity is best placed to identify its own risks in its internal environment, and has delegated authority to implement cybersecurity protocols proportionate in cost to these conceived risks. 105e Joint Investigation of Ashley Madison in 2016 highlights that cybersecurity governance frameworks are assessed with consideration of possible risks faced in the circumstance, and security measures in view of the amount of sensitive personal information held. 106Failure to take reasonable steps may also include having a lack of basic security measures in place which could reasonably have been implemented, such as encryption of passwords. 107Further, the hack of the Sony PlayStation Network in 2011 emphasises that if appropriate security safeguards are in place, an entity may still comply with its data security obligations under APP 11.1 despite a security breach occurring. 108'Misuse, Interference and Loss' and 'Unauthorised Access, Modification or Disclosure' A 'misuse' occurs where personal information is used for a purpose not permitted by the Privacy Act. 109'Interference' with personal information arises where the integrity and security of the personal information is compromised, but does not necessarily require modification of its content. 110This would have application where smart home devices are hijacked but the hacker does not change the basic functionality of the device. 111The same scenario could also be applied to establish an 'unauthorised access'. 112A 'loss' is established in this context where there is either a physical or electronic loss of personal information. 113Destroy or De-Identify Information Under APP 11.2, where personal information is no longer needed by the entity 'for any purpose for which the information may be used or disclosed' the entity must take reasonable steps to 'destroy the information or to ensure that the information is de-identified.' 114 De-identification requires removal of personal identifiers and removing or altering information which may allow an individual to be identified. 115The costs involved in this process are generally high, so entities may opt rather to destroy information through secure methods, but must avoid unauthorised disclosure during the destruction process. 116
C Mandatory Data Breach Notification Scheme
The Australian privacy model previously operated on a voluntary notification scheme, whereby there was no requirement under the Privacy Act to notify affected individuals or the Information Commissioner when a data security breach occurred. 117This voluntary notification scheme was criticised for underreporting instances of serious data breaches and for excessive delays in notification. 118The introduction of mandatory data breach notification scheme, which took effect from 22 February 2018, is the result of numerous recommendations by the Australian Law Reform Commission ('ALRC') and the OAIC to provide increased transparency to consumers. 119Mandatory Data Breach Notification ('DBN') emanates principally from California in the United States of America, 120 but has been adopted worldwide from Canada to the European Union and New Zealand. 121ngela Daly highlights the comparative regimes in the US, which operate in sectors as a 'patchwork of unharmonised data breach notification legislation', to that of the European Union, where, similar to Australia, data breach notification laws operate alongside existing comprehensive data protection laws. 122It is predicted that notification rates should double in Australia with the introduction of the new scheme for mandatory notification. 123e rationale of DBN in Australia is twofold.First, it is so individuals may personally take remedial steps if personal information is compromised, such as by changing passwords to mitigate the potential for identity theft. 124Second, it encourages entities to be proactive in taking steps to address data breaches and have readily available data breach response plans. 125DBN recognises that the absence of notification to individuals of data breaches which involve personal information 'does not align with the almost universal agreement from the Australian public that an organisation should inform them if their personal information is lost [or breached].' 126Australia's new mandatory DBN scheme is enacted under the Privacy Amendment (Notifiable Data Breaches) Act 2017. 127The amendments insert a new Part IIIC into the Privacy Act. 128This was done deliberately in an attempt to streamline the regulatory process. 129The DBN scheme places an obligation on entities subject to the Privacy Act to notify the OAIC and 'affected individuals' as soon as practicable when an entity has reasonable grounds to 'believe' that an 'eligible data breach' has occurred. 130An 'eligible data breach' occurs where there is a 'data breach', the 'data breach' is likely to result in 'serious harm' to one or more individuals from the perspective of a 'reasonable person', and an exception to the requirement for notification cannot be established. 131As the relevant entity ultimately determines whether or not an 'eligible data breach' occurs and mandatory notification is required, the DBN scheme is based on the PBR notion of delegated authority. 132It relies on entities acting responsibly through a detailed 'risk-based analysis' and maintaining regulatory conversation with the OAIC. 133 Data Breach A 'data breach' occurs under section 26WE(2) of the Privacy Act where there is 'unauthorised access to or unauthorised disclosure of personal information, or a loss of personal information.' 134These terms are not defined in the current Privacy Act or new provisions, but are to be given their ordinary meanings. 135'Unauthorised access' has been described to occur where personal information is accessed by someone who is not permitted access to that information. 136This definition is generally intended for external interferences with an individual's personal information stored by an entity. 137he Data Breach Guide refers to unauthorised access as 'databases containing personal information being "hacked" into or otherwise illegally accessed by individuals outside of the agency or organisation.' 138The terms 'unauthorised disclosure' and 'loss' are generally intended for internal interferences with personal information, and may arise from inadvertence on the part of the entity. 139here the entity does not have reasonable grounds to believe an eligible data breach has occurred but 'suspects' one may have, the entity must, within thirty days of developing the suspicion, perform a 'reasonable and expeditious assessment' of the suspected breach under section 26WH. 140Wilful ignorance will not circumvent an entity's obligations or liability under the new provisions. 141 2 Serious Harm In order to balance individual and corporate interests, the compliance burden in DBN is reduced to eligible data breaches likely to cause 'serious harm'. 142The legislative intention for this requirement is to minimise the risk of 'notification fatigue' on the part of individuals and the administrative burden this may place on entities. 143'Serious harm' is not defined in the Privacy Act but is considered a high threshold. 144The Explanatory Memorandum to the Privacy Amendment (Notifiable Data Breaches) Bill 2016 (Cth) notes that serious harm may include 'serious physical, psychological, emotional, economic and financial harm, as well as serious harm to reputation as well as other forms of serious harm'. 145'Serious harm' is measured from the perspective of a 'reasonable person in the entity's position' and what they would 'identify as a possible outcome of the data breach.' 146Section 26WG of the Privacy Act identifies a non-exhaustive list of matters relevant in assessing the likelihood of serious harm, including the kind and sensitivity of the information. 147In determining whether an unauthorised access or disclosure will cause serious harm, the phrase 'likely to occur' is interpreted as 134 Privacy Amendment (Notifiable Data Breaches) Bill 2017 (Cth) 5-7.
QUT Law Review -Vol 18, No 2 | 282 parties. 159Once an entity has complied with the obligation, the other entities are relieved of that same duty.
Existing Criticisms of Mandatory DBN: Enforcement and Compliance
The mandatory data breach notification model has been criticised for its focus on reputational sanctions as its principal regulatory mechanism and has been described as failing to adequately address the aftermath of a data breach in a practical manner. 160Greenleaf and Clarke have identified a nearuniversal failure internationally of compliance authorities, including the OAIC, in documenting and publishing statements of data breach complaints as a major contributing factor to issues of transparency in the enforcement and compliance process of such models. 161Though organisations are required under the scheme to publish of data breaches with respect to affected individuals and the OAIC, it has been argued this alone is insufficient to make details of data breaches available for public attention. 162Further supplementation of notification under the scheme has been advised in the form of publication on the OAIC website, as part of a permanent, browsable and searchable database, to allow recurrent aspects of breach notification to be identified by interested parties. 163Such a searchable data base is promoted as likely to exhibit more of a deterrent effect on organisations and more effectively induce improvement of data security measures than is currently observed in compliance activities of regulated parties. 164Without a public forum in which OAIC can publish statements by affected entities, the 'light touch regulation' envisioned by the PBR may be imbalanced given the lack of 'feedback loops' available to allow consumers to become aware of data breaches, encourage organisational compliance and complaints, and discourage data security breaches. 165 ANALYSIS OF LEGAL RESPONSES This part of the paper will analyse the legal responses identified in Part III and determine whether those responses are capable of sufficiently addressing cybersecurity concerns in the smart home.
A Does Data in a Smart Home Device Constitute 'Personal Information'?
In order for cybersecurity breaches to be regulated under APP 11 or the DBN scheme, the data collected by the relevant smart home device must constitute 'personal information' within the meaning of section 6(1). 166If the information is not capable of identifying or reasonably identifying an individual, it is outside the ambit of the Privacy Act. 167This question can only be answered on a case- by-case analysis.Clearly, the Google Home, which synthesises a user's web history and emails, and is further capable of integrating with third party applications, will constitute 'information … about an identified individual' within the meaning of section 6(1).Even if the information did not explicitly name the individual, 168 the context and sheer volume of information stored about the user would that individual 'reasonably identifiable'. 169Additionally, information relayed from a FitBit to the Google Home exposes the device to 'health information' within the meaning of 'sensitive information' under section 6(1). 170In contrast, the information collected on a smart lightbulb may only store preferences for remote-control lighting, which makes identifying or reasonably identifying an individual challenging.Thus, data stored by some smart home devices such as a smart lightbulb may not necessarily, alone, constitute 'personal information' under the Privacy Act.This is particularly so given the Federal Court's recent consideration of the definition and what constitutes information 'about' an individual for this purpose in Privacy Commissioner v Telstra Corporation Limited, 171 where, according to some commentators, the qualification may have been narrowed. 172e situation of a breached smart lightbulb may change regarding the interpretation of the kind of data involved, however, where a hacker infiltrates a smart bulb and patches through a 'current status' update. 173The individual would then be reasonably identifiable, as the presence of their physical location can be determined by the hacker.The situation changes again where a hacker uses a breached lightbulb to access other devices in the smart home network, where the other devices carry similar firmware with shared vulnerabilities to the smart bulb.The hacker would then theoretically be able to access a user's home control inputs and devices and commit further attack. 174rk Burdon highlights that while the ALRC's 2008 Report recommended a limited definition of 'personal information', it recognised the purpose of DBN in Australia, alike that in the EU, is more extensive in application than in mitigation of identity theft, the principal approach of the US. 175As such, Burdon argues that, to achieve this more comprehensive application, the Australian DBN approach must seek to incorporate rather than negate circumstances which are context-dependent. 176nder this approach, circumstances of breach constituting personal information triggering an obligation to notify may change when a device, which is interconnected with other devices in a smart home network, is breached.This may be so even where the obligation would not exist for breach of the device alone had it not been interconnected.the volume of information stored in the smart home network almost guarantees that the information constitutes personal information under section 6(1) as it is readily identifiable to an individual.Telstra has estimated that the average Australian household contains thirteen internet-connected devices. 177his figure is set to increase to over thirty devices by 2021. 178Given the rise of the smart home market in Australia, it is increasingly likely that the data stored in an individual smart device will either constitute 'personal information' alone, or, if not, it will fall within the definition as part of the smart network, due to the greater context provided by additional information from an increased number of devices. 179Do the Legal Responses Address Cybersecurity Threats to the Smart Home?
Compliance with APP 11 is ultimately delegated to the regulated entity to interpret and implement protocols in a smart home device. 180Assuming that smart home device data constitutes 'personal information' within the meaning of section 6(1), an entity may be liable for failure to take 'such steps as are reasonable in the circumstances' in relation to a cybersecurity breach of a device. 181e expansion of the smart home market has raised concerns that some manufacturers of smart home devices are prioritising profitability over product development at the expense of product safety in the commercial drive for an increased profit margin. 182A study which interviewed IoT designers and developers in Australia regarding their perspectives on the growth of the market identified that there are entities in Australia that focus purely on 'innovation' rather than 'privacy in the design of IoT devices.' 183 It was found that these entities aimed for 'quick innovation and pushing new products'; the legal framework 'a lagging indicator into what innovation offers.' 184 If this reasoning resulted in a market consensus or trend of implementing poor cybersecurity protocols in smart home devices at the design phase in favour of innovation, 185 and that device was breached by a hacker, 'reasonable steps' would be construed in relation to the steps, or lack of steps, the entity had taken to prevent the breach.The focus of the terminology in APP 11.1 is not on the design infrastructure of a breached device, but rather on analysing the security measures at the point of the breach. 186The use of the terms 'misuse, interference and loss' as well as 'unauthorised access, modification or disclosure' concentrate the analysis on reasonable steps taken at the point of breach, such as incorporating mutual authentication or secure communication. 187ile the OAIC may determine that an entity did not adequately secure personal information, such as in failing to encrypt data as it is transmitted and transferred to other smart home devices, this does not materially prevent the breach of the device from occurring in the first place.The effectiveness of APP QUT Law Review -Vol 18, No 2 | 286 notification obligations has been calculated to be a much higher amount overseas. 196Under PBR reasoning, the obligation of mandatory notification incentivises entities to further invest in data security measures in their devices.This is to prevent cybersecurity breaches from occurring and the need for notification ever arising, as this could cause significant reputational damage on top of the surface and hidden costs that would result from a notifiable breach. 197The PBR reasoning underpinning DBN does not always align smoothly in the context of smart home devices as establishing the requirement for notification is not always clear.
By focusing on 'data' breaches, an entity may comply with the 'letter of the law' in not reporting 'data' breaches even if a smart home device has been hacked.Two concepts can be distinguished: a hack of a device and a hack of data.A clear hack of data, such as widespread ransomware or physical attack on numerous smart homes, would likely trigger the obligations of the scheme.On the contrary, a hack of a single smart home device is not strictly notifiable as it may not fulfil the requirements of an 'eligible data breach'. 198While the first limb in establishing a data breach is likely fulfilled given that 'unauthorised access' is interpreted liberally, proving 'serious harm' is considered a high threshold. 199 hacker may obtain 'unauthorised access' to a smart device, but where they do not modify the content of the device and merely observe the use of information by the inhabitants, it may be difficult to establish 'serious harm'.This sort of breach would have to be established as either 'psychological', 'emotional' or, a more probable than not threat of 'physical' harm to the affected individuals. 200ere a hacker obtains unauthorised access to surveillance cameras, 'serious harm' may be established as the private nature of the home and the reasonable expectation of privacy within it is compromised, and the inhabitants are at greater risk of serious physical or psychological harm. 201Thus, whether breaches of smart home devices that do not necessarily modify 'data' will be notifiable is inherently contextual.The content has to be 'defined by individuals themselves according to context' and not delegated upon an entity to determine from the standard of a 'reasonable person in the entity's position'. 202The entity may obfuscate its obligation under the DBN scheme in these situations by either remedially acting to shut down the hacker, or avoiding notification to comply strictly with the letter of 'serious harm', but not the spirit of the term. 203Depending on the method used to infiltrate a smart home or a particular device, these situations would allow hackers who breach smart home devices for stalking purposes to continue without the risk of being compromised.Neither of these situations result in the potentially affected individuals from being able to take remedial steps to protect themselves or increase transparency.This is counter to the intention and purpose of the scheme. 204rther, there are issues with quantifying an 'eligible data breach'.The use of the words 'one or more individuals' implies that an 'eligible data breach' may apply to a small household which establishes a smart home network. 205At the same time, the provisions also militate against the risk of 'notification fatigue' from entities and the corresponding lack of utility for individuals in constant notification. 206his would suggest the scale of the breach and number of individuals affected remains the primary indicator of whether the eligible data breach is notifiable in the circumstances.Paradoxically, the increase in the scale of a data breach may decrease or diminish the chance of 'serious harm' to each particular individual, 207 and thereby fail the requirement for notification on the second limb of the criterion.Hence, 'eligible data breach' potentially may be inapplicable to both breaches of small smart home networks and large-scale breaches, such as of cloud service providers in the smart home. 208
(b) 'Jointly and Simultaneously' Held Information
The concept of jointly-held information may have application to interconnected devices in the smart home and the requirement for notification provided the devices 'hold' personal information within its meaning under section 6(1). 209The application of 'jointly-held information' will inevitably depend on the individual devices in a smart home network and whether the data transfer between these devices constitute outsourcing, joint ventures, shared service arrangements or potentially an 'online platform'. 210The concept may apply where data is held jointly and simultaneously on a smart home network and a hacker uses a single smart home device to breach the entire network. 211'Man-in-the middle' attacks could also trigger notification requirements in these scenarios. 212For example, 'a data breach involving an individual's name may [increase the risk of serious harm] if the entity's name links the individual with a particular form of physical or mental health care.' 213 The interconnected nature of smart home data places tensions on the conceptions of 'serious harm' and 'personal information', as when information is combined and concurrently accessible through various smart home devices through 'communication' via protocols, the sensitivity of the information increases the risk of 'serious harm'.
C Are Smart Home Devices Conceptually and Practically Compatible with Australia's Existing
Legal Framework?
The DBN scheme attempts a balancing act between individual and corporate interests. 214The scheme asserts that individuals have a 'right to know' about unauthorised access to devices storing their information to facilitate mitigation of identity theft and other kinds of access likely to give rise to 'serious harm'.It is designed to protect those adversely affected by security breaches, 215 by letting 'individuals know that their data has slipped into unauthorised hands.' 216 The auxiliary aim is for mandatory DBN to act as a public information disclosure mechanism which improves organisational security control by encouraging sound informational and cybersecurity management as an organisational priority.The regulatory tool is framed with the consequence of reputational sanction. 217UT Law Review -Vol 18, No 2 | 288 In applying these aims, the contextual environment and its 'social application' is crucial. 218Consumers of smart home devices have a reasonable expectation of end-to-end secure connectivity. 219In a smart home network, a hacker may infiltrate a home automation system and manipulate appliances to cause physical and emotional attack.As more devices are connected to a smart home network, the accumulation of risks increases, and the interconnectivity between these devices is capable of causing problems such as uncoordinated administrators and differences in administrator preferences. 220nformation stored on these devices is also potentially accessible to an unlimited number of devices. 221his raises tensions as to whether the consumer expectation that a single smart home device will not 'create a backdoor to other devices in their home' remains achievable. 222CONCLUSION Since the enactment of mandatory DBN in February 2018, there have been 550 notifications reported as at the June to September 2018 Quarter. 223Of this total, 57 per cent were from malicious or criminal attacks, with the largest sources subject to notification being health service providers and the finance sector. 224To date, there have been no specific reported instances of notification in relation to smart home devices and as such security standards are yet to be enforced by the Commissioner.
It still remains to be seen how the DBN scheme and the introduction of the concept of 'jointly-held' information will be applied to breaches of smart home devices in Australia.Entities may no longer conceal cybersecurity breaches that have compromised their networks where an eligible data breach can be established and no relevant exceptions apply. 225In this regard, mandatory DBN may provide greater potential relief to affected consumers of smart home devices by creating a trend of increased transparency.Despite early criticisms of its practicality and enforceability, the scheme is therefore a welcome contribution to Australia's data breach notification regime. 226The benefits would be increasingly realised if, as recommended by various commentators, the OAIC were to implement a searchable public database recording notifiable data breaches to encourage organisational compliance to the scheme. 227ndatory DBN attempts to advance overarching objectives of deterrence, mitigation, transparency through information and public confidence. 228There are clear advantages and disadvantages of the merging of the DBN scheme with the current legal and regulatory privacy framework, and these are brought to the forefront in the context of smart home device breaches.The merge highlights 'vertical tensions' and 'shared horizontal weaknesses' between the current privacy law framework and the introduction of the DBN scheme. 229Vertically, there are inconsistencies in application of the DBN scheme, as an increase in the number of affected individuals may decrease the risk of 'serious harm' to each individual.The risk of 'serious harm' increases, however, in situations where information is 'jointly-held' and each individual is deemed to be more at risk.Horizontally, the adherence to PBR and 'light touch regulation' allows entities subject to the Privacy Act to dictate the terms in which smart home devices are designed and administered potentially without regard to cybersecurity protocols.While mandatory data breach notification may help to foster organisation culture and corporate social responsibility centred around privacy and 230 it may simultaneously encourage increased risk-taking, poor design level security protocols and 'creative compliance'. 231e extent to which the DBN scheme may apply to smart home device breaches is uncertain, but it is also unlikely to have much, if any, impact for breaches of small smart home networks.This is because the introduction of the scheme, whilst enforcing notification for serious breaches of some devices, may not prevent individual data breaches for other devices in a smart home network from becoming notifiable.There are issues with the conceptualisation of the definition of 'personal information' under the Privacy Act. 232The data collected by a smart home device, in its retention of personal preferences for automation of certain functions of the home, does not neatly fit under 'personal information' or 'sensitive information' within the meaning of section 6(1). 233The focus of the DBN scheme on 'data' may also allow an entity to obfuscate its obligations of mandatory notification by complying with the 'letter' of the scheme rather than its spirit.The most viable use for the scheme in relation to the smart home is its application to the concept of 'jointly-held information', which lacks historical legal basis.The interconnectivity of devices highlights unchartered territory in cybersecurity and an attempt to achieve legal certainty in an inherently uncertain area. 234Questions surrounding obligation and liability will inevitably arise as the concept of 'jointly-held information' gains traction and smart home devices become outdated and extend beyond their intended product life-cycle. 235is paper has argued that, while attempting to balance conflicting interests between individuals and entities, 236 the DBN scheme raises questions over the continuing viability of PBR in the wake of digital disruption.Gartner Consulting predicts that the smart home market is between five to ten years away from maturity. 237The global smart home market is projected to be worth around forty billion dollars by 2020. 238It is possible that a principle-based approach which allows overwhelming flexibility to the regulated entity is no longer feasible, as there may be no market-based solution to the issue of poor cybersecurity. 239A more prescriptive approach which specifies mandatory minimums and is less focused on ensuring flexibility for entities, or an approach which focuses on cybersecurity more generally rather than an emphasis on 'data', may be more tenable alternatives to traditional PBR in the rise of the smart home. 240 Kaman Tsoi and Mandy Milner, ''What Can I help You With?': Privacy and the Digital Assistant' (2016) 13(9) Privacy Law Bulletin 190.See generally the introduction of the DragonDictate released in 1997 and the release of the iPhone in 2007; See generally Sivaraman et al, above n 3. QUT Law Review -General Issue QUT Law Review -Vol 18, No 2 | 270 | 2019-05-12T14:24:57.454Z | 2019-03-01T00:00:00.000 | {
"year": 2019,
"sha1": "7aa1b2659170728f9ca986daa577eb25ed3f3c67",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.5204/qutlr.v18i2.752",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "7aa1b2659170728f9ca986daa577eb25ed3f3c67",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Sociology"
]
} |
52181706 | pes2o/s2orc | v3-fos-license | Gut Bacterial Communities in Geographically Distant Populations of Farmed Sea Bream (Sparus aurata) and Sea Bass (Dicentrarchus labrax)
This study investigated the profile of the autochthonous gut bacterial communities in adult individuals of Sparus aurata and Dicentrarchus labrax reared in sea cages in five distantly located aquaculture farms in Greece and determine the impact of geographic location on them in order to detect the core gut microbiota of these commercially important fish species. Data analyses resulted in no significant geographic impact in the gut microbial communities within the two host species, while strong similarities between them were also present. Our survey revealed the existence of a core gut microbiota within and between the two host species independent of diet and geographic location consisting of the Delftia, Pseudomonas, Pelomonas, Propionibacterium, and Atopostipes genera.
Introduction
Studies on fish gastrointestinal tract microbiota (GITM) are mostly focused on the isolation, identification and evaluation of microorganisms in farmed species. The main target of such studies is the possible use of these microorganisms as potential probiotics in order to promote fish growth and health [1]. With the advent of next generation sequencing technologies, results have demonstrated that fish GITM diversity shows higher complexity than originally considered [2]. Knowing the core microbiota (sensu [3]) is pivotal in predicting and further investigating the provided microbial services to the host [4], since these communities are important for the ecological understanding of the gut habitat and the functions of its microbes [5]. The investigation of co-occurrence patterns, including core and less frequent occurring microbes, has been shown to be extremely useful for depicting fundamental and keystone microbial species across same types of habitats-host in spatial and temporal scales [6]. Such approaches have shown that correlations between microbes and latitude can exist even for the human gut [7].
While dietary studies profiling the human gut microbiota pose certain limitations [8], sea cage farmed fish species can be a good model system to investigate fish core GITM since these populations are genetically homogeneous and consume a well-balanced diet that meets their nutritional requirements throughout their life cycle, while populations of the same species are reared in similar environmental conditions. For fish GITM, it has been suggested that these communities are not mere reflections of their host's habitat but are rather shaped by host-specific selective forces [9]. In this study, we compared the GITM of Sparus aurata and Dicentrarchus labrax individuals originating from five
Results and Discussion
In this study, we analyzed the midgut bacterial diversity of farmed Sparus aurata and Dicentrarchus labrax individuals in order to determine members of the adult core microbiota of these commercially important fish species. Taking into account that microbiota are important in health and disease, revealing the core microbiota of a species would be important in order to explore how to achieve a beneficial collaboration between host and microbiota (for a review see [5]). The analyzed fish individuals had the following features: (a) common genetic origin, (b) very similar supplied commercial feed (Table S2), (c) origin from distant aquaculture farms (23-554 km between them), (d) similar age, and (e) were sampled within days. These criteria allowed us to assess the core microbiota of these animals by minimizing the effects of host genetics, nutritional state and environmental stressors (e.g., salinity, temperature) variability. In this study, core OTUs refer to the ones found in each individual midgut sample. The effect of the surrounding water was not studied since it is expected to be insignificant for the GITM diversity as shown previously [20][21][22][23]. A single water sampling on the same day of the midgut sampling of the investigated fish individuals, would not be so informative due to the following two factors: 1) marine bacterioplankton is characterized by strong variation in short (e.g., [24,25]) and longer [26][27][28][29] time scales, and 2) the life cycle of farmed S. aurata and D. labrax spans over several months. To the best of our knowledge, this is the first study combining all the above features for the gut bacterial communities of S. aurata and D. labrax. Floris et al. [30] investigated the gut microbiota of S. aurata at two coastal lagoons in Sardinia, Italy, but their study was based on older techniques with limited power to uncover the full extent of biodiversity. For D. labrax, there are a few relevant studies but were mainly focused on candidate probiotic's evaluation [31][32][33][34] and the effect of alternative feed ingredients in gut microbiome [35].
Despite the low reads numbers in some samples (Table 1) rarefaction curves have reached a plateau ( Figure S2), indicating satisfactory coverage of the existing bacterial OTUs. The effect of different aquaculture location on bacterial species richness was not important since OTUs richness between locations did not vary significantly ( Figure S3). Each species had a rather defined bacterial community, with 10-21 OTUs accounting for ≥80% of the relative abundance per sample ( Table 1).
Within Proteobacteria in S. aurata, Betaproteobacteria was the dominant class in four locations (Yaltra, Chania, Chios, Igoumenitsa; Greece), while in Atalanti, Betaproteobacteria and Gammaproteobacteria co-dominated, with 22.1% and 23.7%, respectively ( Figure S4). Other than this, Gammaproteobacteria was the second most abundant class of Proteobacteria, with Alphaproteobacteria always in low abundances ( Figure S4). On the contrary, in D. labrax, Gammaproteobacteria dominated in three locations (Chania, Yaltra, Atalanti; Greece) followed by Betaproteobacteria and Alphaproteobacteria. In the rest of the locations (Igoumenitsa and Chios; Greece), Betaproteobacteria was the dominant taxon. In general, Alphaproteobacteria abundances in D. labrax were higher than in S. aurata ( Figure S4). The most abundant orders in all locations for both host species were the Micrococcales, Corynebacteriales, Propionibacteriales, Bifidobacteriales, Flavobacteriales, Bacteroidales, Bacillales, Lactobacillales, Burkholderiales and Pseudomonadales.
A small set of OTUs was found to occur in all individuals from all five locations (8 in S. aurata and 10 in D. labrax), i.e., representing the core mid gut microbiota (sensu [3]) for each species (Figure 1). Moreover, five of these OTUs ( Figure 1) were shared between the two species. The closest phylogenetic relatives for these OTUs were Delftia acidovorans (Burkholderiales), Pseudomonas panacis (Pseudomonadales), Pelomonas puraquae (Burkholderiales), Propionibacterium acnes (Propionibacteriales) and Atopostipes suicloacalis (Lactobacillales). (Table S3). The estimation of the shared OTUs doubling time (based on the 16S rDNA gene copy number [38] ranged between 0.8 and 2.0 h -1 (Table S3), implying that they represent bacteria which can grow fast in the fish GIT and thus, they are more likely to outcompete other bacterial taxa. Delftia spp. have been previously retrieved from fish gut of healthy grouper [39], rainbow trout [40], 2012), and Atlantic salmon [41] individuals. The members of this genus are strictly aerobic and chemoorganotrophic but not fermentative [42]. Pseudomonas spp. have been isolated from several fish species and have been evaluated as potential probiotics in aquaculture industry [43][44][45]. Pelomonas sp. could be a resident GITM as it has been found in the gut of farmed fish [46]. Propionibacterium acnes is commonly found in fish [47][48][49][50]21] and snails [51] but its major importance for the human skin microbiome [52] renders it as an uncertain autochthonous gut symbiont for S. aurata and D. labrax. Atopostipes is a fermentative genus and to date it has been associated with fermented flesh of skate (Raja pulchra) [53] but also with the Atlantic salmon (Salmo salar) gut [49]. Thus, it is likely a bacterium with potential fermentative role in farmed S. aurata and D. labrax.
S. aurata shared OTUs belonged to taxa (Burkholderiales, Pseudomonadales, Flavobacteriales, Actinobacteria) reported in wild, organic and conventionally reared S. aurata individuals [48] while the identified closest relatives of these OTUs have been previously retrieved from similar environments (Table S3). This further suggests that these bacteria could be members of S. aurata core bacterial community.
The observed core bacterial community for both species consisted mostly from nonsporulating, mesophilic bacteria, with diverse types of respiration with some of them presenting important features in other animals. For example, Micrococcus luteus possesses anti-Vibrio activity in the freshwater fish Oreochromis niloticus; Pseudomonas panacis degrades cellulose in the gut of the bark beetles Dendroctonus armandi, while P. veronii has been reported to have metabolic pathways related to central carbohydrate metabolism, nutrients uptake and plant hormone auxin production in the grapevine, Vitis vinifera, root [54]. Delftia spp. have been previously retrieved from fish gut of healthy grouper [39], rainbow trout [40], 2012), and Atlantic salmon [41] individuals. The members of this genus are strictly aerobic and chemo-organotrophic but not fermentative [42]. Pseudomonas spp. have been isolated from several fish species and have been evaluated as potential probiotics in aquaculture industry [43][44][45]. Pelomonas sp. could be a resident GITM as it has been found in the gut of farmed fish [46]. Propionibacterium acnes is commonly found in fish [21,[47][48][49][50] and snails [51] but its major importance for the human skin microbiome [52] renders it as an uncertain autochthonous gut symbiont for S. aurata and D. labrax. Atopostipes is a fermentative genus and to date it has been associated with fermented flesh of skate (Raja pulchra) [53] but also with the Atlantic salmon (Salmo salar) gut [49]. Thus, it is likely a bacterium with potential fermentative role in farmed S. aurata and D. labrax.
S. aurata shared OTUs belonged to taxa (Burkholderiales, Pseudomonadales, Flavobacteriales, Actinobacteria) reported in wild, organic and conventionally reared S. aurata individuals [48] while the identified closest relatives of these OTUs have been previously retrieved from similar environments (Table S3). This further suggests that these bacteria could be members of S. aurata core bacterial community.
The observed core bacterial community for both species consisted mostly from nonsporulating, mesophilic bacteria, with diverse types of respiration with some of them presenting important features in other animals. For example, Micrococcus luteus possesses anti-Vibrio activity in the freshwater fish Oreochromis niloticus; Pseudomonas panacis degrades cellulose in the gut of the bark beetles Dendroctonus armandi, while P. veronii has been reported to have metabolic pathways related to central carbohydrate metabolism, nutrients uptake and plant hormone auxin production in the grapevine, Vitis vinifera, root [54]. Most of the rest core gut bacterial OTUs, in S. aurata and D. labrax were assigned to similar orders such as Corynebacteriales, Pseudomonadales and Micrococcales, though in different species (Table S2). Most of them have been retrieved from similar isolation sources (e.g., stool, intestine, manure) (Table S3). The number of OTUs occurring only in one location varied between 11-61 and 22-55 for S. aurata and D. labrax, respectively ( Figure S4).
Geographic distance between the aquaculture farms did not show any correlation with the gut bacterial community structure for both species ( Figure S5) and nonmetric multidimensional scaling (NMDS, Figure S6) based on the Bray-Curtis distance of presence/absence OTUs, showed no clear geographic separation (ANOSIM using Euclidean distances p = 0.391, R = 9.3 −5 ) of the gut bacterial communities for both species as well. This implies, that the observed GITM structure for each of the two fish species investigated in this study are not related to the vicinity of the aquaculture farms.
In the current study, the correlations between the overlap and dissimilarity of GITM communities structure were positive for both fish species considered (r = 0.477 and 0.574 for p < 0.002 in S. aurata and D. labrax, respectively) (Figure 2), suggesting high inter-individual variability in terms of OTUs abundances even in the same location. Similar results have also been observed in fecal microbiota for both S. aurata and D. labrax [55,56]. While in human gut microbiome, the inter-individual variability is more easily understood due to parameters such as dietary patterns and personal interests [57,58], here we concluded that inter-individual variability in the autochthonous gut bacteria of D. labrax and S. aurata, is more likely related with individual genetic factors. The observed inter-individual variability means that the gut microenvironment of these two host fish species promotes selective pressure in the bacterial communities. However, while the overlap of these bacterial communities increases, the same happens with dissimilarity, indicating host-independent parameters also shaping gut bacterial community in human [18,59] and fish GITM [60,61].
The most prominent factors promoting the inter-individual microbiota variation have only recently been taken into account and these are host genotype, gut colonization during the early developmental stages, environmental effects on GITM acquisition, diet, diseases and respective medication [8]. One reason for the GITM inter-individual variability is that caged fish are fed mechanically, a way that does not secure equal food consumption for each fish due to individual differences in their activity. The extent of GITM individual variability is important to know for the following reasons: (a) it dictates the number of replicate samples per species that need to be analyzed [58], (b) it helps distinguishing between autochthonous (resident) bacteria which colonize the gut mucosa and the allochthonous (transient) bacteria occurring mostly in the digesta [9,62]. The demonstrated individual host variability could be the reason for the low number of shared OTUs in both allopatric populations studied here, but larger datasets are required in order to fully unravel this issue. Although one-way analysis of variance (ANOVA) revealed no statistically significant differences between gut bacterial communities for both species (p > 0.05), (Figure S3), the biological relations of the bacterial communities were different. The ratio of positive to total (PT) correlations of the most dominant OTUs of S. aurata and D. labrax individuals was significantly different (p < 0.05), suggesting different biological relationships in the guts of the two species (Figure 3). The high ratio of the positive to total (PT) correlations of the most dominant operational taxonomic units demonstrates that the majority of the dominant bacteria have either cooperative interactions or, at least, they do not participate in competitive nutrition. Such relationships in microbial populations are believed to be beneficial to the host as they ensure high capacity of utilizing the complex array of available substrates found in the gut [63,35,64]. Although one-way analysis of variance (ANOVA) revealed no statistically significant differences between gut bacterial communities for both species (p > 0.05), (Figure S3), the biological relations of the bacterial communities were different. The ratio of positive to total (PT) correlations of the most dominant OTUs of S. aurata and D. labrax individuals was significantly different (p < 0.05), suggesting different biological relationships in the guts of the two species (Figure 3). The high ratio of the positive to total (PT) correlations of the most dominant operational taxonomic units demonstrates that the majority of the dominant bacteria have either cooperative interactions or, at least, they do not participate in competitive nutrition. Such relationships in microbial populations are believed to be beneficial to the host as they ensure high capacity of utilizing the complex array of available substrates found in the gut [35,63,64]. Microorganisms 2018, 6, x FOR PEER REVIEW 8 of 12 8
Conclusions
It is still unknown whether and how the gut microbial communities of fish can contribute nutrients and energy to the host and maintain a balance with the fish's metabolism and immune system. This study presents evidence for core gut bacterial communities within the two examined host species (S. aurata and D. labrax), and also a small set of OTUs that have been found in common between them, indicating that some autochthonous gut bacterial representatives of the Delftia, Pseudomonas, Pelomonas, Propionibacterium and Atopostipes genera can colonize different host species. Despite the inter-individual variability and the distance of each farm location, there is no significant difference between the gut bacterial communities in the two host species. The results also revealed these gut bacterial communities form different biological relations between their members as revealed by their populations association networks.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1, Table S1. Body weight of the Sparus aurata and Dicentrarchus labrax individuals used in this study; Table S2. Ingredients of the diets used at the time of sampling; Table S3. Bacterial 16S rDNA operational taxonomic units (OTU) found in the midgut of commercially reared Sparus aurata and Dicentrarchus labrax individuals from different aquaculture sites in Greece; Figure S1. Aquaculture sampling sites. I: Igoumenitsa, Y, Yaltra, A: Atalanti, Ch: Chios, C: Chania; Figure S2. Rarefaction curves bacterial operational taxonomic units generated by16S rDNA tag pyrosequencing from the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S3. Box-plot of the bacterial operational taxonomic units found in the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S4. Taxonomy (phyla: top row; Proteobacteria sub-phyla: bottom row) of the found bacterial operational taxonomic units found in the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S5. Relationship of the shared operational taxonomic units (OTUs) and the total number of OTUs with the distance between different Sparus aurata and Dicentrarchus labrax aquaculture sites in Greece; Figure S6. Non-metric multidimensional scaling (NMDS) based on the gut bacterial operational taxonomic units between Sparus aurata and Dicentrarchus labrax individuals from different aquaculture sites in Greece. Red and blue lines include all S. aurata and D. labrax samples, respectively.
Author Contributions: E.N.: performed field and laboratory analyses, analyzed the data, and contributed to paper writing; A.M.: analyzed the data and contributed to paper writing; E.A.: conceived the study and contributed to paper writing; E.M.: conceived the study and contributed to paper writing; K.A.K: conceived the study, analyzed the data, and wrote the first draft of the paper.
Conclusions
It is still unknown whether and how the gut microbial communities of fish can contribute nutrients and energy to the host and maintain a balance with the fish's metabolism and immune system. This study presents evidence for core gut bacterial communities within the two examined host species (S. aurata and D. labrax), and also a small set of OTUs that have been found in common between them, indicating that some autochthonous gut bacterial representatives of the Delftia, Pseudomonas, Pelomonas, Propionibacterium and Atopostipes genera can colonize different host species. Despite the inter-individual variability and the distance of each farm location, there is no significant difference between the gut bacterial communities in the two host species. The results also revealed these gut bacterial communities form different biological relations between their members as revealed by their populations association networks.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/6/3/92/s1, Table S1. Body weight of the Sparus aurata and Dicentrarchus labrax individuals used in this study; Table S2. Ingredients of the diets used at the time of sampling; Table S3. Bacterial 16S rDNA operational taxonomic units (OTU) found in the midgut of commercially reared Sparus aurata and Dicentrarchus labrax individuals from different aquaculture sites in Greece; Figure S1. Aquaculture sampling sites. I: Igoumenitsa, Y, Yaltra, A: Atalanti, Ch: Chios, C: Chania; Figure S2. Rarefaction curves bacterial operational taxonomic units generated by16S rDNA tag pyrosequencing from the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S3. Box-plot of the bacterial operational taxonomic units found in the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S4. Taxonomy (phyla: top row; Proteobacteria sub-phyla: bottom row) of the found bacterial operational taxonomic units found in the midgut of Sparus aurata and Dicentrarchus labrax individuals originating from different aquaculture farms in Greece; Figure S5. Relationship of the shared operational taxonomic units (OTUs) and the total number of OTUs with the distance between different Sparus aurata and Dicentrarchus labrax aquaculture sites in Greece; Figure S6. Non-metric multidimensional scaling (NMDS) based on the gut bacterial operational taxonomic units between Sparus aurata and Dicentrarchus labrax individuals from different aquaculture sites in Greece. Red and blue lines include all S. aurata and D. labrax samples, respectively.
Author Contributions: E.N.: performed field and laboratory analyses, analyzed the data, and contributed to paper writing; A.M.: analyzed the data and contributed to paper writing; E.A.: conceived the study and contributed to paper writing; E.M.: conceived the study and contributed to paper writing; K.A.K: conceived the study, analyzed the data, and wrote the first draft of the paper. | 2018-09-16T06:22:59.998Z | 2018-09-01T00:00:00.000 | {
"year": 2018,
"sha1": "298fc1b960ef1e9cc25fbc0c913539f3669ddd92",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3390/microorganisms6030092",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a4bc05b1cd22878c3ad9635a5288d984d341cf88",
"s2fieldsofstudy": [
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.