pmid stringlengths 8 8 | pmcid stringlengths 8 11 ⌀ | source stringclasses 2
values | rank int64 1 9.78k | sections unknown | tokens int64 3 46.7k |
|---|---|---|---|---|---|
35966660 | PMC9363821 | pmc | 880 | {
"abstract": "Microbial biofilms show high phenotypic and genetic diversity, yet the mechanisms underlying diversity generation and maintenance remain unclear. Here, we investigate how spatial patterns of growth activity within a biofilm lead to spatial patterns of genetic diversity. Using individual-based computer simulations, we show that the active layer of growing cells at the biofilm interface controls the distribution of lineages within the biofilm, and therefore the patterns of standing and de novo diversity. Comparing biofilms of equal size, those with a thick active layer retain more standing diversity, while de novo diversity is more evenly distributed within the biofilm. In contrast, equal-sized biofilms with a thin active layer retain less standing diversity, and their de novo diversity is concentrated at the top of the biofilm, and in fewer lineages. In the context of antimicrobial resistance, biofilms with a thin active layer may be more prone to generate lineages with multiple resistance mutations, and to seed new resistant biofilms via sloughing of resistant cells from the upper layers. Our study reveals fundamental “baseline” mechanisms underlying the patterning of diversity within biofilms.",
"introduction": "1. Introduction Understanding how diversity is maintained within populations is one of the most important challenges in ecology and evolution (Barton and Keightley, 2002 ; Gibbons and Gilbert, 2015 ; Shade, 2017 ). Populations can adapt to changing environments via selection on pre-existing diversity (standing variation), and/or via selection on new ( de novo ) mutations, with different implications for the speed and nature of adaptation (Barrett and Schluter, 2008 ). The factors controlling the balance between standing and de novo diversity remain a topic of debate even for well-mixed populations (Barrett and Schluter, 2008 ). For spatially structured populations such as microbial biofilms the picture is more complex, since spatial structure can have drastic effects on evolutionary dynamics (Korona et al., 1994 ; Stewart and Franklin, 2008 ; Stacy et al., 2015 ). Expanding populations are often characterized by genetic drift at the expanding front, leading to lineage loss and spatial segregation of surviving lineages (Habets et al., 2006 ; Hallatschek and Nelson, 2008 , 2010 ; Perfeito et al., 2008 ; Excoffier et al., 2009 ; Nadell et al., 2010 ; Korolev et al., 2011 ; Freese et al., 2014 ; Mitri et al., 2016 ; Giometto et al., 2018 ). This has implications for the evolutionary maintenance of cooperative phenotypes (Ben-Jacob et al., 1994 ; Kreft, 2004 ; Habets et al., 2006 ; Bollback and Huelsenbeck, 2007 ; Park and Krug, 2007 ; Hallatschek and Nelson, 2008 , 2010 ; Perfeito et al., 2008 ; Excoffier et al., 2009 ; Nadell et al., 2010 , 2016 ; Korolev et al., 2011 ; Martens and Hallatschek, 2011 ; Mitri et al., 2011 , 2016 ; Good et al., 2012 ; Mitri and Foster, 2013 ; Frost et al., 2018 ). In addition, some lineages that are located right at the growing front can expand dramatically, in a phenomenon known as gene surfing (Hallatschek et al., 2007 ; Hallatschek and Nelson, 2008 , 2010 ; Gralka et al., 2016 ). Such spatial effects strongly influence the distribution of clone sizes for de novo mutations: bacterial colonies exhibit more jackpot events (large clones) compared to well-mixed populations (Fusco et al., 2016 ). Spatial effects can also lead to fragmentation of the population into independently evolving subpopulations (Fux et al., 2005 ; Steenackers et al., 2016 ). Moreover, evolutionary dynamics feeds back on the spatial structure of the population, for example through changes in growth speed or adhesive capacity (Kim et al., 2014 ; Steenackers et al., 2016 ; Kayser et al., 2018 ). Microbial biofilms are widely observed to be phenotypically and genetically diverse (Hall-Stoodley et al., 2004 ; Stewart and Franklin, 2008 ; Stacy et al., 2015 ). This diversity is ecologically important, and probably contributes to the tolerance of clinical biofilms to antibiotic treatment (Mah and O'Toole, 2001 ; Stewart, 2002 ; Fux et al., 2005 ; Excoffier et al., 2009 ; Hallatschek and Nelson, 2010 ; Kim et al., 2014 ; Nadell et al., 2016 ; Frost et al., 2018 ). In environmental or clinical contexts, biofilms are likely to be seeded from genetically diverse inocula, such as skin, gut, soil, ocean, or river microbiota, so that standing variation may play a significant role. However, biofilms can also act as sources of de novo variation (Korona et al., 1994 ; Stewart and Franklin, 2008 ; Stacy et al., 2015 ). As we discuss below, spatial structure can drastically affect mutant fixation probabilities (Kim et al., 2014 ; Fusco et al., 2016 ). Spatial gradients of selection pressure, such as antibiotic, within the biofilm may also accelerate the emergence of resistant mutants, while the biofilm environment may favor the emergence of mutator strains and/or the horizontal transfer of genetic material (Stewart, 2002 ). In addition, spatial structure may promote the evolution of specific phenotypes that are well-adapted to the biofilm environment (Ben-Jacob et al., 1994 ; Nadell et al., 2010 , 2016 ; Mitri et al., 2011 ; Mitri and Foster, 2013 ; Frost et al., 2018 ). Biofilms are characterized by an uneven distribution of growth activity. Nutrients are rapidly consumed at the growing edge of the biofilm, so that the interior becomes nutrient-depleted. Therefore, growth is limited to a well-defined layer close to the biofilm front, where nutrient has not yet been consumed (Stewart and Franklin, 2008 ; Stacy et al., 2015 ; Stewart et al., 2016 ). This is known as the active layer ; it has been observed in in vitro experiments (Pamp et al., 2008 ; Stewart et al., 2016 ) and in ex vivo clinical lab samples (Stewart et al., 2016 ), as well as in simulations (Xavier et al., 2004 ; Nadell et al., 2010 , 2013 ; Young et al., 2022 ) and theory (Korolev et al., 2010 ). The width of the active layer is controlled by the balance between nutrient supply and consumption (Nadell et al., 2010 ). Hence, nutrient availability, nutrient consumption rate, nutrient diffusivity, biomass density and growth yield all affect the active layer width (Nadell et al., 2010 ). The active layer width is closely coupled to biofilm morphology: biofilms with thin active layers tend to have rough interfaces, while those with thick active layers tend to be smooth (Nadell et al., 2010 ; Farrell et al., 2013 ; Young et al., 2022 )—although dynamical fluctuations of the active layer are also important (Young et al., 2022 ). In this study, we investigate in detail how the spatial pattern of growth activity within biofilms leads to spatial patterns of standing and de novo diversity. Using individual-based biofilm simulations, we track the fate of hundreds of neutral cell lineages in growing biofilms. Our simulations allow direct observation of the loss of standing diversity, and we infer the gain of de novo diversity from patterns of lineage length. In this work, we choose to compare biofilms grown to equal size , under conditions where the active layer thickness is different. Our study complements previous work by Mitri et al. ( 2016 ), who studied diversity in bacterial colonies, grown for equal time with differing nutrient availability. Increasing nutrient availability increases the active layer width (Nadell et al., 2010 ). Mitri et al. ( 2016 ) observed that well-fed colonies retain standing diversity over more generations than poorly fed colonies; however over a similar timescale, well-fed colonies undergo more generations of growth than poorly-fed ones. Therefore, comparing colonies over the same timescale, well-fed and poorly-fed colonies retain similar amounts of standing diversity since the differences in colony size compensate for the differences in active layer thickness. Here, our aim is to understand the fundamental role of the active layer, for which the picture is clearer when we compare biofilms of equal size. We find that active layer thickness controls both the balance between standing and de novo variation, and the spatial patterns of de novo mutations within the biofilm. For biofilms of equal size, those with a thick active layer retain more standing diversity and their de novo diversity is more evenly distributed across the biofilm. In contrast, biofilms with a thin active layer retain less standing diversity, and their de novo diversity is concentrated close to the growing interface. Since de novo diversity is concentrated in fewer lineages, the occurrence of multiple mutations along the pathway to high-level antibiotic resistance is more likely in biofilms with thinner active layers. In this study, we do not aim to represent biofilm growth and evolution in realistic detail, but rather to provide a baseline model that reveals fundamental mechanisms connecting spatial patterning of growth and diversity, onto which more complex effects can be superposed.",
"discussion": "4. Discussion Biofilms often show high levels of genetic diversity, which is believed to contribute to antibiotic tolerance and resistance (Mah and O'Toole, 2001 ; Stewart, 2002 ). Understanding whether this diversity primarily arises from pre-existing (standing) variation or from newly generated ( de novo ) variation has significant implications. For example, adaptation to environmental challenges is generally faster from a basis of standing variation (Barrett and Schluter, 2008 ). Here, we used an individual-based biofilm model, to show how the spatial patterns of microbial growth within a biofilm lead to spatial patterns of standing and de novo diversity. Our work reveals a central role for the active layer of growing microbes at the biofilm interface. Comparing biofilms of equal size, a biofilm with a thick active layer retains more standing diversity, and its de novo diversity is more evenly distributed, both spatially and among individuals in the population. In contrast, a biofilm with a thin active layer retains less standing diversity, and its de novo diversity is concentrated close to the biofilm interface, with relatively less de novo diversity being located in the deeper parts of the biofilm. This implies that microbes with multiple mutations, leading to high-level antibiotic resistance, are more likely in biofilms with a thin active layer, compared to biofilms of equal size with a thick active layer. We also find that the local dynamics of the active layer plays a role, for example, causing local hot spots of loss of standing variation when interface bulges are lost behind the growing front. Putting our results together, our model predicts contrasting spatial patterns of standing diversity and de novo diversity. Standing diversity is greatest in the lower parts of the biofilm, while de novo diversity is greatest at the top of the biofilm. This could have consequences when biofilms are subjected to environmental challenges. For example, antibiotics that target primarily the active, upper, part of the biofilm would tend to select on de novo diversity, while those that target primarily the lower part of the biofilm might select on standing diversity (Pamp et al., 2008 ). Likewise, sloughing of the upper layers of a biofilm might disperse de novo diversity to the wider environment, while leaving standing diversity in place. In this work, we compared biofilms grown to equal size , with different active layer thickness, achieved by varying the parameters of our individual-based model. In doing this, we follow the work of Drescher et al. ( 2016 ), who also point to biofilm size, rather than age, as a key control parameter. This contrasts with the work of Mitri et al. ( 2016 ), who compared bacterial colonies grown for equal time , on media with varying nutrient availability. Mitri et al. ( 2016 ) found that, overall, nutrient availability had little effect on loss of standing diversity, because the differences in colony size counteracted the effects of the active layer thickness. In this work, we aimed to elucidate the fundamental mechanisms by which growth patterning leads to patterning of diversity. These mechanisms are clearer when we compare biofilms of equal size. One might argue that comparing biofilms of equal size restricts the practical relevance of our conclusions, since slow-growing biofilms will generally be smaller than fast-growing ones. However, in the natural environment, biofilm maturity does not necessarily correspond to increasing size: biofilm growth can be limited by space (e.g., inside a cavity in a medical implant) or by chemical interactions (e.g., the secretion of pulcherrimin which causes growth arrest in Bacillus subtilis colonies; Arnaouteli et al., 2019 ). Bearing in mind that our comparison is made for biofilms of equal size, it would be important to carefully define the conditions for any experimental test of these predictions. To control the active layer thickness in our simulations, we varied two model parameters: the bulk nutrient concentration S bulk and the maximal specific growth rate μ max . We could have chosen to vary a single parameter. For example, increasing S bulk alone (as in the study of Mitri et al., 2016 ) increases the active layer thickness, but it also increases the average activity of microbes within the active layer ( Supplementary Figure 6 and Supplementary Table 1 ). Increasing μ max alone decreases the active layer thickness, while increasing the average activity of microbes within the active layer ( Supplementary Figure 6 and Supplementary Table 1 ). By varying multiple parameters, we can identify the active layer thickness as the controlling factor, rather than other factors, such as the activity of individual microbes, that correlate with individual parameters. Importantly, we have assumed neutrality in this study: a priori , all microbial agents in our simulations have equal fitness and identical traits. This allows us to predict patterns of mutations within the biofilm from lineage lengths, without explicitly simulating mutation events. Neutral models have a distinguished history in ecology and evolution (Volkov et al., 2003 ; Azaele et al., 2006 ); they are useful for predicting baseline phenomena, deviations from which can point to specific biological mechanisms. In this study, the predicted baseline phenomenon is the connection between the active layer and patterns of standing and de novo diversity. Neutral models do not provide a realistic description of the real biological system, but they do provide a useful reference to which to compare biological measurements (Nee, 2005 ). Similarly, our study aims to elucidate baseline mechanisms, rather than to provide a realistic model for an evolving biofilm. Our model neglects many biological and physical factors, including fitness effects of mutations, antibiotic effects on mutation rates, the emergence of hypermutators, persisters, physical effects of exopolysaccharide production, 3D geometric effects and fluid flow. All of these could produce different outcomes for the patterning of standing and de novo diversity within a biofilm, and should be investigated in future work. Feedback between evolutionary processes and the spatial structure of the population (e.g., the formation of biofilm bulges by fitter mutant clones, or a change in the local active layer thickness due to a mutant with a different growth yield) could also have interesting effects. Previous work on evolution in spatially expanding microbial populations has focused on the distribution of clone sizes, i.e., the number of descendants of a mutant that emerges within the population (Hallatschek et al., 2007 ; Hallatschek and Nelson, 2008 , 2010 ; Fusco et al., 2016 ; Gralka et al., 2016 ; Farrell et al., 2017 ; Schreck et al., 2019 ). The clone size distribution is different in a spatially expanding population compared to a well-mixed population; for example, mutants that emerge right at the front can be carried along at the front and produce large clone sizes even in the absence of fitness benefits, in a phenomenon known as gene surfing (Hallatschek et al., 2007 ; Hallatschek and Nelson, 2008 , 2010 ; Gralka et al., 2016 ; Farrell et al., 2017 ). In this work, we consider de novo diversity from a different perspective. While the clone size distribution considers the number of descendants arising from an individual mutation event, here we predict the total number of mutations (of any type) that are located at a particular spatial position within the biofilm. By tracking the lineages of microbes within the biofilm, we can predict patterns of de novo diversity, in terms of predicted mutation density, within the biofilm. However, since we do not connect the lineages of different microbes within the biofilm (i.e., we do not measure relatedness between individuals), we cannot track the fate of particular mutations. Therefore our work provides a different and complementary approach to understanding patterns of de novo diversity. Computer simulations provide a powerful way to investigate phenomena that might be difficult to study experimentally, but they are not a substitute for experimental data. Tracking of lineages within experimental microbial populations is now possible, for well-mixed populations, using barcoding methods, although this has not been used for spatially structured populations (Blundell et al., 2019 ; Jasinska et al., 2020 ). For biofilms, advanced image analysis of growing biofilms allows the tracking of cell lineages in space and time (Jeckel and Drescher, 2021 ). Spatially-resolved detection of point mutations is challenging at present, but may well become possible in future. Therefore, experimental tests of the ideas presented in this work, although difficult, are not out of the question."
} | 4,503 |
33106602 | PMC7610411 | pmc | 881 | {
"abstract": "Eukaryogenesis is one of the most enigmatic evolutionary transitions, during which simple prokaryotic cells gave rise to complex eukaryotic cells. While evolutionary intermediates are lacking, gene duplications provide information on the order of events by which eukaryotes originated. Here we use a phylogenomics approach to reconstruct successive steps during eukaryogenesis. We found that gene duplications roughly doubled the proto-eukaryotic gene repertoire, with families inherited from the Asgard archaea-related host being duplicated most. By relatively timing events using phylogenetic distances we inferred that duplications in cytoskeletal and membrane trafficking families were among the earliest events, whereas most other families expanded predominantly after mitochondrial endosymbiosis. Altogether, we infer that the host that engulfed the proto-mitochondrion had some eukaryote-like complexity, which drastically increased upon mitochondrial acquisition. This scenario bridges the signs of complexity observed in Asgard archaeal genomes to the proposed role of mitochondria in triggering eukaryogenesis.",
"discussion": "Discussion This large-scale analysis of duplications during eukaryogenesis provides compelling evidence for a mitochondria-intermediate eukaryogenesis scenario. The results suggest that the Asgard archaea-related host already had some eukaryote-like cellular complexity, such as a dynamic cytoskeleton and membrane trafficking. Upon mitochondrial acquisition there was an even further increase in complexity with the establishment of a complex signalling and transcription regulation network and by shaping the endomembrane system. These post-endosymbiosis innovations could have been facilitated by the excess of energy allegedly provided by the mitochondrion 28 , 29 . A relatively complex host is in line with the presence of homologues of eukaryotic cytoskeletal and membrane trafficking genes in Asgard archaeal genomes 5 , 6 , 30 . Moreover, some of them, including ESCRT-III homologues, small GTPases and (loki)actins, have duplicated in these archaea as well, either before eukaryogenesis or more recently 5 , 6 , 30 . This indicates that there has already been a tendency for at least the cytoskeleton and membrane remodelling to become more complex in Asgard archaeal lineages. A dynamic cytoskeleton and trafficking system, perhaps enabling primitive phagocytosis 31 , might have been essential for the host to take up the bacterial symbiont. Molecular and cell biology research in these archaea, from which the first results have recently become public 32 , 33 , is highly promising to yield more insight into the nature of the host lineage. In addition to a reconstruction of the host, further exploration of the numerous acquisitions, inventions and duplications during eukaryogenesis is key to fully unravelling the origin of eukaryotes."
} | 718 |
32810127 | PMC7480867 | pmc | 882 | {
"abstract": "Social interaction between microbes can be described at many levels of details: from the biochemistry of cell-cell interactions to the ecological dynamics of populations. Choosing an appropriate level to model microbial communities without losing generality remains a challenge. Here we show that modeling cross-feeding interactions at an intermediate level between genome-scale metabolic models of individual species and consumer-resource models of ecosystems is suitable to experimental data. We applied our modeling framework to three published examples of multi-strain Escherichia coli communities with increasing complexity: uni-, bi-, and multi-directional cross-feeding of either substitutable metabolic byproducts or essential nutrients. The intermediate-scale model accurately fit empirical data and quantified metabolic exchange rates that are hard to measure experimentally, even for a complex community of 14 amino acid auxotrophies. By studying the conditions of species coexistence, the ecological outcomes of cross-feeding interactions, and each community’s robustness to perturbations, we extracted new quantitative insights from these three published experimental datasets. Our analysis provides a foundation to quantify cross-feeding interactions from experimental data, and highlights the importance of metabolic exchanges in the dynamics and stability of microbial communities.",
"introduction": "Introduction Most microorganisms that affect the environments we live in [ 1 ] and that impact our health [ 2 ] do not live in isolation: they live in complex communities where they interact with other strains and species. The past decade has seen a surge of scientific interest in microbial communities, such as the human microbiome, but most studies remain limited to cataloguing community composition [ 3 ]. Our mechanistic understanding of how biochemical processes occurring inside individual microbial cells command interaction between cells, and lead to the emergent properties of multi-species communities remains limited [ 4 ]. Microorganisms consume, transform and secrete many kinds of chemicals, including nutrients, metabolic wastes, extracellular enzymes, antibiotics and cell-cell signaling molecules such as quorum sensing autoinducers [ 5 – 8 ]. The chemicals produced by one microbe can impact the behaviors of others by promoting or inhibiting their growth [ 9 ], creating multi-directional feedbacks that can benefit or harm the partners involved [ 10 , 11 ]. If a community is well-characterized and given sufficient data on population dynamics, it should be possible to parameterize the processes involved in microbe-microbe interactions by fitting mathematical models [ 12 ]. Any model can potentially yield insights [ 13 ], but the complexity of most models so far has been either too high for parameterization [ 14 ], or too low to shed light on cellular mechanisms [ 15 ]. Microbial processes may be modelled across a range of details: At the low end of the spectrum we have population dynamic models such as generalized Lotka-Volterra (gLV) [ 16 ] and Consumer-Resource (C-R) models [ 17 ], which treat each organism as a ‘black-box’. For example, C-R models assume a linear or Monod dependence of microbial growth on resource uptake kinetics. At the high end of the spectrum, we have detailed single-cell models such as dynamic flux balance analysis (dFBA) [ 18 ] and agent-based models [ 19 ] that have too many parameters to be parameterizable by experimental data. For example, the linear equations for fluxes obtained from quasi-steady-state assumption of dFBA are underdetermined. What is an appropriate level of detail to model and constrain microbial processes using data, to produce accurate predictions as well as new mechanistic insights? Here we propose a generalizable framework that couples classical ecological models of population and resource dynamics with coarse-grained intra-species metabolic networks. We show that modeling communities at this intermediate scale can accurately quantify metabolic processes from population dynamics data acquired in the laboratory. We demonstrate the approach on three evolved/engineered communities of Escherichia coli ( E . coli ) strains with increasing levels of complexity: (1) unilateral acetate-mediated cross-feeding [ 20 ], (2) bilateral amino-acid-mediated cross-feeding between leucine and lysine auxotrophies [ 21 ], and (3) multilateral amino-acid-mediated cross-feeding between 14 distinct amino acid autotrophies [ 22 ]. The parameterized models report inferred leakage fractions of metabolic byproducts that are difficult to measure directly by experiments, reveal how resource supply and partitioning alter the coexistence and ecological relationships between cross-feeders, and predict the limits of community robustness against external perturbations.",
"discussion": "Discussion Predicting population dynamics of a microbial community from interactions between its members is difficult because interaction happens across multiple scales of biological organization [ 36 ]. Here we propose a mechanistic ecology model based on a coarse-grained representation of cell metabolism that accurately describes the population dynamics of three laboratory communities with well-defined metabolic exchanges. Previous studies have used genome-scale models and metabolic flux analysis, but these studies require flux measurements by isotope tracing and metabolomics to fit the adjustable flux parameters. Some success was also achieved by fitting the time series data with coarser-grained ecological models [ 37 – 41 ] such as the gLV equations; however, in gLV-type models, interspecific interactions are phenomenologically defined based on density dependency, which gives little mechanistic understanding of the underlying mechanism [ 15 ]. By contrast, our model has explicit formulations of context dependency by representing the chemical flows within and between microbes and thus can explain the metabolic part of microbe-microbe interactions. When we have limited prior knowledge and data on a given community it becomes critical to choose the right level of details. However, by applying our approach to well-defined laboratory systems, we show that a highly detailed metabolic network is not necessary for developing useful ecological models. In single-bacteria studies, coarse-grained metabolic models have been employed to understand the design principles of metabolic networks and their regulation [ 42 ], as well as to predict metabolic flux distributions useful for synthetic biology [ 43 ] and industrial [ 44 ] applications. Compared to genome-scale models, using coarse-grained models linking ecology and metabolism is simple and has recently become popular [ 25 , 45 , 46 ]. Depending on the research question, a coarse-grained metabolic network can be created at any level of granularity from a single reaction to the complete whole genome-scale reconstruction. The choice of granularity and how to derive a simpler model from the more complex one are usually empirical but can be facilitated by more systematic approaches to reduce dimensionality. Our model could extract new insights from those previously published empirical data on well-defined laboratory systems. The analysis shows that unidirectional cross-feeding is equivalent to a commensalism and bidirectional cross-feeding is equivalent to a mutualism. As shown by our study (Figs 2 – 4 ) and previous work [ 27 , 32 ], the actual relationship between cross-feeders, however, can be diverse in simple environments (e.g., glucose minimal medium) with constant resource supply due to a combination of positive effects of cross-feeding with negative effects of competition and toxicity of cross-fed metabolites, suggesting that the exact outcome cannot be precisely delineated by the cross-feeding type alone. For example, we predicted that, without supplementation of amino acids, coexistence of the leucine and lysine auxtrophies can only be achieved when one strain is limited in growth by glucose while the other strain is limited by the amino acid it is auxotrophic for ( Fig 3E ). Although it is theoretically possible that growth of the two auxotrophies is simultaneously limited by the amino acids they are auxotrophic for (i.e., the lysine auxotroph limited by lysine and the leucine auxotroph limited by leucine), this interaction pattern does not occur in the phase diagram because glucose will always be sufficiently depleted to a level that becomes growth limiting to at least one strain. The control of resource pool availability via population dynamics has been demonstrated to be a key mechanism for microbial community to optimize the metabolic strategy of its members to yield resistance to invasions and to achieve maximum biomass [ 46 ]. Mechanistic models including explicit nutrients and other realistic features, such as the models presented in this study, can help identify knowledge gaps [ 47 ]. For example, recent experiments have demonstrated that the coexistence of two carbon source specialists in the unilateral cross-feeding example is mutualistic in the sense that the consortium is fitter than the individuals [ 48 ]. The syntropy can be explained by a null expectation from theoretical ecology models [ 49 ]: the glucose specialist provides acetate in an exchange for a service provided by the acetate specialist which scavenges the acetate down to a level at which growth inhibition is insignificant. Although the mechanism of resource-service exchange has been considered in our model, the coexistence regime in the phase diagram ( Fig 2G ) is competitive, rather than mutualistic. Since mutualism occurs when the reciprocal benefits associated with cross-feeding outweigh competitive costs [ 50 ], our model may predict either or both of lower benefits and higher costs than needed to achieve mutualistic coexistence. Overall, the cost-benefit nature of the cross-feeding interaction between polymorphic E . coli strains is more complex than thought and warrants further research. Our modeling framework explains well the three published experiments but has noteworthy limitations. For example, we assume that the leakage flux is proportional to the conversion rate from substrate to metabolite (proportionality assumption), rather than proportional to the internal metabolite concentration. When does this assumption remain valid and how does it break down? By leveraging our previous experiences in modeling E . coli growth and resource allocation [ 43 , 51 ], we developed a coarse-grained single-strain model that explicitly assumes a linear dependency of leakage rate on metabolite concentration ( S1 Text ). We found that the proportionality assumption remains valid for an internal metabolite when its concentration was perturbed at the upstream, rather than the downstream of the metabolite ( S11 Fig ). This makes sense because the proportionality assumption couples metabolite leakage with upstream biosynthesis but does not take feedback regulation from downstream reactions and metabolites into accounts. When a perturbation is imposed from the downstream side, the proportionality assumption can lead to undesired behavior such as high leakage flux at low metabolite concentration. Although the assumption remains valid in the context of the current study where resource availability is the only varying external condition, it may prevent us from generalizing our modeling framework to different types of perturbations. Future studies may correct this limitation by incorporating metabolite concentration and associated reaction kinetics. So far, the current framework has been applied to well-characterized communities with known chemicals and associated interactions which provided a ground through to assess our model. Can the same approach be applied to infer community structure of complex microbiomes (e.g., human gut microbiome) where most of the metabolic exchanges involved in microbe-microbe interactions are still unknown? Our model has the potential if some technical challenges can be solved. First, direct modeling of a real-world microbiome with hundreds of species would be hurdled by too many unknown model parameters. One way to solve this problem is to simply ignore the rare species [ 38 ]. Another—arguably better—approach might be by grouping species composition into functional guilds using unsupervised methods that infer those groups from the data alone [ 52 ], or to use prior knowledge from genomics or taxonomy to create such functional groups. Second, inferring chemical mediators within a community of interacting populations is a nontrivial task. It can be facilitated by prior knowledge such as searching the literature or leveraging systems biology tools such as community-level metabolic network reconstruction [ 53 ]. Finally, our model is nonlinear, so that an efficient and robust nonlinear regression approach for parameter estimation is essential. For a model with similar size to the 14-auxotroph community we analyzed here, non-linear optimization algorithms may fail to converge to a realistic set of parameters and manual parameter selection is often the only feasible approach. Although we primarily chose the manual method to calibrate our models in this proof-of-concept study, manual fitting is a subjective and time‐consuming process, requires an expert operator with prior knowledge to choose physically and biologically realistic values, and perhaps more importantly, is unable to infer correlations among parameters. These downsides of manual parameter fitting has, at least for now, precluded it from being applied to large-scale microbial communities. On the positive side, the process of trial-and-error was greatly improved by the speed at which the intermediate-scale model runs simulations on a regular desktop computer. Beyond these technical issues, the model itself can be extended in multiple ways such as incorporating mechanisms of resource allocation [ 46 ]. Despite any present limitations, we anticipate that network inference using mechanism-explicit models can open new avenues for microbiome research towards more quantitative, mechanistic, and predictive science."
} | 3,559 |
38055772 | PMC10729037 | pmc | 884 | {
"abstract": "Rare earth elements (REE) are essential ingredients in\nmany modern\ntechnologies, yet their purification remains either environmentally\nharmful or economically unviable. Adsorption, or biosorption, of REE\nonto bacterial cell membranes offers a sustainable alternative to\ntraditional solvent extraction methods. But in order for biosorption-based\nREE purification to compete economically, the capacity and specificity\nof biosorption sites must be enhanced. Although there have been some\nrecent advances in characterizing the genetics of REE-biosorption,\nthe variety and complexity of bacterial membrane surface sites make\ntargeted genetic engineering difficult. Here, we propose using multiple\nrounds of in vivo random mutagenesis induced by the\nMP6 plasmid combined with plate-throughput REE-biosorption screening\nto improve a microbe’s capacity and selectivity for biosorbing\nREE. We engineered a strain of Vibrio natriegens capable\nof biosorbing 210% more dysprosium compared to the wild-type and produced\nselectivity improvements of up to 50% between the lightest (lanthanum)\nand heaviest (lutetium) REE. We believe that mutations we observed\nin ABC transporters as well as a nonessential protein in the BAM outer\nmembrane β-barrel protein insertion complex likely contribute\nto some—but almost certainly not all—of the biosorption\nchanges we observed. Given the ease of finding significant biosorption\nmutants, these results highlight just how many genes likely contribute\nto biosorption as well as the power of random mutagenesis in identifying\ngenes of interest and optimizing a biological system for a task.",
"conclusion": "Conclusions We introduced the MP6 plasmid into V. natriegens and confirmed that it still confers an enhanced\nmutation rate after\nmultiple rounds of mutagenesis—the first work to do this with\nan organism other than E. coli . We\nthen used this plasmid to randomly mutate V. natriegens to have more than a 200% improvement in biosorption of dysprosium\nwhile causing minimal growth defects. These improvements to dysprosium\nbiosorption also incidentally caused changes to V. natriegens ’ membrane’s selectivity for some REE over others—most\nnotably between adjacent heavy REE. In addition to successfully\ngenerating biosorption mutants, we\nalso found new genes to analyze to see if they are responsible for\nthe biosorption changes observed. We identified 26 genes that we mutated\nthat we believe have a good chance of contributing to the biosorption\nchanges we observed. Further analysis and characterization of the\nindividual genetic changes responsible can point to a more detailed\nmechanistic understanding that can in turn help with efforts to rationally\nengineer the biosorption capacity of V. natriegens . Among the more promising genes, we believe that mutation to bamC , which could affect a large portion of the outer membrane\ngenes, is a good candidate for further analysis. We also believe that\nPN96_12170—the tyrosine protein kinase that regulates the production\nof the extracellular polysaccharide colanic acid—is a good\ncandidate due to the previously observed biosorption increases that\nresulted from knocking out polysaccharide-related genes. Although\nthe changes to biosorption that we obtained were significant,\nmore work remains to be done to make it into a viable system for purifying\nREE from other metals and each other. The bacteria need to be optimized\nin complex environments, where we are optimizing for the enhancement\nof REE biosorption over competing metals. The same goes for optimizing\nfor the biosorption of certain REE over others. Since there is likely\na smaller subset of genetic changes that can achieve these more specific\ngoals, 96-well plates may not have a high enough throughput to work\non a reasonable time-scale. In order to make this a viable task, the\nthroughput of these tasks will need to be greatly increased—possibly\nusing high throughput microfluidic systems that are capable of screening\nhundreds of thousands of droplets in a day. Despite the work\nthat remains to be done, we believe that this\nwork is an encouraging sign that with further throughput, bacteria\ncan be engineered to have sufficient capacity and selectivity to replace\nsolvent extraction as the method of choice for producing purified\nREE.",
"introduction": "Introduction Rare earth elements (REE) are essential\ningredients to many technologies\nincluding catalysts, 1 high-efficiency lighting, 2 and lightweight high-strength magnets found in\nhard drives, 3 wind turbines, 4 electric vehicles, 5 and many other applications. 1 , 4 , 6 , 7 These magnets often utilize multiple\nREE such as neodymium, praseodymium, dysprosium, and terbium. The\ndemand for these technologies is rapidly increasing, and the corresponding\nsupply of REE needs to increase with it. 8 Current methods of purifying REE utilize solvent extraction, which\noften requires high temperatures and harsh chemicals, giving these\nimportant elements a high carbon and environmental footprint. 9 Biosorption—or the adsorption of\nelements or molecules onto\na biological surface—is a sustainable alternative to current\nmethods of REE purification. Bacteria can provide low-cost, high-capacity\nmetal binding. 10 They could also potentially\nbe reused over multiple biosorption and desorption cycles. 11 , 12 REE binding sites are found on both gram-negative and gram-positive\nbacteria 13 and—under select conditions—have\nalready been found to have competitive REE separation factors compared\nto current solvent extraction methods. 12 Recent works have sought to characterize the underlying genomics\nof REE biosorption in the gram-positive microbe Bacillus\nsubtilis ( 14 ) and in the gram-negative\nmicrobe Shewanella oneidensis . 15 Furthermore, in our work on S. oneidensis , we were able to produce mutants that increased total biosorption\nby 79% and increased the separation factor of the microbe for Eu over\nLa by 20%. 15 While this improvement in\nthe separation factor may seem small, computational modeling suggests\nthat it could result in a 27% reduction in the length of a successive\nenrichment process needed to achieve 99.9% purity Eu. 15 Although these advancements are promising, much work remains\nto produce bacteria with a truly competitive REE-biosorption capacity\nand selectivity ( i.e. , being capable of purifying\na moderately dense solution of mixed REE within only a few dozen enrichment\nsteps). While genetic tools for S. oneidensis are improving\nrapidly, 16 , 17 it is still much less genetically tractable\nand grows more slowly than laboratory workhorse microbes like E. coli . 18 , 19 These limitations make\ngenetic engineering guided by new knowledge on the genetics of REE-biosorption\nchallenging. Furthermore, mutations that impact fitness could take\nthe already relatively slow growing S. oneidensis and make it even slower. Random mutagenesis is a commonly\nused tool for improving and evolving\nselect biological functions. 20 − 22 In vitro random\nmutagenesis methods such as error-prone PCR have a broad mutational\nspectrum and can achieve a high mutation rate compared to in vivo methods. 20 However, only\na limited sequence length can be targeted and it is often laborious\nto reintroduce these mutations back into target genomes. 20 For the rapid production of less targeted mutations, in vivo mutagenesis methods are preferred. In vivo mutagenesis can be achieved with chemical mutagens such as ethylmethanesulfonate\n(EMS)—which is a human carcinogen—or with UV light.\nBoth these methods have distinct weaknesses as chemical mutagens often\nhave a biased mutational spectrum and UV methods are highly toxic\nto cells. 23 Due to the limitations\nof existing methods, new methods have been\ndeveloped to enhance bacterial mutagenesis in vivo . CRISPR methods, such as EvolvR, 24 can\ngenerate broad spectrum continuous mutagenesis in an organism but\nusually require specific targets. For the introduction of more generalized\nmutations, Badran and Liu developed several inducible plasmids containing\ngenes that were known to interfere with DNA replication fidelity mechanisms\nincluding proofreading, mismatch repair, and base selection in order\nto introduce an inducible mutation rate in E. coli up to 322,000 times the wild-type. 23 A\nplasmid such as this makes it possible to gradually introduce mutations\nover several generations, which could potentially limit our selection\nof mutants with growth deficiencies while still introducing a relatively\nlarge number of mutations. To maximize our chances of evolving\na high-performing biosorption\nmicrobe, we decided to work with the extremely fast-growing and highly\nengineerable microbe Vibrio natriegens . 25 − 29 As increased mutation rates tend to be associated with an accumulation\nof growth defects, 30 we hypothesized that\nstarting with a faster growing organism could make these defects relatively\nless severe. Recent work has also succeeded in making V. natriegens naturally competent, a feature that is very useful for conducting\ntargeted mutagenesis once the ideal genetic targets in V.\nnatriegens for REE biosorption are better understood. 31 However, to date, new advanced tools for in vivo mutagenesis have been used exclusively in the lab\nworkhorse E. coli . 23 Applications\nfor the MP6 plasmid have been largely restricted to phage-assisted\ncontinuous evolution and phage-assisted noncontinuous evolution. 32 There are no published instances of the MP6\nplasmid—or any other system using a small molecule inducer\nto enhance the mutation rate—being used in other organisms.\nIn this work, we make the first demonstration of the use of MP6 for\ninducible in vivo mutagenesis and improvement of\nREE-biosorption in V. natriegens .",
"discussion": "Discussion Improved V. natriegens Strains Have Similar\nTotal Biosorption Capacity to Other REE-Adsorbing Materials V. natriegens has a biosorption capacity comparable\nto those of other REE adsorbents and is superior compared to some\nother suggested REE-adsorbing bacteria. The wild-type V. natriegens has a DCW of.505 mg per 1 mL of 1 OD culture and our best biosorbing\ndysprosium mutant has a DCW of.635 mg per 1 mL of 1 OD. The measured\nwild-type V. natriegens adsorbed 17.4 mg Dy/g DCW.\nOur best mutant adsorbed 43.6 mg Dy/g DCW. E.\ncoli engineered to display REE-binding lanthanide\nbinding tags (LBT) have a capacity to biosorb 28 mg Tb/g DCW. 44 This is less than our randomly mutated V. natriegens is capable of adsorbing, although the LBT\nwill likely be less promiscuous with other metals than the V. natriegens ’ membrane. A system that displayed\nLBT on the extracellular matrix of E. coli performed slightly better, achieving a capacity of 54 mg Tb/g. 45 Our optimized mutant also has similar biosorption\ncompared to another proposed biological adsorbent, salmon milt (50\nmg Nd/g). 46 This is still far less adsorption\ncapacity, however, compared to the potential of other materials Park et al. ( 44 ) established as competitors\nto biosorption like activated carbon (145 mg La/g), 47 ligand-grafted silica (167 mg Eu/g), 48 and cation exchange resin (120 mg Er/g), 49 although with higher throughput and more rounds of mutagenesis,\nwe may be able to achieve these values. Mutant Bacteria Have Competitive Adjacent Heavy REE Separation Although our separation factors for separating heavy from light\nelements are small compared to some existing competitors, 50 , 51 our adjacent heavy REE separations are potentially more competitive.\nFor example, while Mattocks et al. recently demonstrated\nproduction of >98% pure solutions of neodymium and dysprosium in\na\nsingle pass 52 and Nelson et al. have created molecules with separation factors up to 213 for neodymium/dysprosium, 53 our system’s Nd/Dy separation factor\nremains below 2. Part of the inspiration for this work was the\nfinding by Bonificio and Clarke 12 that\nbacteria can separate REE comparable to solvent extraction if the\nbacteria surface was preprotonated or if the REE was fractionated\noff at a certain pH. For example, they found that they produced a\nsuperior separation factor for Yb/Tm if they started with a pH 2.5\nwash to get lighter REE off and then did pH 2 and pH 1.5 rinses to\nremove the heavier REE. Practically, however, this is not an industrially\nviable system because the REE capacity at these low pHs is very low\nand additional neutralization chemicals are required for this system.\nBut if we could engineer our bacteria such that they only had the\nbinding sites that were better at adsorbing Yb over Tm to begin with,\nthen, we could get a much more efficient process. Bonificio\nand Clarke 12 achieved a separation\nfactor of 1.5 with their method, while the solvent benchmark they\nused (RE(III)-HCl-EHEHPA) only had a separation factor of 1.1. 54 Other solvents do perform better than this solvent,\nhowever, but only marginally. For example, the solvent P507 has a\nseparation factor of 2.55. 51 We start with\na separation factor of 1.17, and our best mutant has a separation\nfactor of 1.34, while simultaneously raising the total REE biosorption\ncapacity. We note that we achieved this without directly optimizing\nfor the separation factor. Competing Metals Offer Challenges—and Opportunities—for\nREE Purifications Our results suggest that several other\nmetals—including at least one commonly present in potential\nREE sources—offer competitive or superior binding to REE. Some\nmetals are easier than others to deal with. For example, silver chloride\nis particularly insoluble; therefore, most of the silver can be precipitated\nwith little interference to the REE. Copper, however, is a potentially\nmore problematic contaminant as it is common in some ores, 44 and its biosorption appears correlated with\nthat of REE. LBT had a similar lack of selectivity for REE over copper. 44 It is apparent from our results, however, that\nthe current lack of specificity is not inevitable, and direct optimization\nfor higher REE/copper separation factors could lead to better mutants. One of the most notable results of our assay with competing metals\nwas the striking increase in the heavy–light separation factors\nwith increasing competing metals. We theorize that this was caused\nby the competing metals having a higher preference for the nonspecific\nREE binding sites—the sites that bind to all the rare earths\napproximately equally—than the REE binding sites that prefer\nthe heavier rare earths. This suggests that further genetic changes\nto eliminate more of these nonspecific sites could improve REE separations\neven more. But even in the absence of these genetic improvements,\nthese results suggest that we could also improve REE separations just\nby introducing select competing metals—ideally metals that\nhave chemistry that allow for easy downstream separation from REE. MP6 Effectively Evolves New Functions over Multiple Generations The MP6 plasmid proved to be an effective chassis for optimizing\nparticular functions over multiple generations. Using a mutagenic\nplasmid had a few distinct advantages over other existing methods:\n(1) it accumulated mutations over multiple generations, which could\npotentially help limit the emergence of growth defects and (2) it\ntargeted the entire genome. Given the limited growth defects in our\nevolved mutants—even over multiple rounds—the hypothesis\nthat a slower mutation rate over a period of growth would limit defects\nappears to be validated. The existence of any growth defects can still\nnegatively impact the techno-economic prospects of the mutants for\nREE recovery and purification, although we note that the growth rates\nof all except R3n2 remain competitive with E. coli . One expected disadvantage of our method is the emergence\nof mutations in the MP6 plasmid itself. Loss of functionality of the\nplasmid would limit the number of generations we could apply this\nmethod. Given the expected negative growth impact of emerging mutations,\na lack of functionality of MP6’s key genes is something we\ncould reasonably expect to emerge and fixate in the population fairly\nquickly. This disadvantage seems to have been largely avoided:\nno mutations\nto the MP6 plasmid emerged in any of the mutant strains. Although\nthe mutation rate was notably lower for the round three mutants, we\nsuspect that small differences in how we conducted the assay (amount\nof time bacteria grew overnight and the live/dead ratio after culture\ngrew overnight) could be responsible rather than there being a genetic\nbasis. We hypothesize that the reason for the lack of changes\nin the MP6\nplasmid is that the time between mutations and selection might be\ntoo short for the MP6-deficient mutants to fixate. For each round,\nwe selected strains with improved biosorption. Strains without a functional\nmutagenic plasmid are less likely to develop mutations that improve\nbiosorption. Perhaps it is just too improbable for both those things\nto arise sequentially in a single mutation period (≈7 generations).\nThus, by selecting biosorption mutants, we are also selecting mutants\nthat have fully intact MP6 plasmids. Biosorption Changes Hint at Diverse REE Binding Sites Our experiments further support the idea that bacteria contain numerous\nimportant binding sites for REE on their membrane. When our wild-type\nbacteria interacted with dysprosium alone, they adsorbed 16.3 out\nof 107 total μM (15.3%). With the same concentration of bacteria\nin a mix of REE (excluding Sc), the wild-type adsorbed 38.7 out of\n112 μM (34.6%). This indicates that instead of having a few\nbinding sites with relatively similar preferences for different REE, V. natriegens likely has a large number of different binding\nsites with very different REE preferences. Although we were\noptimizing for dysprosium in our screen (the lightest of the heavy\nREE), the greatest separation changes were generally in favor of other\nmiddle (Sm and Eu) and heavy (Yb and Lu) REE. This indicates that\nthe easiest way of optimizing for Dy biosorption is to optimize for\nthe binding sites that prefer REE with properties adjacent to Dy.\nGiven the relatively weaker increase in REE binding in the mixed REE\ncase than in the single REE, it appears that rather than increasing\nthe total number of binding sites in the system, perhaps the genetic\nchanges that occurred may have traded some generalist REE binding\nthat bound to all REE equally or binding sites that have a preference\nfor light REE for binding sites that have a stronger preference for\nthe middle/heavy REE. It is encouraging, however, that the separation\nchanges were not\nuniversal among all mutants. For example, R3n1 had a relatively increased\npreference for the middle REE as opposed to R3n2’s preference\nfor heavy REE. This indicates that we may be able to engineer the\nselectivity of the bacterial membrane for certain groups or individual\nREE."
} | 4,711 |
33041468 | null | s2 | 885 | {
"abstract": "Superhydrophilic zwitterions have been extensively exploited for surface modification to improve antifouling properties. However, it remains challenging to form layers of < 20 nm with high zwitterion content on the surfaces with different degrees of hydrophilicity. We demonstrate that amine-functionalized sulfobetaine (SBAm) can be co-deposited with dopamine on ultrafiltration (UF) membranes, leading to a thickness of 10 nm to 50 nm and an SBAm content of up to 31 mass% in the coating layers. The covalently grafted SBAm is stable underwater and improves the antifouling properties, as evidenced by the lower trans-membrane pressure required to retain targeted water fluxes than that required for the pristine membranes. The SBAm is also more effective than conventionally used sulfobetaine methacrylate (SBMA) for the zwitterion grafting on the surface to improve antifouling properties."
} | 223 |
35036688 | PMC8756567 | pmc | 887 | {
"abstract": "The rapidly developing\nartificial intelligence (AI) requires revolutionary\ncomputing architectures to break the energy efficiency bottleneck\ncaused by the traditional von Neumann computing architecture. In addition,\nthe emerging brain–machine interface also requires computational\ncircuitry that can conduct large parallel computational tasks with\nlow energy cost and good biocompatibility. Neuromorphic computing,\na novel computational architecture emulating human brains, has drawn\nsignificant interest for the aforementioned applications due to its\nlow energy cost, capability to parallelly process large-scale data,\nand biocompatibility. Most efforts in the domain of neuromorphic computing\nfocus on addressing traditional AI problems, such as handwritten digit\nrecognition and file classification. Here, we demonstrate for the\nfirst time that current neuromorphic computing techniques can be used\nto solve key machine learning questions in cheminformatics. We predict\nthe band gaps of small-molecule organic semiconductors and classify\nchemical reaction types with a simulated neuromorphic circuitry. Our\nwork can potentially guide the design and fabrication of elementary\ndevices and circuitry for neuromorphic computing specialized for chemical\npurposes.",
"conclusion": "Conclusions Neuromorphic\ncomputing simulation is accomplished within a crossbar\nsimulator CrossSim. The energy consumption can be dramatically reduced\nif the widely used von Neumann architecture can be substituted by\nneuromorphic computing devices. Unlike other studies on machine learning\napplications in cheminformatics, we used the CrossSim simulator to\nsolve chemical problems, namely, organic molecular band gap prediction\nand chemical reaction type classification. In the task of predicting\nthe band gaps of small-molecule organic semiconductors, different\nlookup tables of different memristors are applied in CrossSim at the\nconductance changing possibility distribution range of [0.1, 0.9]\nto overcome the divergence when TaOx is applied as the conductance\nupdate policy. Among these memristors, LISTA exhibits the smallest\nloss. This suggests that LISTA will be the best choice for fabricating\na neuromorphic device to predict small organic molecular band gaps.\nAccording to the comparison of different memristors in two crossbar\nnetworks, we obtain the conclusion that TaOx may not be appropriate\nto be used as a conductance update policy for networks with relatively\nsmall dimensions. This phenomenon may be due to the relatively high\nwrite noise of TaOx-based memristors. Regarding the classification\nproblem, the accuracy and prediction amount attained by CrossSim are\nclose to those by Keras, and better performance can be achieved by\nundersampling the reaction types with more abundant data. Our work\ncan potentially inspire the following studies that focus on generating\nlookup tables for novel neuromorphic materials in CrossSim and applying\nthem in other chemical problems.",
"introduction": "Introduction The rapid development\nof artificial intelligence (AI) demands imminent\nresponse to the energy efficiency bottleneck brought by the conflict\nbetween the traditional von Neumann computational architecture and\nthe rapidly evolving deep learning algorithms. 1 In the von Neumann architecture, the data movement between memory\nand processing units causes a large penalty of energy. 2 On the other hand, human brains consume orders of magnitude\nless energy but usually outperform modern computers in tasks such\nas image recognition and natural language analysis. 3 As an example of this energy efficiency gap, the training\nprocess of the first generation of AlphaGo, the famous deep learning\nmodel to play the strategy board game of Go, requires ∼5 ×\n10 5 W 4 as the peak power, while\nhuman brains work at only ∼20 W 5 but can easily defeat the early versions of AlphaGo. Neuromorphic\ncomputing can fill this energy gap by enabling the processing of big\ndata within the memories to eliminate data movement, which can drastically\nreduce the energy consumption by ∼5 orders of magnitude. 6 Recently, neuromorphic computing has been\ndemonstrated with resistive\nmemories or memristors based on various materials, including oxides, 7 phase change materials, 8 and polymers. 9 A memristor is a key device\nwhose conductance can be tuned via programmed applied\nvoltage pulses. In a matrix–vector multiplicator implemented via a crossbar circuit in neuromorphic computing ( Figure 1 a), the conductance\nof memristors acts as the weights of an abstract artificial neural\nnetwork in machine learning algorithms ( Figure 1 b). 10 The input\nvectors are given in the form of voltages (in Figure 1 b). As a result of Ohm’s law and Kirchhoff’s\nlaw, the output of the crossbar memory is naturally the results of\nvector–matrix multiplication operations. Therefore, by tuning\nand reading the conductance of memristors, training and inference\nof a neural network can be accomplished within memristors. Figure 1 Schematic for\nthe structure of (a) neuromorphic computing and its\nrelationship with (b) neural networks. Traditionally, to complete vector–matrix multiplications\nin a conventional computational architecture, the process units need\nto fetch data from the dynamic random-access memory (DRAM) to obtain\nthe weights and inputs, multiply inputs with the weights, and move\ndata back to the DRAM, resulting in high energy consumption. 32 However, crossbar circuitry allows for in-memory\ncomputing and parallel reading. The memristors serve as the memory\nwhere the vector–matrix multiplication take place in\nsitu . Parallel reading allows these operations to occur in\na single step by mimicking the current summary in a parallel-connected\ncircuit. 11 Parallel writing, which is also\nbeneficial for energy usage, is used to update the weights; 11 weights ( W ij ) are updated based on the outer product of the row (length\nof the voltage pulse x i ) and the column (height of the voltage pulse y i ), according to eq 1 . It had been proven that for an N × N crossbar, analog resistive memory crossbars\ncan be O ( N ) more energy-efficient\nthan a conventional digital memory-based architecture. 11 1 To estimate the performance\nof a memristor in a neuromorphic computing\ncircuit quickly without fabricating such a circuit, neuromorphic computing\nsimulation software packages, such as CrossSim 26 and NeuroSim, 27 have attracted\nmuch attention. Fuller et al. ( 7 − 9 ) used CrossSim\nto perform classification in backpropagation simulations on handwritten\ndigits and file classification on a Sandia data set. Yu et\nal. ( 12 , 13 ) added hybrid precision synapse\nand advanced learning algorithms in NeuroSim to increase the Modified\nNational Institute of Standards and Technology (MNIST) 28 handwritten recognition accuracy. The accuracy\nand energy consumption for the training of MNIST images with different\nemerging nonvolatile memories are also compared using NeuroSim in\n2018. 2 Additionally, Chen et al. used MATLAB to simulate the classification performance of their\nmemristive synapse for neuromorphic computing. 14 Wang et al. ( 15 ) used SPICE 29 to simulate an in-memory\ncomputing system based on a resistive random-access memory crossbar\ncircuit, which can perform parallel and general computing tasks, and\nindicated that this architecture could handle with data-intensive\nproblems. Hu et al. ( 16 ) developed\na dot product engine with transistor–memristor arrays and combined\nit with MATLAB to formulate a single-layer neural network for MNIST\nhandwritten classification, achieving 89.9% recognition accuracy. At the same time, machine learning has found wide application in\ncheminformatics, for example, generating molecules with desired properties, 17 , 30 , 31 predicting pharmaceutical properties\nfor drug-like compounds, 18 and so forth.\nSimilar to other AI applications, the exciting applications of machine\nlearning in cheminformatics calls for energy-efficient computational\nhardware. The decent feature of the neuromorphic architecture is suitable\nfor machine learning tasks in cheminformatics. However, few studies\nhave been conducted for this application. Here, we introduce CrossSim\ninto chemistry-related machine learning objectives. As an example\nof regression tasks, we used CrossSim to predict the band gap of small-molecule\norganic semiconductors; as an example of classification tasks, we\nused CrossSim to classify chemical reactions. We conducted experiments\non the band gap prediction with different materials’ lookup\ntables. In CrossSim, the weight update is based on lookup tables that\nare acquired from experiments. The lookup tables provide the information\nof the probability distribution of conductance change at a given conductance\nby the applied voltage pulse. 26 After comparing\nthree materials including a lithium-ion synaptic transistor, 8 tantalum oxide (TaOx)-based resistive memories, 7 and an electrochemical neuromorphic organic device, 9 we found that the lithium-ion synaptic transistor\nhas the lowest loss, while tantalum oxide-based resistive memories\nexhibit a fluctuation with increasing training epochs in the regression\ntask to predict the band gaps for organic semiconductors. To reveal\nthe reason why the fluctuation exists in TaOx-based memristors, we\ntrained two sequential crossbar circuits with different lookup tables\n(i.e., circuits based on different materials). We found that the performance\nof TaOx memristors depends on the dimension of the crossbar circuit,\nand they tend to underperform in relatively small crossbar arrays\n(50 × 1). This phenomenon may be due to the write noise intrinsic\nto TaOx memristors. 7 To demonstrate the\nability of neuromorphic computing for classification applications,\nanother network was trained to classify the chemical reaction types\nfrom a reaction database, USPTO-50k. 24 We\nobtained consistent results between the CrossSim simulation and the\nwidely used deep learning application programming interface (API)\nKeras. Due to the imbalance of the data set, undersampling 19 was applied to data-abundant reaction types.\nCompared to previous outcomes, CrossSim training after undersampling\nshowed higher classification precision and less false positive cases.\nOur work demonstrates that neuromorphic computing techniques can be\nused in cheminformatic problems well and further discusses the material\ndependency on the training performance. Our work can potentially be\nused to guide the design and fabrication of devices and circuitry\nfor neuromorphic computing specialized for chemical purposes.",
"discussion": "Results and Discussion Prediction\nfor the Band Gap of Small Organic Molecules Prediction of\nthe band gap of a small organic molecule was first\nperformed using the four lookup tables with the conductance change\nprobability distribution in the range [0.25, 0.75], which was recommended\nby the manual of CrossSim software. 26 As\nshown in Figure 3 a,\nthe average losses of LISTA_Current (current-controlled pulse measurement\nfor the Li 1– x CoO 2 -based\nlithium-ion synaptic transistor), 8 LISTA_Voltage\n(voltage-controlled pulse measurement for the Li 1– x CoO 2 -based lithium-ion synaptic transistor), 8 and ENODe (PEI/PEDOT:PSS) 9 converged normally, but the loss of TaOx 7 did not converge after five or six epochs. The learning rate was\nexplored for better performance, but it did not change the result\nof TaOx prediction, as shown in Figure S1 . Ideal results were gained by enlarging the possibility distribution\nrange to [0.1, 0.9]. The performances of the available materials are\ncompared in Figure 3 b. Generally, LISTA_Current and LISTA_Voltage had a smaller average\nloss than TaOx and ENODe, which indicates that the material for LISTA\n(Li 1– x CoO 2 ) may be the\nfavorite material to fabricate a physical device to make predictions\non molecular band gaps. Figure 3 c shows the comparison between the actual band gaps and predictions\nfor the test set using the LISTA_Current lookup table, where the coefficient\nof determination is 0.96. This prediction result was comparable to\nthe result obtained from the regression with TensorFlow Keras, where\nthe coefficient of determination is 0.98. In addition, the TaOx prediction\nresult in Figure 3 b\nfluctuated with increasing epochs. To find out the reason of fluctuation,\nresearch on the simulation of different materials for the two crossbars\nwas conducted. The results are discussed in the following paragraph. Figure 3 (a) Average\nloss for different lookup tables with the lookup table\npossibility distribution range [0.25, 0.75] for training; (b) average\nloss for different lookup tables with the lookup table possibility\ndistribution range [0.1, 0.9] for training; (c) actual vs predicted result for the LISTA_Current simulation with the lookup\ntable range from 0.1 to 0.9 at epoch 40 for the test set. Crossbar networks with dimensions of (1024, 50) and (50,\n1) were\nsimulated with different lookup tables in CrossSim, as shown in Figure 2 b. The simulation\nresults are shown in Figure 4 . Fluctuations were observed in each figure when the TaOx\nlookup table was utilized for the (50, 1) crossbar network. In Figure 4 c, the network with\nthe dimension of (1024, 50) was completely simulated with the TaOx\nlookup table, and the network with a dimension of (50, 1) was simulated\nwith other materials. It was demonstrated that the fluctuation in\nthe TaOx simulation was caused by the (50, 1) network. According to\nthe application of TaOx-based resistive memories on handwritten digit\nrecognition, 7 we think this perturbation\non a small-dimension crossbar network is caused by the intrinsic write\nnoise of this material. 33 If the dimension\nof the crossbar network is small, the write noise will exhibit a large\nimpact on the crossbar performance. However, the influence of this\nwrite noise can be neglected when TaOx is applied in large-dimension\ncrossbar networks. Therefore, TaOx can capture features of large-dimension\nnetworks but lacks the ability to fit the results with small-dimension\ncrossbar networks. Figure 4 Using different materials to simulate the band gap prediction\nfor\ntraining. (a) LISTA_Current: both crossbars using LISTA_Current; LCLV:\nfirst crossbar using LISTA_Current and second crossbar using LISTA_Voltage;\nLCTa: first crossbar using LISTA_Current and second crossbar using\nTaOx; and LCEN: first crossbar using LISTA_Current and second crossbar\nusing ENODe; (b) LISTA_Voltage: both crossbars using LISTA_Voltage;\nLVLC: first crossbar using LISTA_Voltage and second crossbar using\nLISTA_Current; LVTa: first crossbar using LISTA_Voltage and second\ncrossbar using TaOx; and LVEN: first crossbar using LISTA_Voltage\nand second crossbar using ENODe; (c) TaOx: both crossbars using TaOx;\nTaLC: first crossbar using TaOx and second crossbar using LISTA_Current;\nTaLV: first crossbar using TaOx and second crossbar using LISTA_Voltage;\nand TaEN: first crossbar using TaOx and second crossbar using ENODe;\nand (d) ENODe: both crossbars using ENODe; ENLC: first crossbar using\nENODe and second crossbar using LISTA_Current; ENLV: first crossbar\nusing ENODe and second crossbar using LISTA_Voltage; and ENTa: first\ncrossbar using ENODe and second crossbar using TaOx. In addition, a two-hidden-layer model for the band gap prediction\ntask was implemented. Dimension of the model was set to be (1024,\n50, 50, 1). Also, the three crossbars were all simulated with the\nLISTA_Current lookup table. The training result is shown in Figure S3 . It took more epochs, about 160 epochs,\nfor the MSE loss to reach a steady state, and the loss was around\n0.00282, which was close to the result we obtained from the one-hidden-layer\nmodel whose loss was 0.00285. After that, the test set was tested\non the two-hidden-layer model, and a coefficient of determination\nof around 0.98 was obtained. As mentioned previously, the coefficient\nof determination for the one-hidden-layer model is 0.96. Upon comparing\nthe results from these two models, we believe that the one-hidden-layer\nmodel was sufficient for the band gap prediction task. Moreover, the\ntraining result for a three-hidden-layer model is provided in the Supporting Information . Classifying Chemical Reactions In Table 1 , classification\nresults obtained\nusing CrossSim (based on LISTA_Current lookup table) and Keras are\nlisted. The precision is defined as the number of correctly classified\nreactions (true positive) divided by the total number of reactions\nbeing classified as belonging to a specific reaction type (true positive\n+ false positive). Higher precision was obtained for reaction types\nwith larger data amount. Table S1 shows\nthe recall of both CrossSim and Keras classification models. Recall\nis defined as the number of true positive cases divided by the summation\nof true positive and false negative cases. As shown in Table 1 and S1 , no obvious difference in the performance could be discovered between\nthe two models. Therefore, we believed that the classification task\naccomplished by CrossSim was comparable to that completed by Keras.\nBecause the USPTO-50k database is not balanced between reaction types,\nwe adopted the undersampling strategy 19 on the majority of classes in the training set to seek for a better\nclassification performance. As a result of undersampling, Rx_1 and\nRx_2 were diminished by 20% in the training set; Rx_3 and Rx_6 were\nreduced by 10%. The classification for CrossSim with an undersampling\nstrategy is shown in Tables 2 and S2 . The numbers of false positive\ncases for data-abundant reaction types are decreased. Therefore, the\nprecision of classification is increased with the application of the\nundersampling strategy. Table 1 Classification Result\nComparison between\nCrossSim and Keras CrossSim Keras precision true positive + false positive precision true positive + false positive number\nof reactions in the test set RX_1 heteroatom alkylation and arylation 0.756 1835 0.763 1935 1500 RX_2 acylation\nand related processes 0.721 1521 0.737 1464 1190 RX_3 C–C bond formation 0.698 357 0.717 346 550 RX_4 heterocycle formation 0.663 70 0.789 61 90 RX_5 protection 0.594 18 0.441 23 65 RX_6 deprotection 0.647 763 0.676 772 825 RX_7 reduction 0.712 276 0.701 238 450 RX_8 oxidation 0.574 48 0.556 57 80 RX_9 functional group interconversion 0.473 110 0.499 103 160 RX_10 functional\ngroup addition 0.5 2 1 1 25 Table 2 Classification Result for CrossSim\nwith Undersampling CrossSim\nwith undersampling precision true positive + false positive RX_1 0.902 1536 RX_2 0.805 1152 RX_3 0.704 597 RX_4 0.788 80 RX_5 0.540 87 RX_6 0.655 875 RX_7 0.903 380 RX_8 0.514 107 RX_9 0.421 183 RX_10 1 3"
} | 4,672 |
40305870 | PMC12043352 | pmc | 888 | {
"abstract": "SUMMARY Cyanobacteria are investigated for fundamental photosynthesis research and sustainable production of valuable biochemicals. However, low product titer and biomass productivities are major bottlenecks to the economical scale‐up. Recent studies have shown that the introduction of a metabolic sink, such as sucrose, 2,3‐butanediol, and 2‐phenyl ethanol, in cyanobacteria improves carbon fixation by relieving the “sink” limitation of photosynthesis. However, the impact of light intensity on the behavior of this sink‐derived enhancement in carbon fixation is not well understood and is necessary for translation to outdoor cultivation. Here, using random mutagenesis, we engineered Synechococcus elongatus PCC 11801 to overproduce 1.24 g L −1 phenylalanine (Phe) in 3 days, identified L531W in the TolC protein as an important driver of Phe efflux, and investigated the effect of light intensity on total carbon fixation. We found that low light results in competition between biomass and Phe, whereas under excess light, a higher flux of fixed carbon is directed to the Phe sink. The introduction of the Phe sink improves the quantum yields of photosystem I and II with a concomitant increase in the total electron flow leading to nearly 70% increase in carbon fixation at high light in the mutant strain. Additionally, the cyclic electron flow decreased, which has implications for the ATP/NADPH production ratio. Our data highlight how light intensity affects the sink‐derived enhancement in carbon fixation, the role of CEF to balance the source‐sink demand for ATP and NADPH, and the enhancement of inorganic carbon fixation in cyanobacteria with an engineered sink.",
"introduction": "INTRODUCTION Cyanobacteria are promising organisms for metabolic engineering due to their ability to use sunlight and CO 2 to produce biochemicals and biofuels sustainably (Knoot et al., 2018 ). However, industrial use of cyanobacteria requires addressing several barriers such as, but not limited to, economical scale up and improving product titers (Huang et al., 2017 ). Improving the rate of carbon fixation and thus the production of growth‐associated biochemicals can increase the feasibility of using cyanobacteria in industrial applications (Jaiswal et al., 2022 ). Approaches to improve carbon fixation and photosynthetic efficiency have thus far been directed at engineering ribulose‐1,5‐bisphosphate carboxylase/oxygenase (RuBisCO) enzyme (Liang et al., 2018 ; Whitney et al., 2011 ), enabling greater light penetration by reduction of antenna size (Kirst et al., 2014 ), manipulation of the Calvin‐Benson‐Bassham (CBB) cycle (Liang et al., 2018 ), engineering photorespiratory bypass mechanisms (Shen et al., 2019 ; Xu et al., 2023 ), or reducing dissipative losses (Kromdijk et al., 2016 ). Recently, engineering of the production of biochemicals (“sink engineering”) in cyanobacteria has been shown to produce a significant augmentation of carbon fixation (Zhang et al., 2017 ). Various cyanobacteria engineered to produce biochemicals such as sucrose (Abramson et al., 2016 ; Santos‐Merino, Torrado, et al., 2021 ), 2,3‐butanediol (2,3 BD) (Oliver & Atsumi, 2015 ), 2‐phenylethanol (Ni et al., 2018 ), glycerol (Savakis et al., 2015 ), phenylpropanoids (Kukil et al., 2023 ), and ethylene (Ungerer et al., 2012 ) have shown increased carbon fixation ability (Table S1 ). However, neither the underlying mechanism of enhanced C‐fixation nor the effect of light intensity on carbon assimilation is well understood. Investigating the dynamics of this source‐sink relationship is particularly important for evaluating and improving the feasibility of cyanobacteria in outdoor cultivation applications in which light intensities can vary substantially. It has been previously hypothesized that under excess light, the introduction of an additional sink can improve the utilization of the excess light energy (Grund et al., 2019 ; Zhou et al., 2016 ). However, it was recently shown that a sucrose sink showed improvement in the quantum efficiency of PSII at all light levels, whereas an electron valve, cytochrome P450, led to photosynthetic enhancement only under moderate to high light (Santos‐Merino, Torrado, et al., 2021 ). Furthermore, the extent of the increase in electron transport was substantially lower compared with the enhancement observed in carbon fixation (Santos‐Merino, Torrado, et al., 2021 ). Introduction of an additional sink or biomass composition shift can result in a change of adenosine triphosphate (ATP) to nicotinamide adenine dinucleotide phosphate (NADPH) demand, which is an important factor in photosynthetic productivity (Erdrich et al., 2014 ; Kramer & Evans, 2011 ; Steichen et al., 2024 ). Linear electron flow through photosystem II (PSII) and photosystem I (PSI) in Synechococcus elongatus generates an ATP/NADPH ratio of 1.38 due to the presence of 13 c‐subunits in its ATP synthase (Pogoryelov et al., 2007 ), whereas the CBB cycle requires an ATP/NADPH ratio of 1.5 (Noctor, 1998 ). Alternate electron flow (AEF), one of which is cyclic electron flow (CEF) around PSI, produces only ATP and can make up the shortfall in ATP (Allen, 2003 ; Nogales et al., 2012 ). However, the effect of additional sinks on LEF and CEF under different light conditions remains unclear in cyanobacteria. In this work, we use random mutagenesis combined with selection on Phe analogues to isolate Phe‐overproduction strains using the fast‐growing cyanobacterial strain S. elongatus PCC 11801 (Jaiswal et al., 2018 ). We further characterize the single‐nucleotide polymorphisms (SNPs) underlying Phe‐overproduction using whole genome sequencing. The mutants with the Phe sink showed increased net carbon fixation. We show that the sink‐derived improvement in carbon fixation is light dependent by studying Phe and biomass accumulation, PSII efficiency, nonphotochemical quenching, and PSI donor and acceptor side limitations. We also examined the effect of the Phe sink on LEF and CEF and thus its ability to alter the ratio of ATP/NADPH production. Figure 1 shows an outline of cyanobacterial photosynthetic machinery, electron transport pathways, Phe sink, ATP and NADPH production, and demands as well as the site of action of the photosynthetic inhibitors used in this study. Figure 1 Electron transport pathways in the light harvesting system of cyanobacteria. Schematic representation of cyanobacterial photosynthetic machinery showing electron transfer pathways, nonphotochemical quenching pathway, action of electron transport inhibitors, Calvin cycle, Biomass and Phe sinks, and ATP/NADPH synthesis and consumption. \n † ATP:NADPH ratio can vary depending on N source, carbon uptake, amino acid reuptake, carbon recycling, etc. The red arrows indicate loss of excitons/electrons to photoprotective mechanisms. CBB, Calvin‐Benson‐Bassham; CEF, cyclic electron flow; Cyt b \n 6 \n f , cytochrome b \n 6 \n f ; DBMIB, dibromothymoquinone; DCMU, 3‐(3,4‐dichlorophenyl)‐1,1‐dimethylurea; E4P, erythrose 4‐phosphate; Fd, ferridoxin; FNR, ferridoxin NADP + oxidoreductase; HA, hydroxylamine; LEF, linear electron flow; NPQ, nonphotochemical quenching; PC, plastocyanin; PEP, phosphoenolpyruvate; PQ, plastoquinone; PSI, photosystem I; PSII, photosystem II; Phe, phenylalanine.\n\nIntroduction of Phe sink suppresses CEF By generating additional ATP molecules, the CEF can aid in balancing the ATP/NADPH production and demand (Kramer & Evans, 2011 ). To understand the effect of the introduction of the Phe sink on linear and CEF, and thus ATP/NADPH production, we utilized P700 difference absorption spectroscopy. CEF can be estimated by blocking the linear electron flow from PSII using PSII specific inhibitors DCMU and HA. The dark re‐reduction rate of P700 was measured in the absence (LEF + CEF) and in the presence of DCMU and HA (CEF) to calculate the percentage of CEF (%CEF) in WT and M14.2 under LL, ML, and HL conditions (Figure 7a ). The %CEF in WT is higher than M14.2 at all light levels. The re‐reduction of P700 is significantly faster in DCMU and HA‐treated WT compared with the corresponding M14.2 sample (Figure 6d–f ). In the mutant, the %CEF is reduced to nearly 1% or to a negligible rate, similar to the DBMIB‐treated sample, due to the introduction of the Phe sink. Since DBMIB treatment should inhibit both linear and CEF, the P700 re‐reduction rate under this inhibitor likely originates from charge recombination events in P700 or, to a minor degree, incomplete inhibition of cyt b \n 6 \n f . Since CEF generates only ATP, the decrease in %CEF by the introduction of the Phe sink will result in a decrease in the ATP/NADPH ratio in M14.2. This indicates that under the growth light intensities tested, Phe sink is likely to be more NADPH‐intensive compared with the biomass sink. Our results do not indicate any change in %CEF in both WT and mutant as a function of light intensity. However, the HL tested here has been shown previously as an optimal growth condition for PCC 11801 (Jaiswal et al., 2018 ) and that this HL illumination did not result in a higher NPQ (Table 2 ). This observation suggests that CEF may yet be induced in PCC 11801 as a photoprotective mechanism under photoinhibitory high light conditions. In WT, the only major sink is biomass and thus the demand for ATP and NADPH should be the same under different light intensities as the biomass compositions are unchanged. In M14.2, the flux toward the Phe sink is light intensity‐dependent (Figure 4c ). Because the ATP and NADPH requirements for biomass and the Phe sink are different (Figure 1 ), the change in flux to the Phe sink under ML and HL will change the net demand for ATP and NADPH. A decrease in %CEF with light intensity might meet any increased demand for NADPH to produce Phe. However, we do not find any change in %CEF with light intensity in M14.2 (Tukey's test for multiple comparison). Figure 6 PSI redox kinetics of WT and M14.2. (a–c) Light‐induced oxidation and dark re‐reduction kinetics of P700 in LL, ML, and HL‐acclimated strains. For each growth light, a corresponding actinic light of comparable intensity was used for determination of the intermediate P700 oxidation state P . Traces for untreated, DCMU and HA, and DBMIB‐treated samples are given. (d–f) The dark re‐reduction kinetics of P700 are shown after normalization to P m ′ as described previously (Holland et al., 2016 ). Traces are representative of three biological replicates. Figure 7(b) shows the relative electron transport through PSI (rETR1) to be significantly higher in the mutant at both ML and HL conditions than WT, with most of the increase in rETR1 arising from the increased LEF. The rETR1 and ФPSI, the photochemical efficiency of PSI, reflect the rate of both linear and cyclic flow, whereas rETR2 and ФPSII report only on the LEF. The enhancement in total electron flow, as apparent from rETR1, is 86% at ML and 83% at HL, comparable to the enhancement in carbon fixation (Figure 3 ). A plot of the ФPSI versus ФPSII also supports the shift toward LEF in M14.2 at HL conditions (Figure 7c ). Figure 7 Partitioning of electrons into linear and cyclic pathways in WT and M14.2. (a) Estimation of the fraction of CEF in WT and M14.2 under different light intensities. (b) rETR1 and the contribution of linear and cyclic flow toward it. (c) ФPSI is plotted against ФPSII to determine the extent of cyclic and linear electron flow. Data represent mean and standard deviation from three biological replicates. Statistical significance is calculated using a two‐tailed t ‐test. The P ‐values are located above the * in the Figure.",
"discussion": "DISCUSSION Cyanobacteria are increasingly investigated as a sustainable biochemical factory to produce high value products due to their ability to use CO 2 as the sole carbon source (Knoot et al., 2018 ). Several studies have engineered cyanobacteria for the production of biofuels and biochemicals; however, there are still hurdles to overcome before their industrial application becomes feasible (Jaiswal et al., 2022 ). One way to improve the feasibility of cyanobacteria for industrial production is to increase their rate of carbon fixation, with the assimilated carbon diverted to the desired product. The recent demonstration of improved carbon fixation upon introduction of a sink in cyanobacteria is an important advance in the field (Santos‐Merino, Singh, & Ducat, 2021 ). It further provides an excellent opportunity to understand and manipulate the carbon fixation process in cyanobacteria. L‐Phe is an essential amino acid synthesized industrially via fermentation and is valuable for the food, cosmetic, and pharmaceutical industries as a precursor for the synthesis of products such as the artificial sweetener aspartame (Bongaerts et al., 2001 ). In this work, we used random mutagenesis on PCC 11801 to develop Phe overproducing strains that can accumulate 3 g Phe L −1 , the highest reported titer in cyanobacteria. Our strain M14.2 shows a higher productivity under both ambient and 3% CO 2 conditions with a maximum productivity of about 400 mg Phe L −1 d −1 under a 3‐day cultivation cycle. This is significantly higher than the 90.4 mg L −1 d −1 (Brey et al., 2020 ) and 152.5 mg L −1 d −1 (Kukil et al., 2023 ) production of Phe reported in Synechocystis 6803, highlighting the value of the choice of a fast‐growing cyanobacteria. To enable a fair comparison with heterotrophs, we compare the two‐step process of sugar production by crops and subsequent conversion to Phe by heterotrophs with the single‐step cyanobacterial process using space–time yield as previously described (Brandenburg et al., 2021 ). The current best Phe‐producing bacterial strain accumulates roughly 72 g Phe L −1 in the media with a yield of 0.26 mol Phe per mol glucose (Liu et al., 2018 ). Assuming the average yield for sugarcane over the last 10 years (70.63 tfw/ha/annum) and a sugar content of 15%, the space–time yield for the two‐step process is 6.91 kg Phe ha −1 d −1 . In the case of cyanobacterial photoautotrophic Phe production, we assume commercial photobioreactors of scale 500 000 L ha −1 (Masojídek & Torzillo, 2014 ) and productivity 66 mg Phe L −1 d −1 observed in diel shake flask experiments (Figure S5 ), which yields 33 kg Phe ha −1 d −1 . This is 4.8‐fold more efficient than the current heterotrophic process. Future work will target further increasing flux to Phe. Previously, it has been shown that nearly 50% of CO 2 fixed can be directed to Phe (Brey et al., 2020 ), which indicates that there is further room for improvement. This can be addressed by combining metabolic engineering approaches as previously described for tryptophan production (Deshpande et al., 2020 ). It is surprising to see that no SNPs were identified in DAHPS, the first step that drives flux into the shikimate pathway. A similar observation was found in a tryptophan overproducing mutant in Synechocystis PCC 6803 (Deshpande et al., 2020 ). In contrast, Synechocystis PCC 6803 mutants that overproduced Phe obtained through adaptive laboratory evolution all had mutations in DAHPS (Kukil et al., 2023 ). Instead, overproduction of Phe was enabled by a SNP in the ACT (feedback regulated domain) of the PD in this study. A similar deregulation of PD was observed in C. glutamicum where a mutation in Arg‐202 and Gly‐224 residues in the ACT domain (Chan & Hsu, 1996 ). Additionally, M14.2 had a SNP in TolC, a protein previously found crucial to cysteine export in Escherichia coli (Wiriyathanawudhiwong et al., 2009 ). TolC forms a channel in the periplasm connecting the outside of the cell with inner membrane bound transporters. TolC‐mediated efflux of Phe has not been previously reported. We speculate that the identified SNP in the TolC protein improves Phe efflux to the extracellular space, thereby improving Phe production in M14.2. The KaiA, KaiB, and KaiC proteins form the central circadian clock oscillator system of cyanobacteria (Ishiura et al., 1998 ). KaiA plays a key role in the phosphorylation cycle of KaiC as it stimulates KaiC's autokinase activity during the day (Nakahira et al., 2004 ). KaiA has also been suggested to entrain the circadian clock by taking input from the photosynthetic metabolism (Rust et al., 2011 ). It is therefore possible that a mutation in KaiA affects the clock oscillator function in M14.2 or makes the circadian clock of the mutant insensitive to the introduction of a powerful Phe sink. However, no growth phenotype was observed in M14.2 under diurnal conditions (Figure S5 ). Whether the mutation in kaiA played any role in improving Phe production in M14.2 should be investigated further by a detailed physiological and metabolic characterization of the mutant. Further research should be directed at identifying the roles of TolC, KaiA, and the Ig‐like domain containing proteins on enabling Phe production. Previously, it has been suggested that an additional sink is necessary to relieve sink limitation (Mellor et al., 2019 ; Zhou et al., 2016 ). We show that an overflow homologous sink, such as Phe in our study, can result in enhanced carbon fixation. In this work, we highlight the role of light intensity in the enhancement of carbon fixation. The Phe sink competes with biomass under LL, whereas under ML and HL, the enhanced carbon fixation is directed entirely to the engineered sink. This is similar to lactate production in Synechocystis sp. PCC 6803, where source limitation does not result in benefit due to lactate sink (Grund et al., 2022 ) but contrasts with studies utilizing 2,3 BD and sucrose sinks, which showed a reduction in biomass accumulation, although the total carbon fixation increased by the introduction of the new sink (Abramson et al., 2016 ; Oliver & Atsumi, 2015 ; Santos‐Merino, Singh, & Ducat, 2021 ). The production of Phe‐derived 2‐phenylethanol also improved total carbon fixation (Ni et al., 2018 ). However, there was a competition between biomass and product sinks in the first 4 days, but after 4 days, the enhancement in carbon fixation was directed solely toward the new product sink (Ni et al., 2018 ). These observations indicate that the nature of the sink product can influence the behavior of the sink. Products such as 2‐phenylethanol (Ni et al., 2018 ) and 2,3 BD (Oliver & Atsumi, 2015 ) can be toxic to cyanobacteria, while the sucrose sink (Abramson et al., 2016 ) can result in osmotic shock due to the addition of NaCl, resulting in competition between biomass and product sink even when light is not limiting (Burnap, 2015 ; Ducat et al., 2012 ). In contrast, Phe does not compete with biomass under light‐sufficient conditions. These observations indicate a complex relationship between the introduced sink and growth. The growth differences under different product sinks may also be attributed to the different requirements for precursors for each sink. Phe biosynthesis draws on PEP and E4P as precursors, whereas 2,3 BD requires pyruvate and sucrose requires hexose phosphates. These precursors may differ in their abundance under differing light availability, in turn affecting the flux toward the engineered product sink. In our work, the change in light intensity significantly affected the carbon partitioning to the Phe sink. This is similar to the enhancement in carbon assimilation attained in 2,3 BD production in S. elongatus 7942 when light intensity was increased from 50 μmol photons m −2 sec −1 to 250 μmol photons m −2 sec −1 (Oliver & Atsumi, 2015 ). It is hypothesized that the excess absorbed light energy that is otherwise dissipated as nonphotochemical quenching can be utilized by the introduction of an additional carbon sink (Abramson et al., 2016 ; Zhou et al., 2016 ). Our data support this idea as we find an improved quantum yield of PSII in Phe overproducing strains under ML and HL conditions (Figure 5 ). The photochemical quenching (qP) also improved substantially under high light in the mutant strains. These results are similar to the observations made in sucrose (Abramson et al., 2016 ) and glycerol‐producing cyanobacteria strains (Savakis et al., 2015 ). The improvement in PSII quantum yield leads to increased LEF (Figure 5b ), which in turn meets the increased demand for ATP and NADPH created by the Phe sink. In S. elongatus PCC 11801, the LEF generates ATP and NADPH in a ratio of 1.38, whereas alternate electron transport pathways such as cyclic and Mehler reaction generate only ATP (Allen, 2003 ; Berla et al., 2015 ). The CBB cycle requires 3 ATP and 2 NADPH to fix one net mol of carbon dioxide. However, the ATP and NADPH requirement for biomass and other sinks can differ, often resulting in a mismatch between the production and consumption of ATP and NADPH. The balanced synthesis of ATP and reductant is therefore a key impetus for the regulation of linear and cyclic electron transport pathways. This aspect of light reaction further represents an important avenue for improvement of photosynthetic efficiency and carbon fixation (Erdrich et al., 2014 ). Our data reveal that %CEF is greatly reduced in response to the introduction of the Phe sink under all light conditions. This suggests that a lower ATP/NADPH demand might be sufficient for the partitioning of carbon to the Phe sink compared with the biomass sink, which is the only carbon sink in WT. Because of unknown and difficult‐to‐measure growth‐associated and nongrowth‐associated maintenance costs, it is difficult to accurately estimate the ATP/NADPH demand for biomass accumulation, but it requires a minimum ATP/NADPH of 1.51, but is likely higher (Erdrich et al., 2014 ). Previous flux balance models predict a higher demand of 1.73 (Shastri & Morgan, 2005 ). The estimate for Phe synthesis is 1.52, but this could be slightly higher if an ATP‐dependent amino acid transporter is taken into account (Montesinos et al., 1997 ). Our results indicate that the ATP/NADPH demand for biomass is higher than the Phe sink, as evident from the lower %CEF in the Phe overproducing strains. This observation is also consistent with the fact that Phe represents a more reduced carbon molecule (degree of reduction = 4.44) than biomass (degree of reduction = 4.2), requiring more reductant than ATP (Figure 1 ; Table S4 ). Similarly, we estimate the ATP/NADPH demand for 2,3 BD, d ‐lactate, and sucrose to be 1.27 (degree of reduction = 5.5), 1.4 (degree of reduction = 4), and 1.54 (degree of reduction = 4) respectively (Tables S3 and S4 ). These demands are lower than the demand for biomass (Shastri & Morgan, 2005 ). The introduction of these sinks, similar to the Phe sink, thus necessitates greater NADPH demand. This is in agreement with the observation of the decreased rate of P700 + re‐reduction in sucrose exporting cultures relative to uninduced cultures, as a decreased P700 + re‐reduction indicates a reduced CEF and cyclic photophosphorylation. Furthermore, the deletion of the CEF genes has enabled enhanced production of lactate (Selão et al., 2020 ). The decrease of CEF in the mutant may lead to the relaxation of an important safety mechanism of photosynthesis known as photosynthetic control (Berla et al., 2015 ; Huang et al., 2015 ). In photosynthetic control, the increased acidification of the thylakoid lumen at high light intensities slows down the oxidation of PQH 2 at the Q o ‐site of the cyt b \n 6 \n f complex, which in turn decreases the inter‐photosystem electron transport. CEF is a key contributor to proton gradient formation and photosynthetic control, which protects PSI from photoinhibition (Huang et al., 2015 ). The establishment of photosynthetic control is evident in the increased donor side inhibition of PSI at ML and HL in both WT and M14.2 (Figure 8b ). However, the substantial decrease of CEF leads to lower photosynthetic control and donor side inhibition of PSI in M14.2 (Figure 8b ). This speeds up the LEF in M14.2 at ML and HL (Figure 7b ). One interesting application of a Phe sink‐driven decrease in CEF in a light intensity‐independent fashion could be the development of an inducible Phe overproduction strain to study the activation mechanism of CEF. A key question that remains to be answered is why evolution has not fully optimized the photosynthetic capacity as the improvement of light utilization efficiency under heterologous and native overflow sink reveals additional underutilized photosynthetic potential. One possibility is that there is a trade‐off between maximizing biomass production and acclimating to different stressors that are encountered in the natural environment. Our low light data suggest that there is a competition for cellular energy between biomass and Phe production (Figure 3 ), which could reduce biomass accumulation and strain robustness under dynamic outdoor growth conditions. Overall, our work adds to emerging reports of engineering aromatic amino acids photosynthetically and also the enhancement in carbon fixation attained by the addition of a carbon sink. With further work in understanding the molecular mechanisms of source‐sink sensing and dynamic regulation under different environmental conditions, accompanied by further metabolic engineering, particularly in recently discovered fast‐growing strains, cyanobacteria are one step closer to being economically feasible factories for the production of valuable biochemicals."
} | 6,391 |
33941698 | PMC8126839 | pmc | 889 | {
"abstract": "Significance Coral reefs are in catastrophic decline worldwide, in part due to increasingly warm surface waters that cause mass coral bleaching and mortality. However, corals in the northern Red Sea and Gulf of Aqaba have shown no sign of bleaching, despite local seawater temperature rising faster than the global average. We show that the exceptional heat tolerance of the common symbiotic reef-building coral Stylophora pistillata from the Gulf of Aqaba is based on a rapid gene expression response and recovery pattern when exposed to heat stress up to 32 °C. Such temperatures are not anticipated to occur in the region within this century, giving real hope for the preservation of at least one major coral reef ecosystem for future generations.",
"conclusion": "Conclusion The northern Red Sea and Gulf of Aqaba constitute a coral reef refugium from global warming ( 11 , 29 , 52 , 53 ), which is thought to stem from selection for thermal tolerance during northbound migration through the warmer waters of the southern Red Sea following the last ice age ( 11 , 29 ). Our study has shown that S. pistillata from the northern Red Sea and Gulf of Aqaba exhibits an exceptional transcriptomic resilience and is capable of producing a fast and pervasive gene expression response to both short-term (hours) and long-term (days and weeks) thermal stress. Indeed, we observed only negligible transcriptomic change(s) following an 11 d exposure to temperatures >2 °C above the local MMM, which is thought to induce bleaching in most other coral populations. Even at 5 °C above their MMM, clear signs of acclimation were observed. Only at 7.5 °C above the local MMM did these corals exhibit irreversible transcriptomic change(s), loss of symbiont algae, an increase of opportunistic (and potentially pathogenic) bacteria in their microbiome, and high mortality. Importantly, our multitemperature/multitimepoint approach indicates a link between the short-term transcriptomic resilience and longer-term acclimation capacity. Overall, our study represents a significant step toward understanding the complex transcriptomic basis of coral thermal tolerance. Assuming that our results for S. pistillata are representative of the general coral population in the Gulf of Aqaba and the northern Red Sea, these corals represent the best chance for humanity to preserve a major, highly biodiverse coral reef ecosystem that, to date, has been unaffected by ocean warming, in contrast with most other reefs that will continue to be decimated by the combined effects of local and global anthropogenic stressors ( 2 ).",
"discussion": "Discussion The capability of a coral holobiont to rapidly increase the number of DEGs in response to environmental stress and subsequently relax this response poststress has been termed transcriptomic resilience ( 20 , 23 ). The main finding from our study is that S. pistillata holobionts from the Gulf of Aqaba can mount an extraordinarily fast and pervasive gene expression response and show extremely strong transcriptional resilience when exposed to thermal stress up to 32 °C, which is ∼5 °C above their maximum monthly mean summer temperatures (MMM) ( 11 , 29 ). Heat exposure to 5 °C above the natural MMM should (in analogy with coral holobionts from almost any other reef locality) cause high levels of stress and bleaching—if not almost instant death. Yet, S. pistillata from the Gulf of Aqaba demonstrates extreme tolerance to these elevated seawater temperatures ( 11 , 29 ). Transcriptomic resilience of the coral host in particular appears key to this exceptional thermal tolerance. Coral and algae transcriptomes were substantially modified and displayed low transcriptomic resilience when subjected to temperatures reaching 34.5 °C, even during short-term exposure, after which only about 34% of coral and 53% of algae DEGs at T1 returned to baseline expression levels at T2 ( Fig. 3 A and B ). At 34.5 °C, the bacterial microbiome showed no resilience but exhibited a shift to an opportunistic bacterial community, even during recovery. Specifically, long-term exposure to 34.5 °C resulted in symbiont bacteria loss, dominance of Vibrio species, commonly regarded as opportunistic pathogens, and coral death within a few days. The gene expression in coral host and symbiont algae showed little similarity between short- and long-term heat stress exposure at 32 °C ( Fig. 3 C ). Yet, modifications to the coral and algal transcriptomes began to exhibit higher levels of similarity in both short- and long-term thermal stress exposures at 34.5 °C, providing the first transcriptomic snapshot of the consequences of exceeding the upper critical thermal threshold temperature for these corals that, based on physiological observations, appears to be between 33.8 °C and 34.2 °C ( 32 ). The rate at which the coral transcriptome can recover to baseline expression levels following thermal stress exposure has previously been linked to thermal tolerance. In the present study, S. pistillata exposed to 32 °C was able to re-establish its baseline gene expression levels in just 11 h following short-term stress, and in just 2 d after long-term stress ( Figs. 1 and 3 A and B ), revealing exceptionally fast transcriptomic resilience for this coral population. In a study of two species of Acropora in the Ofu Island back reef in American Samoa, which underwent a period of more than a month of natural heat stress at ∼1 °C above the regional bleaching threshold (29 °C), Acropora gemmifera colonies went from 3,504 DEGs immediately following heat stress to 12 DEGs (i.e., nearly complete recovery) 4 mo later, with no significant mortality observed. In the same time interval, nearby Acropora hyacinthus maintained roughly half (1,063) of its DEGs and suffered 85% mortality ( 22 ). Relatively fast transcriptomic resilience was also observed in A. hyacinthus corals from the Ofu back reef that were experimentally exposed to 6 °C above the 29 °C regional bleaching threshold for 1 h ( 20 ), with 8,913 DEGs (27% of the transcriptome) observed 1 h after heat exposure, dropping to 3,831 DEGs (a 57% drop) 15 h into recovery at 29 °C. Under natural conditions, Ruiz-Jones and Palumbi ( 39 ) observed strong and fast transcriptomic resilience in A. hyacinthus corals 2 d after a natural temperature spike at ∼31 °C (i.e., 2 °C above the regional bleaching threshold). However, this gene expression response involved only 177 coexpressed genes (∼0.5% of the A. hyacinthus transcriptome), far from the large plasticity observed in S. pistillata in this study. On the other hand, the gene expression observed in these A. hyacinthus corals pointed to the “unfolded protein” response as a first line of defense of corals against thermal stress, consistent with the findings in Acropora sp. from Ofu Island (American Samoa) ( 40 ) and in S. pistillata from the Gulf of Aqaba ( 30 ) (this study) ( Dataset S2 ). In our study, the number of DEGs observed in S. pistillata exclusively at the recovery time point (T2) increased with increasing experimental temperature ( SI Appendix , Fig. S5 ). At 34.5 °C, the transcriptomic pattern in S. pistillata was not only characterized by a low percentage of gene recovery but also by a large increase in new DEGs at T2 ( SI Appendix , Fig. S5 ), consistent with the pattern observed in the less heat-tolerant A. hyacinthus ( 22 ). Thus, low ability to recover normal gene expression and a tendency for new DEGs to appear during recovery seems to characterize a coral that has exceeded its thermal tolerance limit. The transcriptomic resilience of S. microadriaticum was strong and similar to that of its coral host up to 32 °C ( Fig. 3 B ) in both the short-term and long-term experiments. It did, however, decrease at 34.5 °C, especially in the long-term experiments ( Fig. 3 B ), indicating a sharp transition in resilience of the symbiont population, similar to the coral host, between 32 °C and 34.5 °C ( 41 ). In addition, 414 DEGs appeared in the 32 °C treatment at T2 in the short-term experiment ( Dataset S3 ), mainly as a down-regulation of genes involved in photosynthesis, indicating that reduced photosynthetic activity might be beneficial to the symbionts immediately after thermal stress, possibly to reduce the abundance of reactive oxygen species. By analogy to the concept of transcriptomic resilience, we define the ability of the bacterial microbiome to dynamically change its composition in response to heat stress and return to the initial composition during the recovery phase as “microbiome resilience,” which further adds to the notion of microbiome flexibility ( 28 ). In contrast to the strong transcriptomic resilience of the coral and algal symbionts, the bacterial microbiome composition showed no resilience during short-term heat stress experiments, even at temperatures of 29.5 °C and 32 °C ( Fig. 4 ). Rather, microbiome composition shifted significantly during recovery (i.e., at T2) ( Fig. 4 A ), indicating that different dynamics are at play compared to coral host and algal gene expression. Importantly, in contrast to the set of all genes present in the genomes of coral host and symbiotic algae, which is a fixed entity, the microbiome is “fluid” ( 42 ) and an open entity (i.e., there is a dynamic exchange of external and holobiont-associated microbes). This was demonstrated by the appearance of opportunistic bacteria, such as Saprospiraceae and Vibrionaceae, in the short-term experiment at 34.5 °C ( Figs. 2 E and 4 A ). Notably, bacteria of the genus Aureispira (Saprospiraceae) increased in abundance at all experimental temperatures above 27 °C and remained highly abundant in the recovery time point. Members of the Aureispira are assumed to be primarily seawater residents ( 43 ) but have also been found in tissues of Acropora muricata affected by White Syndrome, suggesting their involvement in opportunistic infections ( 44 ). There is also evidence that members of the Aureispira prey on Vibrio sp. in a process linked with the availability of calcium ions ( 45 ) and thus may help to counter an increasing abundance of opportunistic bacteria ( 46 ). Together, our observations indicate that in short-term heat stress events up to 32 °C, the S. pistillata holobiont can achieve near-complete gene expression recovery (98.3% of the genes recovered), but that substantial shifts in the bacterial microbiome composition are seen, consistent with previous work ( 26 ). During long-term heat stress, the bacterial microbiome composition was stable up to 32 °C, with a dominance of the tissue-associated Endozoicomonadaceae ( Fig. 4 B ), commonly present in S. pistillata ( 47 , 48 ) and assumed to be abundant in healthy corals ( 49 ). As such, their decrease in relative abundance at 34.5° is consistent with the notion that these bacteria are important for holobiont functioning ( 48 , 50 , 51 ). Acclimation, defined as a return to a normal state of gene expression during a long-term thermal stress, is an important factor in coral thermal tolerance ( 21 ). In our study, accepting the assumption that the long-term experiments can be considered a continuation of the short-term experiments, we also investigated the capability of the S. pistillata holobiont to acclimate to elevated temperatures. To do so, we compared the number of DEGs at T1 in the short-term (T1 = 6 h) and long-term (T1 = 11 d) heat stress experiments ( Fig. 3 A ). Both the coral host and algal symbiont host showed signs of acclimation at stress temperatures of 29.5 °C and 32 °C, with a much smaller fraction of DEGs at T1 in the long-term heat stress compared with T1 in the short-term heat stress ( Fig. 3 A ). Similarly, the bacterial microbiome showed signs of acclimation up to 32 °C ( Fig. 4 B ), with a more stable community composition among genotypes and between temperatures (27 °C, 29.5 °C, and 32 °C) in the long-term heat stress experiment ( Figs. 2 F and 4 B ). Signs of acclimation in the microbiome disappeared at the experimental temperature of 34.5 °C, along with the appearance of opportunistic bacteria."
} | 3,039 |
36425758 | PMC9679549 | pmc | 891 | {
"abstract": "Summary Green self-powered devices based on biodegradable materials have attracted widespread attention. Here, we propose the construction of the transient biotriboelectric nanogenerator (TENG) using green-in-green bionanocompoites. The green-in-green nanocomposites, cellulose nanocrystal (CNC)/polyhydroxybutyrate (PHB), are prepared with a high-pressure molding method. The CNC promotes the degradation and enhances the dielectric constant of CNC/PHB. It further allows for the significant improvement of the triboelectric output of CNC/PHB-based TENG. The voltage output and current output of CNC/PHB-based TENG are 5.7 and 12.5 times higher than those of pristine PHB-based TENG, respectively. Also, the bio-TENG exhibits admirable signal stability in over 20000 cycles. Despite the high hardness of CNC/PHB, a soft but simple-structured arch sensor is successfully assembled using CNC/PHB-based TENG. It can attain the precise real-time monitoring of various human motions. This study may provide new insights into the design/fabrication of green functional materials, and initiate the next wave of innovations in eco-friendly self-powered devices.",
"conclusion": "Conclusions Here, a green-in-green CNC/PHB bionanocomposite was prepared by high-pressure molding. It was used for the first time to fabricate biotriboelectric nanogenerators. Contrary to the pure PHB sample, the degradation of the CNC/PHB nanocomposite could occur at room temperature with the existence of hydrophilic CNC. Also, the CNC changed the dielectric properties of CNC/PHB, which could induce more charges over the electrode of the bio-nanogenerator. Consequently, the electrical output of CNC/PHB-based single-electrode TENG increased with the increase of CNC contents. The output voltage and output current of 5CNC/PHB-based TENG were approximately 5.7 times and 12.5 times higher than those of pristine PHB-based TENG, respectively. Besides, the CNC/PHB-based TENG demonstrated considerable high performance in instantaneous output power output and charging various external capacitors. Importantly, in more than 20000 continuous cycles, it exhibited a remarkable stability/durability of the output electric signals. Further, despite high hardness of CNC/PHB composite, a flexible but simple arch sensor was assembled using the CNC/PHB-based bio-TENG. It showed significant practical value in athletic monitoring. In a word, we believe this work confirms a simple but effective method to improve the electrical outputs of a bio-polyester-based TENG. It also provides a great template for preparing more green-in-green materials used in environmental-friendly self-powered devices.",
"introduction": "Introduction With the increasing consumption of energy and the gradual depletion of traditional fossil resources, the development of sustainable energy harvesting/supply technology has been attached much attention. 1 , 2 , 3 The self-powered system can convert all kinds of ambient energy into electrical energy to maintain the operation of the device. 4 , 5 , 6 , 7 The technology mainly depends on the special coupling effects that include triboelectric effect, 8 , 9 , 10 piezoelectric effect, 11 , 12 , 13 electromagnetic effect, 14 , 15 , 16 thermoelectric effect 17 , 18 , 19 and pyroelectric effect, 20 , 21 , 22 , 23 and so forth. He et al. conceived a flexible piezoelectric-enhanced triboelectric nanogenerator (TENG), which could harvest ambient energy to drive commercial electronical devices. 24 Ouyang et al. fabricated a bioresorbable pressure sensor based on the triboelectric effect for cardiovascular postoperative care. 25 Natural biodegradable materials have biocompatibility and renewability, 26 , 27 , 28 , 29 which makes these materials be used in TENG for implantable healthcare function devices, 30 transient medical equipment, 31 and wearable sensors. 32 Consequently, there is tremendous potency and value for using the biomaterials in self-powered electronics and systems. 33 , 34 , 35 However, most bio-materials generally have structural or functional drawbacks which limit their applications. Therefore, it is very necessary to employ versatile chemical or physical methods for solving these problems. Polyhydroxybutyrate (PHB) is a thermoplastic, chiral polyester. It is synthesized by microorganisms as a carbon source and energy storage in the case of the imbalance of carbon and nitrogen nutrition. 36 , 37 , 38 PHB has good environment compatibility and biodegradability, and its decomposition products can be safely absorbed by the environment. 39 , 40 In addition, the microbial polyester has some special functions, including piezoelectric property and pyroelectric property. So far, there are some reports and applications about these properties. Chernozem et al. utilized the electrospinning technology to prepare PHB/rGO (graphene oxide) 3D composite biological scaffolds used in various tissue engineering. 34 Zviagin et al. fabricated fibrous scaffolds with ZnO/PHB, which improved scaffolds’ piezoelectric responses. 41 However, these employments merely involve the bio-engineering. To our knowledge, there are few reports involving the PHB in green self-powered devices. Cellulose is a productive and widely sourced biopolymer on the earth. 42 , 43 , 44 One dimensional cellulose nanocrystal (CNC), with high crystallinity and diameter smaller than 100 nm, can be prepared by chemical, physical or biological treatment. 45 The natural CNC has highly ordered crystalline structure, large specific surface area, unique morphological structure, good environmental compatibility, and strong hydrophilicity. As far, CNC has been applied in much advanced applications, such as supercapacitors, 46 batteries, 47 sensors, 48 , 49 and so forth. It is recognized as an attractive component for fabricating functional materials. Consequently, it is conceivable that the combination of CNC and PHB can induce new performances or improve existing performances. As depicted in Figure 1 , we prepare a green-in-green CNC/PHB-based TENG for the first time. Simultaneously, a flexible biological TENG sensor comprising the high hardness bioderived CNC/PHB composite is further manufactured. The structure and performance of CNC/PHB suggest the green nanofiller CNC has considerable effects on the degradation and dielectric property of green PHB. Meanwhile, the CNC/PHB with higher dielectric constant can induce more charges on the back electrode, which greatly enhances the performances of the bio-TENG. Additionally, the TENG exhibits remarkable stability of the electrical signal in over 20000 cycles. Based on the triboelectric effect of CNC/PHB bionanocomposites, a simple but flexible bio-TENG sensor is conceived and assembled to collect information about human motions. This further demonstrates the considerable value of CNC/PHB bio-composites in environment-friendly self-powered sensors. Figure 1 Schematic illustration showing the preparation of CNC/PHB nanocomposite and the application of a flexible sensor based on CNC/PHB nanocomposite",
"discussion": "Results and discussion The dispersion and morphology of CNC in water and in the 5CNC/PHB casting membrane were directly revealed using TEM, as shown in Figure 2 . Clearly, the CNC had good dispersion in water ( Figure 2 A). The length and the diameter of CNC were about 200 nm and 10 nm, respectively. The morphology and structure of CNC did not change in the as-fabricated CNC/PHB composite sample. Most importantly, the green dopant achieved a relatively good dispersion in the PHB matrix ( Figure 2 B). Figure 2 Characterizations of the CNC/PHB composite (A and B) The morphology of CNC in water (A) and PHB matrix (B), respectively. (C–E) FTIR (C), DSC (D), and WAXD (E) of CNC/PHB composite samples of different CNC contents. (F–H) SEM (F-H) of 3CNC/PHB composite. (G and H) are magnified views of the portions highlighted by the red rectangular frames in (F) and (G), respectively. (I and J) The residual weight fraction (I) and sequential photographs (J) of the pristine PHB sample and the 5CNC/PHB composite sample during the hydrolysis at pH 12 and at 25 °C. Sample size: 8 mm in diameter and ∼0.8mm in thickness. Additionally, FTIR spectra of PHB-based composites were obtained to investigate CNC in bio-nanocomposites ( Figure 2 C). The band at 1428 cm −1 and 3342 cm −1 were assigned to -CH 2 - bending and stretching vibration of O-H respectively. The intensity of these two bands increased with the increment of CNC content. This further confirms that the CNC was doped into the composite system successfully. To study the effects of CNC on the PHB crystalline structure, further characterizations were carried out. Figures 2 D and 2E show the DSC and WAXD results of CNC/PHB nanocomposites, exhibiting melting behaviors and information about the crystalline region, respectively. For the pristine PHB sample, its crystalline degree is 76.0% ( Figure S1 ). When the 1% CNC is added, the green-in-green composite has a higher crystalline degree of 78.0%. However, with the further increase of CNC content, the crystallinity of the CNC/PHB nanocomposites decreases from 78.0% to 75.4% ( Figure 2 D). It is generally believed that the crystallization of polymer can be divided into two stage: nucleation and growth. This result reveals that the moderate content of CNC, as a nucleating agent, is beneficial for the formation of anisotropic structures. But the excess CNC may impede the polymer molecular chain from getting into the lattice, which finally dwindles the proportion of PHB crystals. Similar phenomenon also occurred in the crystallization of PDLA/PLLA/CQD composites. 50 WAXD was further employed to examine the microstructure of the PHB crystals in CNC/PHB ( Figure 2 E). It is obvious that the three CNC/PHB samples had similar X-ray diffraction spectra due to high-pressure treatment, which suggests there may be similar crystalline structures. 51 The direct morphological observation of the crystalline structure of 3CNC/PHB samples was acquired through SEM ( Figures 2 F–2H). Apparently, a wrinkled spherulite structure was obtained in the 3CNC/PHB sample. Also, it occurred in other PHB-based nanocomposites ( Figure S2 ). Besides these structural characterizations, the degradation of the bio-materials was discussed. As shown in Figure 2 I, obviously, there is almost no mass loss for the pristine PHB specimens in the alkali aqueous solution. Moreover, no obvious change in appearance was exhibited ( Figure 2 J). Contrary to the PHB sample, the residual weight fraction of the 5CNC/PHB composite decreased with time, which indicated the hydrolysis of bio-composite occurred ( Figure 2 I). When the degradation time reached 29 days, the sample edge whitened and crakes appeared on the sample surface ( Figure 2 J). With the time extended, the white area expanded continually and the sample was disintegrated into pieces in 116 d. After another 60 days, the residual mass fraction of the 5CNC/PHB sample reached about 84%. The reason for almost no degradation of pure PHB in comparison with 5CNC/PHB may be ascribed to the following three points. First, the tight structure in the crystalline region is known to make the solution penetration difficult. The pure PHB sample has higher crystallinity that is more detrimental to hydrolytic degradation ( Figure S1 ). Second, the CNC, a hydrophilic nanomaterial, improves the hydrophilicity of CNC/PHB nanocomposites. This further facilitates the solution diffusion and the hydrolysis in CNC/PHB. Last but not least, lower experiment temperature inhibits the hydrolysis reaction for both PHB and CNC/PHB. In a word, this result confirms the enormous possibility and value of green-in-green nanocomposites in the field of degradable electronic devices. The CNC/PHB composite with different CNC contents was assembled into TENGs respectively. Because the surface potential of CNC/PHB is different from that of polydimethylsiloxane (PDMS), a triboelectric potential is established when the two materials contact and then separate. It can drive the reciprocating flow of electrons between the back electrode and the ground ( Figure S3 ). Also, the pristine PHB-based TENG was prepared as a reference using the same method. Subsequently, kinetic energy collection performance of generators was evaluated ( Figure 3 A). Figures 3 B and 3C show the electrical outputs of the bio-TENGs under the external excitation of 4 N and 1 Hz. It is obvious that the generated energy output of CNC/PHB-based TENG was improved with the increase in the CNC content. Among them, the open-circuit voltage and short-circuit current of the 5CNC/PHB-based TENG could reach 248.8 V/cm 3 and 2.5 μA/cm 3 , respectively, which were about 6.7 times and 13.5 times than those of the pure PHB-based TENG, respectively. Because these composites have similar exterior appearances and internal structures, this result suggests that the addition of CNC might affect the intrinsic property of the nanocomposites, thus improving the performances of the bio-TENG. Figure 3 Working mechanisms of the CNC/PHB-based TENG (A) Impact-measurement system for simultaneously collecting the electrical potential and current outputs. (B and C) Open-circuit voltage (B) and short-circuit current (C) outputs of pure PHB and CNC/PHB, generated at the same stimulating frequency of 1 Hz, and an applied force of 4 N. Sample size: 8.0 mm in diameter and 0.6 mm in thickness. (D) Dielectric constant profiles of CNC/PHB nanocomposites. (E) Schematic drawing of the working mechanism for the single-electrode bio-TENGs. In order to better explain this phenomenon, a characterization of these materials was carried out. The dielectric constant (ε r ) of CNC/PHB nanocomposites at different frequencies was obtained ( Figure 3 D). The dielectric constant reflects the ability of materials in responding to the external electric field. In other words, the material with a higher dielectric constant has greater polarization intensity in the same electric field. At the range of 10 0 ∼ 10 3 Hz, material polarization mainly depends on interfacial polarization. With the increase of frequency of alternating electric field, the relaxation of interfacial polarization occurs. So, the polarization of these CNC/PHB composites is very low at higher frequencies, and the dielectric constants had similar value. However, ε r of the nanocomposite demonstrated a positive correlation with the content of CNC at lower frequencies. Previous reports have proved that materials with a higher dielectric constant can enhance electrification induction and favor higher triboelectric performances. 52 , 53 As shown in Figure 3 E, the open-circuit voltage and short-circuit charges of the single-electrode bio-TENGs can be approximately regarded as 54 : (Equation 1) V o c = S σ 1 2 C (Equation 2) Q s c = S σ 1 2 where V oc is the open-circuit voltage, Q sc is the short-circuit transferred charges, σ 1 is the surface charge density over the back electrode, S is the surface area of the electrode, and C is the capacitance between the back electrode and the grounding terminal. When test conditions remain the same, S and C are usually constant. Thus, V oc and Q sc are related to σ 1 only. For CNC/PHB with higher dielectric constant, it can induce more charges over the back electrode and thereby improve both V oc and Q sc . Consequentially, because the short-circuit current ( I sc ) is defined as the ratio of Q sc to time, I sc also increases with the increment of the CNC amount. Furthermore, the 5CNC/PHB-based TENG was investigated in detail to determine the effect of stimulating frequency and force amplitude over triboelectric outputs. As presented in Figures 4 A and 4B, the electrical outputs of the 5CNC/PHB-based TENG were collected, under the stimulation of a given frequency of 1Hz with the various applied forces from 4 N, 8 N, and 16 N. It is evident that the open-circuit voltage and short-circuit current outputs increased gradually with the increase of impact force. The voltage and current output density increased from 248.8 V/cm 3 and 2.5 μA/cm 3 to 331.4 V/cm 3 and 4.9 μA/cm 3 , respectively. These results indicate the 5CNC/PHB-based TENG is relatively sensitive to the change of the external applied force, thereby proving the green generator can convert diverse environmental mechanical energy into electrical energy. Figure 4 Characters of the 5CNC/PHB-based TENG (A and B) Open-circuit voltage and short-circuit current outputs generated by the 5CNC/PHB-based TENGs, stimulated at different applied forces but with the same frequency of 1 Hz. (C and D) Open-circuit voltage and short-circuit current outputs generated by the 5CNC/PHB-based TENGs, stimulated at different frequencies but with the same applied force of 16 N. Sample size: 8.0 mm in diameter and 0.6 mm in thickness. Figures 4 C and 4D show the electrical outputs of 5CNC/PHB-based TENG stimulated by an impact force of 16 N, with the frequency varied from 1 Hz, 2 Hz to 3 Hz. Similarly, with the increase of stimulation frequency, the bio-TENG can generate higher electrical outputs, and the voltage density could achieve 365.2 V/cm 3 and 5.5 μA/cm 3 . However, the increment obtained was relatively small and a similar phenomenon also occurred in previous reports. 55 In conclusion, this bio-material nanogenerator could respond to the variation of stress and frequency. This indicates that the triboelectric electromechanical conversion device can adapt to various mechanic movements in ambiance. More investigations were implemented on electrical output performance ( Figure 5 ). In order to determine the maximum electric power output, the voltage and current of 5CNC/PHB-based TENG under different exterior resistance were collected. As depicted in Figure 5 A, the current and voltage present positive and negative correlations with external load resistance respectively, because of ohmic loss. When the resistance varies from 1 MΩ to 5 GΩ, there is a dramatic change in output electric signals. Additionally, the dependence of instantaneous power density on external load resistance is calculated and plotted in Figure 5 B. Evidently, the instantaneous peak power is the maximum when the external resistance reaches around 166 MΩ. The corresponding electric power density is about 58.4 μW/cm 3 \n Figure 5 C shows that a series of capacitors are charged by using a 5CNC/PHB-based TENG. The capacitors of 0.1 μF and 0.2 μF are charged to 3 V within about 50 s and 100 s, respectively. As we all know, stable energy output is one of the important criteria to evaluate the performance of electronic devices. Consequently, the short-circuit current of 5CNC/PHB-based TENG under incessant stimulation is selected as the main criterion. Amazingly, as shown in Figure 5 D, the current generated by the apparatus has hardly changed in more than 20000 cycles, demonstrating the great possibility of this device in self-powered sensors. Figure 5 Output performances of the 5CNC/PHB-based TENG (A and B) Dependence of the voltage and current outputs (A) and the calculated power density (B) of 5CNC/PHB-based TENG on the external loads, stimulated at 3 Hz and 16 N. (C) Equivalent circuit of a self-charging power system based on the TENG for powering capacitor, and voltage curves when charging different capacitors at the stimulation of a 3 Hz, 16 N impact. (D) Output current-time profiles of 5CNC/PHB-based TENG working for ≈20000 continuous cycles. Sample size: 8.0 mm in diameter and 0.6 mm in thickness. The previous tests have proved that the 5CNC/PHB-based TENG has considerable energy output performance. However, it is well acknowledged that PHB materials generally have high brittleness that discourages the value of the bio-polyester. Consequentially, because of the small size, a flexible sensor with a simple structure was conceived ( Figure 6 ). It could accurately perceive the human movement according to the electrical signals from this sensor. As depicted in Figure 6 A, PI is used as supporting materials to carry the 5CNC/PHB sample and PDMS film, and conductive carbon cloth is used to transfer electric charges. Figure 6 B exhibits a digital picture of the sensor which has the proper size and could be attached to joints of the limbs. The charges on the friction surface can create a triboelectric potential between the 5CNC/PHB sample and the ground. To balance the generated electric potential drop, the electrons are driven to flow back and forth between the back electrode and the earth ( Figure 6 C). Then, the rotating angle of the joints can affect the stress applied to the arch structure, which makes the sensor generate different electric signals. Consequently, we can analyze and perceive the motion of volunteers. When the sensor is attached to the finger joints or elbow, the sensor can generate characteristic signals in response to joint motions ( Figures 6 D and 6E). The signal induced by a flexion-extension movement has been highlighted with the blue dotted oval frame. Apparently, a positive correlation is unraveled between the signal by this sensor and the rotation angle of the joints. In addition, the sensor was also placed into the shoes to demonstrate the versatility of the sensor. Figure 6 F exhibits the output signals generated by the sensor over a volunteer at different walking speeds. The magnitude of signals will expand with the increase of the velocity of the volunteer. In a word, the sensor based on the bio-TENG can precisely reflect the movement of volunteers. This fact further demonstrates its potential in the field of behavior perception, athletic monitoring, and human-machine interfacing. Figure 6 Working principles and performances of the flexible sensor (A and B) Schematically illustration (A) and a digital picture (B) of a 5CNC/PHB-based sensor for behavior monitoring. (C) A diagram of the working mechanism of the sensor. (D and E) Output currents of bio-TENG sensor attached to a volunteer’s knuckles (D) and elbow (E) with different bending angles. (F) Output currents of bio-TENG sensor fixed in the shoes of a volunteer with different speeds. Conclusions Here, a green-in-green CNC/PHB bionanocomposite was prepared by high-pressure molding. It was used for the first time to fabricate biotriboelectric nanogenerators. Contrary to the pure PHB sample, the degradation of the CNC/PHB nanocomposite could occur at room temperature with the existence of hydrophilic CNC. Also, the CNC changed the dielectric properties of CNC/PHB, which could induce more charges over the electrode of the bio-nanogenerator. Consequently, the electrical output of CNC/PHB-based single-electrode TENG increased with the increase of CNC contents. The output voltage and output current of 5CNC/PHB-based TENG were approximately 5.7 times and 12.5 times higher than those of pristine PHB-based TENG, respectively. Besides, the CNC/PHB-based TENG demonstrated considerable high performance in instantaneous output power output and charging various external capacitors. Importantly, in more than 20000 continuous cycles, it exhibited a remarkable stability/durability of the output electric signals. Further, despite high hardness of CNC/PHB composite, a flexible but simple arch sensor was assembled using the CNC/PHB-based bio-TENG. It showed significant practical value in athletic monitoring. In a word, we believe this work confirms a simple but effective method to improve the electrical outputs of a bio-polyester-based TENG. It also provides a great template for preparing more green-in-green materials used in environmental-friendly self-powered devices. Limitations of the study The uniform dispersion of nanofillers in the polymer matrix is very critical to the final properties of the polymer matrix nanocomposites. It is known nanofillers essentially tend to agglomerate in the polymer matrix. Currently, it remains a great challenge to prepare a uniformly dispersed polymer matrix nanocomposite with a very high nanofiller content. In this study, we prepared a series of PHB-based composites with different amounts of CNC. The positive effect of CNC on the electrical output of CNC/PHB-based TENG was discovered. This phenomenon is ascribed to the improvement of the dielectric constant of biocomposite systems, which is induced by CNC. It is notable that the dispersion of nanofiller in the matrix has immense effects on the dielectric property of the biocomposite system. However, serious aggregation of CNC will occur if the amount of CNC is increased further to more than 5%. Therefore, at the present stage, we conclude that the 5% CNC content is the optimal CNC content for the best output performance of the CNC/PHB-based TENG. Future research efforts will mainly focus on the design and manufacture of green-in-green bionanocomposites with extremely high content of uniformly dispersed CNC. We anticipate that this will further boost the generated energy output of green-in-green bionanocomposites."
} | 6,268 |
35609504 | null | s2 | 892 | {
"abstract": "Synthetic microbial consortia represent a frontier of synthetic biology that promises versatile engineering of cellular functions. They are primarily developed through the design and construction of cellular interactions that coordinate individual dynamics and generate collective behaviors. Here we review recent advances in the engineering of synthetic communities through cellular-interaction programming. We first examine fundamental building blocks for intercellular communication and unidirectional positive and negative interactions. We then recap the assembly of the building blocks for creating bidirectional interactions in two-species ecosystems, which is followed by the discussion of engineering toward complex communities with increasing species numbers, under spatial contexts, and via model-guided design. We conclude by summarizing major challenges and future opportunities of engineered microbial ecosystems."
} | 231 |
31922128 | PMC6946262 | pmc | 894 | {
"abstract": "The von Neumann bottleneck has spawned the rapid expansion of neuromorphic engineering and brain-like networks. Synapses serve as bridges for information transmission and connection in the biological nervous system. The direct implementation of neural networks may depend on novel materials and devices that mimic natural neuronal and synaptic behavior. By exploiting the interfacial effects between MoS 2 and AlOx, we demonstrate that an h-BN-encapsulated MoS 2 artificial synapse transistor can mimic the basic synaptic behaviors, including EPSC, PPF, LTP, and LTD. Efficient optoelectronic spikes enable simulation of synaptic gain, frequency, and weight plasticity. The Pavlov classical conditioning experiment was successfully simulated by electrical tuning, showing associated learning behavior. In addition, h-BN encapsulation effectively improves the environmental time stability of our devices. Our h-BN-encapsulated MoS 2 artificial synapse provides a new paradigm for hardware implementation of neuromorphic engineering.",
"introduction": "1. Introduction The challenges of traditional computing architectures stem from storage capacity limitations and the high cost of specific data transfer speeds between memory and processors, the so-called von Neumann bottleneck [ 1 – 5 ]. With the advent of the artificial intelligence and big data era, this dilemma is becoming more profound. Brain-inspired neuromorphic engineering is different to the von Neumann architecture, combining memory and computation, with efficient energy utilization, and flexible adaptive and massively parallel computing capabilities [ 6 ]. It may achieve unprecedented technological breakthroughs, fundamentally overcoming the von Neumann bottleneck [ 7 , 8 ]. Artificial synapses, just as those in the biological nervous system [ 9 ], play an important role in connecting various neuron blocks as the basic units of neuromorphic engineering [ 10 ]. Constructing new, stable, reliable, and efficient artificial high-performance synaptic devices is essential for neuromorphic engineering and neural network computing [ 11 ]. Many artificial synaptic devices have been reported, including oxide electric double layer [ 12 – 14 ], ionic liquid/gel transistors [ 15 – 20 ], memristors [ 21 – 29 ], phase-changed memory [ 30 – 34 ], and ferroelectric transistors [ 35 – 37 ]. Also, the unique internal and interfacial structures of two-dimensional (2D) materials, as well as their electrical and optical properties [ 38 – 40 ], make them promising candidates for complex neuromorphic engineering [ 41 – 45 ]. In addition, optical modulation can establish a connection between the external environment and the brain through the visual system [ 46 – 48 ], and combining effective optoelectronic modulation is critical for neuromorphic engineering applications, such as artificial eyes and super vision [ 49 – 51 ]. Here, we demonstrate an efficient photoelectrical tunable h-BN-encapsulated MoS 2 synaptic transistor with basic synaptic functions. Furthermore, under electrical modulation, we successfully simulate the impressive Pavlov classical conditioning experiment through V bg tuning, which realizes the acquisition, extinction, and recovery function of associated learning. Due to the h-BN encapsulation, our devices exhibit superior environmental time stability. Our h-BN-encapsulated MoS 2 artificial synaptic transistor provides a novel paradigm for neuromorphic engineering based on 2D materials.",
"discussion": "3. Discussion In conclusion, our breakthrough, efficient, photoelectrical tunable, diverse h-BN-encapsulated MoS 2 synaptic transistor demonstrates basic synaptic functions including EPSC, PPF, LTP, LTD, synaptic gain, frequency, and weight plasticity. In addition, under electrical modulation, we successfully simulated the Pavlov classical conditioning experiment and realized the associated learning function. It is worth mentioning that due to the h-BN encapsulation, our devices have superior environmental stability. Our synaptic transistor provides an unparalleled perspective on novel 2D material-based neuromorphic engineering and brain-like computing."
} | 1,033 |
36367937 | PMC9651858 | pmc | 895 | {
"abstract": "A robust power device for wearable technologies and soft electronics must feature good encapsulation, high deformability, and reliable electrical outputs. Despite substantial progress in materials and architectures for two-dimensional (2D) planar power configurations, fiber-based systems remain limited to relatively simple configurations and low performance due to challenges in processing methods. Here, we extend complex 2D triboelectric nanogenerator configurations to 3D fiber formats based on scalable thermal processing of water-resistant thermoplastic elastomers and composites. We perform mechanical analysis using finite element modeling to understand the fiber’s deformation and the level of control and engineering on its mechanical behavior and thus to guide its dimensional designs for enhanced electrical performance. With microtexture patterned functional surfaces, the resulting fibers can reliably produce state-of-the-art electrical outputs from various mechanical deformations, even under harsh conditions. These mechanical and electrical attributes allow their integration with large and stretchable surfaces for electricity generation of hundreds of microamperes.",
"introduction": "INTRODUCTION The fast development of wearable technologies has triggered a strong interest, both in academia and industry, to develop novel forms of efficient power systems beyond batteries, including energy harvesting from sunlight ( 1 – 3 ), heat ( 4 , 5 ), and movement ( 6 ). Among these developments, triboelectric nanogenerators (TENGs) have proven to be highly efficient to harvest low-frequency mechanical energy. Combined with unique attributes such as the potential for simple and low-cost fabrication schemes and an important materials and architecture flexibility, TENG devices have a great potential for impact in implants ( 7 – 9 ), robotics ( 10 , 11 ), sports ( 12 , 13 ), health care ( 14 , 15 ), and security ( 16 , 17 ). A particularly emerging trend is to develop ultralight weight, large-area, low-profile, and unobtrusive TENG configurations ( 6 ), for the next generation of implantable probes ( 18 ), scaffolds ( 19 ), and wearable devices ( 20 ). Compared with standard two-dimensional (2D) devices, networks of advanced functional fibers represent ideal choices, allowing for highly complex and multimodal deformations, combined with breathability, robustness, and washability. Moreover, fiber technologies have evolved to the point where thin (submillimeter-diameter) fibers integrating different materials with complex micro- and nanostructures can be fabricated with high production yield. These advances have enabled the recent large-scale fabrication of high-performance display textiles and fiber lithium-ion batteries, which represent important achievement for smart electronic textiles and offer innovative ways to interact with electronic devices ( 21 – 23 ). Thus far, however, the most advanced TENG structures combining high efficiency and robust output performance with the proper encapsulation, flexibility, and deformability could not be demonstrated in the fiber form. Current fiber-based TENG devices often operate in the single-electrode mode. This configuration is convenient to fabricate and use and can result in fibers with excellent elastic properties. However, existing systems lack an encapsulating body, which is indispensable for constructing a robust TENG configuration to avoid water vapor–induced deterioration of the triboelectric charges. This feature is imperative for harvesting energy under in vivo ( 8 ) and other complicated conditions ( 24 ), such as in the context of wearables and textiles. Moreover, single-electrode systems will see their output fluctuations depending on which materials they get into contact with, hence can be prone to unstable performance that is difficult to anticipate. Incorporating all the triboelectric active surfaces within a single device also ensures stable electrical outputs independent of material properties of the external mechanical sources, which can be an important requirement for self-powered sensing purposes. Another commonly used configuration is the contact-separation mode, which alleviates the limitations mentioned above, but at the cost of fabrication complexity, which, in turn, markedly limit the performance of current contact-separation TENG fibers. These fibers are often composed of a cylindrical core-sheath structure with a gap between the inner core column and the outer sheath tube ( 25 – 28 ). Thus far, only very short lengths—generally at the scales of a few tens of centimeters—have been realized, by using multistep processing approaches that rely on coating, wrapping, and molding. These challenges have limited this design to laboratory scales, severely precluding their mass production and application. In addition, commonly used metallic wire–based electrodes are neither soft nor elastic, and even with the well-known serpentine configuration ( 29 ), the achieved mechanical deformability usually remains insufficient to meet the requirements for wearable applications. These fabrication challenges have also imposed severe limitations on the achievable device architectures, precluding progress in terms of performance and modes of mechanical actuation. Moreover, there has been a lack of focus on mechanical deformation modeling, partly because of the limited flexibility of the fabrication processes. Proper modeling is key to understanding and designing novel fiber devices with higher performance and robustness. In this work, we alleviate the challenges associated with fabricating 3D TENG designs within ultrasoft and microstructured fibers, via innovative materials, processes, and modeling-based designs. Specifically, we demonstrate an advanced and robust triboelectric fiber composed of two stretchable and conductive nanocomposite materials with microstructured surfaces, separated by a large air gap and surrounded by a water-repellent elastomeric cladding. We show that thermal drawing can be used to produce the targeted fibers with thin, uniform, and complex cross-sectional structures in a scalable, simple, and precise way. We first design the fiber architecture by relying on finite element modeling (FEM) that enables a deeper understanding of the mechanical deformation of the fiber system. Operating in the contact-separation mode, the fiber can reliably generate electricity from reversible compression and stretching, even under the harsh conditions of humidity or repeated cycling. Aligning a 10-m-long fiber on a large and stretchable surface, we demonstrate the fiber’s robust and reliable mechanical and electrical output performance triggered by various levels of compression and output currents of up to hundreds of microamperes. The innovative triboelectric fiber designs, materials, and processing strategy establish an innovative yet scalable approach to realizing robust and efficient TENG systems within soft and wearable constructs, paving the way toward novel devices and applications in health, personal care, and wearable devices.",
"discussion": "DISCUSSION In this work, we proposed conceptual advances to fiber-based energy harvesting systems in the aspects of materials, processes, and device designs. Specifically, we developed a highly soft and stretchable microstructured triboelectric fiber that integrates all the necessary functional components for contact-mode operation, including two triboelectric parts and an encapsulated large gap in between, within a water-repellent elastomeric cladding. We performed FEM simulations on the structure of the demonstrated fiber system to study its deformation under compression. By using the experimentally validated FEM, we explored the level of control and engineering on the mechanical behavior of the fiber that can be achieved on the basis of application-targeted designs. We proposed advanced design unachievable at these scale and feature sizes with other fabrication approaches, which integrated microstructured surfaces and staggered triboelectric components in a single thin fiber. Working in the contact-separation mode, the produced fiber can harvest energy from a variety of mechanical deformations, including compression and stretching. Moreover, the excellent deformability and robust performance of the long fibers allow the functionalization of large area textiles and enable energy harvesting and sensing with multiple modes of mechanical activation. This work brings innovative solutions to the scientific and technological challenges facing the field of TENG devices and advanced fibers and opens novel opportunities in energy harvesting, sensing, soft electronics, and smart textiles."
} | 2,174 |
32929064 | PMC7490352 | pmc | 898 | {
"abstract": "Recently, three-terminal synaptic devices have attracted considerable attention owing to their nondestructive weight-update behavior, which is attributed to the completely separated terminals for reading and writing. However, the structural limitations of these devices, such as a low array density and complex line design, are predicted to result in low processing speeds and high energy consumption of the entire system. Here, we propose a vertical three-terminal synapse featuring a remote weight update via ion gel, which is also extendable to a crossbar array structure. This synaptic device exhibits excellent synaptic characteristics, which are achieved via precise control of ion penetration onto the vertical channel through the weight-control terminal. Especially, the applicability of the developed vertical organic synapse array to neuromorphic computing is demonstrated using a simple crossbar synapse array. The proposed synaptic device technology is expected to be an important steppingstone to the development of high-performance and high-density neural networks.",
"introduction": "Introduction With the rise of the “big data” era, in which there has been an explosion of unstructured data, such as images, text, sound, and video, handling such types of data through using conventional von Neumann computing with separate processing and memory units has become difficult 1 – 5 . Neuromorphic computing‒which mimics the ability of the human brain to perform energy-efficient parallel processing of information using a complex neural network (NN)‒has attracted considerable attention as one of the pathways to meet such technical demands 6 – 10 . The brain processes and memorizes information simultaneously, which makes it free from the bottleneck problem. As such NNs in the brain consist of numerous synapses, the development of high-density and low-power synapse-like devices is essential to the successful implementation of neuromorphic computing 1 , 2 , 4 , 11 – 14 . As pioneering research, extensive studies on an artificial synapse based on a two-terminal resistive memory device have been conducted in recent years 4 , 7 , 11 , 13 , 15 – 17 . These two-terminal synapses are fabricated in a crossbar array structure, whose simplicity and short channel ensure a high integration density and low power consumption. However, nondestructive weight update in the two-terminal synapse is difficult to be accomplished because of its structural nature, i.e., a single shared terminal for reading and writing 7 , 15 – 21 . Very recently, Wang et al. effectively alleviated this issue by applying a significantly low readout voltage pulse, but further researches are still required for resolving this issue fundamentally 17 . In the meantime, three-terminal synaptic devices have attracted considerable interest owing to their nondestructive-weight-update behavior, which is attributed to the completely separated terminals for reading (drain) and writing (gate) 1 , 6 , 9 , 22 – 27 . In recent studies, three-terminal artificial synapses implemented with various inorganic and organic materials showed a desirable weight-controllability property via various charge-storage principles using interfacial traps 28 – 30 , atomic vacancies 14 , ion intercalation 22 , 26 , 28 , 31 , and floating gates 32 – 35 . For example, electric-double-layer transistors and floating-gate transistors have been demonstrated to be able to successfully emulate a biological synapse 12 , 32 , 33 , 36 . However, three-terminal synapses have a lower array density and a structural limitation on line-design compared to the two-terminal crossbar array structure in a complicated circuit configuration; these drawbacks result in a lower processing speed and higher energy consumption of the entire system. Herein we propose a vertical synapse featuring a remote weight update via ion gel, which is also extendable to a crossbar array structure. For the device configuration, a sub-100-nm-thick poly(3-hexylthiophene) (P3HT) channel is positioned at every cross-point of the pre- and postsynaptic terminals, and the ion-gel weight-control (WC) layer is deposited on them. Mobile ions in the ion gel readily penetrate the free volume in the P3HT channel, which results in a nonvolatile change in the channel conductance. Important synaptic properties, such as short-term plasticity (STP), excitatory and inhibitory postsynaptic currents (EPSC and IPSC, respectively), and long-term potentiation/depression (LTP/D) are evaluated via current–voltage measurements. In particular, the dimensional condition of vertical channel for achieving the optimal LTP/D characteristics are investigated via control of the channel length and area of the line cross-point. Finally, the applicability of the developed organic synapse array to the hardware NNs (HW-NNs) is evaluated in two ways: small-scale real-time learning and large-scale theoretical simulation.",
"discussion": "Discussion In this study, we successfully implemented a crossbar synapse array based on a vertical organic transistor with an ion-gel WC layer. This three-terminal synapse array was achieved by adopting the vertical gate-all-around field effect transistor (GAA-FET) concept and securing acceptable gate controllability with the assistance of an ion-gel dielectric. Mobile ions in the ion gel penetrated the free volume in the P3HT vertical organic channel located at every cross-point between the top and bottom electrode lines, which resulted in a nonvolatile change in the channel conductance. By virtue of ion movement, the proposed device exhibited diverse synaptic characteristics, such as STP, EPSC/IPSC, and LTP/D. In particular, optimization of the channel length and area of the line cross-point yielded excellent LTP/D characteristics, such as a large dynamic range (>10), low nonlinearity (<1), sufficient effective number of conductance states (>64), and low cycle variation (<1%). Furthermore, we demonstrated the feasibility of using the proposed vertical organic synapse array for implementing a complex NN through real-time training and classification tasks in a simple 2 × 3 NN. A very high recognition rate of 92.5% for MNIST digit patterns was achieved in a simulated two-layer ANN with a size of 400 × 200 × 10. To implement a hardware ANN with the vertical organic synapses as a follow-up research, the excellent endurance of the synapses is critically required. In this regard, identifying and understanding the failure mechanism for weight update will help in assessing and improving the endurance. Besides, the researches optimizing encapsulation layers, ion-gel dielectrics, and organic semiconductors in the synapses need to be done for the excellent endurance. Notably, this GAA-FET concept has already been considered for 3-nm technology nodes (for a lateral type) and next technology node (for a vertical type) by many global semiconductor companies. Thus, this research is meaningful as a proof-of-concept of a cross-point FET-type synapse array that can be used to implement NNs based on Si CMOS technology. We expect the proposed vertical crossbar synapse array to play a pioneering role in the development of high-performance and high-density NNs in the future."
} | 1,801 |
25359219 | PMC4215303 | pmc | 899 | {
"abstract": "To simplify the architecture of a neuromorphic system, it is extremely desirable to develop synaptic cells with the capacity of low operation power, high density integration, and well controlled synaptic behaviors. In this study, we develop a resistive switching device (ReRAM)-based synaptic cell, fabricated by the CMOS compatible nano-fabrication technology. The developed synaptic cell consists of one vertical gate-all-around Si nano-pillar transistor (1T) and one transition metal-oxide based resistive switching device (1R) stacked on top of the vertical transistor directly. Thanks to the vertical architecture and excellent controllability on the ON/OFF performance of the nano-pillar transistor, the 1T1R synaptic cell shows excellent characteristics such as extremely high-density integration ability with 4F 2 footprint, ultra-low operation current (<2 nA), fast switching speed (<10 ns), multilevel data storage and controllable synaptic switching, which are extremely desirable for simplifying the architecture of neuromorphic system."
} | 262 |
39028812 | PMC11259175 | pmc | 901 | {
"abstract": "The algal endosymbiont Durusdinium trenchii enhances the resilience of coral reefs under thermal stress. D. trenchii can live freely or in endosymbiosis, and the analysis of genetic markers suggests that this species has undergone whole-genome duplication (WGD). However, the evolutionary mechanisms that underpin the thermotolerance of this species are largely unknown. Here, we present genome assemblies for two D. trenchii isolates, confirm WGD in these taxa, and examine how selection has shaped the duplicated genome regions using gene expression data. We assess how the free-living versus endosymbiotic lifestyles have contributed to the retention and divergence of duplicated genes, and how these processes have enhanced the thermotolerance of D. trenchii . Our combined results suggest that lifestyle is the driver of post-WGD evolution in D. trenchii , with the free-living phase being the most important, followed by endosymbiosis. Adaptations to both lifestyles likely enabled D. trenchii to provide enhanced thermal stress protection to the host coral.",
"introduction": "INTRODUCTION Uncovering the foundations of biotic interactions, particularly symbiosis, remains a central goal for research, given that virtually no organism lives in isolation. Coral reefs are marine biodiversity hotspots that are founded upon symbioses involving dinoflagellate algae in the family Symbiodiniaceae ( 1 ). These symbionts are the “solar power plants” of reefs, providing photosynthetically fixed carbon and other metabolites to the coral host ( 2 , 3 ). Breakdown of the coral-dinoflagellate symbiosis (i.e., coral bleaching), often due to ocean warming, puts corals at risk of starvation, disease, and eventual death. Symbiodiniaceae microalgae are diverse, with at least 15 clades including 11 named genera ( 1 , 4 – 6 ), encompassing a broad spectrum of symbiotic associations and host specificity. Most of these taxa are facultative symbionts (i.e., they can live freely or in symbiosis), although exclusively symbiotic or free-living species also exist ( 1 ). Genomes of Symbiodiniaceae are believed to reflect the diversification and specialization of these taxa to inhabit distinct ecological niches ( 7 , 8 ). The genomes of symbionts, due to spatial confinement, are predicted to undergo structural rearrangements, streamlining, and rapid genetic drift (e.g., pseudogenization) ( 7 ). These traits are present in symbiotic Symbiodiniaceae ( 8 ). Whole-genome duplication (WGD) is an evolutionary mechanism that generates functional novelty and genomic innovation ( 9 , 10 ) and can occur due to errors in meiosis, i.e., via autopolyploidy. Following WGD, the evolutionary trajectory of duplicated sequence regions generally proceeds from large-scale purging, temporary retention and/or divergence, to fixation ( 11 ). WGD-derived duplicated genes [i.e., ohnologs ( 12 , 13 )] that are retained can provide a selective advantage and enhance fitness through increased gene dosage, specialization in function, and/or the acquisition of novel functions ( 11 ). WGD has been described in free-living unicellular eukaryotes such as yeast ( 14 – 16 ), ciliates ( 17 , 18 ), and diatoms ( 19 , 20 ), but not in symbiotic species. Evidence of WGD is absent in the Symbiodiniaceae, except for the genus Durusdinium , as observed in microsatellite sequence data ( 21 – 23 ). This genus includes the thermotolerant species Durusdinium trenchii ( Fig. 1A ), a facultative symbiont that confers heat tolerance to corals, thereby enhancing holobiont resilience under thermal stress ( 24 , 25 ). We hypothesize that WGD played a critical role in enhancing the capacity of this symbiont to confer heat tolerance to host species. Specifically, the facultative lifestyle (i.e., free-living or symbiotic) of D. trenchii favored fixation of WGD both during the free-living phase as an adaptation to fluctuating environmental conditions, and the symbiotic phase with an expanded gene inventory being further modified by the coral or other host species ( 26 ). Here, we generated de novo genome assemblies from two isolates of D. trenchii and analyzed their evolutionary trajectories. On the basis of gene expression profiles, we elucidate how the facultative lifestyle has contributed to the fate of ohnologs in these microalgae, and how natural selection acting on gene families has increased thermotolerance of corals hosting D. trenchii symbionts. These data provide strong evidence for the dual lifestyle hypothesis as a driver of post-WGD genome evolution. Fig. 1. WGD in a facultative coral endosymbiont. ( A ) Microscopic images of a free-living D. trenchii cell and an Exaiptasia pallida anemone hosting D. trenchii under fluorescence, with red indicating the presence of D. trenchii . ( B ) Repeat landscapes shown separately for the CCMP2556 and SCF082 genomes. ( C ) Circle plot depicting the location of syntenic blocks containing collinear gene blocks (i.e., ohnologs) between the CCMP2556 and SCF082 genomes. Ribbons indicate syntenic gene blocks identified with MCScanX that overlap with putative WGD-duplicated regions in both isolates (blue; n = 2427), one isolate only (red; n = 612), or neither isolate (black; n = 35). ( D ) The percentage of genes in duplicated collinear gene blocks relative to the number of duplicated collinear gene blocks identified within the genomes of Suessiales species. ( E ) Number of genes and syntenic genes recovered for each gene duplication category for the two isolates. ( F ) Phylogenetic tree of order Suessiales showing the number of lineage-specific gene family duplications at each node.",
"discussion": "DISCUSSION Our results provide strong evidence that the dual lifestyle has been a key driver of post-WGD genome evolution in the dinoflagellate D. trenchii . Our working hypothesis is illustrated in Fig. 4 . Fig. 4. Model of divergence post-WGD in a facultative endosymbiont. Putative selective constraints faced by free-living and symbiotic Symbiodiniaceae under the dual lifestyle are shown, with a focus on post-WGD ohnolog sequence divergence and differential gene expression. Under the null hypothesis of a solely free-living lifestyle, we expect post-WGD adaptations to primarily be driven by fluctuating environmental conditions (e.g., nutrient availability). Under the hypothesis of a dual lifestyle that includes symbiosis, adaptations will also strengthen the maintenance of a stable host-symbiont relationship and efficient nutrient/metabolite exchange within the coral holobiont. Although our results provide stronger support for the free-living phase as the primary driving force behind post-WGD evolution, both lifestyles impact the maintenance and expression divergence of ohnologs. These combined selective forces increase the overall fitness in D. trenchii , with the greater expression divergence of ohnologs under elevated temperatures a contributor to the high thermotolerance of this species when it is in symbiosis with corals ( 47 ). Benefits conferred by WGD to a free-living lifestyle in more variable environments, as well as tailoring of post-WGD duplicates to different lifestyles, primed D. trenchii to persist longer in the coral holobiont when faced with thermal stress. Whether symbiosis may also have negative effects on fitness post-WGD is unknown ( 48 ). It should be noted that the dual lifestyle is widespread in Symbiodiniaceae ( 1 ), but WGD is not. Although other facultative symbionts within Symbiodiniaceae (e.g., Cladocopium thermophilum and Durusdinium glynnii ) are also known for their thermotolerance ( 49 – 51 ), WGD was not implicated in these lineages ( 8 , 27 ). Therefore, the key feature of D. trenchii that we are addressing is not dual lifestyle alone, but rather how the capacity for dynamically switching between the symbiotic versus free-living phase affects post-WGD genome evolution and adaptation. Because Symbiodiniaceae propagate to very high densities in coral tissues (10 5 to 10 6 cells/cm 2 ) ( 52 , 53 ), the symbiotic phase of D. trenchii allows a rapid increase in the population size, particularly of fast-growing genotypes, while resident in host tissues. Consequently, genotypes that have faster growth rates or greater resilience to heat due to WGD-derived adaptations can re-seed free-living populations upon dissociation from the coral due to colony death, bleaching, or other mechanisms of symbiont population control. Repeated cycles of symbiosis followed by the free-living phase may therefore increase the overall fitness of D. trenchii populations under the dual lifestyle ( 26 ). Retention of multiple gene copies combined with fixed, adaptive changes likely makes D. trenchii more capable of metabolic maintenance under dynamic, often stressful environments, and hence a more resilient symbiont. Such factors may, in turn, explain the large geographic and expanded host range of D. trenchii ( 23 ) and its well-known capacity for increasing coral survival under heat waves. Therefore, in an intriguing and unexpected twist, WGD, primarily driven by selection under a free-living life phase has converted D. trenchii into a coral symbiont able to protect the host coral from thermal stress during symbiosis. D. trenchii is also a valuable model for studying the genome-wide impacts of facultative lifestyles."
} | 2,342 |
23390580 | PMC3565167 | pmc | 902 | {
"abstract": "The accurate calculation of decimal fractions is still a challenge for the binary-coded computations that rely on von Neumann paradigm. Here, we report a kind of memristive abacus based on synaptic Ag-Ge-Se device, in which the memristive long-term potentiation and depression are caused by a chemically driven phase transformation. The growth and the rupture of conductive Ag 2 Se dendrites are confirmed via in situ transmission electron microscopy. By detecting the change in memristive synaptic weight, the quantity of input signals applied onto the device can be “counted”. This makes it possible to achieve the functions of abacus that is basically a counting frame. We demonstrate through experimental studies that this kind of memristive abacus can calculate decimal fractions in the light of the abacus algorithms. This approach opens up a new route to do decimal arithmetic in memristive devices without encoding binary-coded decimal.",
"discussion": "Discussion Division of integers can be done using successive subtraction, but such treatment just results in the answer of quotient and remainder. Decimal fractions cannot be calculated directly by this approach, much less recurring decimal. For example, we have the result of 1 remainder 1 for , but it is impossible to obtain 1.25 just using 5 − 4. Furthermore, it is unable to do division by successive subtraction if the dividend is smaller than the divisor. Things are different if such division is implemented in the light of the abacus algorithms. The Chinese abacus algorithms suggest that division can be done by combining subtraction and addition, where the moves like “ replace ” are required. Accordingly, complex external circuitry should be necessary to implement the “ replace ” operations during the division. Several rules should be clarified before the calculation. First, subtraction will continue unless the remainder is smaller than the divisor (by comparing the resultant ΔSW). Second, the quotient is obtained according to the number of the times that subtraction is performed. For , if A − B − B is smaller than B , then the subtraction stops and the quotient will be 2. Third, given that the remainder ( r ) is smaller than the divisor, addition is required to operate the “ replace ”. Considering the decimal system, the remainder should be replaced by r × 10. To achieve this replacement, a multiple of 9 ( r × 9) is added, i.e. r + r × 9. Then successive subtraction can be implemented again. Fourth, every time the “ replace ” operation is carried out, the decimal fraction is getting ten times smaller, e.g. from tenths to hundredths. Fifth, once the remainder equals 0 the computation is fulfilled. Here we take as the example to show the approach. Following the method shown in Figure 4 , after the device is reset to 0, a train of 5 positive pluses is applied to obtain the dividend 5, as indicated by the blue columns in Figure 5 . Then we calculate 5 − 4 using a train of 4 negative pluses (orange columns in Figure 5 ), which leaves 1 as the remainder. The subtraction cannot be carried out again as 1 < 4 and the integer quotient is thus 1. So, a “ replace ” operation is required to continue the computation. We apply a train of 9 positive pulses (1 × 9) to replace the remainder 1, i.e. 1 + 1 × 9 = 10. The subtraction of 10 − 4 − 4 is performed until the remainder 2 is smaller than 4 once again, resulting in a new quotient of 2. We obtain the first number to the right of the decimal point, i.e. 2 tenths. Then the replacement has to be done for a second time, leading to the hundredths position. For this situation, two trains of 9 positive pulses (i.e. 2 × 9) are used to replace remainder 2 by 20 (2 + 9 + 9). After that, we repeat the above subtraction (−4) for 5 times until the remainder becomes 0, as confirmed by the last ΔSW of 0% in Figure 5 . Accordingly, 5 is the number at hundredths position, and a decimal fraction 1.25 is thus calculated. The statistical data of is provided in Supplementary Figure S4 to show the repeatability of this abacus approach. The approach presented herein can be adopted to calculate the decimal fraction smaller than one, recurring decimal (e.g. ), and so on. It is possible to do decimal arithmetic in memristive devices without encoding binary-coded decimal. This kind of memristive abacus can provide new insight into the development of the intelligent computing beyond the binary paradigm. This memristive approach, however, is at a “proof-of-concept” stage, which makes reliability assessment premature, because it is a challenge to completely solve the fluctuations in memristive synaptic responses at this stage 6 14 . Our hope is that the presented prototype will stimulate more systematic studies focusing on memristive computations. It is still a long way to accomplish the intelligent computing beyond the “proof-of-concept” stage. Many obstacles, for instance, the integration of MOSFETs/memristors circuits to improve the signal-to-noise ratio and transfer IMP states, are waiting to be solved. In addition, the in situ TEM observations suggest that the memristive LTP and LTD in many physical systems can be understood in the light of the kinetics of chemically driven phase transformation 27 ."
} | 1,316 |
34539237 | null | s2 | 904 | {
"abstract": "In order to better understand the relationship between Flagelliform (Flag) spider silk molecular structural organization and the mechanisms of fiber assembly, it was designed and produced the "
} | 48 |
33478381 | PMC7819203 | pmc | 906 | {
"abstract": "Background Lignin deposited in plant cell walls negatively affects biomass conversion into advanced bioproducts. There is therefore a strong interest in developing bioenergy crops with reduced lignin content or altered lignin structures. Another desired trait for bioenergy crops is the ability to accumulate novel bioproducts, which would enhance the development of economically sustainable biorefineries. As previously demonstrated in the model plant Arabidopsis, expression of a 3-dehydroshikimate dehydratase in plants offers the potential for decreasing lignin content and overproducing a value-added metabolic coproduct (i.e., protocatechuate) suitable for biological upgrading. Results The 3-dehydroshikimate dehydratase QsuB from Corynebacterium glutamicum was expressed in the bioenergy crop switchgrass ( Panicum virgatum L.) using the stem-specific promoter of an O-methyltransferase gene ( pShOMT ) from sugarcane. The activity of pShOMT was validated in switchgrass after observation in-situ of beta-glucuronidase (GUS) activity in stem nodes of plants carrying a pShOMT::GUS fusion construct. Under controlled growth conditions, engineered switchgrass lines containing a pShOMT::QsuB construct showed reductions of lignin content, improvements of biomass saccharification efficiency, and accumulated higher amount of protocatechuate compared to control plants. Attempts to generate transgenic switchgrass lines carrying the QsuB gene under the control of the constitutive promoter pZmUbi-1 were unsuccessful, suggesting possible toxicity issues associated with ectopic QsuB expression during the plant regeneration process. Conclusion This study validates the transfer of the QsuB engineering approach from a model plant to switchgrass. We have demonstrated altered expression of two important traits: lignin content and accumulation of a co-product. We found that the choice of promoter to drive QsuB expression should be carefully considered when deploying this strategy to other bioenergy crops. Field-testing of engineered QsuB switchgrass are in progress to assess the performance of the introduced traits and agronomic performances of the transgenic plants. Supplementary Information The online version contains supplementary material available at 10.1186/s12870-021-02842-9.",
"conclusion": "Conclusion The QsuB engineering approach has been established in switchgrass. This work highlights the fact that selecting an adequate promoter to drive QsuB expression should be an important parameter for successful engineering of other crops with this gene via tissue culture-dependent transformation methods. Considering that pShOMT activity is induced in the leaf and root by key regulators of biotic and abiotic stress responses such as salicylic acid, jasmonic acid and methyl jasmonate [ 12 ], it will be essential to field test our engineered pShOMT::QsuB switchgrass to assess its agronomic performance and resilience to environmental stress.",
"discussion": "Discussion Here, we describe the successful expression of the bacterial 3-dehydroshikimate dehydratase QsuB gene under the control of pShOMT in switchgrass. We show that the resulting plants display 12–21% reduction in lignin, a 2–3-fold increase in the bioaccumulation of PCA and a 5–30% increase in saccharification efficiency. pShOMT was previously shown to be preferentially active in stem vascular tissues in sugarcane, rice, maize, and sorghum [ 12 ], making it a good promoter candidate to express QsuB specifically in lignifying tissues within vascular bundles. Similar to previous observations made in sugarcane, we were able to detect GUS activity in stem nodes from switchgrass lines carrying a pShOMT::GUS construct. Nevertheless, an apparent reduction of lignin content observed in some discrete regions of leaf blades (i.e., fibers on the adaxial zone) from plants carrying the pShOMT::QsuB construct indicate that pShOMT is also active in leaf cells with secondary wall accumulation (Fig. 4 b). In addition to pShOMT , attempts to generate transgenic switchgrass lines with constructs containing QsuB under the control of the constitutive promoter of the maize ubiquitin1 gene ( pZmUbi-1 ) was unsuccessful, whereas only a single event was obtained with a pZmCesa10::QsuB construct containing the promoter of the maize cellulose synthase gene CESA10 involved in secondary cell wall formation [ 13 ] (Figures S 2 , S 3 ). This is possibly the result of toxicity occurring during the plant regeneration process when using these two pZmUbi-1::QsuB and pZmCesa10::QsuB constructs. Considering that QsuB diverts lignin biosynthesis, using the promoter of a lignin biosynthetic gene to drive QsuB expression may be more suited spatial-temporally during plant development. Interestingly, the single pZmCesa10::QsuB line showed a reduction of total lignin content as well as reduced phloroglucinol staining in leaf fibers (Figure S 2 E, F). Obtaining more switchgrass transgenic events with the pZmCesa10::QsuB construct will be essential to validate the effectiveness of pZmCesa10 in driving QsuB expression to reduce lignin content. The exact mechanism by which QsuB expression reduces lignin in switchgrass is still unresolved; in particular, whether the cytosolic pools of shikimate —required for HCT activity— and p -coumaroyl-shikimate are reduced remain to be demonstrated. Similarly, it would be interesting to determine the lignin monomeric composition in the different QsuB switchgrass lines, especially the relative amount of p -hydroxyphenyl (H) units, which is known to be higher in Arabidopsis QsuB plants and typically increases in HCT down-regulated dicot species [ 8 , 14 – 20 ]. Furthermore, the recent discovery in several plant species —including switchgrass— of genes encoding putative 3-hydroxylases (C3H) that convert p -coumarate to caffeate, as well as genetic evidence of their role in lignin formation in Brachypodium distachyon , question the exclusive role of HCT and the involvement of p -coumarate esters during lignin biosynthesis in monocots [ 21 ]. The overproduction of PCA in switchgrass lines expressing QsuB probably results from a partial conversion of the endogenous pool of 3-dehydroshikimate catalyzed by QsuB activity. Notably, increases in PCA titers (2–3-fold compared to control switchgrass) are smaller than those previously reported in Arabidopsis and tobacco plants containing the QsuB gene under the control of the promoter of the Arabidopsis cinnamate 4-hydroxylase gene ( pAtC4H ), which were at least two orders of magnitude higher compared to controls plants [ 8 , 22 ]. In connection with these observations, it has been demonstrated in vitro that PCA acts as a competitive inhibitor of at least one HCT isoform from switchgrass (i.e., PvHCT2) [ 23 ]. Therefore, it would be informative to attempt to identify putative p -coumaroyl-protocatechuate conjugates in metabolite extracts from pShOMT::QsuB switchgrass to determine if such HCT promiscuous activity —and possibly HCT inhibition— also occurs in vivo. Finally, it is promising to observe that the QsuB engineering strategy has the potential to enhance PCA titers in switchgrass biomass because several techno-economic analyses demonstrated the benefits of producing co-products in planta to render bioenergy crops economically sustainable [ 1 , 24 , 25 ]. In fact, several studies have already reported on the use of PCA as carbon source or pathway intermediate for the biological synthesis of diverse valuable products such as beta-ketoadipic acid, muconolactone, muconic acid, 2-pyrone-4,6-dicarboxylic acid, bisabolene, and methyl ketones [ 22 , 26 – 30 ]."
} | 1,922 |
37546216 | PMC10401328 | pmc | 907 | {
"abstract": "The commercial application of surfaces with superhydrophilic (SHPL) properties is well known as an efficient strategy to address problems such as anti-fogging, anti-frosting, and anti-biological contamination. However, current SHPL coatings are limited by their poor water and abrasion resistances. Thus, herein, to solve these problems active glass was employed as a substrate, and a stable and transparent SHPL solution was prepared via the spraying process. Aqueous polyacrylic resin (PAA), SiO 2 nanoparticles (NPs), tetraethyl orthosilicate (TEOS), and sodium allyl sulfonate (SDS) were utilized as the four main components of the PAA-TEOS-SiO 2 coating. The durability properties including anti-abrasion, resistance to water, and contact component loss were investigated via the Taber abrasion test, boiling water immersion test, and anti-fogging test, respectively. Furthermore, the structure, composition, and wettability of the coating before and after the friction and water immersion tests were compared via water contact angle (WCA) measurements. Furthermore, the effect of the type of resin on the properties of the coating was investigated. The surface morphology of the blended water-based polyacrylic acid (PAA) resin was uniform and flat and its adhesion to the substrate was the highest (4.21 MPa). Considering the durability and optical properties of the coating, the optimal blend was 3 wt% PAA resin, which exhibited a transmittance of 90%. When the content of TEOS, which enhanced the crosslinking in the coating, was increased to 2 wt%, the results showed that the SHPL coating maintained good anti-friction, boiling resistance, and anti-fogging properties under the conditions of 300 cycle Taber friction with 250 g load and soaking in hot water at 100 °C for 1 h. In particular, the excellent durability of strong acid and alkali resistance, heat resistance, and long-term aging resistance will facilitate the commercial viability and expand the application of SHPL coating in various research fields.",
"conclusion": "4 Conclusions By selecting different hydrophilic resins and further optimizing the TEOS content, the dual cross-linking of SiO 2 NPs was achieved, and a transparent, SHPL, and durable anti-fogging coating was obtained via a simple universal spraying method. The repulsion between SiO 2 in the sol solution and its low permeability threshold led to the formation of a fractal nano-porous inorganic surface, while the optimized hydrophilic PAA resin acted both as a surface modifier for the NPs and a chemical binder between the substrates and NPs as well as between the NPs. By adjusting the content of TEOS in the coating, its compact surface structure contributed to higher durability. The prepared SHPL anti-fogging PAA-TEOS-SiO 2 coatings exhibited good corrosion resistance, water resistance and abrasion resistance, high temperature thermal stability, and could maintain their superhydrophilicity after treatment at 180 °C for 25 days. The unique design of double cross-linking of SiO 2 NPs and two organic solvents, as well as the investigation of individual effects to achieve high transparency and SHPL anti-fogging properties will provide new insights into how to design scalable coatings for nanofabrication with enhanced durability.",
"introduction": "1 Introduction The construction of organic polymer–inorganic hybrid materials at the molecular level or nanoscale can combine the unique properties of organic polymers and inorganic materials at the molecular level. 1–3 Due to the unique phase structure and a variety of interfacial functional groups in polymers, composite organic polymer–inorganic hybrid materials at the molecular level exhibit unique properties compared with simple hybrid materials. 4,5 In water dispersion systems, due to their characteristics of long molecular chains, adjustable structure, and easy introduction of polar groups (such as –COOH, –SO 3 H and –NH 2 ), organic polymer, and inorganic oxide particles can provide steric hindrance and electrostatic repulsion for particles through physical adsorption, thus improving the dispersion and stability of the system. 6,7 In this context, cross-linking of inorganic NPs with organic polymers has been the focus of attention since 2000. 8–10 Due to their large surface area and abundant surface hydroxyl groups, SiO 2 NPs have been used to prepare super/high hydrophilic coatings with surface roughness, which are promising and competitive in a variety of applications, such as solar panels, surface self-cleaning, and anti-fogging glass. 11–13 However, due to their rich surface activity, they may be dispersed unevenly in the system and easily aggregated, which affects their use. 14 One of the most challenging and advanced issues in all types of hybrids is the pursuit of good functional characteristics such as anti-fogging, transmittance, anti-abrasion, and durability. 15,16 When heterogeneous components with different transmittance are mixed in the dispersive state, light scattering occurs, which significantly increases the turbidity of the material. 17–19 Nevertheless, when particles smaller than 20 nm are nano-dispersed in the continuous polymer matrix, optically transparent films can be obtained. 20,21 In this case, the uniform dispersion of NPs in polymer matrices is very difficult because of their large surface area and component incompatibility, which can easily cause polymerization and/or phase separation in the order of tens of nanometers or more. 22 Thus, to overcome these problems, several advanced interface designs have been developed using polymerization stabilizers and active surface modifiers. 23,24 For example, by changing the ratio of comonomers, Zhang et al. easily adjusted the interaction between the polymer components (sulfobetaine methacrylate (SBMA) and 2-hydroxyethyl methacrylate (HEMA)) and silica NPs, and prepared coatings with excellent scratch and abrasion resistance. 25 Due to the reversibility of the water-assisted electrostatic and hydrogen bond interactions, the coatings were self-healing and maintained good transparency. To eliminate the optical transparent barrier and realize the versatility of hydrophilic coatings, Zhang et al. prepared a polyvinyl alcohol (PVA)/SiO 2 coating via an ultrasonic enhanced dispersion and immersion-curing process. 26 Combined with a bi-component acrylic polyurethane basecoat, the PVA/SiO 2 coating not only possessed anti-fogging, anti-reflection, self-cleaning and underwater anti-fouling properties, but the universality of its substrates was also further extended. However, these polymers could only adhere to particle surfaces through weak physical interactions, such as van der Waals forces, hydrogen bonds, and electrostatic interactions, leading to the formation of non-uniformly dispersed components in the coating system. 27,28 In particular, the development of robust super-hydrophilic anti-fogging coatings still has not been explored to date. In addition, SHPL coatings usually have very high surface energy, and thus they easily lose their super hydrophilic energy in a hot and humid environment due to the loss of their active groups. 29–31 Therefore, anti-boiling property must be considered in the design and manufacture of next-generation anti-fogging coatings. Excitingly, the double cross-linking strategy is an effective alternative for the preparation of durable SHPL anti-fogging coatings. Herein, we propose a low-cost, scalable strategy for manufacturing SHPL anti-fogging coatings with complex and robust structures, where SiO 2 NPs were confined in a water-based solvent to provide appropriate roughness and an SHPL resin was coated as an outer high surface energy layer to enhance the superhydrophilicity of the system. Also, the interfacial interaction between the SiO 2 NPs and the substrate was further enhanced. Subsequently, double cross-linked TEOS was used as a high surface energy coating, and the “adhesive + NP + adhesive” dual cross-linking strategy was used to achieve a strong ultra-hydrophilic surface. The microstructure, chemical compatibility, thermal energy storage properties, long-term properties, thermal stability and anti-fogging performance of the prepared double cross-linking coating were also evaluated in detail. The designed SiO 2 -based SHPL anti-fogging coating provides a surface double cross-linking and highly selective enrichment strategy based on PAA and TEOS, which provides a new strategy to realize SHPL anti-fogging properties and reveals the interaction between unique surface properties and superwettability at the macro level.",
"discussion": "3 Results and discussion 3.1 Design strategy and characterization of the composite coating The molecular weight and viscosity of a polymer solution are two key factors affecting the morphology of crosslinked coatings due to the existence of covalent bonds between the activated glass substrate and the high viscosity and high transparency of water-based organic resins, based on which the adhesion of the coating can be improved and its durability can be guaranteed. 32,33 In our design, TEOS was used as a double crosslinking binder to further improve the robustness of the coating. Several water-based organic resins (including polyacrylic acid (PAA), polyurethane (PU), epoxy resin A (EP A), epoxy resin B (EP B), polyester resin (PAN), and organic silicon resin (SI)) were selected to optimize the resin. A schematic diagram of the synthetic SiO 2 -resin coating is shown in Fig. 2(a) . Briefly, using SDS as the stabilizing surfactant, after successively adding resin, deionized water, and nano SiO 2 sol, the organic polymer macromolecules self-assembled around the SiO 2 NPs to obtain SiO 2 -resin composite micelles in aqueous solution. It is well known that the adhesion can be further enhanced by adding a hydrophilic resin as an adhesive in the construction of SHPL composite coatings and SiO 2 NPs. 34 As shown in Fig. 2(b) , compared with the uncoated glass, the WCA of the coated surface was less than 5°, reaching a super hydrophilic state. According to the GB/T 31726-2015 standard, after the slide was placed on a water bath heating device for 60 s, as shown in Fig. 2(c) , it still maintained high transparency and no visible fog drops appeared, indicating its excellent anti-fogging property. 35,36 In addition, the coating had a good appearance. Furthermore, when the coating sample was placed parallel on colored paper, the image and text below were clearly visible, indicating that the coating has good optical properties. To ensure that the introduction of the resin did not damage the hydrophilic properties of the coating, it was beneficial to construct a hydrophilic resin through screening based on the wettability of the resin. Consequently, the introduction of the optimized resin ensured that the hydrophilicity of the coating was not damaged, contributing to the construction of a super hydrophilic coating. The wettability results are shown in Fig. S1. † PAA and polyurethane resin (PU) had good hydrophilic properties, with a WCA of less than 20°, the while other resins had a WCA greater than 60°. According to our previous work, EP B is a fluorocarbon resin with a C–F bond and low surface energy, and thus its intrinsic contact angle is large. In contrast, EP A, PAN, and SI have higher surface energies. However, PU and PAA contain abundant hydrophilic hydroxyl groups. 30,37 Therefore, PAA and PU were selected as organic components to prepare the organic and inorganic hybrid coatings. Fig. 2 (a) Schematic diagram of the double cross-linking design of the anti-fogging SHPL coating. (b) Wettability and (c) anti-fogging performance of the glass substrate and SHPL SiO 2 coating. SEM surface images of (d) SiO 2 , (e) PU-SiO 2 , and (f) PAA-SiO 2 coating. (g) Transparence, (h) WCA, and (i) adhesion on glass of SiO 2 , PU-SiO 2 , and PAA-SiO 2 coating. (j) WCA of PU-SiO 2 and PAA-SiO 2 coating after different Taber abrasion cycles. To further study the influence of different components on the surface morphology of the coating, the SEM morphology of the SiO 2 NPs after crosslinking with two types of resins was compared. As shown in Fig. 2(d) , the surface of the coating with 0.5 wt% SiO 2 NPs was observed to possess the obvious convex morphology of the NPs, in which the size of the raised NPs was about 3–10 nm. The surface morphology of the initial coating was uniform, and the spherical particles were closely arranged without gaps, holes and other defects. A comparison of the surface morphology of the PU-SiO 2 and PAA-SiO 2 coatings is shown in Fig. 2(e and f) , respectively. Meanwhile, a comparison of the surface morphologies of the polyurethane and PAA crosslinking coatings at different magnifications is shown in Fig. S2. † As shown in Fig. 2(e) and S2(a–c), † there is an obvious porous structure on the surface of the PU-SiO 2 coating. At a higher magnification, as shown in Fig. S2(a), † a large number of spherical NPs exposed on the coating surface through these holes can be observed. In contrast, the PAA-SiO 2 coating was more uniform and denser, as shown in Fig. 2(f) and S2(d–f), † indicating that the PAA resin resulted in better encapsulation without significant exposure of the spherical NPs. The surface of the PAA resin became relatively rough, and the SiO 2 NPs were adsorbed on the surface of PAA, indicating the formation of strong interfacial bonding between the organic phase and inorganic phase. 38 Considering transparent anti-fogging applications, it was necessary to investigate about optical transparency, surface wettability and abrasion resistance of the coating. As shown in Fig. 2(g and h) , the transparency of the SiO 2 coating without crosslinking remained at 90%. Subsequently, after crosslinking with PU and PAA, the transparency of the coating was 91% and 92%, respectively. Furthermore, droplets (3 μL) were dropped on the ceramic coating and composite coating, and the average value of the WCA was taken as the WCA of the coating surface. The hydrostatic contact angles of the ceramic coating and composite coating are 4.2°, 4.4° and 4.6°, respectively. The results indicated that the three coatings had a small contact angle, which is consistent with the fact that there are abundant –OH and –COOH hydrophilic groups on the surface of the SiO 2 NPs, PU, and PAA. 39–41 The adhesion test results of the three coatings are presented in Fig. 2(i) , where the adhesion of the inorganic coating without resin is 2.59 MPa. The adhesion of the organic and inorganic hybrid coating with PU resin increased to 3.98 MPa. Meanwhile, the adhesion of the organic and inorganic hybrid coating with PAA resin increased to 4.21 MPa. It can be seen that the prepared PU-SiO 2 coating and PAA-SiO 2 coating had a good appearance and initial anti-fogging performance, and the adhesion between the coating and the substrate also improve. In the case of the abrasion resistance of the PU-SiO 2 coating and PAA-SiO 2 coating, as shown in Fig. 2(j) , the abrasion resistance of latter was obviously better. After the 240 r Taber abrasion test, the WCA of the PU-SiO 2 coating was close to 15°, which greatly damaged the superhydrophilicity and anti-fogging effect of the coating. However, under the same condition, the WCA of the PAA-SiO 2 coating was only about 6°, which still maintained a good SHPL and anti-fogging effect (Fig. S3 † ). Combined with the adhesion comparison results, the surface of the PAA-SiO 2 coating was more uniform and denser than that of the PU-SiO 2 coating, and thus its strength was relatively higher and it had better abrasion resistance. Therefore, the PAA water-based transparent resin was selected to construct the organic and inorganic hybrid coating. To further optimize the content of PAA resin to obtain the coating with the best performance, coatings with a PAA content of 1 wt%, 2 wt%, 3 wt% and 4 wt% were prepared, and organic and inorganic hybrid coatings were prepared to study the changes in light transmittance, abrasion resistance and boiling resistance with an increase in the resin content. As mentioned above, the particle size of the NPs in the coating was closely related to its final transmittance. 42 Therefore, the relationship between particle size and PAA resin content in the coating was quantitatively analyzed using a laser particle size analyzer, and the transparency of the coating was further characterized, as shown in Fig. 3(a) . The particle size of the NPs increased with an increase in the content of PAA. The particle size of the coating with 1 wt% PAA was 48 nm, while that of the coating with 2 wt% PAA was 164 nm. When the PPA content was further increased to 3 wt% and 4 wt%, the particle size reached 186 nm and 285 nm, respectively. Thus, it can be concluded that there is a positive correlation between the PAA content and particle size. With an increase in the PAA content, the transparency of the 1 wt%, 2 wt% and 3 wt% coatings corresponded to 91.1%, 90.6% and 90.0%, respectively. Subsequently, with a further increase in the resin content to 4 wt%, the transparency of the coating was significantly reduced to 88.9%, as shown in Fig. 3(b) ; meanwhile, the actual sample was whiter and foggy than the bare glass slide. On the one hand, the increase in particle size increased the roughness of the coating surface, thus enhancing the light scattering effect; on the other hand, it also enhanced the light scattering effect of the particles themselves, and finally the transmittance of the coating decreased significantly. 43 Thus, to balance the transmittance and durability of the coating, the organic and inorganic hybrid coating was prepared by mixing 3 wt% PAA resin. Under the condition that the transparency of the coating was maintained at higher than 90%, the results of the Taber abrasion resistance under 250 g load and the boiling resistance test are illustrated in Fig. 3(c) . The corresponding abrasion resistance of the coatings with a PAA content of 1 wt%, 2 wt%, 3 wt% and 4 wt% reached 210 r, 220 r, 240 r, and 260 r cycles of abrasion resistance, respectively. Meanwhile, the boiling time of the coatings reached 25 min, 30 min, 40 min and 45 min, respectively. The presence of PAA provided a good protective “armor” for the coating, preventing the penetration, diffusion, migration, and loss of hydrophilic groups in the SHPL coating when contacted with water molecules. 44 To observe the crosslinking and redispersion effect of the PAA resin on the TEOS-SiO 2 medium, the surface morphology of the PAA resin was measured, as shown in Fig. 3(d) . With an increase in PAA content, the coating became denser and the surface particles were completely coated. Fig. 3 (a) Average size and transparency of the PAA-SiO 2 SHPL coating at different mass ratios of PAA. (b) Anti-abrasion cycles and anti-boiling time of PAA-SiO 2 coatings at different mass ratios of PAA. SEM surface image of PAA-SiO 2 coating at (c) 3 wt% and (d) 4 wt% PAA mass ratio. To determine the abrasion resistance durability and degradation mechanism of the as-deposited coatings, the coating morphology, anti-fogging performance and composition changes were characterized with respect to time. Fig. 4(a) and S4(a–c) † shows the morphological changes in the PAA hybrid coating at the scratch and the interface between abrasion and non-abrasion during the process of 0–300 r cycles. As shown in Fig. S4(a), † after 60 r of abrasion, a single scratch was observed to be shallow and narrow, and no obvious scratch marks were seen in the abrasion area. After 120–180 r, as shown in Fig. S4(b and c), † an imprint gradually appeared, and the size of a single scratch became wider, but the scratches in the abrasion area were still less in general. After 240 r of abrasion, as shown in Fig. 4(a) , an obvious deep scratch could be observed and the appearance of the scratch was clearly visible, but no abrasion phenomenon occurred. After 300 r of abrasion, a significant change with wide and deep scratches appeared on the surface, as shown in Fig. 4(b) . There is a significant difference between the scratches and the surrounding morphology. A smooth surface without the presence of SiO 2 NPs indicates that the coating in the worn area was completely damaged. Fig. 4 Surface SEM images of the PAA-SiO 2 coatings after the Taber abrasion test for (a) 240 r cycles and (b) 300 r cycles. Surface SEM images of the PAA-SiO 2 coatings after boiling immersion test for (c) 40 min and (d) 50 min. Change in (e) sulfur content and (f) wettability of surfactant during Taber abrasion test. Change in (g) sulfur content and (h) wettability of surfactant during boiling immersion test. To analyze the degradation mechanism of the coating after boiling for 40 min, the changes in its surface morphology were observed during the whole process of boiling from 10 min to 50 min. The change in morphology during the boiling process is shown in Fig. 4(c and d) and S4(d–f). † As shown in Fig. S4(d and e), † during the first 20 min of boiling, the spherical SiO 2 NPs were gradually exposed, indicating the dissolution of the resin, accompanied by a degradation in the comprehensive performance of the coating. Slight spalling occurred after the coating was boiled for 30–40 min, as shown in Fig. S4(e) † and 4(c) , and exposed areas of the substrate on the nanometer scale could be observed. After boiling for 50 min, the coating structure was obviously damaged, and multi-zone and large area spalling occurred ( Fig. 4(d) ). The superhydrophilicity of the PAA-SiO 2 coating is mainly attributed to the hydrophilic group of the SDS surfactant. Given that it is a characteristic element, a variation in the content S element corresponds to the loss of SDS. The variation in the characteristic S content on the surface of the coatings at different stages of abrasion is illustrated in Fig. 4(e) . During the abrasion process from 0 to 180 r, the content of S element on the surface of the coating did not decrease significantly compared with the initial stage. Regarding the surface morphology, due to the slight abrasion and good structural integrity at this stage, the surface activity was also completely distributed on the surface of the coating. When the abrasion revolution continued to 240 r, the content of S element decreased to a certain extent, which is consistent with the change in the characteristics of the surface topography of the coating. The surface of the coating worn at 240 r was damaged due to the deep scratches, leading to the loss of hydrophilic substances in a small area. After reaching 300 r, the content of S element decreased significantly, which was only 0.2%. This is because the coating had an obvious spalling phenomenon at this time, which removed a lot of its surface activity, and thus the content of S element on the surface dropped sharply. The WCA and actual anti-fogging effect of the coating at different abrasion stages are shown in Fig. 4(f) . Combined with the variation in the surface morphology, chemical composition and actual anti-fogging effect of the coating in the abrasion stage, it can be seen that the coating surface structure was complete within the first 240 r of abrasion, and the hydrophilic substances on the coating surface were also completely distributed, and thus the coating still had a good anti-fogging effect after abrasion. After 300 r of abrasion, the coating structure was obviously damaged, followed by the synchronous loss of hydrophilic substances, and thus the coating degraded with the loss of its anti-fogging effect. The variation in the surface characteristic element S content in the boiling stage of the coating is shown in Fig. 4(g) . Different from the morphology, the S element was rapidly lost within 20 min of boiling, where its content decreased from 0.88% to 0.3%. With an increase in the boiling time, and the content of S element remained at about 0.2% and no obvious fluctuation occurred. This is because after boiling for 20 min, most of the hydrophilic substances in the coating were lost to the water environment and formed a dynamic balance, and consequently the S element on the surface of the coating did not change in the subsequent boiling process. The change in the WCA on the coating surface and the change in the actual anti-fogging effect in the corresponding boiling stage were further characterized, as shown in Fig. 4(h) . In the first 40 min of boiling, the WCA of the coating was maintained at less than 6°, which could maintain good superhydrophilicity and anti-fogging performance. After boiling for 50 min, the WCA increased to about 10°, and the anti-fogging effect decreased significantly. Based on the changes in surface morphology, elements and actual anti-fogging effect of the coating during the whole boiling stage, it can be seen that although the hydrophilic components were rapidly lost within the first 40 min of boiling, the structural integrity of the coating was relatively high. This is because the synergistic effect of the hydrophilic substances with the relatively complete surface structure caused the coating to maintain a good super-hydrophilic anti-fogging effect. 45 However, after boiling for 50 min, the structure of the coating was obviously damaged with a decrease in hydrophilic substance content, and thus the coating degraded and lost its anti-fogging characteristics. To obtain higher abrasion resistance and boiling resistance, an appropriate amount of TEOS crosslinking agent was added to the hybrid coatings to improve their strength. Based on this strategy, coatings containing 0.5 wt%, 1.0 wt%, 1.5 wt% and 2.0 wt% TEOS were prepared, and their transparency and surface morphology were determined. As shown in Fig. 5(a–d) , with an increase in TEOS content, the coating gradually presented a dense network structure. In addition, as shown in Fig. 5(e) , the increase in TEOS content has no significant effect on the transmittance of the coating. To further understand the effect of microstructure changes on the coating properties, the wettability of five coatings after boiling for 60 min was tested to measure the boiling resistance of each coating. As shown in Fig. 5(f) , with an increase in TEOS content, the WCA of the coating gradually decreased after boiling for 60 min. When the TEOS increased to 1.5 wt%, the WCA of the coating after boiling for 60 min was about 5°, which had a good actual anti-fogging effect. With different TEOS contents in the coating and after 300 r of abrasion test, the change in actual anti-fogging effect is illustrated in Fig. 5(g) . After boiling at 60 °C for 1 h, the coating with a TEOS content of 2 wt% could maintain a good anti-fogging performance. Fig. 5 Surface SEM images of the PAA-TEOS-SiO 2 coating with different mass ratios of TEOS: (a) 0.5 wt%, (b) 1 wt%, (c) 1.5 wt%, and (d) 2 wt%. (e) Transparency of the PAA-TEOS-SiO 2 coating with different TEOS contents. (f) Wettability changes after anti-boiling and anti-abrasion tests and (g) anti-fogging performance of the PAA-TEOS-SiO 2 coating with different TEOS contents. To explore the reasons for the enhanced durability, the micro-morphologies of the coatings were compared and observed after 300 r of abrasion and 60 min of boiling. As shown in Fig. 6(a 1 ) , when the TEOS content was 0.5 wt% or less, the hybrid coating was seriously damaged after 300 r of abrasion, forming scratches about 1–2 μm wide, and the scratches were obviously worn and almost no SiO 2 NPs existed. With an increase in TEOS content to 1–1.5 wt%, no clear scratch edge was seen, and nanoparticles were observed in the scratch and not completely worn through ( Fig. 6(a 2 and a 3 ) ). As shown in Fig. 6(a 4 ) , for the coating with a TEOS content of 2 wt%, only blots were visible without obvious abrasion, and the SiO 2 NPs in the scratches were only flattened without peeling off. These results proved that with an increase in the TEOS content, the abrasion resistance of the coating gradually increased, and the coating with 2 wt% TEOS content could still maintain a relatively complete structure after 300 r of abrasion. Meanwhile, compared with similar SiO 2 -based SHPL coatings (Table S1 † ), the current work shows good abrasion resistance. Fig. 6 Surface SEM images of the PAA-TEOS-SiO 2 coating after the Taber abrasion test for 300 r cycles with different mass ratios of TEOS: (a 1 ) 0.5 wt%, (a 2 ) 1 wt%, (a 3 ) 1.5 wt%, and (a 4 ) 2 wt%. Surface SEM images of the PAA-TEOS-SiO 2 coating after boiling immersion test for 60 min with different mass ratios of TEOS: (b 1 ) 0.5 wt%, (b 2 ) 1 wt%, (b 3 ) 1.5 wt%, and (b 4 ) 2 wt%. Change in (c) wettability and (d) sulfur content of the surfactant during the Taber abrasion test. Change in (e) wettability and (f) sulfur content of surfactant during boiling immersion test. A comparison of the surface morphology of each coating after 60 min of boiling process is shown in Fig. 6(b 1 –b 4 ) . When the TEOS content was 0.5–1.0 wt%, the coating could be evenly peeled off after boiling, and the peeling area gradually decreased with an increase in TEOS content ( Fig. 6(b 1 and b 2 ) ). Combined with the initial morphology of the coating, it can be seen that the exfoliation resulted in some large-sized SiO 2 NPs aggregating in the coating. Due to their small specific surface area, the binding force of the large NPs on the surrounding NPs and the matrix was weakened. As shown in Fig. 6(b 3 and b 4 ) , when the content of TEOS increased to 1.5–2.0 wt% or above, the difference between the coating morphology and the initial surface morphology after boiling was small, and there was almost no exposed substrate. The results show that with an increase in TEOS content, the degree of crosslinking of the coating improved, and the coating had higher structural stability. To further understand the influence of the microstructure changes on the coating performance, the wettability of five coatings was tested. By comparing Fig. 6(c) and (e) , with an increase in TEOS content to 2.0 wt%, the WCA of the coating could be maintained at about 5° after 60 min of boiling and 300 r of abrasion, proving that improved durability was obtained. The variations in the characteristic S content on the surface of the coatings at different stages of anti-abrasion and anti-boiling tests are illustrated in Fig. 6(d) and (f) , respectively. After 300 r and 60 min, the content of S element decreased to about 0.2%. Under the external force of abrasion and boiling, good structural integrity could be maintained for a longer time, thus improving the durability of the coating. Based on the above-mentioned results and discussion, the enhanced comprehensive performance of double cross-linking can be illustrated by the hydrolysis of PAA and TEOS, including different bonding cases such as electrostatic interaction between SiO 2 NPs and SDS, and covalent bonding between the active glass substrate and PAA. A schematic illustration of the structure of the SiO 2 -PAA-TEOS coating is shown in Fig. 7(a) . By controlling the pH value of the solution and the electrostatic interaction between the anionic polymer chains, a series of anionic polymers was dispersed in SiO 2 solution on the nanoscale, and transparent and homogeneous SiO 2 -PAA-TEOS double-cross-linked materials were synthesized. The phase separation in the system was reduced by controlling the degree of hydrolysis of TEOS, slowing down the condensation of silanol and introducing coupling sites between the silicon phase and polymer chain in the in crosslinking process instead of using coupling agents. Specifically, the extent to which TEOS hydrolyzed was controlled by the limited availability of water released during the formation of PAA prior to mixing with the silica solution, which should be nearly the same for all the solutions studied. Therefore, the more TEOS in the starting material, the lower the degree of hydrolysis. In addition, the water released during the dehydration of PAA may also be involved in hydrolysis. It has been mentioned in the literature that due to the direct chemical bonding between the silyl alcohol group and the carboxylic acid group, the carboxylic group of PAA will act as the coupling site between the polymer and the silica phase. 46 To understand the process for the formation of the SiO 2 -PAA-TEOS coating, FT-IR spectroscopy was performed. As shown in Fig. S5, † the peaks at 1107 cm −1 , 967 cm −1 , 796 cm −1 , and 1096 cm −1 are attributed to the bending vibration of Si–O, symmetric stretching vibration of Si–O, bending vibration of Si–OH, and asymmetric stretching vibration of Si–O–Si, respectively. The peak at 1727 cm −1 is assigned to the stretching vibration of C \n \n\n<svg xmlns=\"http://www.w3.org/2000/svg\" version=\"1.0\" width=\"13.200000pt\" height=\"16.000000pt\" viewBox=\"0 0 13.200000 16.000000\" preserveAspectRatio=\"xMidYMid meet\"><metadata>\nCreated by potrace 1.16, written by Peter Selinger 2001-2019\n</metadata><g transform=\"translate(1.000000,15.000000) scale(0.017500,-0.017500)\" fill=\"currentColor\" stroke=\"none\"><path d=\"M0 440 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z M0 280 l0 -40 320 0 320 0 0 40 0 40 -320 0 -320 0 0 -40z\"/></g></svg>\n\n O of the PAA resins. Compared with that of the SiO 2 -PAA coating, a new absorption peak at around 1000 cm −1 could be observed for the SiO 2 -PAA-TEOS coating, which is assigned to the stretching vibration of the C O group. This shows that the carboxyl group (–COOH) in the hydrolysis process is cross-linked again with the hydroxyl group (–OH) in the silica phase. 30 Thus, the slow polycondensation of the silica phase occurs simultaneously with hydrolyzation, inhibiting the growth of large SiO 2 particles, preventing their agglomeration and increasing the compatibility between the polymer and inorganic phases. Using the above-mentioned three raw materials as the main components, the SHPL anti-fogging coating was prepared via the spray method. Fig. 7 (a) Schematic diagram of the double cross-linking mechanism of the PAA-TEOS-SiO 2 coating. (b) WCA of the SHPL coating after immersion in 1 M acid solution and 1 M alkali solution for different times. (c) Optical pictures of SHPL coating after immersion in 1 M acid solution and 1 M alkali solution for 36 h and 72 h, respectively. (d) WCA of SHPL coating after heating resistance test at 120 °C, 160 °C, 200 °C, 240 °C, 280 °C for 1.5 h. (e) Aging resistance of organic and inorganic hybrid super hydrophilic anti-fog coating. Inset is a photo of the appearance of the coating after 50 days of aging resistance test. To study the decay resistance of the organic and inorganic hybrid SHPL coating, the above-mentioned optimized organic and inorganic hybrid coating was immersed in hydrochloric acid solution with pH = 2 and sodium hydroxide solution with pH = 13, the acid and alkaline resistance of the coating was tested, and the change in the WCA on the coating surface after soaking for different times was recorded. As shown in Fig. 7(b) , the WCA of the coating remained below 6° after soaking in sodium hydroxide solution for 18 h, while that of the coating remained about 6° after soaking in hydrochloric acid solution for 72 h. The coating had good acid resistance and it possessed better acid resistance than alkaline resistance, which can be attributed to the acidic environment during solution preparation, while it was susceptible to oxidative degradation in alkaline environments. However, as shown in Fig. 7(c) , after the alkaline resistance test, the SiO 2 -PAA-TEOS coating did not show bleaching phenomenon compared with the SiO 2 inorganic coating, which indicates that the acid and alkali resistance further improved. Therefore, the addition of resin and the improvement in stable structure are beneficial for the construction of coatings with anti-oxidation and decomposition properties, thereby improving their acid and alkaline resistance. An accelerated aging experiment was performed to test the aging resistance. Specifically, the coating was stored at 180 °C, and the wettability of the coating surface was measured and recorded every 5 days. As shown in Fig. 7(d) , after 25 days of aging resistance test, the coating still maintained an SHPL state with a WCA of <6°. After 50 days of aging resistance test, the WCA increased to 13.8° with no changes in its appearance such as the appearance of yellow and spots on its surface. Also, the slight increase in WCA can be attributed to the oxidative decomposition at high temperature for the organic PAA resin with carbon chains as a skeleton. Therefore, the coatings containing resins were often more sensitive to temperature and more likely to degrade under continuous high temperature treatment. To test the heat resistance, the coating was placed at different temperatures for 90 min, then cooled naturally to room temperature, and its surface WCA measured, as shown in Fig. 7(e) . After thermal treatment at 120 °C, 160 °C, 200 °C, 240 °C and 280 °C, the WCA of the coating was less than 6°. The results show that the coating has good aging and heat resistance and it maintained its SHPL and anti-fogging properties in the temperature range of 0–280 °C."
} | 9,354 |
32670405 | PMC7341569 | pmc | 908 | {
"abstract": "Lignocellulose is the most abundant biomass on earth with an annual production of about 2 × 10 11 tons. It is an inedible renewable carbonaceous resource that is very rich in pentose and hexose sugars. The ability of microorganisms to use lignocellulosic sugars can be exploited for the production of biofuels and chemicals, and their concurrent biotechnological processes could advantageously replace petrochemicals’ processes in a medium to long term, sustaining the emerging of a new economy based on bio-based products from renewable carbon sources. One of the major issues to reach this objective is to rewire the microbial metabolism to optimally configure conversion of these lignocellulosic-derived sugars into bio-based products in a sustainable and competitive manner. Systems’ metabolic engineering encompassing synthetic biology and evolutionary engineering appears to be the most promising scientific and technological approaches to meet this challenge. In this review, we examine the most recent advances and strategies to redesign natural and to implement non-natural pathways in microbial metabolic framework for the assimilation and conversion of pentose and hexose sugars derived from lignocellulosic material into industrial relevant chemical compounds leading to maximal yield, titer and productivity. These include glycolic, glutaric, mesaconic and 3,4-dihydroxybutyric acid as organic acids, monoethylene glycol, 1,4-butanediol and 1,2,4-butanetriol, as alcohols. We also discuss the big challenges that still remain to enable microbial processes to become industrially attractive and economically profitable.",
"conclusion": "Conclusion and prospective The sustainability and competitiveness of biotechnological processes for the production of bio-based chemicals from renewable carbon sources critically depend on the need to equip the microbial platform with efficient metabolic pathways able to readily transform these carbon resources into bio-based products at titer, rate and yield (TRY) that are economically viable. For high-volume, low-priced chemicals such as those considered here, the higher these three performance indices are, the more competitive biotechnological processes become compared to petrochemical processes. Even though many microorganisms on earth are able to assimilate pentose and hexose sugars, their carbon metabolism is not optimally fashioned to achieve these performances. This necessitates to rewire their actual metabolic network, to plug (non-natural) pathways that exist in other organisms or to create new ones (artificial or synthetic), while ensuring cellular homoeostasis. Metabolic diversity found in Nature, combined with enzymes’ promiscuity, provides inspiration for implementing non-natural or constructing synthetic metabolic pathways that could meet these goals. In this review, we showed that plugging the non-phosphorylating pentose pathway or implementing synthetic pentose-1-P pathway in the industrially tractable organism E. coli , Corynebacterium glutamicum or S. cerevisiae empower these microbial cells with strong capability to convert lignocellulosic sugars into a wide range of added-value chemicals. Although these microbial processes are far from industrial performances, there are strong indications that this objective could be soon reached. First, these pathways are orthogonal to the natural metabolic networks, which in theory make them insensitive to endogenous metabolic and genetic regulation, while being deliberately controllable by mean of biosensors/biocontrollers [ 124 ]. This orthogonality may allow to partition carbon utilization into the production and growth purpose, enabling tunable dynamic control to favor one purpose over the other, as it was demonstrated for the production of glucaric acid from glucose [ 125 ]. Second, the number of reaction steps from the substrate to the product in the constructed pathway has to be as short as possible. The synthetic pathway for BDO from xylose is a best example since it only encompasses 6 steps, instead of the 21 required from glucose. Less reaction steps go in hand with less optimization workload and less intermediates to be siphoned off in branching pathways. Consequently, higher yield and higher productivity should be expected for production of bio-based chemicals. There still remain many challenges ahead to better improve microbial cell factories and concurrent biotechnology processes to make them industrially more appealing. We anticipate at least three major actions that must be actively worked on. The first one dealt with further rewiring of carbon metabolism for the conversion of all available renewable carbon resources, which include no solely lignocellulosic biomass, but CO 2 and methane from landfills. Although the conversion of these carbon sources into bio-based chemicals still largely depends on traditional metabolic paths, new routes must be designed and constructed that should be orthogonal to the central carbon metabolism without perturbing the cellular homoeostasis. As for instance, new route for methane and CO 2 assimilation into ethanol has been proposed by Liao and coworkers [ 126 ]. The same authors also provided various pathways enabling rewiring carbon metabolism to achieve maximal carbon conservation (reviewed in [ 127 ]). Innovation in yet inexistent pathways can be guided from exploration of the huge metagenomics data using chemoinformatics tools. This strategy has been recently applied to explore novel biosynthetic pathways for EG synthesis from C1 carbon source [ 128 ]. A second action that is critically determinant for high productivity and yield will be to increase catalytic efficiency and substrate specificity of pathway enzymes. However, getting these criteria can be a daunting task especially for enzymes in non-natural or synthetic pathways for which the substrates are not natural. This problem was clearly illustrated with α-ketoacid decarboxylase of Lactococcus lactis , whose promiscuity has been exploited to decarboxylate KdxD, DOP and α-adipate, but concomitantly generated metabolic side products due to this promiscuous activity [ 93 ]. Nevertheless, enzyme promiscuity provides an evolution starting point to yield specialized enzymes that catalyze the desired reactions on non-natural substrate. Rational enzyme engineering is a first approach to investigate the possibility to create enzyme with new function or with higher affinity to non-natural substrate than on its natural substrate. This strategy has been successful to completely change the specificity of an E. coli aspartate kinase encoded by lysC from its natural substrate—aspartate—to malate [ 129 ]. However, this rational approach is insufficient in many cases, asking for enzyme evolution either by random mutagenesis or by in vivo evolutionary engineering. Direct enzyme evolution can be very powerful if an appropriate high-throughput assay can be set up. Droplet-based microfluidic is thus well appropriate for this evolution, and its combination with IVTT (in vitro transcription–translation system) enables to screen for even larger libraries as it avoids transformation step in E. coli prior to the screen [ 130 , 131 ]. Alternative to this method is the direct in vivo enzyme evolution using automated micro-reactors as developed by Altar company ( http://www.altar.bio/about-us/ ). The use of this technique requires a phenotypic screening assay that exclusively relies on the vital function of the enzyme of interest. As an example, this strategy has been used to evolve an amino acid transaminase that can convert l -homoserine into the non-natural intermediate 2-keto-4-hydroxybutyrate (HOB), which thereafter could transfers its one carbon unit (CHO) to tetrahydrofolate (H4F), replacing the natural transfer C1-moieties from serine and glycine to H4F in an E. coli [ 132 ]. Fermentation process is likely a third target to be addressed early in the development of the microbial platforms as nicely reviewed by [ 133 ]. This aspect is really of the outmost importance when it deals with lignocellulosic hydrolysates or any raw materials that are on the one hand very cheap but on the other hand ill characterized, causing harsh fermentation conditions. Hence, key for industrial success, these new engineered microbial systems must be readily challenged with these harsh conditions, such as increasing their tolerance to toxic compounds and their global robustness by adapted laboratory evolution techniques [ 134 , 135 ]. Besides these aspects, the choice between production phases coupled or decoupled to growth can be decisive in setting the fermentation process as it should be mostly determined by the capacity to reach maximal production yield, while assuring redox and energy balance."
} | 2,200 |
22320432 | null | s2 | 909 | {
"abstract": "Silkworms and spiders generate fibers that exhibit high strength and extensibility. The underlying mechanisms involved in processing silk proteins into fiber form remain incompletely understood, resulting in the failure to fully recapitulate the remarkable properties of native fibers in vitro from regenerated silk solutions. In the present study, the extensibility and high strength of regenerated silks were achieved by mimicking the natural spinning process. Conformational transitions inside micelles, followed by aggregation of micelles and their stabilization as they relate to the metastable structure of silk are described. Subsequently, the mechanisms to control the formation of nanofibrous structures were elucidated. The results clarify that the self-assembly of silk in aqueous solution is a thermodynamically driven process where kinetics also play a key role. Four key factors, molecular mobility, charge, hydrophilic interactions, and concentration underlie the process. Adjusting these factors can balance nanostructure and conformational composition, and be used to achieve silk-based materials with properties comparable to native fibers. These mechanisms suggest new directions to design silk-based multifunctional materials."
} | 311 |
34751128 | PMC8579308 | pmc | 910 | {
"abstract": "Bacteria commonly live in spatially structured biofilm assemblages, which are\nencased by an extracellular matrix. Metabolic activity of the cells inside\nbiofilms causes gradients in local environmental conditions, which leads to the\nemergence of physiologically differentiated subpopulations. Information about\nthe properties and spatial arrangement of such metabolic subpopulations, as well\nas their interaction strength and interaction length scales are lacking, even\nfor model systems like Escherichia coli colony biofilms grown\non agar-solidified media. Here, we use an unbiased approach, based on temporal\nand spatial transcriptome and metabolome data acquired during E.\ncoli colony biofilm growth, to study the spatial organization of\nmetabolism. We discovered that alanine displays a unique pattern among amino\nacids and that alanine metabolism is spatially and temporally heterogeneous. At\nthe anoxic base of the colony, where carbon and nitrogen sources are abundant,\ncells secrete alanine via the transporter AlaE. In contrast,\ncells utilize alanine as a carbon and nitrogen source in the oxic\nnutrient-deprived region at the colony mid-height, via the\nenzymes DadA and DadX. This spatially structured alanine cross-feeding\ninfluences cellular viability and growth in the cross-feeding-dependent region,\nwhich shapes the overall colony morphology. More generally, our results on this\nprecisely controllable biofilm model system demonstrate a remarkable\nspatiotemporal complexity of metabolism in biofilms. A better characterization\nof the spatiotemporal metabolic heterogeneities and dependencies is essential\nfor understanding the physiology, architecture, and function of biofilms.",
"conclusion": "Conclusion In this study, we employed an unbiased approach based on temporal and spatial\ntranscriptomes and metabolomes to reveal that a multitude of amino acids and\nmixed acid fermentation pathways display profiles that are consistent with\ncross-feeding. Particularly strong regulation was displayed by alanine\nmetabolism, and we showed that alanine is a cross-fed metabolite inside\n E. coli colonies, between two spatially segregated\nsubpopulations, with an interaction length scale of tens of microns. Alanine\nconsumption supports growth in the cross-feeding-dependent region of the colony\nas a carbon and nitrogen source. Although many aspects of metabolism in biofilms\nare still unknown, methods for improved spatial and temporal analyses of\nmetabolite profiles and transcriptome data promise the possibility to discover\nnew metabolic interactions, and more generally understand the stability and\nfunctions of microbial communities.",
"introduction": "Introduction After bacterial cell division on surfaces, daughter cells often remain in close\nproximity to their mother cells. This process can yield closely packed populations\nwith spatial structure, which are often held together by an extracellular matrix.\nSuch spatially structured assemblages, called biofilms ( Flemming et al., 2016 ), are estimated to be the most abundant\nform of microbial life on Earth ( Flemming and\nWuertz, 2019 ). The metabolic activity of cells inside these dense\npopulations leads to spatial gradients of oxygen, carbon, and nitrogen sources, as\nwell as many other nutrients and waste products ( Ackermann, 2015 ; Evans et al.,\n2020 ; Pacheco et al., 2019 ;\n Stewart and Franklin, 2008 ). Cells in\ndifferent locations within biofilms therefore inhabit distinct microenvironments.\nThe physiological responses to these microenvironmental conditions result in\nspatially segregated subpopulations of cells with different metabolism ( Barroso-Batista et al., 2020 ; D’Souza et al., 2018 ; Stewart and Franklin, 2008 ). Bacterial growth into densely\npacked spatially structured communities, and metabolic activity of the constituent\ncells, therefore naturally lead to physiological differentiation ( Evans et al., 2020 ; Røder et al., 2020 ; Serra\net al., 2013a ; Stewart and Franklin,\n2008 ). Metabolic and phenotypic heterogeneities are frequently observed in multi-species\ncommunities ( Garg et al., 2016 ; Henson et al., 2019 ; Kim et al., 2020 ; Pacheco\net al., 2021 ; Serra and Hengge,\n2021 ) and in single-species biofilm populations ( Cole et al., 2015 ; Dal Co et\nal., 2019 ; Lin et al., 2018 ;\n Moree et al., 2012 ; Rani et al., 2007 ; Serra et al., 2013b ; Teal\net al., 2006 ). Identifying the origins of these heterogeneous\nsubpopulations, and how they interact with each other, is important for\nunderstanding the development and function of biofilms ( Cole et al., 2015 ; Lin et\nal., 2018 ; Liu et al., 2015 ;\n Prindle et al., 2015 ). Multi-species\nbiofilms predominate in the environment, yet they are highly complex and feature\nmany concurrent intra-species and inter-species interaction processes, which may be\norganized in space and time ( Brislawn et al.,\n2019 ; Garg et al., 2016 ; Kim et al., 2020 ). Due to this complexity of\nmulti-species biofilms, it is often difficult to disentangle whether spatiotemporal\nphysiological differentiation or a particular interaction between subpopulations is\ncaused by the relative position of the species, or the position of each species in\nthe context of the entire community, or the mutual response of different species to\neach other. In contrast, single-species biofilms offer precisely controllable model\nsystems with reduced complexity for understanding basic mechanisms of metabolic\ndifferentiation and the interaction of subpopulations. Investigations of phenotypic\nheterogeneity in single-species assemblages have already revealed fundamental\ninsights into metabolic interactions of subpopulations with consequences for the\noverall fitness and growth dynamics of the assemblages ( Arjes et al., 2021 ; Cole et\nal., 2015 ; Evans et al., 2020 ;\n Lin et al., 2018 ; Liu et al., 2017 ; Liu et\nal., 2015 ; Wolfsberg et al.,\n2018 ). However, even for single-species biofilms, the extent of metabolic\nheterogeneity and metabolic dependencies of subpopulations are unclear. To obtain an unbiased insight into the spatial organization of metabolism inside\nbiofilms, we measured metabolome and transcriptome dynamics during the development\nof E. coli colony biofilms on a defined minimal medium that was\nsolidified with agar. Our model system enabled highly reproducible colony growth and\nprecise control of environmental conditions, which allowed us to detect phenotypic\nsignatures of subpopulations. The temporally and spatially resolved data revealed\nthat alanine metabolism displays a unique pattern during colony growth. We\ndetermined that secretion of alanine occurs in a part of the anoxic region of the\ncolony, where the carbon and nitrogen sources are abundant. The secreted alanine is\nthen consumed in a part of the oxic region of the colony, where glucose and ammonium\nfrom the minimal medium are lacking. This spatially organized alanine cross-feeding\ninteraction occurs over a distance of tens of microns, and has important\nconsequences for the viability and growth of the localized cross-feeding-dependent\nsubpopulation, and for the global colony morphology.",
"discussion": "Discussion We showed that alanine metabolism is spatiotemporally regulated inside E.\ncoli colony biofilms. The high degree of control and reproducibility of\nour model system enabled us to determine that interference with the export and\nconsumption of alanine causes phenotypes in cell viability and cell growth rate.\nThese viability and growth rate phenotypes are localized in the region of the colony\nthat depends on alanine as a nutrient source, but they affect the global morphology\nof the colony. Based on our results, we propose the following model for the spatial organization of\nalanine metabolism in colonies that have grown for 72 hr ( Figure 6A ): Cells at the bottom periphery of the colony (red\nregion in Figure 6A ) have access to oxygen,\nglucose, and ammonium, and perform either aerobic respiration or fermentation by\noverflow metabolism ( Basan et al., 2015 ;\n Cole et al., 2015 ) – these two possible\nmetabolic states cannot be distinguished with our current approaches. Cells at the\nbottom center of the colony (orange region in Figure\n6A ) are anaerobic yet they have access to glucose and ammonium from the\nagar-solidified medium. These cells ferment glucose and secrete alanine, primarily\n via AlaE. Although many amino acid exporters have been\ndescribed for E. coli ( Pacheco et\nal., 2021 ; Prindle et al., 2015 ;\n Rani et al., 2007 ), their functions\nhave remained elusive under regular physiological conditions. Our data now reveal a\nfunction for the alanine exporter AlaE during biofilm growth. Secreted alanine\ndiffuses through the colony and can only be utilized by oxic nutrient-deprived cells\n(blue region in Figure 6A ). Alanine\nconsumption in the mid-height oxic region also has a detoxification effect, by\nreducing otherwise inhibitory levels of extracellular alanine. Alanine consumption\nat the oxic top of the colony is not significant, perhaps because the extracellular\nalanine is consumed before it reaches this region. Figure 6. Model for alanine cross-feeding in E. coli colony\nbiofilms. This model applies to colonies grown for 72 hr on solid M9 agar containing\nglucose and ammonium. ( A ) In cross-feeding capable colonies,\ncells in the bottom layer of the biofilm have access to glucose and\nammonium. Only cells in the outer periphery of the biofilm (green dashed\nline) have access to oxygen. Cells in the red region can use ammonium,\nglucose, and oxygen to perform aerobic respiration or fermentation by\noverflow metabolism. Cells in the orange region have access to glucose and\nammonium, but no oxygen. These cells secrete alanine. The secreted alanine\ncan be consumed by cells in the oxic region above this layer (depicted as\nblue), which perform aerobic respiration and convert alanine into pyruvate\nand ammonium that can be used for growth and to maintain cell viability.\n( B ) Colonies of\nΔ alaE Δ dadAX cells have a reduced\nability to consume and export alanine. These colonies have a region\nperforming aerobic respiration or overflow metabolism (red), similar to the\nparental strain. These colonies also have an anoxic fermentation region\n(orange), yet this region displays significantly less alanine secretion\ncompared to the parental strain. Furthermore, these colonies lack an alanine\nconsuming population in the oxic region. Due to the limited alanine\nsecretion and alanine consumption of this strain,\nΔ alaE Δ dadAX colonies display higher\ncell death and less growth in the otherwise cross-feeding-dependent oxic\nregion, resulting in a more conical colony shape in comparison to parental\ncolonies. Colonies that are impaired in alanine cross-feeding because of a reduced ability to\nsecrete and consume alanine display less growth and more cell death in the oxic\nglucose-deprived region, leading to a conical-shaped colony ( Figure 6B ). Furthermore, cross-feeding impaired cells are more\nsusceptible to growth inhibition by high extracellular alanine levels ( Figure 4C ) and are outcompeted by the parental\nstrain during colony growth. The spatiotemporal organization of alanine metabolism during colony growth is unique\namong amino acids ( Figure 1D and E ), and is\nbased on the secretion of alanine in the anaerobic, glucose-rich, and ammonium-rich\nbase of the colony. Why is alanine secreted in this region? We speculate that this\nis not an altruistic trait evolved to support a starving oxic subpopulation at a\ndifferent location, because such a trait would be highly susceptible to social\ncheaters in a multi-species community. Instead, alanine secretion in this region may\nbe a necessity to avoid high intracellular alanine levels that presumably result\nfrom anaerobic fermentation of glucose in the presence of ammonium. High alanine\nlevels are inhibitory ( Figure 4C ), so that\nsecretion of alanine and transport of alanine away from this population is\nbeneficial to this population. On the other side of the cross-feeding interaction, the alanine-consuming\nsubpopulation in the aerobic mid-height region of the colony strongly benefits from\nthe secreted alanine originating from the base of the colony. We note that the\nalanine consumption in the aerobic mid-height region of the colony necessarily\ncauses a steeper alanine concentration gradient between the two interacting\nsubpopulations, compared with a case in which no alanine is consumed. Due to Fick’s\nlaw of diffusion, a steeper concentration gradient causes a higher diffusive flux.\nTherefore, the presence of the alanine-consuming subpopulation results in a benefit\nfor the alanine-secreting subpopulation, by causing a higher diffusive flux of\nalanine away from the alanine-secreting population. It is unclear whether this\nbenefit for the alanine-secreting subpopulation is significant, as it is not\npossible for us to measure local growth rates or cell viability in the anoxic base\nof the colony due to optical limitations of confocal fluorescence microscopy in this\nregion. In summary, the interaction between the two cross-feeding subpopulations is\nlikely mutualistic. The spatial organization of alanine cross-feeding between two subpopulations we\ndescribed in this study is analogous to the carbon cross-feeding in E.\ncoli colonies via acetate, because it involves\nmetabolite secretion in the anaerobic population and consumption in the\ncarbon-starved oxic population ( Cole et al.,\n2015 ; Dal Co et al., 2019 ; Wolfsberg et al., 2018 ). However, the\nlocation of the population that is proposed to consume acetate as a carbon source\nspans most of the oxic region of the colony ( Cole\net al., 2015 ), whereas alanine is primarily consumed only in the oxic\nmid-height region of the colony. Interestingly, our spatial transcriptomes did not\nreveal a signature for acetate cross-feeding between the anaerobic and oxic regions\nof the colony ( Figure 1—figure supplement\n4B ), yet transcripts coding for enzymes involved in lactate, formate, and\nsuccinate metabolism display patterns that are indicative of spatially organized\nmetabolism that could be the basis of carbon cross-feeding. Whether acetate,\nlactate, formate, and succinate are in fact cross-fed in our system remains to be\ntested in future work. In contrast to these metabolites, alanine can not only be\nused as carbon source, but also as a nitrogen source in the cross-feeding-dependent\nregion, and we note that it is currently not clear what limits growth in the higher\nregions of the colony – whether it is carbon, nitrogen, or other elements such as\niron, sulfur, or phosphorous. Very recently, it was shown that alanine can be shared\nin colonies of the Gram-positive bacterium Bacillus subtilis , and\nthat this effect required three-dimensional colonies for unknown reasons ( Arjes et al., 2021 ). In our E.\ncoli model system, three-dimensional growth is required to create an\nanoxic region replete with carbon and nitrogen sources that causes alanine\nsecretion, and we speculate that this effect may also be required in B.\nsubtilis, and in other species. Several cross-feeding interactions have been described in detail for multi-species\ncommunities ( Henson et al., 2019 ; Kim et al., 2017 ; Moree et al., 2012 ; Pande\net al., 2016 ; Røder et al.,\n2020 ; Smith et al., 2019 ; Watrous et al., 2013 ; Yang et al., 2009 ). Even though cross-feeding interactions\nare likely a ubiquitous process in single-species bacterial multicellular structures\n( San Roman and Wagner, 2018 ), only a\nfew metabolic interactions between subpopulations have been documented for\nsingle-species biofilms ( Arjes et al., 2021 ;\n Cole et al., 2015 ; Evans et al., 2020 ; Lin et\nal., 2018 ; Liu et al., 2015 ).\nFor E. coli colonies grown on agar-solidified medium, and for\nbacterial communities in general, it is still unclear how many subpopulations\ninteract metabolically, and on which length and time scales these interactions take\nplace, and how significant these many interactions are for the community growth and\nstability. Conclusion In this study, we employed an unbiased approach based on temporal and spatial\ntranscriptomes and metabolomes to reveal that a multitude of amino acids and\nmixed acid fermentation pathways display profiles that are consistent with\ncross-feeding. Particularly strong regulation was displayed by alanine\nmetabolism, and we showed that alanine is a cross-fed metabolite inside\n E. coli colonies, between two spatially segregated\nsubpopulations, with an interaction length scale of tens of microns. Alanine\nconsumption supports growth in the cross-feeding-dependent region of the colony\nas a carbon and nitrogen source. Although many aspects of metabolism in biofilms\nare still unknown, methods for improved spatial and temporal analyses of\nmetabolite profiles and transcriptome data promise the possibility to discover\nnew metabolic interactions, and more generally understand the stability and\nfunctions of microbial communities."
} | 4,246 |
35729336 | PMC9213395 | pmc | 911 | {
"abstract": "Memristors, when utilized as electronic components in circuits, can offer opportunities for the implementation of novel reconfigurable electronics. While they have been used in large arrays, studies in ensembles of devices are comparatively limited. Here we propose a vertically stacked memristor configuration with a shared middle electrode. We study the compound resistive states presented by the combined in-series devices and we alter them either by controlling each device separately, or by altering the full configuration, which depends on selective usage of the middle floating electrode. The shared middle electrode enables a rare look into the combined system, which is not normally available in vertically stacked devices. In the course of this study, it was found that separate switching of individual devices carries over its effects to the Complete device (albeit non-linearly), enabling increased resistive state range, which leads to a larger number of distinguishable states (above SNR variance limits) and hence enhanced device memory. Additionally, by applying a switching stimulus to the external electrodes it is possible to switch both devices simultaneously, making the entire configuration a voltage divider with individual memristive components. Through usage of this type of configuration and by taking advantage of the voltage division, it is possible to surge-protect fragile devices, while it was also found that simultaneous reset of stacked devices is possible, significantly reducing the required reset time in larger arrays.",
"conclusion": "Conclusion A double stack (M-I-M-I-M) memristive configuration was fabricated which combines two devices and the ability to individually switch them by using a middle electrode (ME), or the ability to simultaneously switch them by applying voltage pulses to the top (TE) and bottom (BE) electrodes of the device. Each individual device can be separately tuned, with resulting resistive changes carried over to the complete device. Complete device resistance is dependent on the nonlinearity of individual device IVs and thus is generally not equal to the sum of constituent device resistances, but rather the sum of their static resistances at the voltage shares they receive from the divider. Individually switching the devices and then registering the resistive state of the Complete device can lead to increased resistive state resolution and consequently device memory density, even though the intrinsic state resolution of certain device technologies may exceed measurement resolution. Completed device switching can either switch both devices at the same time if they have similar resistivities, or only the most resistive one. Constituent device resistive state will influence the voltage balance in the voltage divider, thus deciding the amount of stimulus they will receive. In the case of serially connected devices of similar resistive range, switching will take place in both devices as long as the stimulus they receive is enough to change their state. For antiserially connected devices, evidence points towards a “memristive fuse” type of behaviour whereby due to opposing changes in resistive states across the devices, the allocation of voltage in the divider can swing much more widely than for serial devices -in principle-. In the case of similar resistive states (and switching threshold voltages) in antiserially connected devices, the Complete device may function as an attractor with unchanged resistive state, while the individual devices are impacted by the voltage stimulus applied. In either case, the existence of a second memristor in serial connection may, under the right conditions, function as a surge protection mechanism, in the same way as using a resistance in series, protecting each other from catastrophic failure. Finally, in the case of devices with the same polarity, simultaneous reset has been observed, which is a powerful tool to cut reset times in memristor arrays by half. In conclusion, we have demonstrated in silico a 3-terminal component consisting of 2 serially connected, vertically stacked, memristors, which can be used both as a “fuse” and as individual devices. We have shown that: a) the cointegration of the two devices and b) the fact that they share an electrode does not cause any significant deviation from the theoretical behaviours that are expected when examining independent devices, either in isolation or in serial/antiserial configuration as was done in previous work. Furthermore, we show that it is possible to obtain full-stack switching in these co-integrated devices; the co-integration process does not skew the required switching conditions sufficiently to introduce catastrophic complications. We hope that this work will help the community develop such 3-terminal devices as basic components for applications that can use either their individual or their fuse-configuration properties in the future.",
"introduction": "Introduction Since their conception 1 and eventual fabrication 2 memristive devices have been extensively studied in an effort to incorporate them in circuit designs 3 , in novel neuromoprhic computing setups 4 , as artificial electronic synapses in an attempt to replicate synaptic functions normally only available to biological brain structures 5 and as electronic nociceptors for artificial skin applications 6 . Increased integration density, which resulted from incorporation in circuits, uncovered issues which impede nominal device operation, such as sneak path currents. A popular mitigation strategy has been the usage of new topologies, such as complementary resistive switching (CRS) 7 , 1 selector 1 memristor (1S1M) 8 , 1 diode 1 memristor (1D1M) 9 configurations. Other theoretical propositions include using a complementary memristor array composed by two anti-serially connected memristors 10 . At the same time, these technologies also protect from unwanted switching or destructive breakdown of fragile devices, by enforcing a (soft) compliance current through the multiple layers used to fabricate them. On a separate front, efforts are being directed towards increased range in tuneability of devices 11 , especially in settings of neuromorphic computing, where synaptic weight management is of vital importance. By using memristors as standalone components, analogue reconfigurable circuits 12 and threshold logic gates 3 can be fabricated, among other possible circuitry. One commonly referenced “device ensemble” is the memristive fuse 13 , 14 , which elaborates on the expected behaviour of two memristors in series. Specifically, it has been previously pointed out 15 that memristive fuse is the natural extension of complementary switching when states in both devices exhibit analogue behaviour. In all of these “hardwired ensemble” cases it is not possible to gain insight into the workings of the separate layers of the ensemble, and fine tuning of states is a critical issue which has not yet been resolved. Here, we propose a three-terminal device, 1M1M topology, which consists of two memristors fabricated vertically, with an interleaved, but accessible common middle electrode. The case for three terminal connected devices has already been theoretically made where this configuration is used to propose neuromorphic or conventional circuits 10 , 16 . This configuration allows probing of the whole stack, while the ability to tune both memristors separately allows fine tuning of the compound stack’s resistive state. Other expectations from this type of configuration include increased tuneability, simultaneous switching of both devices and an inherent surge protection of devices due to the stack functioning as a voltage divider, improving resilience. With regards to fine tuneability, “super-resolution” memristive ensembles represent an idea that has been explored theoretically, with cells consisting of two memristors connected in parallel 17 , while here we are implementing it in a serial connection topology. This configuration allows both higher resolution and higher memory density per unit area. Both are achieved via manipulation of individual devices with different switching ranges, enabling the set-up of a coarsely-set resistive state originating from one device plus fine resistive tuning around that state by utilizing the second device. Notably, even for devices with similar resistive ranges and tuning tolerances, independent control of resistive states of two devices can achieve better control of the aggregate resistive state. Finally, the superposition of resistive states across both devices, broadens the overall tuning range of the aggregate state, providing further memory density (more available states). For simplicity we shall refer to these effects simply as “super-resolution” in the rest of the paper. Control of forming 18 and/or device manufacture can lead to both “serially” and “anti-serially” connected devices which can exhibit distinct computational behaviours. The difference being that in a “serial” stack, application of voltage of some chosen polarity will elicit the same response from both devices in the stack (both will either increase or decrease resistance)."
} | 2,298 |
36846531 | PMC9939870 | pmc | 912 | {
"abstract": "An artificial neural network (ANN) is a computational system that is designed to replicate and process the behavior of the human brain using neuron nodes. ANNs are made up of thousands of processing neurons with input and output modules that self-learn and compute data to offer the best results. The hardware realization of the massive neuron system is a difficult task. The research article emphasizes the design and realization of multiple input perceptron chips in Xilinx integrated system environment (ISE) 14.7 software. The proposed single-layer ANN architecture is scalable and accepts variable 64 inputs. The design is distributed in eight parallel blocks of ANN in which one block consists of eight neurons. The performance of the chip is analyzed based on the hardware utilization, memory, combinational delay, and different processing elements with targeted hardware Virtex-5 field-programmable gate array (FPGA). The chip simulation is performed in Modelsim 10.0 software. Artificial intelligence has a wide range of applications, and cutting-edge computing technology has a vast market. Hardware processors that are fast, affordable, and suited for ANN applications and accelerators are being developed by the industries. The novelty of the work is that it provides a parallel and scalable design platform on FPGA for fast switching, which is the current need in the forthcoming neuromorphic hardware.",
"conclusion": "Conclusions ANNs are known for their high degree of connectedness and massive data volumes. For the realization of the single-layer networks, neuron-level parallelism is more effective. The intrinsic distributed component of ANNs is in both memory and computational logic, suggesting that the implementation will be done directly in hardware, allowing for significant benefits as network sizes grow. The hardware chip The scalable chip design of 8 input ANN and 64 input ANN is performed successfully in Xilinx ISE 14.7. The Modelsim simulation is verified under different test cases and hardware parameters are extracted from the targeted device of Virtex-5 FPGA. The number of multipliers/adders for ANN-8, ANN-16, ANN-24, ANN-32, ANN-40, ANN-48, ANN-56 and ANN-64 are 8, 16, 24, 32, 40, 48, 56, and 64 respectively. The number of slices for ANN-8, ANN-16, ANN-24, ANN-32, ANN-40, ANN-48, ANN-56, and ANN-64 are 379, 726, 1073, 1420, 1766, 2113, 2460, and 2807 respectively. The number of LUTs for ANN-8, ANN-16, ANN-24, ANN-32, ANN-40, ANN-48, ANN-56, and ANN-64 are 648, 1280, 1928, 2560, 3208, 3840, 4488, and 5120 respectively. In the same way, the reported number of IOBs are 147, 275, 403, 531, 657, 787, 915, and 1043 for ANN-8 to ANN-64 respectively. The combinational path delay is 37.091 ns, common to all scalable modules. The hardware efficiency of the design is greater than 90.00% with the MSE = 0.00500 for ANN-64. The hardware usage summary concludes that the ANN chip hardware utilization is increasing with the ANN cluster size. The memory is also increasing from 116,736 kB to 165,892 kB. The chip hardware requirements will increase definitely with the number of neuron inputs. The biggest challenge for the hardware is to develop an embedded chip that can be compatible to support the specific hardware. The limitation of the work is that the chip design supports the 64 neurons processing ANN hardware and the chip functionality is verified in Virtex-5 FPGA. Therefore, the device resources utilization and timing parameters will change on another series of FPGA. The design can be extended further for large-scale ANN using pipelined and parallel processing that supports maximum hardware resources count and combinational blocks on the targeted FPGA. In the research work, we have followed the concept of scalable computing and modular design that can be used to support the design and development of the large-scale neuromorphic embedded chip. In the future, the research can be focused on the hardware chip design and synthesis for multilayer neural network architecture.",
"introduction": "Introduction Neurons are the cells in the nervous system that carry information to the other cells in the nerve and communicate with each other in distinctive ways. The neurons [ 10 ] are the elementary functioning unit in the brain. The nerve cell or the neurons communicate [ 47 ] with each other using a dedicated connection called synapses. The neurons are categorized into three types based on their functionality, which are sensory neurons, motor neurons, and interneurons. The sensory neurons [ 1 ] send the signals to the brain or the spinal cord. The sensory neurons are responsible for the response of the different stimuli of the human such as sound, light, or touch which is affected by the sensory organs by the cells. The motor neurons get the signals to form the brain and spinal cord [ 2 ] to derive the output based on muscle contractions to glandular output. The interneurons are connecting multiple neurons with the brain region or the spinal cord. The connections of these neurons form a circuit called a neural circuit. The neurons comprise a cell body called soma, dendrites, and an axon [ 21 ]. The soma is typically compact. The dendrites and axons are the filaments that extrude from them. The dendrites can extend freely from the soma maybe a hundred micrometers. The axon hillock is the swelling point at which the axon leaves soma, which can go for 1 m in human beings or larger in other species. The axon terminals pass [ 57 ] the signals to synapses and the other cells in the body. It may be that the neurons do not have axons or dendrites in the case of the undifferentiated cells. Typically, neurons are having a cell body, dendrites, and an axon. The cell body comprises the cytoplasm and the nucleus. The axon prolongs from the cell body and regularly provides growth to minor outlets or branches before termination at nerve points. Dendrites cover the neuron cell body and accept the signals from other neurons. The main contact points are synapses responsible for the communication among neurons, which may connect one dendrite to another dendrite, one axon to another axon. The dendrites [ 50 ] are covered with synapses formed by the ends of axons from other neurons. In general nature, the neurons are electrically excitable and maintain the voltage gradients within their membranes. Therefore, the signaling mechanism is electrical and partly chemical. The general-purpose hardware is based on the arithmetic blocks for simple in-memory calculations. Serial processing does not provide fast and sufficient performance for deep learning applications. The ANN architectures are based on parallel computation and operations. Ordinary chips cannot support a large number of highly and simultaneous operations for neuron processing. The AI-based hardware chip includes different chips that enable parallel processing. The main motivation for using the ANN and AI-based hardware accelerators is to get higher bandwidth memory chips and faster computation in comparison to general-purpose hardware. Digital tools and simulators are appropriate and applied for discovering the measurable behavior of neural networks. Silicon neuron systems [ 7 ] are a mix of analog and digital signals that may be used to analyze behavior using VLSI integrated circuits, and simulate electrophysiological behavior for actual neuron processing at various levels of abstraction. The most recent FPGAs can handle a huge number of physical memory and logic gates [ 22 ], allowing large-scale neural networks to be implemented on hardware and at a reasonable cost. The current level of simulation and synthesis technology is that research laboratories can easily afford FPGAs. The hardware synthesis method allows researchers to work on parallel brain cell structures. Digital models will be used for cell-based controls, and digital stem coding techniques will be used to facilitate communication across the medium across vast distances. Subsequently, it is well known that the neurons can be used to module ANN of the earlier generations by equating mean firing rates of processing neurons and hardware for proficient, scalable, and low-power implementations [ 6 ] of single-layer feed-forward networks. Human brain activity can be observed in both the local and delocal domains. The activities are linked to several functions such as vision and hearing, which are linked to specific brain regions. When a brain injury or accident occurs, the behavior of the brain neurons changes. The brain is a miniature network environment in which each portion has its own set of neural connections that are segregated from one another and confections. The local response is merged into a global understanding that causes the entire brain activity to become distressed. Machine learning and ANN-based intelligent methods have been proposed in the medical and health care industry to enhance security and train the models to improve patient treatment, diagnostics, rights, prevention, autonomy, and equality [ 55 ]. The research was offered based on deep learning-based Mobile Net V2 and long short-term memory (LSTM) to automate the process of identifying and classification [ 27 ] skin diseases. Oversampling techniques [ 32 ] can be used to determine cervical cancer based on feature extraction and spatial clustering. The synthetic minority over-sampling is used for hypertension, disease identification [ 31 ], and predictions based on the random forest machine learning method. The Wrapper filter [ 41 ] was used for disease classification and features selection. Neural networks have been applied for the CT images of the human liver for accurate diagnosis [ 56 ] of the disease related to the liver. ANNs have several advantages that make them ideal for solving specific scenarios and difficulties. ANN systems can learn and model non-linear functions as well as construct complicated associations, which is critical for real-world solutions and associates between non-linear and complex function inputs and outputs. The sense inputs and outputs cause the neural networks to alter or learn. ANN is a term that refers to several deep learning technologies that fall under the umbrella of artificial intelligence [ 18 ]. These technologies are mostly used in commercial applications to handle pattern recognition and sophisticated signal processing difficulties. For addressing nonlinear excitation functions, the development and realization of a single neural network require computing logic such as adders, multipliers, and a complex function evaluator [ 40 ]. The precision of the computational blocks is the most significant quality in the digital implementation of a single neural network [ 45 ]. It is acquired by determining their word length, which aids in the selection of a higher resolution. The fulfillment of the function necessitates appropriate mathematical matching, as the better resolution may result in higher system costs. As a result, implementing a single neural network in hardware will necessitate the multiplier, addition, and excitation function realization blocks [ 49 ]. The testing of the advanced neural networks and machine learning algorithms will require an advanced level of FPGA and simulation tools. The FPGA provides the platform in which high performance can be achieved using data processing blocks. The most powerful and mature neuro-chips are digital neural ASICs. High computational precision, great dependability, and high programmability are all advantages of digital technology. Furthermore, advanced design tools for digital full and semi-custom design are accessible. The weights of synaptic connections can be stored on or off the chip. The trade-off between speed and size determines this decision. The organization of the article is as follows: section 2 presents the related work, section 3 presents the structure of the single-layer neural network, and section 4 presents the design of the logarithmic multi-neuron system. The results & discussions are presented in section 5 , followed by conclusions in section 5 .",
"discussion": "Results and discussions The hardware chip of the 8-point ANN and 64-point ANN is designed using VHDL coding in Xilinx ISE 14.7. Figure 5 presents the register transfer level (RTL) block diagram for the 8-point to 64-point ANN chip. The RTL depicts all inputs and outputs of the designed chip.\n Fig. 5 RTL of ANN X 1 < 7:0 > to X 64 < 7:0 > presents the inputs (8-bit) of 64 neuron inputs ANN architecture with std_logic_vector data type. W 1 < 7:0 > to W 64 < 7:0 > presents the weight inputs (8-bit) corresponding to neuron inputs X 1 to X 64 of std_logic_vector data type. B_i < 15:0 > is the bias input treated as the perceptron of the ANN architecture of 16-bit with std_logic_vector data type. X_A < 15:0 > It is the activation function output ANN architecture with the 16-bit size of std_logic_vector data type. Y < 15:0 > It is the actual output with weighted sum and bias input, processed with an activation function of 16-bit of std_logic_vector data type. Modelsim simulation of 8 input ANN in binary and integer is shown in Figs. 6 and 7 respectively. Table 3 lists the test cases used for the functional simulation of the designed ANN chip. Modelsim simulation of 64 input ANN in binary and integer is shown in Figs. 8 and 9 . Table 4 lists the test cases used for the functional simulation of the designed ANN-64 with test case-1 to test case-8.\n Fig. 6 Modelsim simulation of 8 input ANN in binary Fig. 7 Modelsim simulation of 8 input ANN in integer Table 3 Test cases for the simulation waveform Pins Detail Test case-1 Test case-2 Test case-3 Integer Binary Integer Binary Integer Binary X1 < 7:0> Input 2 00000010 1 00000001 4 00000100 X2 < 7:0> Input 3 00000011 2 00000010 4 00000100 X3 < 7:0> Input 4 00000100 3 00000011 6 00000110 X4 < 7:0> Input 5 00000101 3 00000011 8 00001000 X5 < 7:0> Input 6 00000110 4 00000100 3 00000011 X6 < 7:0> Input 7 00000111 4 00000100 5 00000100 X7 < 7:0> Input 8 00001000 5 00000101 7 00000111 X8 < 7:0> Input 9 00001001 5 00000101 9 00001001 W1 < 7:0> Input 1 00000001 3 00000011 2 00000010 W2 < 7:0> Input 2 00000010 1 00000001 2 00000010 W3 < 7:0> Input 3 00000011 2 00000010 3 00000011 W4 < 7:0> Input 2 00000010 4 00000100 3 00000011 W5 < 7:0> Input 2 00000010 5 00000101 4 00000100 W6 < 7:0> Input 1 00000001 4 00000100 5 00000101 W7 < 7:0> Input 1 00000001 3 00000011 5 00000101 W8 < 7:0> Input 2 00000010 1 00000001 4 00000100 B_i < 15:0> Input 150 0000000010010110 250 0000000011111010 300 0000000100101100 Y < 15:0> output 225 0000000011100001 329 0000000101001001 466 0000000111010010 Fig. 8 Modelsim simulation of 64 input ANN in binary and integer (inputs) Fig. 9 Modelsim simulation of 64 input ANN in binary (weights and outputs) Table 4 Test cases for the simulation waveform ANN-64 point Pin Direction Binary Integer Pin Direction Binary Integer Test Case-1 X 1 < 7:0> Input 00000001 1 W 1 < 7:0> Input 00001000 8 X 2 < 7:0> Input 00000010 2 W 2 < 7:0> Input 00001000 8 X 3 < 7:0> Input 00000011 3 W 3 < 7:0> Input 00001000 8 X 4 < 7:0> Input 00000100 4 W 4 < 7:0> Input 00001000 8 X 5 < 7:0> Input 00000101 5 W 5 < 7:0> Input 00001000 8 X 6 < 7:0> Input 00000110 6 W 6 < 7:0> Input 00001000 8 X 7 < 7:0> Input 0000011 7 W 7 < 7:0> Input 00001000 8 X 8 < 7:0> Input 00001000 8 W 8 < 7:0> Input 00001000 8 Sel < 2:0> Input 000 b_i 1 < 15:0> Input 0000000001111000 120 Y 1 < 15:0> output 0000001100110000 408 Test Case-2 X 9 < 7:0> Input 00001001 9 W 9 < 7:0> Input 00000111 7 X 10 < 7:0> Input 00001010 10 W 10 < 7:0> Input 00000111 7 X 11 < 7:0> Input 00001011 11 W 11 < 7:0> Input 00000111 7 X 12 < 7:0> Input 00001100 12 W 12 < 7:0> Input 00000111 7 X 13 < 7:0> Input 00001101 13 W 13 < 7:0> Input 00000111 7 X 14 < 7:0> Input 00001110 14 W 14 < 7:0> Input 00000111 7 X 15 < 7:0> Input 00001111 15 W 15 < 7:0> Input 00000111 7 X 16 < 7:0> Input 00010000 16 W 16 < 7:0> Input 00000111 7 Sel < 2:0> Input 001 b_i 2 < 15:0> Input 0000000000100010 130 Y 2 < 15:0> output 0000001100111110 830 Test Case-3 X 17 < 7:0> Input 00010001 17 W 17 < 7:0> Input 00000110 6 X 18 < 7:0> Input 00010010 18 W 18 < 7:0> Input 00000110 6 X 19 < 7:0> Input 00010011 19 W 19 < 7:0> Input 00000110 6 X 20 < 7:0> Input 00010100 20 W 20 < 7:0> Input 00000110 6 X 21 < 7:0> Input 00010101 21 W 21 < 7:0> Input 00000110 6 X 22 < 7:0> Input 00010110 22 W 22 < 7:0> Input 00000110 6 X 23 < 7:0> Input 00010111 23 W 23 < 7:0> Input 00000110 6 X 24 < 7:0> Input 00011000 24 W 24 < 7:0> Input 00000110 6 Sel < 2:0> Input 010 b_i 3 < 15:0> Input 0000000010001100 140 Y 3 < 15:0> output 0000010001100100 1124 Test Case-4 X 25 < 7:0> Input 00011001 25 W 25 < 7:0> Input 00000101 5 X 26 < 7:0> Input 00011010 26 W 26 < 7:0> Input 00000101 5 X 27 < 7:0> Input 00011011 27 W 27 < 7:0> Input 00000101 5 X 28 < 7:0> Input 00011100 28 W 28 < 7:0> Input 00000101 5 X 29 < 7:0> Input 00011101 29 W 29 < 7:0> Input 00000101 5 X 30 < 7:0> Input 00011110 30 W 30 < 7:0> Input 00000101 5 X 31 < 7:0> Input 00011111 31 W 31 < 7:0> Input 00000101 5 X 32 < 7:0> Input 00100000 32 W 32 < 7:0> Input 00000101 5 Sel <2:0> Input 011 b_i 4 < 15:0> Input 0000000010010110 150 Y 4 < 15:0> output 0000010100001010 1290 Test Case-5 X 33 < 7:0> Input 00100001 33 W 33 < 7:0> Input 00000100 4 X 34 < 7:0> Input 00100010 34 W 34 < 7:0> Input 00000100 4 X 35 < 7:0> Input 00100011 35 W 35 < 7:0> Input 00000100 4 X 36 < 7:0> Input 00100100 36 W 36 < 7:0> Input 00000100 4 X 37 < 7:0> Input 00100101 37 W 37 < 7:0> Input 00000100 4 X 38 < 7:0> Input 00100110 38 W 38 < 7:0> Input 00000100 4 X 39 < 7:0> Input 00100111 39 W 39 < 7:0> Input 00000100 4 X 40 < 7:0> Input 00101000 40 W 40 < 7:0> Input 00000100 4 Sel <2:0> Input 100 b_i 5 < 15:0> Input 0000000010100000 160 Y 5 < 15:0> output 0000010100110000 1328 Test Case-6 X 41 < 7:0> Input 00101001 41 W 41 < 7:0> Input 00000011 3 X 42 < 7:0> Input 00101010 42 W 42 < 7:0> Input 00000011 3 X 43 < 7:0> Input 00101011 43 W 43 < 7:0> Input 00000011 3 X 44 < 7:0> Input 00101100 44 W 44 < 7:0> Input 00000011 3 X 45 < 7:0> Input 00101101 45 W 45 < 7:0> Input 00000011 3 X 46 < 7:0> Input 00101110 46 W 46 < 7:0> Input 00000011 3 X 47 < 7:0> Input 00101111 47 W 47 < 7:0> Input 00000011 3 X 48 < 7:0> Input 00110000 48 W 48 < 7:0> Input 00000011 3 Sel <2:0> Input 101 b_i 6 < 15:0> Input 0000000010101010 170 Y 6 < 15:0> output 0000010011010110 1238 Test Case-7 X 49 < 7:0> Input 00110001 W 49 < 7:0> Input 00000010 2 X 50 < 7:0> Input 00110010 W 50 < 7:0> Input 00000010 2 X 51 < 7:0> Input 00110011 W 51 < 7:0> Input 00000010 2 X 52 < 7:0> Input 00110100 W 52 < 7:0> Input 00000010 2 X 53 < 7:0> Input 00110101 W 53 < 7:0> Input 00000010 2 X 54 < 7:0> Input 00110110 W 54 < 7:0> Input 00000010 2 X 55 < 7:0> Input 00110111 W 55 < 7:0> Input 00000010 2 X 56 < 7:0> Input 00111000 W 56 < 7:0> Input 00000010 2 Sel <2:0> Input 110 b_i 7 < 15:0> Input 0000000010110100 180 Y 6 < 15:0> output 0000001111111100 1020 Test Case-8 X 57 < 7:0> Input 00111001 W 57 < 7:0> Input 00000001 1 X 58 < 7:0> Input 00111010 W 58 < 7:0> Input 00000001 1 X 59 < 7:0> Input 00111011 W 59 < 7:0> Input 00000001 1 X 60 < 7:0> Input 00111100 W 60 < 7:0> Input 00000001 1 X 61 < 7:0> Input 00111101 W 61 < 7:0> Input 00000001 1 X 62 < 7:0> Input 00111110 W 62 < 7:0> Input 00000001 1 X 63 < 7:0> Input 00111111 W 63 < 7:0> Input 00000001 1 X 64 < 7:0> Input 01000000 W 64 < 7:0> Input 00000001 1 Sel <2:0> Input 111 b_i 8 < 15:0> Input 0000000010111110 190 Y 7 < 15:0> output 0000001010100010 674 The percentage of hardware that is used by the device is given by the device utilization report [ 37 ] for the implementation of the chip. The report is taken directly from the Xilinx software as the device utilization report. The report presents the number of adders, multipliers, slices, 4 input lookup tables (LUT) [ 36 ], input/output blocks (IOB), total memory usage (kB), combinational delay (ns) that includes path delay and routing delay. The Xilinx device summary for ANN-8, ANN-16, ANN-24, ANN-32, ANN-40 ANN-48, ANN-66, and ANN-64 is given in Table 5 . The target device is Virtex-5 FPGA with device xc5vlx20t-2-ff323 used for simulation and synthesis [ 24 ]. Figure 10 presents the hardware utilization curve for ANN-8 to ANN-64 hardware chips.\n Table 5 Xilinx software parameters for ANN-8 point to ANN-64 point Size/Parameters Multipliers 16- bit Adders Slices LUTs IOB Delay(ns) Memory (kB) ANN-8 8 8 379 648 147 37.091 116,736 ANN-16 16 16 726 1280 275 37.091 124,480 ANN-24 24 24 1073 1928 403 37.091 130,048 ANN-32 32 32 1420 2560 531 37.091 137,280 ANN-40 40 40 1766 3208 659 37.091 143,424 ANN-48 48 48 2113 3840 787 37.091 151,556 ANN-56 56 56 2460 4488 915 37.091 158,724 ANN-64 64 64 2807 5120 1043 37.091 165,892 Fig. 10 Hardware utilization for ANN-8 to ANN-64 hardware chip In the simulation of ANN-64, the hardware and memory usage depends on the utilizations of multipliers and adders. The detail of these units is reported directly by the software and change with the number of neurons and weight inputs. The hardware utilization will increase with the increase in cluster inputs of the ANN chip. The simulation results show that the number of multipliers, adders, slices, LUTs, memory is increasing as the number of neurons are increasing in the multi-input ANN design. The reason for this is that the adders and multipliers blocks increase the number of gates and concurrent logic modules, which takes up more memory and resources on the FPGA. The report predicts that the number of multipliers and 16-bit adders are increasing with the number of neurons inputs. The predicting of mean squared error (MSE), mean absolute percentage error (MAPE), root mean squared error (RMSE) is done for the FPGA hardware resources [ 25 , 42 ] based on the training and validation sample neurons with different cluster inputs of ANN design. In the training (X 1 to X 40 ) are considered and (X 41 to X 64 ) for validation. The values are determined using the equations [ 19 , 20 ].\n 9 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ MSE=\\frac{1}{n}\\sum \\limits_{i=1}^n{\\left|{y}_i-\\hat{y_i}\\right|}^2 $$\\end{document} MSE = 1 n ∑ i = 1 n y i − y i ^ 2 10 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ RMSE=\\sqrt{\\frac{1}{n}\\sum \\limits_{i=1}^n{\\left|{y}_i-\\hat{y_i}\\right|}^2} $$\\end{document} RMSE = 1 n ∑ i = 1 n y i − y i ^ 2 11 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ MAPE=\\frac{1}{n}\\sum \\limits_{i=1}^n\\frac{\\left|{y}_i-\\hat{y_i}\\right|}{y_i}.100\\% $$\\end{document} MAPE = 1 n ∑ i = 1 n y i − y i ^ y i . 100 % y i is the actual value and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ \\hat{y_i} $$\\end{document} y i ^ is the predicted value for ‘n’ number of predications. Based on linear regression model and 200 estimations for 64 number of neurons the value of MSE = 0.00500, RMSE = 0.07071, and MAPE = − 0.003906%. The efficiency of the hardware simulation depends on the resources utilization such as logic gates, input/output block, combinational logic, memory, and delay. For the complex nonlinear application, multilayer perceptron architecture is beneficial in comparison to single-layer multiple input ANN. On the other hand, it is simple to build up and train a single layer perceptron. The neural network model can be explicitly linked to statistical models, allowing it to share the covariance Gaussian density function. The realization of the MLP will provide more delay in comparison to single-layer multiple input ANN. Figure 11 shows the hardware efficiency with targeted FPGA- Virtex-5 for simulation and synthesis of the binary data. The efficiency variations are noticed with the different test cases in which 8 neurons are processed at a time and parallel processing modular design-based approach is followed to realize the 64 input ANN. The single-layer ANN hardware is used to solve simple problems and parallel processing provides fast computation time. In terms of hardware efficiency, the single-layer will provide faster response and computation time in comparison to MLP. The MLP requires more delay to compute the logic as it is processed by different hidden layers. The output function of the ANN hardware chip is the throughput that depends on the length of binary input, weights, bias input, and hardware and timing parameters. The output layer receives the inputs from the layers above it, executes the calculations using its neurons, and then computes the output.\n Fig. 11 Hardware efficiency with targeted FPGA The hardware delay depends on two components of propagation delay are logic delay and routing delay. The logic delay is a function of the number and kind of logic gates the signal passes through. Because the FPGA compiler tries to cluster the components of a combinatorial path as tightly as possible on the FPGA. The routing delay is a function of the length of the wire path the signal travels, which is often modest. In the simulation, the total path delay is 37.091 ns, in which 89.00% delay is from logic and 11.00% from routing that help to maintain the FPGA efficiency greater than 90.00% in most cases."
} | 6,608 |
30439494 | null | s2 | 913 | {
"abstract": "There is great interest in engineering photoautotrophic metabolism to generate bioproducts of societal importance. Despite the success in employing genome-scale modeling coupled with flux balance analysis to engineer heterotrophic metabolism, the lack of proper constraints necessary to generate biologically realistic predictions has hindered broad application of this methodology to phototrophic metabolism. Here we describe a methodology for constraining genome-scale models of photoautotrophy in the cyanobacteria Synechococcus elongatus PCC 7942. Experimental photophysiology parameters coupled to genome-scale flux balance analysis resulted in accurate predictions of growth rates and metabolic reaction fluxes at low and high light conditions. Additionally, by constraining photon uptake fluxes, we characterized the metabolic cost of excess excitation energy. The predicted energy fluxes were consistent with known light-adapted phenotypes in cyanobacteria. Finally, we leveraged the modeling framework to characterize existing photoautotrophic and photomixtotrophic engineering strategies for 2,3-butanediol production in S. elongatus. This methodology, applicable to genome-scale modeling of all phototrophic microorganisms, can facilitate the use of flux balance analysis in the engineering of light-driven metabolism."
} | 332 |
27345370 | PMC4930481 | pmc | 914 | {
"abstract": "Summary Can a heterotrophic organism be evolved to synthesize biomass from CO 2 directly? So far, non-native carbon fixation in which biomass precursors are synthesized solely from CO 2 has remained an elusive grand challenge. Here, we demonstrate how a combination of rational metabolic rewiring, recombinant expression, and laboratory evolution has led to the biosynthesis of sugars and other major biomass constituents by a fully functional Calvin-Benson-Bassham (CBB) cycle in E. coli . In the evolved bacteria, carbon fixation is performed via a non-native CBB cycle, while reducing power and energy are obtained by oxidizing a supplied organic compound (e.g., pyruvate). Genome sequencing reveals that mutations in flux branchpoints, connecting the non-native CBB cycle to biosynthetic pathways, are essential for this phenotype. The successful evolution of a non-native carbon fixation pathway, though not yet resulting in net carbon gain, strikingly demonstrates the capacity for rapid trophic-mode evolution of metabolism applicable to biotechnology. PaperClip",
"introduction": "Introduction Whether CO 2 can or cannot be transformed into sugar and biomass by carbon fixation is arguably the most basic distinction we make in defining the metabolism of an organism. Carbon fixation by autotrophs is the biochemical gateway to the organic world, as obligate heterotrophs are dependent on this supply of organic carbon. How difficult is it to evolve from one trophic mode of growth to another? Specifically, can the ability to synthesize biomass from CO 2 be introduced into a heterotrophic organism? Exploring the process in which a heterotrophic bacterium, such as Escherichia coli , is evolved to synthesize sugars from CO 2 can serve as a model system to tackle these questions. The utilization of one-carbon compounds, such as CO 2 , methanol, and methane, has been recently drawing attention as a low-cost, abundant feedstock option for biochemical production ( Li et al., 2012 , Müller et al., 2015 , Siegel et al., 2015 ). Recent studies have demonstrated that a wide variety of non-native metabolic pathways can be integrated into model microorganisms, allowing the synthesis of value-added chemicals using sugar as a feedstock ( Galanie et al., 2015 , Yim et al., 2011 ). However, efforts to synthesize sugar from inorganic CO 2 by introducing a non-native carbon fixation cycle have never been successful. Carbohydrate biosynthesis through carbon fixation in E. coli would not only open exciting avenues to directly utilize CO 2 for chemical production, but could also serve as a platform to rapidly optimize carbon fixation enzymes and pathways for subsequent implementation in agricultural crops ( Lin et al., 2014 , Mueller-Cajar and Whitney, 2008 , Parikh et al., 2006 , Shih et al., 2014 ). Furthermore, this experimental approach could shed light on cellular adaptations associated with horizontal-gene-transfer events, on the plasticity of metabolic networks, and on the evolutionary emergence of a biological novelty. The achievement of novel biological phenotypes on laboratory timescales is at the heart of efforts such as the long-term evolutionary experiment, which studied how E. coli developed the ability to utilize citrate throughout several tens of thousands of generations ( Blount et al., 2008 , Maddamsetti et al., 2015 , Wiser et al., 2013 ). These studies show the intricate dynamics of potentiating, actualizing, and refining steps during the evolutionary process ( Quandt et al., 2015 ) and shed light on the interplay between selection, historical contingency, and epistatic effects. In parallel to lab evolution, which forces the selective conditions, synthetic biology efforts manipulate the genetic makeup with the aim of rationally designing desired phenotypes ( Church et al., 2014 , Galanie et al., 2015 ). In spite of significant progress, nothing as extreme as expressing a fully functional pathway that changes the trophic mode of an organism has been ever shown to be achievable. The Calvin-Benson-Bassham cycle ( Bassham et al., 1954 ) is, by far, the most dominant carbon fixation pathway in the biosphere out of all six known natural alternatives ( Fuchs, 2011 , Bar-Even et al., 2012 ). Previous efforts to show functional expression of the CBB cycle carboxylating enzyme, ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), in E. coli have relied on the supply of a glycolytic carbon source (e.g., glucose) to replenish RuBisCO’s substrate, ribulose-bisphosphate (RuBP) ( Durão et al., 2015 , Gong et al., 2015 , Mueller-Cajar and Whitney, 2008 , Parikh et al., 2006 , Zhuang and Li, 2013 ). These important studies therefore did not achieve the defining function of the CBB cycle of autocatalytic sugar synthesis from inorganic carbon. While this approach led to the elegant construction of a RuBisCO-dependent E. coli as a platform for directed evolution of RuBisCO activity ( Durão et al., 2015 , Mueller-Cajar and Whitney, 2008 , Parikh et al., 2006 ), evolving a complete carbon fixation cycle capable of autocatalytically generating biomass from CO 2 has remained an open challenge. Here, we report achieving a fully functional and autocatalytic carbon fixation cycle in E. coli , capable of hexose, pentose and triose sugar synthesis with no input of organic carbon into the cycle. Energy and reducing power are supplied by the oxidation of an organic acid (pyruvate) in an isolated metabolic module and thus no net carbon gain is achieved at this point. This is the first example of a non-native carbon assimilation cycle in which all of the pathway intermediates and products are solely synthesized from CO 2 and the required co-factors.",
"discussion": "Discussion While several synthetic metabolic pathways have been successfully integrated into model microorganisms, such as E. coli and S. cerevisiae , the introduction of an autocatalytic carbon fixation pathway into a heterotrophic host has remained a standing challenge ( Müller et al., 2015 , Siegel et al., 2015 ). In this study, we demonstrate that the combination of rational design and laboratory evolution can rapidly evolve a strain to convert CO 2 into sugars and other biomass components via non-native carbon fixation machinery. The striking ability of E. coli to rapidly change its growth mode demonstrates that central carbon metabolism can be extremely adaptive when subjected to selective conditions and extends the possibilities for achieving novel metabolic phenotypes. Successful introduction of carbon fixation into a heterotrophic model organism suggests that closely related radical modulations of C1 central carbon metabolism, such as the desired synthetic methylotrophy ( Whitaker et al., 2015 ), synthetic mixotrophy ( Fast et al., 2015 ), and synthetic electrotrophy ( Lovley, 2011 ), can be well within reach with a similar approach. From an industrial perspective, even though not translated to immediate industrial applications, a fully functional non-native CO 2 fixation cycle is a critical milestone toward synthetic lithoautotrophy. Our strategy to decouple energy and carbon fixation metabolism enables powerful modularity in the choice of energy module used to energize the synthetic cycle. The methodology we present of selection-based evolutionary transitions can be directly extended to establish a synthetic methylotrophic strain, utilizing industrially relevant energy sources, such as methanol, to drive CO 2 fixation. Furthermore, previous studies described the “paper biochemistry” of several fully synthetic carbon fixation cycles, not known to occur in nature, with promising predicted kinetic properties ( Bar-Even et al., 2010 ). Our results suggest an experimental route to implement these pathways in a fast growing, genetically malleable model organism, such as E. coli. In contrast to a traditional design scheme, in which individual components or sub-modules are tested and then integrated, our metabolic design focused on the coupling of cellular fitness to the desired functionality. While the initial rational design, arising from our in silico analysis, was not sufficient to achieve the desired phenotype, it formed the genetic basis and suggested the selective conditions under which a complex metabolic transition evolved. The ability to create synergy between rational metabolic design and laboratory evolution by tailoring the appropriate experimental setup is essential for harnessing natural selection to fine-tuning the metabolic network. Quantitative a priori prediction of the interactions between endogenous and newly introduced metabolic functions is often challenging due to the lack of sufficient information regarding the kinetics, energetics, and regulation of the relevant components. A design that couples cellular fitness to the activity of a non-native pathway is therefore useful. Such coupling allows the harnessing of natural selection in a controlled environment to obtain the necessary fine-tuning between native and non-native metabolic functions to achieve the desired phenotype. Since chemostats inherently implement a feedback mechanism that continuously increases the selection stringency in response to improvements in the desired pathway activity, they can serve as an ideal platform for evolving pathways that utilize non-native substrates, such as inorganic carbon. A dynamic selection stringency, as in the case of chemostats, enables growth during the early phases of laboratory evolution by supplying limiting amounts of a surrogate sugar (i.e., xylose), which compensates for the lack of full pathway activity, even if the carbon fixation pathway performs suboptimally ( Kleeb et al., 2007 ). The hybrid approach demonstrated here of rational design combined with laboratory evolution has allowed us, for the first time, to achieve a fully functional carbon fixation cycle in a heterologous host. Our results make evident the remarkable plasticity of metabolism and provide a malleable platform for deciphering the biochemistry and evolution of carbon fixation. The rapid emergence of a novel metabolic phenotype in laboratory timescales suggests a route for synthetic biology efforts to optimize pathways of biotechnological importance that can drive transformative advances in our ability to tackle the grand challenge of resource sustainability in years to come."
} | 2,599 |
36909754 | PMC9997075 | pmc | 916 | {
"abstract": "A hybrid piezo/triboelectric nanogenerator (H/P-TENG) is designed for mechanical energy harvesting using polymer ceramic composite films; polydimethylsiloxane/Ba(Zr 0.2 Ti 0.8 )O 3 –0.5(Ba 0.7 Ca 0.3 )TiO 3 (PDMS/BZT–BCT) and polyvinyl alcohol (PVA). A lead-free BZT–BCT piezoelectric ceramic was prepared via solid-state method and blended into PDMS to form a series of polymer-ceramic composite films, ranging from 5% to 30% by weight. The films were forward/reverse poled with corona poling and their electrical properties were compared to non-poled samples. The H/P-TENG constructed with forward-poled 15 wt% BZT–BCT in PDMS achieved the highest open-circuit voltage, V oc of 127 V, short-circuit current density, J sc of 67 mA m −2 , short-circuit charge density, Q sc of 118 μC m −2 , and peak power density of 7.5 W m −2 , an increase of 190% over pristine PDMS-based TENG. It was discovered that incorporating BZT–BCT into the PDMS matrix improved the triboelectric properties of PDMS. The overlapping electron cloud (OEC) model was used to explain the enhancement and the effect of poling direction of the PDMS/BZT–BCT composite used in H/P-TENG, providing fundamental knowledge of the influence of piezoelectric polarisation on contact electrification.",
"conclusion": "Conclusion We have shown that the inclusion of BZT–BCT piezoelectric ceramic is able to enhance the electrical output of a PDMS-based TENG. With the optimal inclusion of BZT–BCT at 15 wt%, the H/P-TENG is able to achieve open-circuit voltage, V oc , short-circuit current density, J sc , and short-circuit charge density, Q sc of 113.2 V, 61.1 mA m −2 , and 100.7 μC m −2 respectively. The increase in the dielectric constant of the polymer composite films is one of the reasons for the enhancement of the electrical output of the H/P-TENG. In addition, we have shown that the forward-poling direction is optimal for the electrical output of H/P-TENG's where the forward-poled 15 wt% polymer composite recorded the highest V oc (127.1 V), J sc (66.6 mA m −2 ), Q sc (117.5 μC m −2 ) and a maximum power density of 7.5 W m −2 , which is equivalent to ∼190% enhancement when compared to the PDMS-based TENG. The overlapping electron cloud (OEC) model is proposed to explain this phenomenon. Our study shows that the inclusion of piezoelectric BZT–BCT is a viable method of enhancing the electrical output of a TENG.",
"introduction": "Introduction In the last decade, the internet of things (IoT) has given rise to a significant increase in the use of smart devices and sensors. 1–3 Understandably, the need for a clean, safe power source for these devices and sensors has risen as well. 4,5 Mechanical energy, being one of the most widely available energies in the environment, provides a more viable means to produce self-powered devices. In that context, triboelectric nanogenerators (TENG) stand out as a promising option among the various conceivable alternatives. 6–8 TENGs transform mechanical energy into electrical energy by utilising the triboelectric effect. 9,10 Contact electrification and electrostatic induction play an important role in the operation of the TENG, 11,12 therefore any modifications and improvements to the TENG would revolve around these two phenomena. Despite its high open-circuit voltage output, TENG has a high internal resistance and a low short-circuit current output, resulting in a poorer and unstable power density. 13–15 On the other hand, piezoelectric nanogenerators (PENG) are able to convert mechanical energy into electrical energy via the piezoelectric effect. 16–18 In comparison to TENG, PENG typically has higher short-circuit current but lower open-circuit voltage. TENG and PENG are both capable of converting irregular, low frequency, and distributed mechanical energy into electrical energy. Combining the benefits of PENG and TENG, a hybrid piezoelectric–triboelectric nanogenerator (H/P-TENG) can be expected to create impressive open-circuit voltage and short-circuit current at the same time, thus it is worth exploring. 19 In recent years, several H/P-TENGs have been produced and demonstrated to have promising features. 20 The type of materials utilised for the triboelectric layers and the number of terminals are important aspects to consider when developing H/P-TENG. 21–25 PDMS is commonly used in TENG systems, as it is a flexible, inexpensive material that has high tribo-negativity. 26,27 Its tendency to attract electrons when in contact with a tribo-positive surface makes it very desirable as TENGs. Conversely, PVA film has high tribo-positivity, allowing it to lose electrons when coming into contact with tribo-negative materials. 28,29 This is a perfect combination of materials for a TENG device, but the energy output is still minimal. Prior research are centered on merging TENG and PENG in either series or parallel modes, or making use of organic piezoelectric materials such as PVDF and its copolymer as the tribo- and piezo-materials simultaneously. 28 When compared to organic counterparts, the piezoelectric characteristics of inorganic piezoelectric materials are significantly superior. The drawback is that there is a restricted amount of flexibility. One further possibility for getting over the aforementioned restriction is to use interconnected ferroelectric films i.e. , lead zirconate titanate (PZT) 30 or BiFeO 3 (ref. 31 ) on flexible glass fibre fabric, ZnO nanorod/PVDF-PTFE, 32,33 ZnO nanowire/parylene C, 34 BTO/PDMS, 35,36 PDMS/carbon coated BT 37 and etc. Ba(Zr 0.2 Ti 0.8 )O 3 –0.5(Ba 0.7 Ca 0.3 )TiO 3 (BZT–BCT) is a lead-free binary piezoelectric ceramic which has gained popularity due to its unique characteristics such as high dielectric constant ∼2400, high piezoelectric constant ∼471 pC N −1 and high polarizability ∼103 mC m −2 . 38,39 BZT–BCT has been utilised as the piezoelectric material in numerous PENG devices, 40,41 but it has not been utilised in any notable TENG-based systems. In this study, we mixed BZT–BCT ceramic powder into the PDMS matrix to form a PDMS/BZT–BCT composite film. The high piezoelectric and dielectric properties of the BZT–BCT ceramic powder is hypothesized to enhance the charge concentration in PDMS due to synergistic effect of piezoelectricity and triboelectricity, thus enhancing the overall electrical output of the H/P-TENG. We studied the wt% and poling direction of the ceramic filler to determine the optimized polymer ceramic composite film that leads to the maximum enhancement to the power density output of the H/P-TENG. By means of its synergistic design, this hybrid generator may compensate for the shortcomings of each transducing mechanism, allowing it to be employed as energy supply units in a wide range of applications.",
"discussion": "Discussion Herein, an overlapping electron cloud (OEC) model is adopted to explain the formation of charges during the contact electrification in H/P-TENG. 28,44,50 This model is able to illustrate how charges are formed within the H/P-TENG during contact electrification and how factors like piezoelectric inclusion and the poling orientation of the PDMS/BZT–BCT polymer composite film affect the charge generation. Fig. 6 shows the OEC model for the H/P-TENG where the PVA film, PDMS film and PDMS/BZT–BCT polymer composite film are depicted as molecules with their respective potential wells. Δ E is used to show the difference in the highest occupied energy level of the electrons in both triboelectric layers, and the dashed lines depict the vacuum energy level. Fig. 6 Depiction of overlapping electron cloud (OEC) model for PDMS/BZT–BCT polymer composite-based H/P-TENGs under the (a) non-poled, (b) forward-poled, and (c) reverse-poled conditions, where the blue sphere is the electron cloud of the tribo-positive PVA molecule, and the red sphere is the electron cloud of the tribo-negative PDMS/BZT–BCT molecule. First, we look at the non-poled H/P-TENG operation. When the two triboelectric layers come into close proximity during the contact phase of the H/P-TENG, the electron clouds of the PVA film and PDMS/BZT–BCT film are able to overlap. When this happens, the potential wells of both triboelectric layers merge, forming a double-well potential as shown in Fig. 6(a) . Due to the higher occupied energy levels of the electrons in PVA, this forms an asymmetric double-well, which provide an energy gradient to drive the electrons from the tribo-positive PVA to the tribo-negative PDMS/BZT–BCT energy well. As a result, the PVA film has net positive charges on its surface, whereas the PDMS film has net negative charges. As the polymer composites are forward-poled, the energy levels shifted downwards, the Δ E increased and subsequently increases the energy gradient of the merged potential well. With an even higher energy gradient, more electrons are driven from the PVA film towards the forward-poled PDMS/BZT–BCT film. This is reflected in the increase of electrical output of the H/P-TENG when the polymer composite films are forward-poled. Under the reverse-poled condition, we predict the energy levels of the electrons are shifted upwards. This has the opposite effect, lowering the Δ E between the PVA film and reverse-poled PDMS/BZT–BCT film. With a narrower Δ E , the energy gradient is also relatively lower, which results in fewer electrons being driven from the PVA film to the PDMS/BZT–BCT film. We charged several capacitors ranging from 1 μF to 330 μF with our H/P-TENG. The voltage storage versus capacitor charging time for the various capacitors that are measured is depicted with solid lines in Fig. 7(a) . When compared to the ranges of capacitors, the 1 μF and 2.2 μF capacitors had a faster charging rate. The 1 μF capacitor is able to charge up to 7 V whereas 2.2 μF capacitor charge up to 6 V within 3 minutes. The total stored energy and charging rates drop significantly when connected to the 6.8 μF capacitor, and even further when the capacitance is ≥10 μF. The voltage stored in the capacitor V C is calculated using the eqn (2) shown below: 12 2 where V S is the voltage supplied by the H/P-TENG, t is the charging time, R is the total resistance and C is the capacitance in the circuit. Since the H/P-TENG has a high internal resistance, it may be assumed that R is constant throughout all capacitors. Thus, as the capacitance increased the stored voltage decreased over time. This is consistent with the findings in Fig. 7(a) , where the higher capacitance exhibits slower charging rate and stored voltage after 180 s. The observed results also fit well with the fitted data which are calculated using eqn (2) (shown as dashed lines in Fig. 7(a) ). The H/P-TENG is utilised to illuminate commercial LEDs. Fig. 7(b) is an image showing fifty blue commercial LEDs connected in series lit up to their maximum brightness. In ESI, † we have also included a video demonstrating how the H/P-TENG is used to illuminate LEDs for 10 seconds at varying frequencies. Table 2 compares our H/P-maximum TENG's output power density to that of other contemporary polymer composite-based H/P-TENGs. The current H/P-TENG has demonstrated relatively higher power outputs than the recent similar PMMA-based H/P-TENG, indicating the potential for micro/nano power sources in self-powered systems. Fig. 7 (a) Measured (solid lines) and theoretical (dashed lines) charging rate of various capacitors by H/P-TENG constructed with polymer composite film of 15 wt% BZT–BCT inclusion and, (b) H/P-TENG constructed with polymer composite film of 15 wt% BZT–BCT inclusion lighting up 50 commercial blue LEDs. Comparison of hybrid piezo/triboelectric nanogenerators power density output from recent years Tribo-positive material Tribo-negative material Power density (W m −2 ) Ref. PVA PDMS/BZT–BCT 7.5 This work PTFE ZnO/PVDF 0.3 \n 32 \n BCZT/PVDF-HFP Silicon rubber 0.2 \n 51 \n Stainless steel fabric PDMS/PVDF-HFP 3 \n 52 \n PET PDMS/BTO 0.4 \n 53 \n Silk nanofibers BFO-GFF/PDMS 3 \n 54 \n Copper BZTO:PDMS 4 \n 55 \n Aluminium PDMS/BiTO 0.2 \n 56 \n Polypyrrole electrodeposited GFP PDMS/BTO-GFP 4 \n 57 \n Paper ZnO/PDMS 6 \n 43 \n Chitosan/BT PTFE 8 \n 58 \n PTFE C-PS/P(VDF-TrFE) 8 \n 59 \n PVDF/BTO Natural rubber 0.4 \n 60 \n PVA/ZnR Silicon rubber 15 \n 44"
} | 3,066 |
34195439 | PMC8233142 | pmc | 918 | {
"abstract": "The columbic efficiency, removal efficiency and voltage production of seven different combinations of carbon (acetic acid, albumin and sucrose) with nutrients (C:N, C:P, C:S, C:N:S, C:P:S, C:N:P and C: N:S:P) were investigated at three different ratios (20:1, 15:1 and 10:1). The effects of various pH values were also explored for these combinations of carbon, and sulfur compounds (pH 6–8). The highest columbic efficiency (75.8%), COD removal efficiency (86%) and voltage (667 mV) were recorded when the acetic acid was used in the MFC and the lowest columbic efficiency (12.8%), removal efficiency (37.6%) and voltage (145 mV) were observed in case of albumin. A marked increase in columbic efficiency, removal efficiency and voltage production were seen with the rise in the pH value from 6 to 8. The lowest columbic efficiency, removal efficiency and voltage production were seen at pH 6 and highest at pH 8. At each investigated pH, the highest removal efficiency, columbic efficiency, and voltage were found at substrate ratio of 20:1 while lower at 10:1. At all pH values, the carbon to nutrient ratios seemed to have followed a similar trend i.e., the COD removal efficiency, columbic efficiency and voltage generation was found in the order C:N > C:N:S > C:N:S:P > C:N:P > C:S > C:P:S > C:P. In all cases, nitrogen showed a higher removal as compared to phosphorous and sulfur.",
"conclusion": "4 Conclusion The MFC exhibited highest columbic efficiency, COD removal efficiency, nutrient removal efficiency and voltage when acetic acid was used in the MFC and lowest was observed for albumin. Increasing the pH from pH 6 to pH 8 resulted in an increased in the parameters being studied. At each pH, the highest columbic efficiency, COD removal efficiency, nutrient removal efficiency and voltage was observed at 20:1 and lowest was seen at 10:1. In all experiments best results were seen when C: N was used in the system and least promising ones were observed for C:P. In all experiment's nitrogen showed the highest removal efficiency and phosphorous exhibited the least removal efficiency.",
"introduction": "1 Introduction Industrial wastewater contains large amounts of organic matter, inorganic nutrients (like nitrogen, phosphorus, and sulfur) along with a wide array of other harmful pollutants that could adversely affect the environment. Hence, it is very important to treat such pollutant laden wastewater before releasing it into the natural water bodies. There are several ways to treat wastewater (chemical, physical, and biological processes), but sustainable wastewater treatment is the need of the day. Sustainable wastewater treatment not only aims at water reuse but also energy recovery and nutrient management ( Goswami et al., 2019 ). Simultaneous removal of carbon, nitrogen and sulfur is possible using conventional wastewater treatment systems ( Abeysiriwardana-Arachchige et al., 2020 ; Diaz-Elsayed et al., 2019 ; Castellanos et al., 2021 ). Although conventional wastewater treatment systems can remove/recover nutrients, it cannot produce electricity. Studies have shown that MFCs can use NO 3 - as a cathodic electron acceptor, allowing the simultaneous C removal at the anode and N at the cathode ( Ge et al., 2020 ; Vijay et al., 2020 ; Deng et al., 2018 ; Kelly and He, 2014 ). Biological sulfur removal can also be accomplished in a MFC ( Li et al., 2021 ; Cai et al., 2017 ; Luo et al., 2020 ). Phosphorous was demonstrated to be recovered in a MFC ( Huang et al., 2017 ; Geng et al., 2018 ; Liu et al., 2018 ). Simultaneous anaerobic sulfide and nitrate removal was coupled with electricity generation as suggested by Cai et al. (2013) . Removal of individual or a combination of two nutrients have been studied, however, no prior study has focused on the simultaneous removal of carbon, nitrogen, phosphorous and Sulfur in a MFC using an abiotic cathode. In the current experiment, nitrogen, phosphorous and sulfur were individually used as well as in combinations (N, P, S, N:P, N:S, P: N and N:P:S) as electron acceptors in the cathodic chamber. Oxygen is the most commonly used electron acceptor in the cathodic compartment of aerobic MFCs; however, high aeration costs make it a less feasible option ( Strik et al., 2011 ). Oxygen from air can be directly used by an aerobic cathode but they require catalysts which are usually expensive ( TerHeijne et al., 2008 ). Hence, various nutrients can be a good alternative to oxygen as final electron acceptors. Although above studies have focused on the removal of single nutrient or a combination of two nutrients, but none of the studies have been done on the simultaneous removal of carbon, nitrogen, sulfur, and phosphorous removal in MFC. Nutrient removal has been studied in the anodic chamber containing microbial communities in most of the previous studies; however, a simultaneous removal of nutrients has not been studied in the cathodic chamber of MFC containing abiotic cathode. In the current experiment nitrogen, phosphorous and sulfur were treated not only individually but in their mutual combinations (N, P, S, N:P, N:S, P: N and N:P:S) also as electron acceptors in the abiotic cathodic chamber. The purpose of this study was to find the most suitable ratio of carbon and nitrogen, phosphorus, and sulfur at a suitable pH where a maximum carbon and nutrient removal could be realized along electricity production.",
"discussion": "3 Results and discussion 3.1 Removal efficiency, columbic efficiency and voltage production The CE, RE and voltage production of seven combinations of carbon with nutrients (C:N, C:P, C:S, C: N:S, C:P:S, C:N:P and C:N:S:P) were studied at different ratios (20:1, 15:1 and 10:1). The best performance of MFC was evident for 20:1 and at pH 8; hence, the following result section will describe results for pH 8 at 20:1 ratio. For the carbon nutrient ratio of 10:1 and varying the pH; it was seen that at pH 6 ( Figure 2 ), C:P showed the least promising results when COD RE, CE and voltage was measured that is 29.2%, 8.4% and 127 mV, respectively. C: N exhibited the highest COD removal (35.1%), CE (10.3%) and voltage 128 mV. At pH 7 C:P showed the lowest voltage value i.e., 221 mV. The lowest RE and CE i.e. 41.9% and 20.7%, respectively, were also shown by C:P. The highest COD RE (50.9%), CE (25.9%) and voltage (229 mV) was exhibited by C: N ( Figure 2 B). When the pH was kept at 8, the C:P showed a lowest COD RE (55.3%) and CE (42.2%). The voltage output was also the lowest (348 mV). C: N showed highest voltage value i.e., 351 mV along with the highest RE and CE i.e., 67.4% and 45.8% ( Figure 2 C). Figure 2 Effect of pH on removal efficiency, columbic efficiency and voltage using acetic acid at 10:1. A: The performance of MFC using acetic acid at pH 6; B: The performance of MFC using acetic acid at pH 7; C: The performance of MFC using acetic acid at pH 8. Figure 2 For the C:N ratio of 20:1 at pH 6; the CE, RE and voltage were 18.2%, 42.7% and 187 mV, respectively ( Figure 3 ). On the other hand, C:P showed the lowest CE (15.3%) with RE (36%) and voltage output (171 mV) ( Figure 3 G). At pH 7, the CE was in the range of 44.7%–54%, RE ranged from 66.8%-76.7% and voltage was in the range of 509 mV–530 mV ( Figure 3 H). Increasing the pH to 8 improved the CE, RE and voltage production ( Figure 3 I). C:N showed the highest CE (75.8%), RE (86%) and voltage (667 mV) while C:P showed the lowest CE (70%), RE (83%) and voltage (642 mV). The results were found in the order C: N > C: N:S > C: N:S:P > C:N:P > C:S > C:P:S > C:P. Figure 3 Effect of pH on removal efficiency, columbic efficiency and voltage using acetic acid at 20:1. G: The performance of MFC using acetic acid at pH 6; H: The performance of MFC using acetic acid at pH 7; I: The performance of MFC using acetic acid at pH 8. Figure 3 The comparison of current and power density has been presented in Tables 1 A, 1 B, 1 C. Table 1A The relationship of current (I) and power density (PD) for acetic acid at various ratios and pH values. Table 1A Acetic acid Ratio pH 6 pH 7 pH 8 I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) 10:01 CN 0.128 9.60 × 10 −2 0.224 2.95 × 10 −1 0.351 7.25 × 10 −1 CNS 0.127 9.48 × 10 −2 0.221 2.87 × 10 −1 0.349 7.16 × 10 −1 CNSP 0.126 9.33 × 10 −2 0.218 2.8 × 10- 1 0.347 7.08 × 10 −1 CNP 0.126 9.33 × 10 −2 0.215 2.72 × 10 −1 0.346 7.04 × 10 −1 CS 0.123 8.89 × 10 −2 0.213 2.67 × 10 −1 0.344 6.96 × 10 −1 CPS 0.121 8.61 × 10 −2 0.209 2.57 × 10 −1 0.342 6.88 × 10 −1 CP 0.119 8.33 × 10 −2 0.202 2.40 × 10 −1 0.34 6.80 × 10 −1 15:01 CN 0.193 2.19 × 10 −1 0.38 8.49 × 10 −1 0.527 1.63 CNS 0.19 2.12 × 10 −1 0.377 8.36 × 10 −1 0.527 1.63 CNSP 0.188 2.07 × 10 −1 0.367 7.92 × 10 −1 0.526 1.63 CNP 0.186 2.03 × 10 −1 0.365 7.84 × 10 −1 0.524 1.62 CS 0.185 2.01 × 10 −1 0.364 7.79 × 10 −1 0.52 1.59 CPS 0.183 1.96 × 10 −1 0.362 7.71 × 10 −1 0.518 1.58 CP 0.182 1.94 × 10 −1 0.359 7.58 × 10 −1 0.516 1.57 20:01 CN 0.187 2.05 × 10 −1 0.529 1.65 0.667 2.62 CNS 0.185 2.01 × 10 −1 0.526 1.63 0.657 2.54 CNSP 0.184 1.99 × 10 −1 0.525 1.62 0.653 2.51 CNP 0.182 1.94 × 10 −1 0.52 1.59 0.651 2.49 CS 0.18 1.90 × 10 −1 0.51 1.53 0.65 2.49 CPS 0.175 1.80 × 10 −1 0.506 1.51 0.645 2.45 CP 0.171 1.72 × 10 −1 0.495 1.44 0.642 2.42 Table 1B The relationship of current (I) and power density (PD) for sucrose at various ratios and pH values. Table 1B Sucrose Ratio pH6 pH7 pH8 I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) 10:01 CN 0.099 5.76 × 10 −2 0.201 2.38 × 10 −1 0.341 6.84 × 10 −1 CNS 0.091 4.87 × 10 −2 0.187 2.06 × 10 −1 0.341 6.84 × 10 −1 CNSP 0.088 4.55 × 10 −2 0.181 1.93 × 10 −1 0.335 6.60 × 10 −1 CNP 0.086 4.35 × 10 −2 0.175 1.80 × 10 −1 0.32 6.02 × 10 −1 CS 0.084 4.15 × 10 −2 0.161 1.52 × 10 −1 0.312 5.73 × 10 −1 CPS 0.08 3.76 × 10 −2 0.156 1.43 × 10 −1 0.225 2.98 × 10 −1 CP 0.076 3.39 × 10 −2 0.15 1.32 × 10 −1 0.221 2.87 × 10 −1 15:01 CN 0.135 1.07 × 10 −1 0.33 6.41 × 10 −1 0.576 1.95 CNS 0.134 1.05 × 10 −1 0.324 6.18 × 10 −1 0.571 1.92 CNSP 0.132 1.02 × 10 −1 0.316 5.87 × 10 −1 0.563 1.86 CNP 0.129 9.78 × 10 −2 0.252 3.74 × 10 −1 0.559 1.84 CS 0.12 8.47 × 10 −2 0.224 2.95 × 10 −1 0.51 1.53 CPS 0.119 8.33 × 10 −2 0.214 2.69 × 10 −1 0.487 1.40 CP 0.116 7.91 × 10 −2 0.21 2.59 × 10 −1 0.469 1.29 20:01 CN 0.194 2.21 × 10 −1 0.565 1.88 0.624 2.29 CNS 0.186 2.03 × 10 −1 0.564 1.87 0.62 2.26 CNSP 0.184 1.99 × 10 −1 0.56 1.84 0.618 2.25 CNP 0.172 1.74 × 10 −1 0.555 1.81 0.614 2.22 CS 0.169 1.68 × 10 −1 0.542 1.73 0.61 2.19 CPS 0.163 1.56 × 10 −1 0.535 1.68 0.607 2.17 CP 0.143 1.20 × 10 −1 0.532 1.66 0.589 2.04 Table 1C The relationship of current (I) and power density (PD) for albumin at various ratios and pH values. Table 1C Albumin Ratio pH6 pH7 pH8 I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) I (mA) P.D (mW/m 2 ) 10:01 CN 0.082 3.95 × 10 −2 0.147 1.27 × 10 −1 0.172 1.74 × 10 −1 CNS 0.078 3.57 0.141 1.17 × 10 −1 0.172 1.74 × 10 −1 CNSP 0.07 2.88 × 10 −2 0.136 1.09 × 10 −1 0.17 1.70 × 10 −1 CNP 0.064 2.40 × 10 −2 0.134 1.06 × 10 −1 0.165 1.60 × 10 −1 CS 0.06 2.11 × 10 −2 0.131 1.01 × 10 −1 0.159 1.48 × 10 −1 CPS 0.055 1.77 × 10 −2 0.127 9.49 × 10 −2 0.151 1.34 × 10 −1 CP 0.033 6.40E-06 0.121 8.61 × 10 −2 0.145 1.23 × 10 −1 15:01 CN 0.136 1.08 × 10 −1 0.203 2.42 × 10 −1 0.206 2.49 × 10 −1 CNS 0.134 1.05 × 10 −1 0.185 2.01 × 10 −1 0.205 2.47 × 10 −1 CNSP 0.133 1.04 × 10 −1 0.188 2.08 × 10 −1 0.204 2.44 × 10 −1 CNP 0.132 1.02 × 10 −1 0.176 1.82 × 10 −1 0.203 2.42 × 10 −1 CS 0.131 1.00 × 10 −1 0.141 1.17 × 10 −1 0.199 2.37 × 10 −1 CPS 0.129 9.78 × 10 −2 0.129 9.79 × 10 −2 0.197 2.28 × 10 −1 CP 0.125 9.19 × 10 −2 0.122 8.76 × 10 −1 0.192 2.16 × 10 −1 20:01 CN 0.178 1.86 × 10 −1 3.25 6.21 × 10 −1 0.479 1.34 CNS 0.167 1.64 × 10 −1 0.32 6.02 × 10 −1 0.471 1.30 CNSP 0.16 1.50 × 10 −1 0.299 5.26 × 10 −1 0.469 1.29 CNP 0.156 1.43 × 10 −1 0.282 4.68 × 10 −1 0.467 1.28 CS 0.151 1.34 × 10 −1 0.28 4.61 × 10 −1 0.464 1.26 CPS 0.153 1.37 × 10 −1 0.279 4.58 × 10 −1 0.462 1.25 CP 0.143 1.20 × 10 −1 0.273 4.38 × 10 −1 0.46 1.24 3.1.1 Sucrose removal Using sucrose as carbon source, at carbon to nutrient ratio of 20:1 and pH 6, the C:P yielded the lowest COD RE (38%), CE (14.5%) and voltage production (151 mV) and the results were shown in Figure 4 . Figure 4 Effect of pH on Removal Efficiency, Columbic Efficiency and Voltage using Sucrose at 20:1. G: The performance of MFC using sucrose at pH 6; H: The performance of MFC using sucrose at pH 7; I: The performance of MFC using sucrose at pH 8. Figure 4 Whereas the highest RE, CE, and voltage production of 42.7%,18.3% and 194 mV respectively was achieved with C: N ( Figure 4 G). When the pH was increased from pH 6 to pH 7, an increase in the CE, RE and amount of voltage was seen. Similarly, C:P again yielded the lowest COD RE, CE, and voltage production of 66.2%, 43 % and 532 mV, respectively, at pH 7 ( Figure 4 H). The highest COD RE of 78.4%, CE of 61.1% and voltage production of 565 mV was recorded for C: N at pH 7. An increase in pH from pH 7 to pH 8 resulted in the increase of the three parameters being studied namely COD RE, CE, and voltage production. In the case of pH 8, the lowest COD RE, CE, and voltage production of 69%, 58% and 589 mV respectively were recorded for C:P, whereas the highest of 86%, 73.9% and 624 mV respectively was observed for C: N ( Figure 4 I). At all the three ratios and pH COD RE was found in the following order C: N > C: N:S > C: N:S:P > C: N:P > C:S > C:P:S > C:P. CE and voltage production followed the same order. 3.1.2 Albumin Keeping carbon nutrient ratio constant at 10:1 and pH was varied then at pH 6 ( Figure 5 A) C:P showed the least COD RE, CE and voltage that is 16.2%, 2.3 % and 33 mV, respectively. C: N exhibited the highest COD removal (22.2%), CE (6.7%) and voltage 83 mV. At pH 7, C:P again yielded the lowest results for COD RE, CE, and voltage with values of 27%, 7.7% and 129 mV, respectively. At pH 7, C: N exhibited the highest COD RE, CE, and voltage that is 35.8%, 12.8% and 147 mV respectively ( Figure 5 B). Increasing the pH to 8 further increased the CE, RE and voltage production ( Figure 5 C). C: N showed the highest CE (27 %), RE (50.6%) and voltage (174 mV) while C:P showed the lowest CE (12.8%), RE (37.6%) and voltage (145 mV). Figure 5 Effect of pH on Removal Efficiency, Columbic Efficiency and Voltage using Albumin at 10:1. A: The performance of MFC using albumin at pH 6; B: The performance of MFC using albumin at pH 7; C: The performance of MFC using albumin at pH 8. Figure 5 At carbon nutrient ratio 15:1, C:P yielded the lowest COD RE of 14.5%, CE of 6.9% and voltage production of 124 mV at pH 6. Whereas the highest RE, CE, and voltage production of 30.8%, 9.1% and 137 mV respectively were achieved with C: N ( Figure 6 D). When the pH was increased from pH 6 to pH 7 an increase in the CE, RE and amount of voltage was seen. When the pH was increased to 7 the CE ranged from 11.1% - 14.7%, RE ranged from 32.7 % - 44.4 % and voltage was in the range of 129 mV–180 mV ( Figure 6 E). When the pH was kept at 8 C:P showed the lowest COD RE (40.7 %) and CE (16.5 %). The voltage output was also the lowest for C:P (192 mV). C: N showed highest voltage value i.e., 210 mV along with the highest RE and CE i.e., 58% and 39.6% ( Figure 6 F). Figure 6 Effect of pH on Removal Efficiency, Columbic Efficiency and Voltage using Albumin at 15:1. D: The performance of MFC using albumin at pH 6; E: The performance of MFC using albumin at pH 7; F: The performance of MFC using albumin at pH 8. Figure 6 In the next series of experiments carbon nutrient ratio was kept constant at 20:1 and pH was varied ( Figure 7 ). When the pH was kept at pH 6 the highest CE (15.3%), RE (40%) and voltage (160 mV) was found for C: N. On the other hand, C:P showed the lowest CE (10%), RE (32%) and voltage output (143 mV) ( Figure 7 G). Similarly, at pH 7, C:P again yielded the lowest COD RE, CE, and voltage production of 44.9%, 17.6% and 273 mV respectively. The highest COD RE of 55.5 %, CE of 17.6 % and voltage production of 325 mV was recorded for C: N at pH 7 ( Figure 7 H). When the pH was kept at 8 C:P showed the lowest COD RE (60.4%) and CE (35%). The voltage output was also the lowest for C:P (460 mV). C:N showed highest voltage value i.e. 472 mV along with the highest RE and CE i.e. 66.1% and 42.5% ( Figure 7 I). For all the three ratios CE, COD RE and voltage generation was in the order C: N > C: N:S > C: N:S:P > C: N:P > C:S > C:P:S > C:P. The highest was exhibited by C: N and the lowest by C:P. Figure 7 Effect of pH on Removal Efficiency, Columbic Efficiency and Voltage using Albumin at 20:1. G: The performance of MFC using albumin at pH 6; H: The performance of MFC using albumin at pH 7; I: The performance of MFC using albumin at pH 8. Figure 7 The type of cathodic electron acceptor seems to exert a great influence on the functioning of MFC. Voltage generation is greatly influenced by the cathode electron acceptor ( Gurung and Oh, 2012 ). How much voltage can be generated depends on the redox potential of the nutrient being used as electron acceptor. Nitrate has lower redox potential as compared to sulphate and phosphate hence, more voltage output was seen in the cases where nitrate was used a final electron acceptor. Hence, electron acceptors with lower redox potential help in greater voltage output ( Cai et al., 2016 ). Electron acceptors are the accelerants that help in speeding up the forward reaction of biodegradation ( Pandit et al., 2011 ). Higher rate of biodegradation translates into higher removal and columbic efficiencies. A marked increase in CE, RE and voltage production were seen when the pH was increased from 6 to 8. The lowest CE, RE and voltage production were seen at pH 6 and highest at pH 8. At lower pH 6 the difference between RE and CE was significant. The pH of MFC can greatly affect MFC performance. The voltage output as well as RE of COD is greatly influenced by pH ( Puig et al., 2010 ). Studies have found that MFC shows lowest current density at pH 6 and that an increase in pH results in an increase in the voltage production ( He et al., 2008 ). The pH value of 8 seems to favor the anaerobic communities in MFC. When the pH is increased the conditions become favorable for the growth and production of electrogenic bacteria ( Mohamed et al., 2020 ). This is the reason why a significant difference could be seen at various applied pH values. 3.2 Substrate transformation in MFC 3.2.1 Nitrogen removal The lowest nitrogen RE was recorded using albumin as the substrate at pH 6 and ration of 10:1, whereas the highest nitrogen RE was observed using acetic acid as the substrate at a pH of 8 and 20:1. When acetic acid was used as the substrate highest nitrogen RE (85%) was achieved at pH 8 and 20:1 and the lowest RE was observed at pH 6 and 10:1 When sucrose was used as a substrate highest nitrogen RE (72.8%) was seen at pH 8 and 20:1 and the lowest RE (29.6%) was seen at pH 6 and 10:1. Using albumin resulted in the lowest of the three substrates with a highest and lowest RE of 60.7% and 14.3% respectively. The highest RE using albumin was observed at pH 8 and a ratio of 20:1 and the lowest RE at pH 6 with a ratio of 10:1 ( Figure 8 A). Figure 8 Nitrogen, phosphorus and sulfur removal at various pH and ratios. A: Nitrogen removal at various pH and ratios; B: Phosphorus removal at various pH and ratios; C: Sulfur removal at various pH and ratios. Figure 8 3.2.2 Phosphorus removal The highest phosphorus removal was achieved using acetic acid at a pH of 8 with a ratio of 20:1, whereas the lowest phosphorus removal was observed using albumin as a substrate with a ratio of 10:1 at pH 6. When acetic acid was used as substrate, the highest removal of the substrate was seen at pH 6 and ratio 20:1, whereas the lowest removal percentage of phosphorus using acetic acid was observed at pH 6 with a ratio of 10:1. In the case of sucrose, the highest RE of phosphorus (47.4%) was also achieved at a pH of 8 and 20:1 and the lowest RE (18.8%) was observed at pH 6 with and 10:1. Using albumin as a substrate resulted in the lowest RE of all substrates. The lowest and highest RE of 5% and 25% respectively were recorded while using pH of 6 and ratio 10:1 and pH 8 and ratio 20:1 respectively ( Figure 8 B). 3.2.3 Sulfur removal using MFC The highest sulfur removal (74%) was achieved using acetic acid at pH 8 and a ratio of 20:1 and the lowest sulfur removal (11.2%) was observed using albumin as a substrate at pH 6 with a ratio of 10:1. While using acetic acid as the substrate the lowest sulfur RE (32%) was achieved at pH 6 and ratio 10:1. The highest removal in the case of acetic acid was achieved at pH 8 and a ratio of 20:1. The lowest RE (28%) using sucrose was observed at 10:1 and pH6 and the highest (70%) RE was seen at pH 8 and 20:1. In the case of sucrose, the lowest RE was achieved using pH 6 and 10:1 whereas the highest RE with sucrose was achieved at pH 8 and 20:1. Using albumin the lowest and highest sulfur removal achieved was 11.2% and 63.4% respectively. The lowest RE using albumin was observed at pH 6 and 10:1 and the highest RE was achieved using pH 8 and a ratio of 20:1 ( Figure 8 C). The redox potentials of nitrogen and oxygen are very close to each other and hence it can be used as an electron acceptor in the cathodic cell ( Han et al., 2020 ). Virdis et al. (2011) found that in a MFC treating synthetic wastewater containing acetate nitrate removal of 94.1% could be achieved. When cathodic nitrate accepts electrons, it turns into nitrogen gas. A coupled MFC system comprising of an oxic-biocathode MFC (O-MFC) and an anoxic-biocathode MFC (A-MFC) was implemented for simultaneous removal of C and N from wastewater ( Xie et al., 2011 ). The MFC system obtained a maximum COD, NH 4 + -N and total nitrogen (TN) removal rate of 98.8%, 97.4% and 97.3%, respectively, at an A-MFC external resistance of 5 Ω. Sulphates in the cathodic chambers can act as electron acceptors. When Sulfide and acetate (C source) was treated using MFC ( Rabaey et al., 2006 ). Ninety eight percent of the sulfide and 46% of the acetate was removed. MFC removed SO 4 −2 via sulfide. This demonstrates that effluents can be polished by a MFC for both residual C and S compounds. Izadi and Rahimnejad (2014) used a dual chamber MFC to investigate the removal of S. The initial concentration of sulfide in the anode compartment was 0.4 g/L and it was completely removed after 3 days of MFC operation. The maximum generated voltage, power and current density were 988.915 mV, 346.746 mW.m- 2 , 1285.64 mA m −2 . Amount of phosphorus removal was different for each substate and it was in the order acetic acid > sucrose > albumin. Varying nutrient ratio and pH influenced the removal percentage of phosphorous. For each substrate, most phosphorous removal was recorded at 20:1 and pH 8. At pH 6 and 10:1 the least phosphorous removal was seen. Phosphate can act an electron acceptor but requires a large amount of energy to accept electrons because of its high redox potential. An increase in pH can help in the precipitation of phosphates and hence result it its removal ( Tao et al., 2015 ). Not too many studies have been done on phosphorous removal using MFC. However, there are studies showing that phosphorous removal is possible using MFC. Air-cathode single chamber MFCs were operated with swine wastewater and 70%–82% of the phosphorous was removed from the influent ( Ichihashi and Hirooka, 2012 ). Using MFC ( Zang et al., 2012 ) were able to remove C and P from urine. The removal efficiencies for PO 4 3- and COD were found out to be 42.6% and 62.4% respectively. The power density of 0.9 W m 3 was obtained. 3.2.4 Simultaneous Nitrogen, phosphorous and sulfur removal Simultaneous removal of nutrients namely nitrogen, phosphorous and sulfur was studied using acetic acid as carbon source in the MFC and was presented in Figure 9 . At pH 6 the lowest nitrogen (39%), sulfur (35%) and phosphorous (25%) RE was seen when the carbon nutrient ratio was kept 10:1. The RE increased when the carbon nutrient ratio 15:1 was used and the highest was noted when 20:1 carbon nutrient ratio was used in the MFC i.e., 48% nitrogen removal, 41% sulfur removal and 36% phosphorous removal was noted. When the pH was increased to 7, an increase was seen in the removal efficiencies was noted. At carbon nutrient ratio of 10:1 and pH 7 the lowest nutrient removal efficiencies were noted. When the carbon nutrient ratio of 15:1 was used, an increase in the removal efficiencies was noted. The highest nitrogen (61%), sulfur (56%), and phosphorous (47%) removal efficiencies were noted when the carbon nutrient ratio 20:1 was used in the experiment. Increasing the pH to pH 8 showed the best results (higher nutrient removal efficiencies as compared to pH 6 and pH 7). At pH 8 the highest RE was noted when the carbon nutrient ratio was kept 20:1 i.e., nitrogen RE was 81%, sulfur RE was 77% and phosphorous RE was 64% ( Figure 9 ). Figure 9 Simultaneous Nitrogen, Phosphorous and Sulfur removal at various pH and ratios. Figure 9 When sucrose was used as a carbon source in the MFC and simultaneous removal of nitrogen, phosphorous and sulfur was studied it was observed that the lowest nitrogen (32%), sulfur (30%) and phosphorous (21%) RE was seen when the carbon nutrient ratio was kept 10:1. The RE became higher when the carbon nutrient ratio 15:1 was used. The highest nitrogen removal (42%), sulfur removal (39%) and phosphorous removal (27%) was noted when 20:1 carbon nutrient ratio was used in the MFC ( Figure 9 ). At carbon nutrient ratio of 10:1 and pH 7 the lowest nutrient removal efficiencies were noted. When the carbon nutrient ratio of 15:1 was used, an increase in the removal efficiencies was noted. The highest nitrogen (60%), sulfur (51%), and phosphorous (42%) removal efficiencies were noted when the carbon nutrient ratio 20:1 was used in the experiment. At pH 8 the highest RE was noted when the carbon nutrient ratio was kept 20:1 i.e., nitrogen RE was 72%, sulfur RE was 67% and phosphorous RE was 59% ( Figure 9 ). In the last set of experiments albumin was used as a substrate and simultaneous removal of the nutrients was studied. The lowest nitrogen (14%), sulfur (12%) and phosphorous (10%) RE was seen when the carbon nutrient ratio was kept 10:1. The RE became higher when the carbon nutrient ratio 15:1 was used. The highest nitrogen removal (22%), sulfur removal (20%) and phosphorous removal (19%) was noted when 20:1 carbon nutrient ratio was used in the MFC. When the pH was increased to 7, an increase was seen in the removal efficiencies was seen. At carbon nutrient ratio of 10:1 and pH 7 the lowest nutrient removal efficiencies were noted. When the carbon nutrient ratio of 15:1 was used, an increase in the removal efficiencies was noted. The highest nitrogen (36%), sulfur (32%), and phosphorous (29%) removal efficiencies were noted when the carbon nutrient ratio 20:1 was used in the experiment. At pH 8, the lowest nitrogen (40%), sulfur (35%) and phosphorous (34%) RE was seen when the carbon nutrient ratio was kept 10:1. The RE increased when the carbon nutrient ratio 15:1 was used and the highest was noted when 20:1 carbon nutrient ratio was used in the MFC i.e., 56% nitrogen removal, 49% sulfur removal and 42% phosphorous removal was noted. The results show that all the three substrates showed best results at pH 8 and carbon nutrient ratio of 20:1. Among the three substrates acetic acid showed the highest nutrient removal. The second highest nutrient RE was shown by sucrose. The least nutrient RE was shown by albumin. The nutrient removal followed a similar trend for all substrates i.e., nitrogen showed the highest RE followed by sulfur. The least was shown by phosphorous. Experiments carried out using nutrients in combination (C: N:S, C:P:S, C:P: N, C:P:N:S) showed that more than one nutrient can be removed simultaneously using an MFC. The results were found in the order C: N > C: N:S > C: N:S:P > C: N:P > C:S > C:P:S > C:P. The best simultaneous removal of nutrients was achieved when acetic acid was used in the system and pH 8 was maintained while keeping the carbon nutrient ratio at 20:1. Simultaneous anaerobic sulfide and nitrate removal in the anodic chamber coupled with electricity generation has been extensively studied ( Cai et al., 2013 , 2014 , 2020 ). However, simultaneous removal of nitrogen and sulfur, phosphorous and sulfur, phosphorus and nitrogen and the combination of the three nutrients in the cathodic chamber of MFC has not been researched extensively till now. This study showed that the simultaneous removal of these combinations of nutrients can be achieved in the cathodic chamber of MFC. Out of the three nutrients, nitrogen showed the highest removal followed by sulfur. The least removal was shown by phosphorous. The reason behind this is their redox potentials which are in the order phosphate > sulphate > nitrate. Nitrate has the lowest redox potential out of the three i.e. (0.74 V). Low redox potential means that it most easily accepts the electrons coming from the anodic chamber ( Sun et al., 2013 ) When nitrate accepts electrons, it gets reduced and turns into nitrogen gas. Sulphate can also accept electrons and turn in to elemental sulfur. Phosphorus removal was the lowest because it is not a very good electron acceptor because of endergonic reduction potential i.e., it requires a large amount of energy. The best nutrient removal was obtained at pH 8. At pH 8 electricity producing thrive resulting in more generation of electrons in the anode that could be accepted by the electron acceptors in the cathodic chamber. Law et al. (2011) investigated how pH range effects removal of N. They found that the nitrogen removal increased as the pH was increased. The maximum removal was found at pH 8.0. Similarly ( Guštin &Marinšek, 2011 ), found that ammonia stripping bench plant removed 92.8% of ammonium and 88.3% of total nitrogen from the anaerobic digestion effluent at high pH. It is because a high pH changes ammonia/ammonium ratio in favor of ammonia. Swine wastewater has been studied in a single chamber MFC for struvite precipitation ( Ichihashi and Hirooka, 2012 ). It was found that 70–82% of phosphorus was removed, and struvite precipitation only occurred on the cathode surface when electrolyte pH was around 8 ( Zhai et al., 2012 ). studied the removal of sulfur in MFC. They found that highest sulfur recovery efficiency (78.6 ± 8.3%) and CE (58.6 ± 1.6%) occurred at a pH 8. MFC showed the highest nutrient removal when acetic acid was used as a carbon source. The results were found in the order Acetic Acid > Sucrose > Albumin. Carbon source greatly influences microorganism metabolism ( Mitra and Mishra, 2018 ). Acetic acid is a simple compound and can be degraded easily by microbes resulting in the release of electrons to the anode. These electrons then move to the cathode and reduce the nutrients there acting as electron acceptors hence resulting in nutrient removal. Sucrose is more complex than acetic acid and its degradation releases fewer electrons as compared to acetic acid. Albumin is the most complex among the three and its degradation offers the least number of electrons. Number of electrons released greatly effects how much nutrient removal takes place. Hence, simpler compounds like acetic acid are more easily used up by exoelectrogens as compared to macromolecules ( Yang et al., 2019 ). Studies have shown that acetic acid is a better electron donor for exoelectrogens as compared to longer chain compounds ( Freguia et al., 2010 ). It has been reported that MFCs are best option to recover nutrients like ammonium and phosphate from wastewater ( Ye et al., 2019 ). In case of each substrate, the nutrient removal was found in the order 20:1 > 15:1 > 10:1. Most nutrient removal was found when the ratio was 20:1. It might be due to a higher amount of organic matter at 20:1 that can be degraded by microorganisms. More organic matter degradation means higher number of electrons available to reduce nutrient ions acting as final electron acceptors. Electrons and protons released by redox reactions in the anode move to the cathode resulting in bio-potential which helps in voltage generation ( Mohan et al., 2009 ). The increase in loading rate results in an increase power generation ( Goud et al., 2011 ). The energy generated by MFC is different with different electron donors ( Sun et al., 2009 )."
} | 8,144 |
35639688 | PMC9191640 | pmc | 920 | {
"abstract": "Significance Biological carbon fixation provides opportunities to directly utilize CO 2 to synthesize a broad range of value-added compounds, potentially displacing petroleum feedstock use in industry. Chemoautotrophs are particularly interesting as their carbon fixation can be driven chemically by renewable H 2 in place of light, which can limit industrial fermentation of photosynthetic organisms. We describe the development of a methanogenic host, Methanococcus maripaludis , for metabolic engineering. Since redox cofactors used in upstream archaeal carbon fixation pathways are orthogonal to typical downstream biosynthetic pathways, it was necessary to engineer both NADH biosynthesis and turnover. In doing so, we are able to show that methanogenic archaea can, indeed, serve as a platform for the high-yield production of bioplastics and monomers from CO 2 and H 2 .",
"discussion": "Discussion Carbon fixation or primary production plays an essential part of the global carbon cycle, providing the chemical basis for life on Earth by transforming inorganic to organic carbon. Despite the enormous scale, biological carbon fixation remains an environmental process and relatively untapped for industrial chemical production. Although less well-characterized than photosynthesis, non-phototrophic CO 2 assimilation is thought to contribute 5 to 22% of ocean primary production and occurs without light input ( 54 ). As such, chemoautotrophs allow sustainable synthesis using CO 2 as the carbon building block with a broader range of sustainable energy inputs, such as electrocatalytically or photocatalytically generated H 2 . Given the low efficiency of photosynthetic carbon fixation (∼8 to 9%) ( 55 ), it is also possible that coupling sustainable H 2 or formate production to engineered chemoautotrophic hosts could match the yields and efficiencies of photosynthetic organisms. In this work, we focused on developing methods to interface the carbon-fixation abilities of methanogenic archaea with downstream biosynthetic pathways derived from heterotrophs. M. maripaludis uses the modified Wood–Ljungdahl pathway for CO 2 assimilation, where two equivalents of CO 2 are used to generate one acetyl-CoA in the most energy-efficient carbon-fixation pathway known in nature in terms of ATP required per mole of fixed CO 2 . One challenge in utilizing methanogenic archaea as a metabolic engineering host platform is that they have evolved to use orthogonal redox cofactors, compared with typical downstream biosynthetic pathways that utilize NAD(P)(H). Toward this end, we utilized 3HB and PHB production as a model heterotrophic pathway to examine its potential as a host for metabolic engineering, as there are many other systems to benchmark its production. Heterotrophs fed sugar-beet molasses, sucrose, cooking oils, glucose, and other carbon sources, of course, achieve the highest yields of more than 50% of cell dry weight in pilot and larger-scale production ( 56 , 57 ). The native bacterial PHB producer, Cupriavadus necator , which can grow heterotrophically or chemoautotrophically via the Calvin–Benson–Bassham cycle, can produce similar yields of 80% cell dry weight of PHB using CO 2 and H 2 ( 58 ). However, other native chemoautotrophic and phototrophic producers of PHB that utilize the Wood–Ljundahl pathway produce ∼20 to 30% cell dry weight under optimized scale-up conditions. There are also other chemoautotrophs, such as acetogens, that can be engineered to produce 3HB and PHB, as they contain Rnf complexes that allow for equilibration between Fd and NAD pools ( 59 ). In comparison, our engineered strain of M. maripaludis can achieve similar yields of PHB (25%) as these native producers under unoptimized, laboratory-scale growth, suggesting that it could serve as a reasonable host for metabolic engineering, given its other attributes, such as the high ATP utilization efficiency for carbon fixation, the production of methane as a valuable by-product, and the potential to use formate as a carbon and electron source to avoid gas mass-transfer issues with H 2 . At this time, M. maripaludis is still in the relatively nascent stages of development as an industrial host and further work to identify sequence determinants for tunable gene expression, the use of an expanding genetic toolkit for genomic engineering, and the development of more robust metabolic models will assist with the continuing domestication of this host for metabolic engineering. A long-term goal is to work toward domesticating new hosts for metabolic engineering, as the broad range of chemical phenotypes in nature can serve as advantages in accelerating the formation of a fermentation-based industry for chemical production. Given the scale of the challenge, many different solutions are needed to aggregately reduce environmental impact and increase efficiency of energy and resource utilization. Compared with canonical fermentation approaches using photoautotrophs or heterotrophs, a hybrid fermentation approach that uses complementary microbial and electrochemical catalysts could have lower land requirements and provide soluble carbon and energy sources in situ for fermentation. In this approach, purified CO 2 , water, and renewable electricity could be provided as inputs, and value-added chemicals derived from acetyl-CoA could be produced, such as alcohols, amino acids, and isoprenoids. In addition, methane could be harnessed using similar processes as those typically used in anaerobic digesters to generate a renewable fuel and energy source. Through a range of physiological studies, we were able to design a highly expressed engineered pathway in M. maripaludis . With further transcriptomic and proteomic studies in addition to other experiments, we showed that NADH availability was a limiting factor for small-molecule production. Using a combination of a synthetic nicotinamide salvage pathway and a formate dehydrogenase to recycle the NADH consumed in our pathway, we were able to achieve titers of PHB and its monomer of up to 171 ± 4 mg/L and 24.0 ± 1.9% of cell biomass, which is two orders of magnitude more than previous efforts in its use as a host ( 23 ). Taken together, we hope that insights presented in this work provide a foundation for more extensive metabolic-engineering efforts in M. maripaludis and other archaea, allowing us to better tap the diverse chemical abilities found in nature."
} | 1,613 |
27047476 | PMC4797314 | pmc | 921 | {
"abstract": "Deep-sea oceanic crust constitutes the largest region of the earth’s surface. Accumulating evidence suggests that unique microbial communities are supported by iron cycling processes, particularly in the young (<10 million-year old), cool (<25°C) subsurface oceanic crust. To test this hypothesis, we investigated the microbial abundance, diversity, and metabolic potentials in the sediment-buried crust from “North Pond” on western flank of the Mid-Atlantic Ridge. Three lithologic units along basement Hole U1383C were found, which typically hosted ∼10 4 cells cm -3 of basaltic rock, with higher cell densities occurring between 115 and 145 m below seafloor. Similar bacterial community structures, which are dominated by Gammaproteobacterial and Sphingobacterial species closely related to iron oxidizers, were detected regardless of variations in sampling depth. The metabolic potentials of the crust microbiota were assayed by metagenomic analysis of two basalt enrichments which showed similar bacterial structure with the original sample. Genes coding for energy metabolism involved in hydrocarbon degradation, dissimilatory nitrate reduction to ammonium, denitrification and hydrogen oxidation were identified. Compared with other marine environments, the metagenomes from the basalt-hosted environments were enriched in pathways for Fe 3+ uptake, siderophore synthesis and uptake, and Fe transport, suggesting that iron metabolism is an important energy production and conservation mechanism in this system. Overall, we provide evidence that the North Pond crustal biosphere is dominated by unique bacterial groups with the potential for iron-related biogeochemical cycles.",
"introduction": "Introduction Oceanic crust microbiology has long been ignored and is not well studied due to technical constraints; however, the crust has been assumed to harbor active microorganisms that may significantly contribute to global biogeochemical cycles and weathering of the seafloor landscape ( Schrenk et al., 2010 ; Wang et al., 2013 ). Several lines of evidence have revealed the presence of microorganisms in this dark, oligotrophic biosphere ( Fisk et al., 1998 ; Cowen et al., 2003 ; Santelli et al., 2008 ; Lever et al., 2013 ); however, some fundamental questions remain, including (1) how much microbial biomass is present in the oceanic crust, (2) where do the microorganisms originate, and (3) what are their metabolic functions. The recent Integrated Ocean Drilling Program (IODP) expeditions dedicated to microbiology ( Expedition 327 Scientists, 2010 ; Expedition 329 Scientists, 2011 ; Expedition 330 Scientists, 2011 ; Expedition 336 Scientists, 2012b ) support the investigation of the basalt-hosted oceanic crust and the collection of uncontaminated samples for microscopic and molecular analysis. Previous studies attempted to count cells from seafloor-exposed basalts ( Einen et al., 2008 ; Santelli et al., 2008 ; Jacobson Meyers et al., 2014 ), subsurface gabbros ( Mason et al., 2010 ) and crustal fluids ( Jungbluth et al., 2013 ). The results showed that cell densities in the seafloor-exposed crust were between 10 6 and 10 9 cells cm -3 , while those in the subsurface had lower cell densities (<10 5 cell cm -3 ). Diverse microbial communities from crustal environments have been detected by culture-dependent and -independent techniques spanning a large range of bacterial phyla. For example, Deltaproteobacteria, Firmicutes, Gammaproteobacteria, and Bacteroidetes are present in the flanks of the Juan de Fuca Ridge (JdFR) and the Costa Rica Rift ( Nigro et al., 2012 ; Jungbluth et al., 2013 , 2014 ). Seafloor basaltic glass from the East Pacific Rise ( Santelli et al., 2008 , 2009 ) and the Arctic spreading ridges ( Lysnes et al., 2004 ), altered basalts from the Hawaiian Loihi Seamount ( Templeton et al., 2005 ; Santelli et al., 2008 ; Jacobson Meyers et al., 2014 ) and the Mid-Atlantic Ridge ( Rathsack et al., 2009 ; Mason et al., 2010 ) are dominated by Gammaproteobacteria and Alphaproteobacteria. Extracellular enzyme activity tests, functional gene analysis, carbon and sulfur isotopic signatures and laboratory incubations demonstrated the presence of active microorganisms involved in methane- and sulfur-cycling and organic matter transformations ( Mason et al., 2010 ; Lever et al., 2013 ; Jacobson Meyers et al., 2014 ; Robador et al., 2015 ; Supplementary Table S1). However, these studies were restricted to seafloor-exposed basaltic habitats ( Templeton et al., 2005 ; Einen et al., 2008 ; Santelli et al., 2008 ), subsurface crustal environments with high temperature basalts ( Nigro et al., 2012 ; Jungbluth et al., 2013 ), and mantle-type rock ( Brazelton et al., 2010 ; Mason et al., 2010 ). The microbial life of the young, cool subsurface basalts in ridge flank systems, which represent a more common hydrologically active type of ocean crust ( Edwards et al., 2012 ), has not been characterized yet. Integrated Ocean Drilling Program Expedition 336 drilled the basaltic basement at “North Pond” (NP), which is located on the western flank of the Mid-Atlantic Ridge ( Expedition 336 Scientists, 2012b ). Numerous hydrological, geological, and geochemical data have been collected at this site from previous ocean drilling ( Becker et al., 2001 ) and site surveys ( Langseth et al., 1992 ; Picard and Ferdelman, 2011 ; Ziebis et al., 2012 ). The data indicated that this area was characterized by vigorous, oxic seawater circulation within the young basaltic crust under a <300 m sedimentary pile ( Expedition 336 Scientists, 2012b ; Ziebis et al., 2012 ). NP was thus suggested as a model system for studying subsurface basalt-hosted microorganisms ( Bach and Edwards, 2003 ; Edwards et al., 2012 ). Modeling approaches suggested the presence of significant biotic oxygen consumption in the upper oceanic crust ( Orcutt et al., 2013b ). Collectively, the few explorations of the NP crustal biosphere suggested the existence of a unique subsurface biosphere in this system, probably supported by energy produced through iron cycling processes ( Thorseth et al., 2001 ; Bach and Edwards, 2003 ; Edwards et al., 2012 ; Scott et al., 2015 ). To test these assumptions, we analyzed the microbial abundance, diversity and metagenomic properties of basalts collected from basement Hole U1383C with a penetration depth of 324 m below seafloor (mbsf). This is the first study of the vertical distribution of microbial communities in the cool, oxic subsurface oceanic crust, and it provides direct evidence to support the hypothesis that the NP crust hosts a unique biosphere with iron metabolizing potential.",
"discussion": "Discussion Using microscopic cell enumeration and molecular techniques, we characterized the microbiota of subsurface basalts from the young, cool oceanic crust at NP. Our results demonstrate that the microbial abundances in the basalts are less than 6.1 × 10 4 cells cm -3 , with the microbial communities dominated by Gammaproteobacteria and Sphingobacteria ( Figures 1 and 2 ). The microbial abundances of the NP basalts are similar to those of previous measurements of cell densities on subsurface basaltic environments ( Fisk et al., 2003 ; Mason et al., 2010 ; Jungbluth et al., 2013 ), which are more than two orders of magnitude lower than those on seafloor-exposed basalts ( Einen et al., 2008 ; Santelli et al., 2008 ; Jacobson Meyers et al., 2014 ). Higher cell densities are observed at 115–145 mbsf, indicating potential correlations with in situ physical and geochemical characteristics as discussed below. The basalt bacterial community structures at different sampling depths are relatively uniform with numerous bacterial species closely related to cultured iron/manganese oxidizers and environmental clones from various oceanic crustal habitats. Moreover, we identified some bacterial lineages that appear to be localized in NP, indicating that a unique microbial biosphere is hosted in this system. Finally, we suggest that iron-related metabolisms are significant processes in basalt-hosted environments based on comparative metagenomics. Distribution of Microbial Abundance and Composition In contrast to previous studies (e.g., Santelli et al., 2008 ; Jungbluth et al., 2013 ), we provide a more detailed characterization of microbial life in the subsurface oceanic crust across a 254 m core. The distribution of microbial abundance in Hole U1383C did not follow the general trend observed in global subsurface marine sedimentary environments, where cell densities decrease logarithmically with increasing sediment depth ( Kallmeyer et al., 2012 ). This trend in sediments was principally controlled by the availability of energy sources, including buried organic matter ( D’Hondt et al., 2004 ; Lipp et al., 2008 ). Along the core analyzed in our study, higher cell densities were found at depths where higher contents of phosphorus oxide (P 2 O 5 ) and porosity occurred in the rock ( Figure 1 ). For example, the cell density at 145 mbsf was fivefold higher than in the top sample at 72 mbsf and the bottom sample at 324 mbsf; the content of P 2 O 5 and the porosity of the basalt at this depth reached 0.3 and 16.6%, respectively, which were among the highest values in the core ( Expedition 336 Scientists, 2012c ). The co-occurrence of higher cell densities and P 2 O 5 at 115–145 mbsf suggests that phosphorus is an important nutrient that may control the endolithic microbial biomass. Phosphorus is known to be an essential element for microbial growth. This is consistent with recent observations of low phosphate content in subsurface crustal fluids compared to bottom seawater, suggesting that phosphorus is a limiting nutrient in the subsurface crustal biosphere ( Lin et al., 2012 ). Meanwhile, the co-occurrence of higher cell densities and porosity at these depth intervals could be explained by the fact that (1) high porosity provides more potential surface area available for microbial colonization ( Nielsen and Fisk, 2010 ), and (2) high porosity facilitates higher rates of fluid flow through the basalts ( Orcutt et al., 2013b ), supplying higher contents of bioavailable nutrients and/or energy from crustal fluids or bottom seawater. In summary, we see a strong correlation of microbial abundance with basalt P 2 O 5 and porosity, which suggests that the variation in microbial abundance in subsurface basalts is controlled by geochemical and/or physical changes. However, we cannot preclude the possibility of other parameters in situ , due to the challenges of obtaining samples and collecting data as well as the heterogeneous nature of the basalts. For example, because the nitrogen content in the basalt is extremely low (<0.01%; Marty, 1995 ; Busigny et al., 2005 ), nitrogen in the crustal fluids, which is the main source of nitrogen, may decrease over time due to consumption to levels that limit microbial growth. Nitrogen limitation is also indicated in altered basalts from Costa Rica Rift ( Torsvik et al., 1998 ). Considering the large volume of the oceanic crustal habitats ( Orcutt et al., 2011b ), even the relatively low microbial abundances determined in this study suggest that basalt-hosted microorganisms contribute significantly to global biogeochemical cycles. Extrapolating from this limited dataset of microbial abundance on seafloor-exposed basalts and in subsurface basalts (Supplementary Table S1) to the global volume of this habitat, we estimate that the total microbial biomass could match or exceed the total cells estimated in subseafloor sediments ( Kallmeyer et al., 2012 ), which is consistent with a recent hypothesis ( Orcutt et al., 2013a ). Modeling approaches based on assumptions of assumed pore space available in the crust suggest that a much higher cell density in the global crust is possible ( Heberling et al., 2010 ), but more investigations of microbial abundance in the crust are needed to constrain these estimates, including microbiological samples from a broader range of crust type, crustal age and permeability conditions. A similar microbial community was found colonizing the basaltic crust of NP regardless of variations in sampling depth ( Figure 2 ). This result was suggested from a previous study of the subsurface gabbroic crust at the Atlantis Massif ( Mason et al., 2010 ), although the study used a low resolution method based on Denaturing Gradient Gel Electrophoresis (DGGE). This suggests that (1) the geochemical redox zonations of the basaltic rock and its surrounding crustal fluids are similar, showing relatively stable ratios of major electron donors (e.g., reduced iron, hydrogen gas and trace amount of dissolved organic carbon) and acceptors (e.g., oxygen and nitrate; Expedition 336 Scientists, 2012c ; Ziebis et al., 2012 ; Orcutt et al., 2013b ); (2) the microbial communities are homogeneously distributed within the porous and permeable basaltic crust by the advection dominated crustal fluid flow at NP ( Edwards et al., 2012 ), where the basalts are continuously seeded with microbial cells by crustal fluids; (3) the dominance of these groups showed their potential importance to the dynamics of the basaltic microbial community. In addition, we retrieved sequences that are not found in other crustal environments and/or without any cultured representatives (the “NP lineage” in the tree), covering diverse bacterial groups ( Figures 3 – 5 ). This is mainly owing to the recent high-throughput sequencing technology, which substantially extended our view of microbial diversity and potential metabolic functions inhabiting the cool, oxic subsurface basalt-hosted biosphere. For example, we identified a single lineage of Marinobacter Group I, which is distinct from any cultured Marinobacter species, indicating a unique lineage localized in the NP basalts. The first detection of Bacteriovorax -related sequences provides clues that the predatory lifestyle may be an important survival strategy and contribute to microbial biogeochemical cycles in nutrient-starved environments exemplified by NP. Furthermore, the identification of some OTUs that show positive correlation with sampling depth may suggest that they are characteristic of deep biosphere lineages, especially for those forming monophyletic lineages without cultured representatives. Carbon and Energy Metabolism in the Cool, Oxic Subsurface Crust The exact mechanism of autotrophic fixation of CO 2 by basalt microorganisms is uncertain due to the lack of key genes involved in the main autotrophic carbon fixation pathways ( Hugler and Sievert, 2011 ). The presence of O 2 and the δ 13 C-TOC values of basalts (approximately –25‰; Sakata et al., 2015 ) suggest carbon fixation by the Calvin–Benson–Bassham cycle, an aerobic pathway found in most Fe oxidizers ( Emerson et al., 2010 ; Lever et al., 2013 ). However, based on the detection of carbon monoxide dehydrogenase, phosphoenolpyruvate (PEP) carboxylase and pyruvate carboxylase, together with the dominance of facultative chemoautotrophs/mixotrophs, we speculate that the subsurface microorganisms at NP use a mixotrophic pathway to assimilate CO 2 into cellular materials, as proposed elsewhere ( Swingley et al., 2007 ). The metagenomes from the basalt-hosted environments (e.g., Loihi Seamount, JdFR, and NP) are enriched in genes for Fe 3+ uptake, siderophore synthesis and uptake, and unspecified Fe transport pathways (Supplementary Figure S6), suggesting that iron-related metabolism could be significant processes supporting life in subsurface basalts from the cool, oxic subsurface crust at NP. Notably, iron oxidation could be an important energy producing process in the basalts. Our diversity analysis showed that some OTUs obtained from NP basalts were closely related to cultured iron oxidizers ( Figures 3 – 5 ; Edwards et al., 2003 ; Blothe and Roden, 2009 ; Smith et al., 2011 ; Swanner et al., 2011 ; Wang et al., 2012 ; Hirayama et al., 2015 ). For example, a monophyletic clade, Marinobacter Group II, may represent Fe-oxidizing facultative chemoautotrophs based on the phylogenetic data here and elsewhere ( Kaye et al., 2011 ), and an iron-oxidizing Marinobacter strain was isolated from 30R-1A basalts at 304 mbsf (GenBank accession number KJ914666), although known iron oxidation genes [e.g., iro , fox, cyc1 , cyc2 , cox , pio , and rus (as summarized by Ilbert and Bonnefoy, 2013 )] and potential candidate genes [e.g., mtoA ( Liu et al., 2012 ), actB ( Refojo et al., 2012 ), and cyc2 PV -1 ( Barco et al., 2015 )], were not detected in the metagenomes. If iron oxidation is a dominant process as discussed previously, aqueous Fe 2+ is transformed to Fe 3+ at the outer membrane or in the periplasm of microbial cells ( Emerson et al., 2010 ; Ilbert and Bonnefoy, 2013 ). Abundant genes involved in Fe 3+ uptake, siderophore uptake and unspecified Fe transport may facilitate the transport of the Fe 3+ into intracellular materials or the binding to organic complexes, including siderophores ( Sandy and Butler, 2009 ). This may avoid the accumulation of insoluble Fe oxyhydroxide and sulfide minerals on the surface of microbial cells due to the rapid chemical precipitation of Fe 2+ at circumneutral pH. Excess Fe minerals would cause encrustation and cell death for lack of energy and nutrient availability. Moreover, the relatively high abundances of siderophore synthesis genes in basalt-hosted metagenomes helps to produce more siderophores, which would facilitate the dissolution of solid-phase iron minerals ( Kraemer et al., 2005 ) and, in turn, provide more bioavailable iron for microbial energy-yielding activities. Previous work hypothesized that a significant fraction of the iron oxidation in the young upper oceanic crust (<20 million-year old) could support microbial biomass production in subsurface basalts, given that (1) iron is assumed to be the quantitatively most important bioavailable element in the basalt (8 wt%; Melson et al., 2002 ), and (2) the kinetic favorability in low-temperature ridge flank systems ( Bach and Edwards, 2003 ; Edwards et al., 2012 ). Other energy producing processes may also exist in this system. The detection of a hydrogen oxidation gene in the 10R-1B-1 metagenome ( Table 2 ) indicates that microbial life is supported by H 2 sources generated from water-rock reactions (e.g., serpentinization; McCollom and Bach, 2009 ). Heterotrophic metabolism is predicted by our metagenome data. Notably, genes involved in hydrocarbon degradation (e.g., alkane monooxygenase, cytochrome P450, flavin-binding monooxygenase, and catechol-2,3-dioxygenase) were identified, indicating that the microorganisms could use hydrocarbons that originated from crustal fluids or bottom seawater ( Lin et al., 2012 ), diffusion from the overlying sediment ( D’Hondt et al., 2009 ), serpentinization reactions ( Proskurowski et al., 2008 ) or even cell lysates ( Jover et al., 2014 ). In addition, heterotrophic metabolism is supported by the enrichment of functional categories, including fermentation and catabolism of aromatic compounds in NP basalts compared to the rest of the metagenomes listed in Supplementary Figure S5. Heterotrophy is also suggested by previous observations of depleted dissolved organic carbon in subsurface crustal fluids compared with the surrounding bottom seawater in the JdFR flank ( Lin et al., 2012 ) and the detection of hydrocarbon degradation genes in the subsurface gabbroic crust at the Atlantis Massif ( Mason et al., 2010 ). In summary, this study demonstrates that similar microbial communities with relatively low abundance are colonizing the cool, oxic subsurface oceanic crust at NP. Unique microbial communities dominated by Gammaproteobacteria and Sphingobacteria have the potential to play a major role in iron metabolism, which appears to be a significant process in this ecosystem. Although the correlation between microbial abundance and in situ physical and geochemical characteristics is indicated in this study, it is still an open question due to the limited data. In addition, the specific contributions of autotrophy versus heterotrophy in the crustal biosphere are still unclear. Ongoing studies at NP (including laboratory incubations, CORK borehole observatory, RNA-based microbial diversity analyses) and future expeditions [e.g., IODP Expedition 357 at the Atlantis Massif ( Früh-Green et al., 2015 )] may elucidate the variability of microbial abundance and diversity and the balance of autotrophy versus heterotrophy in the oceanic crustal biosphere."
} | 5,169 |
21427286 | PMC3063381 | pmc | 922 | {
"abstract": "Hydrogen gas (H 2 ) is a possible future transportation fuel that can be produced by anoxygenic phototrophic bacteria via nitrogenase. The electrons for H 2 are usually derived from organic compounds. Thus, one would expect more H 2 to be produced when anoxygenic phototrophs are supplied with increasingly reduced (electron-rich) organic compounds. However, the H 2 yield does not always differ according to the substrate oxidation state. To understand other factors that influence the H 2 yield, we determined metabolic fluxes in Rhodopseudomonas palustris grown on 13 C-labeled fumarate, succinate, acetate, and butyrate (in order from most oxidized to most reduced). The flux maps revealed that the H 2 yield was influenced by two main factors in addition to substrate oxidation state. The first factor was the route that a substrate took to biosynthetic precursors. For example, succinate took a different route to acetyl-coenzyme A (CoA) than acetate. As a result, R. palustris generated similar amounts of reducing equivalents and similar amounts of H 2 from both succinate and acetate, even though succinate is more oxidized than acetate. The second factor affecting the H 2 yield was the amount of Calvin cycle flux competing for electrons. When nitrogenase was active, electrons were diverted away from the Calvin cycle towards H 2 , but to various extents, depending on the substrate. When Calvin cycle flux was blocked, the H 2 yield increased during growth on all substrates. In general, this increase in H 2 yield could be predicted from the initial Calvin cycle flux.",
"introduction": "INTRODUCTION Hydrogen gas (H 2 ) is a promising transportation fuel that can be used in hydrogen fuel cells to generate an electric current with water as the only waste product. Anoxygenic phototrophic bacteria, including purple nonsulfur bacteria (PNSB), produce H 2 via nitrogenase ( 1 ). H 2 production is an obligate product of the nitrogenase reaction, which is better known for converting N 2 gas to NH 3 . In fact, nitrogenase will produce H 2 as the sole product in the absence of N 2 . To invoke H 2 production, PNSB are grown under conditions that induce nitrogenase activity, such as by supplying N 2 , or in some cases glutamate, as the sole nitrogen source ( 2 – 5 ). Also, several PNSB mutants have been identified that produce H 2 when grown with NH 4 + as a nitrogen source, a condition that normally represses nitrogenase synthesis ( 4 , 6 – 8 ). These mutants typically have activating mutations in nifA , encoding the master transcriptional activator of nitrogenase, and are termed NifA* strains ( 4 , 6 , 7 ). The preferred mode of growth for PNSB is photoheterotrophy, where light provides energy by photophosphorylation and organic compounds are used for carbon. In a recent study, we found that Rhodopseudomonas palustris cells grown with 13 C-labeled acetate incorporated most of the acetate into cell material but that only half of the reducing equivalents that were generated during acetate oxidation were used in biosynthetic reactions. The bacteria were required to oxidize the other half of the reduced carriers of reducing equivalents (e.g., NADH, NADPH, and ferredoxins, here collectively referred to as electron carriers) by some other means. In the case of acetate, cells accomplished this by carrying out CO 2 fixation via the Calvin cycle or by producing H 2 ( 7 ). Others have shown that the Calvin cycle is essential during photoheterotrophic growth on other substrates, even substrates that are substantially more oxidized than biomass ( 9 , 10 ). PNSB mutants lacking the CO 2 -fixing enzyme of the Calvin cycle, ribulose 1,5-bisphosphate carboxylase (RuBisCO), were unable to grow on malate, succinate, or acetate unless cells were grown under nitrogen-fixing conditions to allow H 2 production ( 7 , 11 ) or unless the electron acceptor dimethyl sulfoxide was provided ( 9 , 10 ). Given the important role for H 2 production in oxidizing electron carriers, one would expect PNSB to produce more H 2 from more-reduced substrates and less H 2 from more-oxidized substrates. However, it has long been known that H 2 yields from PNSB do not always differ accordingly with the substrate oxidation state. In 1977, Hillmer and Gest reported that Rhodobacter capsulatus produced about twice as much H 2 from pyruvate as from glucose, a more reduced substrate ( 2 ). Similar results have been reported for other PNSB ( 3 , 5 ). One factor that certainly affects H 2 yields from different substrates is the amounts of storage products, such as polyhydroxybutyrate, and excreted organic acids produced ( 5 , 12 ). H 2 yields from Rhodobacter sphaeroides also appeared to correlate with the substrate free energy ( 5 ), a surprising result given that H 2 production is not expected to be limited by energy during photosynthetic growth. To identify factors other than substrate oxidation state that influence H 2 production, we performed 13 C metabolic flux analysis with R. palustris provided with organic compounds with a range of oxidation states. We determined metabolic fluxes for the wild type (WT) and a NifA* strain grown anaerobically in light in mineral medium containing NH 4 + as the nitrogen source. The wild-type strain does not produce H 2 under these conditions, whereas the NifA* strain expresses nitrogenase and produces H 2 constitutively. This comparison allowed us to determine the contribution of the Calvin cycle to the growth obtained when nitrogenase is not present compared to the growth obtained when nitrogenase is active and competing against the Calvin cycle for electrons by producing H 2 . Our results illustrate how the biochemical constraints of a metabolic network can affect the H 2 yield when meeting demands for biosynthetic precursors. Our results also show that Calvin cycle activity decreases to different extents, depending on the organic substrate supplied, thereby affecting the H 2 yield by competing for electrons.",
"discussion": "DISCUSSION Balancing electrons is a challenge for a PNSB like R. palustris growing photoheterotrophically because energy is obtained by cycling electrons in cyclic photophosphorylation and not by transferring them to a terminal electron acceptor. During this mode of growth, reducing equivalents that are generated during the oxidation of an organic carbon source, but which cannot be put towards biosynthesis, can be used to fix CO 2 via the Calvin cycle or released as H 2 via nitrogenase. Using 13 C metabolic flux analysis, we found that in the absence of H 2 production, a significant proportion of electron carriers (38 to 55%) were oxidized by the Calvin cycle regardless of the substrate oxidation state. It is interesting that fluxes through the oxidative pentose phosphate pathway ( Fig. 1 ) (G6P → R5P + CO 2 ), the TCA cycle, and PDH/POR were very low for all substrates unless a substrate needed to be processed by one of these pathways to make biosynthetic precursors ( Fig. 1 ). Given the challenge associated with photoheterotrophic growth in maintaining electron balance, the low flux through these reactions makes sense in order to maintain a low rate of electron carrier reduction. In some cases, the biochemical constraints of a metabolic network affect the overall need for electron carrier oxidation, and thereby the amount of CO 2 fixed or H 2 produced, by dictating the route that must be taken towards biosynthetic precursors. Specifically, the route succinate took to generate acetyl-CoA produced a level of reducing equivalents similar to that produced by growth on a more reduced substrate, acetate ( Fig. 2 ). This led to Calvin cycle fluxes and H 2 yields that were unexpectedly high with succinate relative to those with acetate. This effect is expected to be more pronounced when bacteria that have different metabolic inventories are compared. For example, R. sphaeroides assimilates acetate using the reductive ethylmalonyl-CoA pathway ( 14 ), unlike R. palustris , which uses the oxidative glyoxylate shunt. It was recently shown that the ethylmalonyl-CoA pathway oxidizes enough electron carriers during acetate assimilation such that the Calvin cycle and H 2 production were dispensable for photoheterotrophic growth ( 15 ). Given the obligate nature of this reductive pathway in R. sphaeroides for growth on acetate, one would expect that H 2 yields would be lower than those of a bacterium using the glyoxylate shunt. To produce H 2 , the NifA* strain shifted electrons away from CO 2 fixation to H 2 production, such that the necessary electron carrier oxidation was shared by the two activities. However, the Calvin cycle flux decreased to different extents, depending on the substrate ( Fig. 1 and 2 ), and thereby affected the H 2 yield. We verified this by showing that a NifA* ΔRuBisCO strain that is incapable of Calvin cycle flux had higher H 2 yields than the NifA* parent ( Fig. 4 ). Preventing Calvin cycle flux in a Rhodospirillum rubrum NifA* mutant was also recently shown to improve the H 2 yield ( 11 ). It is not clear why there was a greater decrease in Calvin cycle flux in response to H 2 production during growth on acetate than during growth on other substrates. One possibility is that the higher levels of CO 2 produced during growth on fumarate and succinate and the addition of NaHCO 3 to the butyrate cultures allowed for greater participation of RuBisCO type II, which has a low affinity for CO 2 . There are other factors that can also affect the H 2 yield. It is well documented that excretion of organic acids or synthesis of electron-rich polymers such as polyhydroxybutyrate can influence the H 2 yield ( 5 , 12 ). Under our growth conditions, organic acid excretion does not affect the final H 2 yield, since excreted compounds were eventually consumed. Changes to polyhydroxybutyrate and other biomass components (in addition to the potential effects of biphasic growth on labeling patterns) could help explain why we were unable to account for 36% of the electrons in H 2 produced from fumarate by the NifA* strain. We assumed that the biomass composition on fumarate was the same as that observed for growth on succinate and acetate ( 7 ). It was also suggested that the free energy of a substrate can influence the H 2 yield from PNSB ( 5 ). However, our results (and data from others) do not show the same correlation (see Fig. S2 in the supplemental material). Rather, there appears to be a large variability in H 2 yields among PNSB ( Fig. S2 ), barring any influence from the different experimental procedures used. In this paper, we identified two metabolic factors that help explain variable H 2 yields among different PNSB: (i) the route taken to generate biosynthetic precursors and (ii) the amount of competing Calvin cycle flux."
} | 2,718 |
35541202 | PMC9077550 | pmc | 923 | {
"abstract": "A mesophilic (37 °C) and a thermophilic (55 °C) two-chamber microbial fuel cell (MFC) were studied and compared for their power production from xylose and the microbial communities involved. The anode-attached, membrane-attached, and planktonic microbial communities, and their respective active subpopulations, were determined by next generation sequencing (Illumina MiSeq), based on the presence and expression of the 16S rRNA gene. Geobacteraceae accounted for 65% of the anode-attached active microbial community in the mesophilic MFC, and were associated to electricity generation likely through direct electron transfer, resulting in the highest power production of 1.1 W m −3 . A lower maximum power was generated in the thermophilic MFC (0.2 W m −3 ), likely due to limited acetate oxidation and the competition for electrons by hydrogen oxidizing bacteria and hydrogenotrophic methanogenic archaea. Aerobic microorganisms, detected among the membrane-attached active community in both the mesophilic and thermophilic MFC, likely acted as a barrier for oxygen flowing from the cathodic chamber through the membrane, favoring the strictly anaerobic exoelectrogenic microorganisms, but competing with them for xylose and its degradation products. This study provides novel information on the active microbial communities populating the anodic chamber of mesophilic and thermophilic xylose-fed MFCs, which may help in developing strategies to favor exoelectrogenic microorganisms at the expenses of competing microorganisms.",
"conclusion": "5. Conclusions The composition of the anode-attached, planktonic and membrane-attached microbial community, and the active subpopulation, was evaluated in a mesophilic (37 °C) and a thermophilic (55 °C) xylose-fed MFC. This study contributes in understanding of the microbial communities directly and indirectly involved in mesophilic and thermophilic electricity generation. An active microbial community dominated by Geobacteraceae was enriched and shown to sustain power production in mesophilic (37 °C) MFCs, whereas thermophilic (55 °C) power production was hampered by the development of competitors such as hydrogenotrophic methanogens and hydrogen oxidizers. A RNA-based analysis is required to understand the role of the microorganisms in MFCs, as a DNA-based analysis may lead to overestimation or underestimation of the contribution of certain species on power production. A different inoculum source, possibly from thermophilic anaerobic processes, and a different start-up strategy, for example by using a poised anode potential or by suppressing the methanogenic archaea e.g. by addition of bromoethanesulphonic acid (BESA), could be viable alternatives to facilitate the establishment of an efficient thermophilic exoelectrogenic biofilm in future studies. The power production from pure cultures of potentially exoelectrogenic thermophilic microorganisms, for example species of the Thermodesulfobiaceae family detected from the thermophilic anodes in this study, must also be evaluated to confirm their role in electricity production.",
"introduction": "1. Introduction The microbial fuel cell (MFC) is an emerging technology for the direct bioconversion of chemical energy of organic substrates to electrical energy. MFCs consist of two electrodes (anode and cathode) connected through an external electrical circuit. The anode acts as electron acceptor in the bioelectrochemical redox reactions of microbial metabolism, whereas the cathode acts as electron donor for biotic or abiotic reactions. The combination of anodic and cathodic reactions creates a potential difference between the electrodes which drives the electrons to migrate from the anode to the cathode, thus generating electrical current (for a review, see Butti et al. 1 ). Biological electricity production in MFCs requires microorganisms capable to oxidize the substrates and transfer the electrons exogenously to the solid anode electrode. Electrons can be transferred to the anode essentially through three mechanisms: short range, long range, and mediated electron transfer (for reviews, see Kumar et al. 2 and Kalathil et al. 3 ). Some microorganisms, such as Geobacter sulfurreducens , can transfer electrons to a surface directly via redox-active proteins present on the outer surface of their cell membrane, such as c-type cytochromes, or via conductive pili called nanowires. 4,5 G. sulfurreducens develops multi-layer structured biofilms, in which nanowires connect the different cells, enabling the electron transfer to the anode. 6 Mediators, in their oxidized form, penetrate the microbial cell and become reduced during cellular metabolism. They then diffuse out of the cell and release the electrons at the anode, becoming oxidized again and thus reusable. 5 Some species, such as Pseudomonas , produce mediators such as pyocyanin endogenously. 7 Once mediators are produced, also other microorganisms present in the mixed culture system can use them to transfer the electrons to the anode. 8 Pure cultures of electrochemically active microorganisms, such as Geobacter sp. 9–11 and Shewanella sp., 12,13 have shown power production from simple substrates such as volatile fatty acids and sugars at mesophilic conditions (25–37 °C) and neutral pH (6.8–7.3). Mixed cultures are more practical for wastewater treatment, as they contain a consortium of hydrolytic, fermentative and electroactive microorganisms able to produce electricity from complex substrates. 9 However, due to the competition for electron donor with non-exoelectrogenic microorganisms such as methanogenic archaea, 14 power production can remain low, and operational conditions must be optimized to favor exoelectrogenic microorganisms. Catal et al. 15 compared electricity production from 12 monosaccharides present in lignocellulosic biomass, including pentoses and hexoses, in a mesophilic (30 °C) MFC inoculated with a mixed culture adapted to acetate. Xylose resulted in the highest potential for electricity production over the other hexoses and pentoses tested. Thermophilic electricity production could be advantageous because of the high rate of biochemical reactions, and thus high electron production rates, of thermophilic microorganisms. 16 MFCs have been operated at temperatures up to 98 °C. 17 However, although over 20 species of microorganisms, mainly belonging to the Proteobacteria phylum, have been reported to produce electricity under mesophilic conditions, the number of known thermophilic exoelectrogenic microorganisms is much lower. 18 To date, only few species have been reported to produce electricity at thermophilic conditions, including Firmicutes such as Caloramator australicus , 18 Thermincola potens , 19 Thermincola ferriacetica , 20 and Thermoanaerobacter pseudethanolicus , 21 as well as Deferribacteres such as Calditerrivibrio nitroreducens . 22 Investigating the composition of the active subpopulation, rather than the whole microbial community, is crucial in understanding the role of microorganisms in MFCs. DNA-based methods may drive to erroneous conclusions in the detection of the key species in bioreactors. 23 Previously performed microbial community analyses have, nevertheless, mainly targeted the presence of the 16S rRNA gene (DNA) whereas, to our knowledge, only one study 19 has also focused on 16S rRNA gene expression (RNA), which is an indicator of the microbial activity. 23 Furthermore, especially in studies on thermophilic MFCs, microbial community analyses have mainly focused on the anode-attached microbial community, lacking information on the planktonic microbial community. The latter community could be involved in electricity generation as well, either directly, by performing mediated electron transfer to the anode 24 or indirectly, by converting the substrates to compounds readily available for the exoelectrogenic microorganisms. In addition, the membrane is a suitable surface for the establishment of a biofilm. Although biofouling of the membrane has been reported in MFC studies, 14,25 only Lu et al. 26 have reported the composition of a membrane-attached microbial community in two brewery wastewater-fed MFCs operated in series at ambient temperature (20–22 °C). However, the microbial community analysis was performed only at DNA level, and the role of the membrane-attached microorganisms detected on the MFC performance was not discussed. 26 Although likely not directly involved in electricity generation, membrane-attached microorganisms may have a role in the functioning of MFCs, which must be elucidated. Therefore, the aim of this study was to investigate the microbial communities growing (i) as anodic biofilm, (ii) in suspended form in the anodic solution (planktonic), and (iii) as biofilm on the membrane of a mesophilic (37 °C) and a thermophilic (55 °C) xylose-fed MFC. Both presence and expression of the 16S rRNA gene were determined with the aim to investigate both the composition of the overall microbial community and the active subpopulation. Power production, as well as xylose and metabolite concentration profiles were also analyzed to determine the possible differences in the electricity production pathways at 37 and 55 °C.",
"discussion": "4. Discussion 4.1 Bioelectricity production and microbial dynamics in the mesophilic MFC An active microbial community mainly composed of Proteobacteria ( Fig. 4 ) generated a relatively high power density in the mesophilic xylose-fed MFC ( Fig. 1c ). Indeed, most of the known mesophilic exoelectrogens belong to the phylum Proteobacteria . 36 The diversity of the active anode-attached subpopulation (cDNA) was remarkably lower than the diversity of the whole community (DNA) ( Table 1 ), confirming that the presence of microorganisms in a bioreactor does not relate to their activity. 23 In particular, Geobacteraceae accounted only for 2% of the anode-attached microbial community, but was the prevalent (65%) active family ( Fig. 5 ), and likely played a major role in power production. In fact, the Geobacteraceae family includes known exoelectrogenic microorganisms which have been widely reported to dominate the anodic microbial community in mesophilic MFCs, regardless of the inoculum source, substrate, and the MFC set-up. 28,37–39 For example, Mei et al. 40 showed that different microbial communities could develop in mesophilic (30 °C) MFCs started-up with different inocula, but Geobacter was found regardless of the inoculum. In this study, the remarkably higher diversity of the anode-attached community (DNA) than the active subpopulation (cDNA) ( Table 1 ) suggests the presence of inactive or dead microorganisms, which could have hampered the activity of the Geobacteraceae , thus lowering power production. 41 The relative abundance of active planktonic Geobacteraceae was only 3% ( Fig. 5 ), suggesting that they were mainly growing attached to the anode. In fact, Geobacter sp. transfers electrons to the anode by direct contact transfer, but is unable to conduct long-range electron transfer. 42 This is confirmed by the prompt power increase after the addition of xylose at the beginning of each fed-batch cycle ( Fig. 1c ), which is common in MFCs dominated by microorganisms performing direct electron transfer. 20 Sphingobacteriales , found among both the active anode-attached and planktonic subpopulations in the mesophilic MFC (14 and 11% relative abundance, respectively), have been previously reported as part of the anodic microbial community, 39,43 but further studies are required to assess their role in electricity generation. No dominant family was detected in the active mesophilic planktonic subpopulation, but instead 6–7 families were present with a similar relative abundance ( Fig. 5 ). Among them, both Desulfovibrionaceae 44 and Rikenellaceae 45 have been reported to produce electricity as pure cultures in MFCs. Rikenella sp. can perform glycolysis and mediated electron transfer to the anode, 45 which likely explains its presence among the active mesophilic planktonic microbial community in this study ( Fig. 5 ). The Rhodocyclaceae family includes Fe( iii ) reducers, such as Ferribacterium , which can be involved in bioelectricity production 46 and has also been found in an anodic biofilm of an acetate-fed MFC. 47 Porphyromonadeceae , which accounted for 18% of the active mesophilic planktonic subpopulation, have been previously detected both in the anode-attached and planktonic population in a mesophilic MFC treating starch, peptone, and fish extract. 48 Although likely not directly involved in bioelectricity production, other microorganisms may also have contributed to the overall performance of the MFC. For example, Synergistaceae (8% of the relative abundance in the mesophilic active planktonic community) may be involved in the recycling of nutrients by quickly digesting the proteins of dead microorganisms. 38 The membrane-attached active microbial community in the mesophilic MFC was highly diverse ( Table 1 ). Comamonadaceae , which accounted for 20% of the active population, include facultative anaerobic microorganisms capable of using short chain volatile fatty acids as a source of carbon for their metabolism. 49 Species belonging to the Comamonadaceae family, such as Comamonas denitrificans , have been previously found in the anodic biofilm of MFCs, and even shown to produce electricity in the absence of oxygen. 50 However, Comamonadaceae were found in this study exclusively on the membrane, suggesting that they had a minor role in bioelectricity generation. Oxygen can flow from the cathodic to the anodic chamber through the AMI-7001 anion exchange membrane with a diffusivity coefficient of 4.3 × 10 −6 cm 2 s −1 , 51 thus exposing the anodic microorganisms to oxygen. The aerobic or facultative membrane-attached microorganisms may consume the oxygen crossing the membrane, favoring the strictly anaerobic exoelectrogens, but also competing with them for the substrates. Kim et al. 51 estimated that, due to the higher biomass yield of aerobes compared to anaerobes, about 10% of the substrate was consumed through aerobic metabolism, reducing the CE of their acetate-fed (1.2 g L −1 ) MFCs. However, they did not perform microbial community analysis to confirm their hypothesis. Besides, membrane-attached microorganisms may reduce power output also by forming a thick biofilm which limits proton transfer from the anodic to the cathodic chamber. 25 4.2 Bioelectricity production and microbial dynamics in the thermophilic MFC In the thermophilic MFC, the relatively low number of active anode-attached microbial families ( Table 1 ) suggests the scarcity of thermophilic exoelectrogenic species. The inoculum selected for the experiment, which was not previously enriched for thermophilic electricity production, can be one of the causes hindering the establishment of an active exoelectrogenic community. However, the same activated sludge was successfully used to enrich dark fermentative hydrogen producers at 55 °C in a previous study. 27 In addition, 20% of the anode-attached active subpopulation was composed by Firmicutes , which have been previously reported to generate electricity in thermophilic, acetate-fed MFCs. 19 About 66% of Firmicutes found in the thermophilic anode-attached community belonged to the family Thermodesulfobiaceae , which includes Coprothermobacter sp., a proteolytic microorganism involved in the fermentation of organic substrates, with production of pyruvate, formate and acetate, and also in syntrophic acetate oxidation (for a review, see Gagliano et al. 52 ). The activity of Coprothermobacter is enhanced by establishing a syntrophy with hydrogenotrophic methanogenic archaea such as Methanothermobacter . 53 Methanothermobacter belongs to the family of Methanobacteraceae , which was indeed among the most abundant active anode-attached families in the thermophilic MFC in this study ( Fig. 5 ). Although Coprothermobacter was previously found among the anode-attached microbial community of thermophilic acetate-fed MFCs, 19,54 and is thus a possible acetate-utilizing anode respiring bacterium, its electrochemical activity as a pure culture has not yet been investigated. Also microorganisms belonging to the order of Chlorobiales , despite being mainly phototrophs, can perform heterotrophic anaerobic respiration, and have been reported as part of the anodic biofilm in MFCs. 46,55 Anaerolineaceae , also found among the thermophilic anode-attached microbial community, is a family of filamentous bacteria involved in the fermentation of various sugars. 56 They are also involved in the syntrophic oxidation of butyrate, and, similarly to Coprothermobacter , grow better in the presence of H 2 -consuming microorganisms, such as methanogenic archaea. 57 The lower power production in the thermophilic MFC is likely due to the lack of effective exoelectrogens and to the consequent high activity of non-exoelectrogenic microorganisms, which consumed part of the electrons through pathways competitive to electricity generation. In fact, the methanogenic archaeal family of Methanobacteriaceae , belonging to the order of Methanobacteriales , accounted for 38% of the active anode-attached community in the thermophilic MFC. Methanobacteriales lack cytochromes and methanophenazine, and are thus able to perform hydrogenotrophic, but not acetoclastic, methanogenesis. 58 Therefore, Methanobacteriaceae cannot compete for the substrate with exoelectrogenic microorganisms, but their metabolism decreases the availability of electrons for electricity production. Methanobacteriaceae have been previously found in a glucose-fed (1.8 g L −1 ) MFC operated at room temperature, and indicated as one of the causes for low bioelectricity production, as about 16% of the electrons were directed to methane production. 14 Rismani-Yazdi et al. 59 reported methane production by Methanobacteriaceae in a mesophilic (39 °C) cellulose-fed MFC only at the beginning of the operation, whereas Hussain et al. 60 reported Methanobacteriaceae in a thermophilic (50 °C) syngas-fed MFC. Such microorganisms likely decreased the efficiency of their MFC by performing hydrogenotrophic methanogenesis. The family of Hydrogenophilaceae , which accounted for 46% of the active planktonic community in the thermophilic MFC, includes the thermophilic Hydrogenophylus sp., which could have consumed a share of electrons by H 2 oxidation, 61 lowering power production in the thermophilic MFC. Thermodesulfobiaceae , found among the anode-attached families, were also found among the planktonic community ( Fig. 5 ). Coprothermobacter is able to perform extracellular electron transfer, 52 but further studies are required to understand its possible involvement in long-range electron transfer to the anode. In the thermophilic MFC, the family of Comamonadaceae was the most abundant membrane-attached family and, similarly to the mesophilic MFC, it was likely related to aerobic metabolism and thus, oxygen consumption. Armatimonadetes , which accounted for 17% of the active membrane-attached community, is also an order of aerobic microorganisms. 62 4.3 Xylose degradation pathways In the mesophilic MFC, the xylose consumption and metabolite production profiles ( Fig. 2 ) suggest that xylose was firstly converted to volatile fatty acids, which were subsequently oxidized to CO 2 and H 2 O likely mainly by Geobacteraceae , which dominated the anode-attached active community. Interestingly, the power density remained stable for about 30 hours after the depletion of acetate and butyrate. A possible explanation is that acetate and butyrate were accumulated and oxidized intracellularly, thus not detectable in the anolyte and resulting in a flow of electrons directed outside the cell to the anode. 4 In fact, after substrate depletion, the soluble COD remained stable ( Fig. 2 ), suggesting that the electron donor was not in the anolyte but likely inside the cells. Also Marshall and May 20 observed the same phenomenon and decided to starve a pure culture of Thermincola for two cycles before electrochemical measurements to avoid interferences from the intracellularly accumulated acetate, and its associated storage products. In the thermophilic MFC, xylose was consumed relatively fast, but acetate, the only metabolite found in the anolyte, was not fully consumed even after 144 hours. The power density peak obtained just after the xylose depletion suggests that exoelectrogenic thermophiles were growing on xylose, but the microbial community was lacking effective acetate-utilizing microorganisms. However, it should be noted that the profiles in Fig. 2 were obtained in the feeding cycle “IX”, whereas the samples for microbial community analysis were collected at the end of cycle “XI”. The anodic potential, which increased from cycle IX to cycle XI in the thermophilic MFC ( Fig. 1b ), suggests a possible shift in the microbial community. 4.4 Performance of the MFCs In the mesophilic MFC, the shape of the polarization curve (the stable slope in the last part of the curve) suggests low mass transfer limitation, as expected in MFCs using soluble sugars as the substrate. The low CE (12 and 3% for the mesophilic and thermophilic MFC, respectively) was attributed to the MFC design, which was not optimized for power production. The slow rate of oxygen reduction in the cathodic surface and the low proton conduction through the membrane are often the main causes of low power production in air-cathode MFCs. 63 In fact, a CE up to 82% was obtained in a xylose-fed, two-chamber MFC (75 mL anodic chamber volume) using 50 mM ferricyanide for the cathodic reaction and a cation exchange membrane. 64 Haavisto et al. , 28 with a similar inoculum and substrate, obtained an 18% higher CE than the one obtained in this study operating a mesophilic (37 °C) upflow microbial fuel cell in continuous mode using ferricyanide at the cathode. Huang and Logan 65 obtained a power production of 13 W m −3 (61% CE) using a xylose-fed air cathode MFC, against the 1.1 W m −3 (12% CE) obtained in this study. However, the anodic chamber of their MFC was equipped with four carbon brushes (6 cm diameter and 7 cm length each), against the single carbon brush (1.5 cm diameter and 5 cm length) used in this study, and their xylose load was three times higher. The structure of the active microbial community in the thermophilic MFC, lacking a known effective exoelectrogen such as Geobacter and including competitors such as methanogenic archaea, was likely the main cause for the lower power produced from the thermophilic MFC in comparison to the mesophilic MFC ( Fig. 3a ). In fact, the non-exoelectrogenic anode-attached microbial community in the thermophilic MFC likely caused a high internal resistance (560 Ω). Temperature also affects oxygen solubility in water, resulting in a decreased availability of oxygen at high temperature. In fact, the oxygen concentration at the cathode was about 7.0 and 5.6 g L −1 in the mesophilic and thermophilic MFC, respectively. In the thermophilic MFC, the power overshoot curve ( Fig. 3b ), previously reported in MFCs, 66,67 prevented the detection of possible mass transfer limitations. A multiple-cycle method, consisting in running the MFC at a fixed resistance for one entire batch cycle, can be applied to avoid overshoot. 68"
} | 5,894 |
38534834 | PMC10968063 | pmc | 925 | {
"abstract": "The gecko can achieve flexible climbing on various vertical walls and even ceilings, which is closely related to its unique foot adhesion system. In the past two decades, the mechanism of the gecko adhesion system has been studied in-depth, and a verity of gecko-inspired adhesives have been proposed. In addition to its strong adhesion, its easy detachment is also the key to achieving efficient climbing locomotion for geckos. A similar controllable adhesion characteristic is also key to the research into artificial gecko-inspired adhesives. In this paper, the structures, fabrication methods, and applications of gecko-inspired controllable adhesives are summarized for future reference in adhesive development. Firstly, the controllable adhesion mechanism of geckos is introduced. Then, the control mechanism, adhesion performance, and preparation methods of gecko-inspired controllable adhesives are described. Subsequently, various successful applications of gecko-inspired controllable adhesives are presented. Finally, future challenges and opportunities to develop gecko-inspired controllable adhesive are presented.",
"conclusion": "6. Conclusions and Outlook Since the discovery of the fact that the gecko’s excellent wall-climbing ability comes from the van der Waals forces generated by the microscopic setae of the paws in intimate contact with the wall surface, people have been devoted to the study of gecko-inspired dry adhesives. In this review, the microstructure of the gecko’s feet was introduced, a hierarchical composite adhesive structure consisting of thin plates, setae, and spatulas that can be adapted to the microscopic morphology of the wall surface and generate strong van der Waals forces by making intimate contact with the wall surface. Next, the controlled adhesion properties of the gecko were introduced in relation to the gecko’s macroscopic regulation of the toe muscles and the shear motion of the setae. Then, the design and preparation of gecko-inspired controllable adhesives were highlighted. Finally, we presented several important applications of gecko-inspired controllable adhesives. As shown in Figure 19 , for the structural design of gecko-inspired controllable adhesives, the design of anisotropic microstructures (such as wedges) is a direct means to achieve controllable adhesion. Adding tip features such as spatulas and mushrooms to the microstructures can improve the adhesion force. In addition, adding magnetizable particles in polymers, shape memory polymers as materials, other methods of preparing gecko-inspired controllable adhesives that can be in the magnetic field, temperature, and other active control methods to make the adhesive micro-geometry deformation or backing layer stiffness change, thus realizing the gecko-inspired adhesive state of active switching. Gecko-inspired controllable adhesives have been developed for more than a decade, and their adhesion and controllable performance have been greatly improved, but there are still many challenges to fully reproduce the gecko’s strong wall-climbing ability. First, the current design for the microscopic geometry of gecko-inspired controllable adhesives is mostly determined by experience and intuition, and lacks a scientifically optimized structural design scheme to optimize the adhesion performance of the adhesion surface. Therefore, by combining machine learning and the design of adhesion surface microstructures, people can study the optimal design of an adhesive with more types of microstructures in the future. Second, the preparation efficiency of the existing gecko-inspired controllable adhesives is low, so how to realize the efficient preparation of large areas is an issue that should be considered in the future. Finally, the adhesion system of geckos is complex. Therefore, how to integrate all the adhesion mechanisms of the gecko into one product will be the future research direction of gecko-inspired controllable adhesives. Gecko-inspired controllable adhesives have a wide range of applications in the field of climbing robots and robotic gripping tasks. However, existing gecko-inspired climbing robots are large compared to geckos in terms of their structural size and weight, and can only climb on smooth wall surfaces. The gripper’s grasping objects also have more than just smooth surfaces. Therefore, improving the self-cleaning, durability, and adaptability to rough wall surfaces of gecko-inspired controllable adhesives will be a long-term challenge.",
"introduction": "1. Introduction Since ancient times, the natural world has been the source of various technological ideas, engineering principles, and major inventions for human beings. For example, frogs have highly regular toe pad microstructures that allow them to climb on slippery surfaces like moist leaves or tree trunks [ 1 , 2 , 3 ], octopuses can capture prey of various sizes stably underwater [ 4 , 5 , 6 ], and mussels can adhere to rocks [ 7 , 8 , 9 ]. By gaining an in-depth understanding of the physiological characteristics of these organisms, researchers have developed synthetic adhesives with unique properties, which enable them to replicate the surface characteristics of living organisms [ 10 ]. In the animal kingdom, geckos are renowned for their excellent climbing ability. They can crawl or run effortlessly on the ground, walls, and ceilings. However, it was only a decade ago that the extraordinary climbing ability of geckos was revealed. Geckos can achieve adhesion and detachment of their feet to the surface in a matter of milliseconds, a capability known as controllable adhesion. Since the remarkable controllable adhesion ability of gecko feet was discovered, scientists have extensively studied their microstructure and controllable adhesion mechanism. Drawing inspiration from the structure and controllable adhesion mechanism of gecko feet, researchers have developed imitation-gecko controllable adhesives and conducted extensive research on the key factors affecting their controllable adhesion performance. Despite the significant research achievements made by scientists in this highly innovative field, bio-inspired controllable adhesives still face many challenges and unresolved issues. In order to further promote the research of gecko-inspired controllable adhesives and improve their practical applicability, the latest progress in gecko-inspired controllable adhesives, along with their application, is reviewed. First, the adhesion mechanism of the gecko’s feet is introduced from three aspects: the structure of the gecko’s feet, the source of the adhesion force, and the behavior of the gecko’s adhesion system in the process of attachment and detachment. Then, design methods of gecko-inspired controllable adhesives based on shear adhesion are summarized from the perspective of structure, and their adhesion performances are summarized. Later, active controllable adhesion strategies based on SMP (shape memory polymers) microstructures, magnetic microstructures, and controllable back layers are introduced, and the representative fabrication methods of the gecko-inspired controllable adhesive including photolithography, ultraprecision machining, and 3D (three dimensional) printing are presented. After that, the applications of gecko-inspired controllable adhesives in climbing robots and gecko grippers are demonstrated. Finally, the future development direction of gecko-inspired controllable adhesives is predicted."
} | 1,863 |
29021783 | PMC5623813 | pmc | 927 | {
"abstract": "The production of extracellular polymeric substance (EPS) is important for the survival of biofilms. However, EPS production is costly for bacteria and the bacterial strains that produce EPS (EPS+) grow in the same environment as non-producers (EPS−) leading to competition between these strains for nutrients and space. The outcome of this competition is likely to be dependent on factors such as initial attachment, EPS production rate, ambient nutrient levels and quorum sensing. We use an Individual-based Model (IbM) to study the competition between EPS+ and EPS− strains by varying the nature of initial colonizers which can either be in the form of single cells or multicellular aggregates. The microbes with EPS+ characteristics obtain a competitive advantage if they initially colonize the surface as smaller aggregates and are widely spread-out between the cells of EPS−, when both are deposited on the substratum. Furthermore, the results show that quorum sensing-regulated EPS production may significantly reduce the fitness of EPS producers when they initially deposit as aggregates. The results provide insights into how the distribution of bacterial aggregates during initial colonization could be a deciding factor in the competition among different strains in biofilms.",
"conclusion": "Conclusions Microbial competition between two bacterial strains with differing EPS producing characteristics (EPS+/QS+/QS− vs. EPS−), has been studied using an IbM, with one strain initially deposited on the substratum as aggregate(s) and the other as individual cells. The results show that when there is no quorum sensing and if EPS− cells attach as relatively large aggregates; then the EPS+ cells gain the maximum competitive advantage if they attach on the substratum as single cells (under the condition that the EPS+ strain invests about half of their energy in EPS production). Xavier and Foster ( 2007 ) and Nadell et al. ( 2008 ) also showed that the optimum investment in EPS is around 0.5, when EPS+ compete with others that invest either more or less in EPS production (in these studies, both strains are initially deposited as single cells on the surface, similar to the control case in this paper). However, when the EPS+ strain is deposited in relatively small clusters and the EPS− strain is deposited as single cells, then the EPS+ bacteria always benefit from producing EPS regardless of the level of energy invested in EPS. According to this simulation, as the EPS+ aggregate size decreases they need to expend less energy on EPS production ( f < 0.5) to gain the maximum fitness advantage. Quorum sensing-regulated EPS production is found to provide no significant advantage over continuous EPS production for all of the cell deposition scenarios, for the range of parameters chosen for the present study. Our numerical results indicate that quorum sensing-regulated EPS production significantly reduces the competitive advantage gained by matrix producers when they deposit as aggregates and compete with single cells of EPS− or vice-versa. Laser-diffraction particle-size scanning tests have shown that 90% of the total planktonic biomass of Pseudomonas aeruginosa consist of cellular aggregates in the size range of 10–400 μm (Schleheck et al., 2009 ). Therefore, it is inevitable that single cells deposited on a surface will compete with different sizes of bacterial aggregates of P. aeruginosa which are deposited on the same surface. Our simulation results may give an insight into this competition because the present results indicate that the aggregate size plays a significant role in the competition with single cells. In vitro experiments of Kragh et al. ( 2016 ) have shown that aggregates of P. aeruginosa gain a competitive advantage over their single cells when competing in the same environment. These experiments could be extended to investigate the effects of different EPS production characteristics and different aggregate sizes on microbial competitions in biofilms, and then our predictions could be tested. Wessel et al. ( 2014 ) used a gelatin based three-dimensional printing strategy to make different sizes of P. aeruginosa aggregate and showed that when the aggregate size exceeds a critical size, localized oxygen depletion regions were found inside the aggregate. These in vitro experimental results show that the growth rate decreases as the aggregate size increases which is consistent with our findings. Although the experimental and simulation results based on continuous model have general agreements, there was some discrepancy due to simplifying assumptions including uniform oxygen consumption throughout the aggregate. However, an Individual-based modeling technique similar to the present study should give more comparable results to these experiments because the IbM can capture heterogeneities inside aggregates more accurately. The present simulation techniques can also be adapted to study the interaction of bacteria such as Sinorhizobium meliloti that forms aggregates (Dorken et al., 2012 ) with other species during the wastewater treatment process (Ben Rebah et al., 2002 ). Even though it is widely believed that public goods producing bacteria are benefited by quorum sensing-regulated gene modulations, our numerical results show that quorum sensing can also have detrimental effects on public good producers. However, these numerical simulations need to be extended to cover a wider range of parameters and be experimentally tested to draw a solid conclusion about these findings. In the present Individual-based Model, factors such as bacteria motility, founder cell density, detachment and attachment etc. are not considered and these developments can form the basis for future work. Moreover, for biofilms growing in a flow environment, the mechanical strength of the biofilm mediated by the EPS composition can provide insights into the biological evolution of polymer producing strains. The flow can also advect quorum sensing signals which can cause the bacteria to misread their local cell density, thereby influencing bacterial competition in constricted geometries, for example in the pores of the soil.",
"introduction": "Introduction Biofilms are surface associated communities of bacteria that are surrounded by adhesive extracellular polymeric substance (EPS) (Davey and O'toole, 2000 ) which not only provides them with mechanical integrity but also allows resistance against attack from foreign entities. Understanding the dynamics of growth and competition between several microbial species in a biofilm is crucial for our understanding of chronic diseases such as cystic fibrosis, infection in medical devices, biofouling and various processes used in wastewater treatment. Mathematical models such as Cellular Automaton (CA) and Individual-based Models (IbMs) (Kreft et al., 2001 ; Picioreanu et al., 2004 ; Xavier et al., 2005 ; Nadell et al., 2008 ; Lardon et al., 2011 ; Jayathilake et al., 2017 ) have been instrumental in providing insights into the spatiotemporal growth and competition of microbes under varying conditions. Kreft et al. ( 1998 ) proposed the use of IbM as a bottom-up approach which attempts to predict community behavior based on the actions and characteristics of the constituent individuals. The IbM was introduced to cope with artifacts which occurred due to the discrete displacement of biomass in CA (Picioreanu et al., 2004 ; Tang and Valocchi, 2013 ). As Ib modeling leads to more realistic biofilm structures (Kreft et al., 2001 ), it has been widely used to study social evolution in biofilms (Kreft, 2004 ; Xavier and Foster, 2007 ; Nadell et al., 2008 ; Mitri et al., 2011 ). Kreft ( 2004 ) used IbM to study competition between the rate and yield strategists in biofilms and concluded that certain spatial structures are needed for maintenance of yield strategists. The rate strategists are found to dominate the biofilm in the short-term due to their high growth rates, while in the long run the yield strategists dominate since they consume nutrients more economically. Nadell et al. ( 2010 ) studied competition between enzyme secreting and non-secreting bacteria under different ratios between nutrient provision and nutrient consumption, and found that if the ratio is small, cell (bacteria) lineage segregation occurs and consequently the cooperative cells (i.e., enzyme-secreting cells) dominate within the biofilm. The cell lineage segregation confers an advantage to the cooperative cells because they are not exploited by non-cooperative ones. Mitri et al. ( 2011 ) found that addition of new species in a multispecies biofilm especially in resource limited scenarios would reduce the fitness of existing cooperative cells that secrete public goods. In addition, the ecological advantages of quorum sensing (QS) -regulated enzyme production (Schluter et al., 2016 ), QS inhibition (Wei et al., 2016 ) and evolution of bacteriocin production (Bucci et al., 2011 ) in biofilms have also been investigated using IbM. EPS mediated adhesion is known to be very important for bacterial biofilm development as it affects both the initial attachment to surfaces and the subsequent resistance to shear flows. However, bacterial adhesion to surfaces ought to be costly because it restricts bacteria mobility and hinders movement to nutrient rich environments. Schluter et al. ( 2015 ) studied the effect of EPS mediated adhesion and found that cells with greater adhesive capabilities gained a competitive advantage when nutrients are abundant. Xavier and Foster ( 2007 ) showed that cells that constitutively produce EPS (EPS+) outcompete non-producers (EPS−) in the presence of significant nutrient gradients. When the EPS+ and EPS− strains are co-cultured in a biofilm, the EPS+ cells initially grow slower than EPS− cells because the EPS+ cells spend a fraction of energy on EPS production, and therefore the EPS− bacteria would initially dominate in the biofilm. However, eventually the production of EPS would help the EPS+ cells to push their descendants into nutrient rich top layers and hence the progeny of EPS+ bacteria would get more access to nutrients and would dominate in the biofilm in long run. Quorum sensing (QS) is a cell-cell communication mechanism used to regulate gene expression and production of public goods in biofilms (Fuqua and Greenberg, 2002 ). Nadell et al. ( 2008 ) investigated the competitive advantage of quorum sensing-mediated down regulation of EPS production. They found that EPS producers under negative quorum sensing control (i.e., EPS production by bacteria stops at high cell densities, referred to as the QS− strain), would dominate when competing with EPS+ strain. However, this effect only lasts for a limited time and the EPS+ cells dominate in the long-term because EPS+ cells suffocate the QS− cells by continuously secreting polymeric substance thereby separating QS− cells from nutrients. These studies demonstrate that spatial distribution of microbes influences the microbial competition in biofilms. In addition nutrient gradients have been known to cause cell lineage segregation in biofilms and the effect has been addressed in many papers (Xavier and Foster, 2007 ; Nadell et al., 2010 ). Generally, low nutrient conditions favor cooperative strains (or species) that produce public goods such as EPS and enzymes. The biofilm structure is also influenced by other factors including microbial mobility, adhesion, initial attachment frequency and bacteria re-attachment to the biofilm (van Gestel et al., 2014 ); however, the effect of these factors on microbial competition in biofilms has not been extensively investigated. For example, when a biofilm grows in a reactor, it can experience erosion and sloughing due to hydrodynamic shearing and the detached biofilm clusters can re-colonize new surfaces and develop into biofilms. Similarly, the aerobic granular sludge aggregates found in sequencing batch reactors can be transported to new locations and have the ability to colonize new surfaces (McSwain et al., 2005 ). It is therefore very likely that bacterial aggregates deposit on new surfaces, hence biofilms originate from both individual cells (single cells) and cell clusters (aggregates). Only recently, Melaugh et al. ( 2016 ) and Kragh et al. ( 2016 ) addressed a similar problem by performing IbM simulations to understand the trade-off between aggregate surface area and relative height compared to single cell colonizers. The findings suggest that single cells perform better when competition is low (i.e., at low single cell densities) and multicellular rounded aggregates perform better when competition is high (i.e., at high single cell densities). In more competitive environments the aggregates perform better because they have access to nutrient rich areas due to their initial height advantage compared to single cells. This trade-off is likely to be influenced by EPS production characteristics of cells because EPS provides cells with sufficient structure to reach high nutrient layers. Moreover, multispecies biofilms may contain strains of bacteria that can either be EPS+ or EPS−. Therefore, EPS production characteristics of cells might offset the competitive advantage gained by bacterial aggregates due to their height. In the present study, we develop a two-dimensional biofilm model based on IbM principles to understand competition between cells and aggregates which express a combination of characteristics (EPS+, EPS−, QS+ and QS−, described under “Methods” below). We simulate the spatiotemporal dynamics of competition under various scenarios of attachment (i.e., as single cells or multiple aggregates) and for different values of energy invested in EPS production by the microbes. The maximum competitive advantage is obtained when the EPS+ cells are initially deposited on the substratum as smaller aggregates and are randomly distributed among individual cells of the EPS−. We also study the effect of quorum sensing- regulated EPS production on competition between single cells and aggregates for different values of QS signal threshold. Overall, the work demonstrates the role of EPS production in conferring an advantage to either single cells or aggregates as they form biofilms under differing conditions.",
"discussion": "Results and discussion In the following sections, the competition between various strains of bacteria (EPS+/QS+/QS−/EPS−) are investigated for a period of 12 days given that they initially attach on the surfaces as either cells or aggregates. Competition between EPS producing (EPS+) and EPS non-producing (EPS−) strains Competition between EPS+ strain and EPS− strain when they initially deposit on the substratum as individual cells has been studied by Xavier and Foster ( 2007 ). A similar case is reproduced here as a control. The bacteria (50 EPS+ and 50 EPS−) are randomly inoculated on the substratum and all of the cells have equal access to substrate (at t = 0 s). Figure 1 shows the biofilm formation for different values of investment in EPS production (different f -values). It is seen that if there is no investment in EPS ( f = 0, Figure 1B ) both species grow identically and there is no competitive advantage for either. However, if energy investment in EPS production is relatively high ( f = 0.6, Figure 1D ), EPS+ strain is outcompeted by EPS− strain. At an intermediate fraction of energy investment ( f = 0.2, Figure 1C ), EPS+ cells dominate in the biofilm. The variation of relative fitness of EPS+ cells as a function of investment in EPS ( f ) and EPS material density ρ EPS is shown in Figure 2A . If the density of EPS decreases compared to the density of bacteria (i.e., ratio ρ/ρ EPS increases) it is advantageous for EPS+ cells since the volume of polymeric substances expands faster. This results in the EPS+ strain being pushed into substrate rich environments while EPS− cells are starved. Xavier and Foster ( 2007 ) also briefly demonstrated that the amount of substrate plays a vital role in the competition between EPS+ and EPS− strains. We find that the relationship between the ratio of the fitness of EPS producers to non-producers for different values of EPS investment (0 < f < 0.6) is unimodal for density ratio ρ/ρ EPS > 2.2 ( t = 6.745, P = 0 and t = −9.809, P = 0 respectively for the linear and quadratic terms for investment in EPS, f , more details about GLM are in Supporting information), indicating that above a certain threshold of investment in EPS the relative fitness of EPS producers declines. For low density ratio conditions (ρ/ρ EPS < 2.2), the relative fitness of the EPS+ strain declines with increased investment in EPS. Figure 1 Competition between EPS+ and EPS− strains when both strains are initially randomly inoculated on the substratum: (A) initial inoculation of bacteria; (B) biofilm after 12 days at f = 0; (C) biofilm after 12 days at f = 0.2; (D) biofilm after 12 days at f = 0.6. It is seen that both strains co-exist in the biofilm if there is no EPS production, EPS+ strain dominates at f = 0.2 and EPS− strain dominates at f = 0.6. The contour plot shows the nutrient level from low to high as white to black. All values are in SI units. Figure 2 Effect of different parameters on the fitness of EPS+: (A) Fitness of EPS+ strain relative to EPS− strain as a function of investment in EPS ( f ) and biomass to EPS density ratio (ρ/ρ EPS ). It is seen that if the density ratio is high it is advantageous for EPS+ strain and also there would be an optimum f -value which gives the maximum benefit for EPS+ strain. If the EPS density is relatively low, EPS+ cells are easily outcompeted by EPS− cells since EPS+ cells cannot push their progeny fast into the nutrient rich upper levels. The lines are the polynomial fits to the corresponding data points and the error bars indicate the standard deviations; (B) relative fitness of EPS+ strain relative to EPS− strain as a function of δ and κ which are two non-dimensional parameters appeared in the nutrient transport equation. It is clear that EPS+ are not beneficial at high values of δ and κ since the heterogeneity of nutrient concentration is less in this case and hence both strains are mixed in the biofilms rather than making own lineages. The lines are the polynomial fits to the corresponding data points and the error bars indicate the standard deviations. To better understand the trade-off due to substrate limitation and bacteria growth, we direct our attention to the nutrient transport equation. (The density ratio for the following simulations is ρ/ρ EPS = 6.67 which is estimated from the parameters in Table 2 ). For our model, the substrate gradients are determined by Equation (2), which can be re-written in non-dimensional form as: (4) ∂ S * ∂ t * = δ 2 ∇ 2 S * - S * κ + S * X * where S * = S / S b is the non-dimensional concentration and S b denotes the bulk substrate concentration. δ = D S Y X S S b μ max ρ L 2 and κ = K S S b are non-dimensional parameters, and ρ and L are biomass density and substrate concentration boundary layer thickness, respectively. The dimensionless parameter δ (Nadell et al., 2010 ) represents the ratio between the maximum rate of substrate transport and maximum rate at which substrate is consumed by bacteria. The biological meaning of κ is subtle: it expresses the affinity of the bacteria for a substrate in the context of given bulk substrate concentration. It can be deduced from Equation (4) (if we only consider the y direction) that the steady state substrate transport is given by d 2 S * d y * 2 = 1 δ 2 S * κ + S * X * , and thus the substrate gradient across the biofilm is d S ∗ d y ∗ = 1 δ 2 X ∗ ( S ∗ − κ ln ( S ∗ + κ ) + C , where C is a constant. It is obvious that the substrate gradients are negatively correlated with κ and δ. Increasing the value of either parameter would decrease substrate gradients and therefore result in substrate rich conditions throughout the biofilm. Figure 2B shows that when κ and δ increase, the EPS− strain easily outcompetes the EPS+ strain due to smaller substrate gradients across the biofilm. If κ is very high (κ = 7), the EPS− strain outcompetes the EPS+ strain regardless of δ. Increasing either parameter results in substrate rich conditions throughout the biofilm and results in a lack of lineage segregation in the biofilm. Since the EPS+ and EPS− strains are well mixed in the biofilm network the EPS+ cells can be exploited by EPS− cells. We were inspired by Nadell et al. ( 2013 ) to derive a simple relationship analogous to Hamilton's rule for the competition between EPS producers and non-producers to show that our model predictions (Figures 1 , 2 ) are consistent with this rule. According to Hamilton's rule (Hamilton, 1964 ), a cooperative strategy, such as EPS production, will evolve if rB > C , where r, B and C are relatedness (measure of genetic similarity of the neighboring cells to the focal cell), fitness benefit, and fitness cost, respectively. The growth rate of a EPS+ cell can be written as d m E P S + d t = [ μ 0 ( 1 + B ) - f ] m E P S + , where B is the additional benefit gained by the cell because the cell is advected to high nutrient layers by the polymeric substances and μ 0 is the specific growth rate of the cell. A nearby EPS− cell will also be benefited by EPS production depending on how far that cell resides from the EPS+ cell. If we assume this EPS-mediated benefit is inversely proportional to the distance from the EPS+ cell ( d ), the growth rate of EPS− cell can be written as d m E P S - d t = [ μ 0 ( 1 + B r E P S + / d ) ] m E P S - , where r EPS + is the radius of EPS+ cell. The EPS+ cell will outcompete EPS− cell if EPS+ cell has higher fitness and therefore: (5) 1 m E P S + d m E P S + d t > 1 m E P S - d m E P S - d t which gives that if [ μ 0 ( 1 + B ) − f ] > [ μ 0 ( 1 + B r E P S + / d ] . Therefore, the cooperative strategy will evolve if: (6) ( 1 - r E P S + / d ) B > f / μ 0 The condition given in Equation (6) is analogous to Hamilton's rule, rB > C , with r = 1 − r EPS + / d and C = f / μ 0 . According to Equation (6), when f increases the relationship will fail at a point where the EPS− strain would outcompete the EPS+ strain. Figures 1 , 2 clearly show this behavior. Equation (6) also indicates that EPS+ cells will dominate if EPS− cells are far away from the growing EPS+ cells ( d >> r EPS + , meaning that relatedness is high). Figure 2B shows similar behavior, the EPS+ strain dominates when there is lineage segregation (for low κ and δ) and EPS− strain dominates when the two strains are mixed (for high κ and δ). Despite the simplicity of the current Ib model, it can predict the competition between polymer producers and non-producers in biofilms which is akin to Hamilton's rule. Competition between aggregates and cells (with EPS+/EPS− characteristics) In reality, biofilms can be initiated by a mixture of single cells and aggregates. If there is a steep nutrient gradient across the biofilm (i.e., small κ and δ values), the initial colonization pattern (i.e., excess of aggregates or single cells) could have a profound effect on the fate of the biofilm inhabitants. Two recent studies (Kragh et al., 2016 ; Melaugh et al., 2016 ) that did not consider EPS production, found that bacteria attaching as aggregates would have a competitive advantage over single cells; as the height of the former gives better access to resources. This competition can be directly influenced by over expression of EPS in the aggregates which can provide them with even greater access to resources and thereby an even greater advantage. To investigate such scenarios we modeled the competition between EPS+ and EPS− bacteria when they attach on the substratum as either circular aggregates or individual cells. We start the investigation by considering two different scenarios for the initial cell and aggregate attachment on the substratum:\n (i) Case 1: EPS+ bacteria are deposited as aggregates and EPS− bacteria are distributed as single cells. We consider the case in which EPS+ and EPS− cells deposit on the substratum as aggregates and single cells, respectively. The initial number of aggregates is varied between 1, 2, and 5 such that the cell number ratio between two strains is always 1:1. Therefore, as the number of aggregates increases, the size of each aggregate decreases accordingly (Figure S1 ). Given the pattern of initial colonization, EPS+ aggregates should have two distinct advantages: as the aggregates produce EPS they can suffocate EPS−, and they can use their height advantage to obtain improved access to substrate. Figure 3A shows biofilm growth when EPS+ strain deposits as a single aggregate. We find that EPS+ strain grows as a single tower and the growth of EPS− cells is inhibited. The population density of EPS+ cells in the EPS matrix decreases as f (the fraction of energy devoted to EPS production) increases. (ii) Case 2: EPS+ bacteria are spread out as single cells and EPS− bacteria are deposited as aggregates. Figure 3 Initial colonization of one aggregate and biofilm after 12 days for different level of energy investment in EPS ( f ): (A) Case 1, EPS+ cells are initially aggregated while EPS− cells are randomly spread out on the substratum. It is seen that EPS+ strain dominates in the biofilm for all the cases; (B) Case 2, EPS+ cells are randomly spread out on the substratum while EPS− cells are initially aggregated. It is seen that EPS+ strain dominates in the biofilm for all the cases. All values are in SI units. We consider the situation in which the EPS− and EPS+ cells deposit on the substratum as aggregates and single cells, respectively. Similar to Case 1, the number of aggregates is varied as 1, 2, and 5 while maintaining the 1:1 ratio between the strains. Even though EPS− cells do not produce EPS, they are still likely to aggregate due to pili-pili interactions between bacterial cells (Ponisch et al., 2017 ). Aggregates of EPS− may have a competitive advantage over EPS+ cells due their height and better access to nutrients, however the EPS+ cells may gain a competitive advantage by producing EPS. Figure 3B shows the results when EPS− strain deposits as a single aggregate. We find that, although EPS− cells are initially aggregated and have some competitive advantage due to height, EPS+ cells always dominate in the biofilm. As the energy invested in EPS production is relatively high ( f = 0.6), the EPS− tower is surrounded by the polymeric matrix due to rapid EPS production and hence EPS− aggregate is not able to access nutrients. The variation in the relative fitness of EPS+ cells for both cases (i and ii) is shown in Figure 4 . At relatively low values of EPS investment ( f < 0.25), starting as a single aggregate (EPS+ or EPS−) decreases the relative fitness of EPS+ bacteria when compared to both strains starting as single cells (Figure 4A ). However, with greater EPS investment ( f > 0.45), the relative fitness of EPS+ strain is significantly enhanced even though EPS− cells gain a height advantage by starting out as an aggregate. An increase in the number of aggregates results in the relative fitness curves moving upward and downward for Case 1 or Case 2, respectively (Figure S2 ), indicating that the number of aggregates have a significant effect on the competition between these two strains. As the number of aggregates increases (i.e., size of each aggregate decreases), the initially aggregated strain receives competitive advantage over the other strain. Figure 4 Fitness of EPS+ relative to EPS− as a function of f and initial inoculation and number (or size) of aggregates (see Figure S2 for different cases): (A) one aggregate; (B) two aggregates; (C) five aggregates. EPS+ strains get the maximum benefit if EPS− and EPS+ strains are initially spread out and aggregated, respectively. The lines are the polynomial fits to the corresponding data points and the error bars indicate the standard deviations. EPS production ( f > 0), no matter how modest, is better than no EPS production in nearly all situations as it allows better access to nutrients, suggesting that, if bacteria can produce EPS, they should. Our results show that the EPS+ strain obtains the maximum competitive advantage (Figures 4A,B ) at f = 0.5 ± 0.1 ( P = 0.0131) when EPS− strain is initially deposited as one/two aggregates and EPS+ strain is deposited as single cells. However, as the number of aggregates increases to five (Figure 4C ) the EPS+ strain obtains the maximum competitive advantage at around f = 0.3 ± 0.2 ( P = 0.0146) when EPS+ cells are initially deposited as aggregates and EPS− strain as single cells. Generalized linear modeling for the data shown in Figure 4 was also performed to test the statistical significant of the results as detailed in the Supporting information. Case 1 with higher numbers of aggregates have higher relative fitness for EPS+ strain than either control or Case 2 ( t = 9.737, P < 2 × 10 −16 ). This indicates that EPS+ aggregates that are spread out more widely across the substratum relative to the non-EPS producers have a fitness advantage compared to when they are clumped into one colony, relative to non-EPS producers or when they are both distributed as single cells on the surface. The optimum EPS investment to maximize the relative fitness of EPS producers is clearly dependent on the spread and size of aggregates in the initial population. The variation in fitness curves seen for Case 1 and Case 2 (Figure 4 ) for different numbers of aggregates can be further explained by scrutinizing the contribution of the initial aggregate to the biomass (mass of EPS+/EPS− cells) and EPS production over time. Bacteria at the bottom of the aggregate do not contribute to biomass production, irrespective of their status (EPS+ or EPS−) because they do not get sufficient substrate (Figure 5A ). This limits the ability of a tall aggregate to compete with the singleton cells that surround it. When EPS investment increases from f = 0–0.6, the fraction of aggregate which contributes to EPS production increases from zero to around 0.55; while the fraction of aggregate which contributes to bacteria production hovers around a value of 0.2 (Figure 5B ). Since the production of a unit volume of EPS is less expensive than the production of biomass (EPS+ strain) (material density of EPS is smaller than that of biomass, Table 2 ); it is easier for cells to directly invest in EPS production rather than creating new EPS+ cells. Separately, Figure 4 shows that the initially aggregated strain can obtain a fitness advantage by a greater margin if that strain deposits as smaller aggregates (Figures 4B,C ). When the aggregate size decreases, the inactive bacteria seen in the initial aggregate (Figure 5 ) also decreases and hence a greater number of cells of the aggregated strain are available to actively compete with the other strain. Figure 5 Inactive bacteria in the aggregate for Case 1: (A) As the biofilm grow some inactive bacteria are seen at the bottom of the aggregate. The inactive bacteria of the aggregate are shown in green color and these inactive bacteria would not contribute to biomass production ( f = 0). Even though the aggregate gets a competitive advantage through its height the inactive cells in the bottom of the aggregate would be costly for it: (B) fraction of the active bacteria in the initially aggregated EPS+ strain as a function of investment in EPS production. As the investment in EPS increases from 0 to 0.6 the fraction contributes to bacteria production remains fairly constant around 0.2 and the fraction contributes to EPS production increases to around 0.55. Mass of the bacteria is the sum of the mass of EPS+ and EPS− strains. All values are in SI units. For Case 1, when EPS+ and EPS− strains are deposited as aggregates and single cells respectively, the EPS− cells are outcompeted by EPS+ cells over the whole range of f values (0 < f < 0.6). However, in Case 2, the distributed EPS+ cells can be outcompeted by EPS− aggregates if they do not produce enough EPS ( f < 0.1, Figure 4C ). This is in contrast to the control case (single cell attachment) where the EPS− cells can “catch a ride” on the polymeric material only when EPS+ cells heavily invest on EPS (when f > 0.5) and get lifted toward the nutrient rich surface (see Figures 1 , 4 ), thereby gaining an advantage over the EPS+ cells. The segregation of the EPS+ and EPS− strains, as observed in Case 1 and Case 2, prevents the non-producers from being pushed to the top by the EPS+ neighbors investing heavily ( f > 0.5) in production of polymeric material. For smaller aggregates (Figure 4C ) the greatest relative fitness for EPS+ is observed when EPS+ cells (around f = 0.3) are initially deposited as aggregates and EPS− strains are deposited as single cells (Case 1). This is expected because the EPS+ strain gains competitive advantage owing to its moderate height (even though they have a fraction of inactive bacteria) and ability to produce EPS. However, for larger aggregates (Figures 4A,B ), the greatest relative advantage is observed when the EPS producers (around f = 0.5) are single cells and the non-producers are aggregated (Case 2) and this seems counter-intuitive. On closer inspection of time dependent relative fitness of EPS+ ( f = 0.5, Case2, one aggregate) we find that it increases rapidly to around 1.5 at time < 1day, and then decreases transiently, before increasing again at the same rate as the other two scenarios (Figure 6 ). Therefore, the overall superior fitness of the EPS+ strain at f = 0.5 (Figures 4A,B ) at day 12 can be attributed to the initial boost in fitness for the cells as seen in Figure 6 . The reason for this initial fitness boost for EPS+ cells is two-fold: larger EPS− aggregates can have many inactive cells, and all EPS+ cells initially have good access to nutrients, hence they grow well. The initial fitness boost increases as f decreases since the EPS+ strain can invest more energy on production of EPS+ cells than polymeric substance. However, at low f -values, the EPS+ strain cannot maintain this initial boost for long since the EPS+ colonies cannot expand quickly due to lack of EPS. Figure 6 Transient variation of the fitness of EPS+ relative to EPS− at f = 0.5. The general trends for control case and Case 1 are similar. However, the fitness of EPS+ rapidly increases to 1.5 for Case 2 and then temporarily decreases and again increases at the same rate as other cases. Overall, the microbes with EPS+ characteristics gain a better competitive advantage if they initially colonize the surface as smaller aggregates and are widely spaced between the cells of EPS−. As the aggregate size decreases the EPS producing strain dominates in the biofilm even with lower levels of EPS production. Competition between QS+/QS− and EPS− strains While EPS production is advantageous it is also metabolically expensive, and therefore it should be beneficial for its production in bacteria to be regulated through a feedback control mechanism such as quorum sensing. Two quorum sensing settings are considered in this work. In the first setting, QS− cells compete with EPS− cells and in the second setting QS+ cells compete with EPS− cells. For the sake of simplicity, the study is carried out at f = 0.5 which gives better fitness for EPS producers for a single aggregate deposition (Figure 4A ). We examine the effects of different QS threshold values on the relative fitness of QS− and QS+ strains for the three scenarios mentioned above (control, Case1 and Case 2) but focus only on the case of a single aggregate (Figure S1 ). Figure 7 shows the diffusion of AI from the biofilm to the surrounding liquid and the resulting QS−regulation for the control case. Starting with single cells being deposited on the substratum (Figure 7A1 ), the population of QS− increases (Figure 7A2 ) and reaches the threshold for AI, τ = 5 × 10 −7 kg/m 3 (Figure 7A3 ), then EPS production is terminated but the QS− cells (colored red in Figure 7A4 ) continue to proliferate under the negative QS control. For the positive QS control, starting from single cells (Figure 7B1 ), initial growth (duration < 1.3 days) of both strains is similar because their characteristics are identical when there is no EPS production (Figure 7B2 ). The QS+ strain then starts to produce EPS when AI reaches its threshold of τ = 5 × 10 −7 kg/m 3 (Figure 7B3 ) and subsequently the QS+ strain dominates in the biofilm because they gain a competitive advantage due to formation of EPS matrix (Figure 7B4 ). Figure 7 EPS production is negatively (A1–A4) and positively (B1–B4) controlled through quorum sensing ( f = 0.5, τ = 5 × 10 −7 kg/m 3 ): (A1) both QS− and EPS− strains are randomly spread out on the substratum as individual cells; (A2) biofilm after 1.7 days; (A3) biofilm after 2.3 days, the QS− cells on the top of the biofilm gradually terminate the production of EPS; (A4) biofilm after 7 days, there is no EPS on the top of QS− linages. (B1) both QS+ and EPS− are randomly spread out on the substratum as individual cells; (B2) biofilm after 1.0 days; (B3) biofilm after 1.3 days, QS+ strain starts to produce EPS; (B4) biofilm after 7 days, QS+ strain dominates in the biofilm due to EPS matrix. All values are in SI units. We compare three cell deposition scenarios (control, Case 1 and Case 2) when EPS production is regulated through QS, and examined the effect of threshold concentration of the auto-inducer (τ = 1 × 10 −7 , 5 × 10 −7 , 8 × 10 −7 , and 10 × 10 −7 kg/m 3 ) on the fitness of different strains. Figure 8 shows how the negative QS control affects the relative fitness of QS− strains for three cell deposition scenarios. The relationship between the fitness of EPS producers relative to non-producers under different initial deposition scenarios is significantly related to the threshold value of the auto-inducer concentration ( t = 12.141, P = 0). Figure 8 The effect of negative quorum sensing regulation on relative fitness of QS− strain at f = 0.5: (A) control case, both QS− and EPS− strains are initially spread out on the substratum; (B) Case 1, QS− cells are initially aggregated while EPS− cells are spread out; (C) Case 2, QS− cells are initially spread out while EPS− strains are aggregated. At the lowest threshold, the relative fitness rapidly increases to 8.5 and then decreases; (D) relative fitness of QS− strain on day 12. The dotted lines in (D) are the relative fitness without QS for respective cases. The standard deviations of plots (A–C) are shown only at the end for clarity. The standard deviation increases over time. For the control case (Figure 8A ), at the lowest threshold (τ = 1 × 10 −7 kg/m 3 ) the relative fitness is around 1 over time. This is because the QS− strain quickly reaches the QS threshold and terminates EPS production. Both QS− and EPS− strains become biologically identical and hence the relative fitness is around 1. At moderate thresholds (τ = 5 × 10 −7 kg/m 3 ) the fitness of QS− is only enhanced in the early stages of biofilm growth. However, higher thresholds (τ = 8 × 10 −7 and 10 × 10 −7 kg/m 3 ) consistently improve QS− strain fitness. Low thresholds confer an initial short-term advantage followed by consistent reduction in fitness; higher thresholds confer a long-term advantage which is consistent with the findings of Nadell et al. ( 2008 ). The QS influence for Case 1 (QS− aggregate vs. EPS− single cells, Figure 8B ) is analogous to the control case, except at the lowest threshold. At the lowest threshold the QS− strain stops production at the onset of growth; they only have the height advantage (Case 1) and therefore take longer to dominate the biofilm. The benefit of QS in Case 2 (EPS− aggregate vs. QS− single cells, Figure 8C ) is either negative or marginally positive. Quorum sensing is only of long-term value at the highest threshold (τ = 10 × 10 −7 kg/m 3 ). At the lowest threshold (τ = 1 × 10 −7 kg/m 3 ) the relative fitness of QS− strain rapidly increases to 8.5 and then gradually decreases and finally becomes negative. QS− cells stop producing polymeric substance in the beginning, and then the growth of the QS− strain is boosted (τ = 1 × 10 −7 kg/m 3 ). The reason behind this initial fitness boost for the QS− strain has already been explained in Figure 6 . At the lowest threshold, EPS− cells dominate in the long run because they have a competitive advantage due to their initial height. Figure 8D indicates that the long-term fitness of the QS− strain is more sensitive to the QS threshold for Case 2, but less sensitive for the other two cases. Moreover, for all the cell deposition scenarios, the relative fitness of the QS− strain is positively correlated with the QS threshold (the Pearson's correlation coefficients are 0.9512, 0.9921, and 0.9927 for control, Case 1 and Case 2, respectively). QS− cells are not outcompeted for the whole range of QS thresholds for the control and Case 1. Since QS− cells terminate EPS production at the onset of growth at the lowest threshold, both strains are identical for the control case and QS− cells have the height advantage for Case 1 (see Figure 4 , control and Case 1 at f = 0). However, in Case 2, when QS− cells stop EPS production at the onset of growth, the QS− strain is easily outcompeted by EPS− due to the height advantage of the EPS− (see Figure 4 , Case 2 at f = 0). At the highest QS threshold (τ = 10 × 10 −7 kg/m 3 ), although the relative fitness of QS− strain is at least slightly enhanced compared to the strains without QS for all attachment scenarios, only for Case 1 can we guarantee that QS− benefits from quorum sensing ( P = 2.38E-9). This is because at higher thresholds, the QS− strain terminates EPS production after they dominate the biofilm, hence stopping EPS production may give a definite advantage to QS− strain for Case 1 because this strain also has the height advantage due to its initial aggregate nature. Figure 9 shows how the positive QS control affects the relative fitness of EPS producing strains for the three cell deposition scenarios. For the control (Figure 9A ), the quorum sensing- regulated EPS production marginally enhances the relative fitness of QS+ strain for the whole range of QS thresholds (10 −7 < τ < 10 −6 kg/m 3 ). At higher thresholds (τ >1 × 10 −7 kg/m 3 ), the QS+ strain does not produce EPS for a long time and thus the relative fitness of QS+ strain is around 1 until they start to produce EPS, and subsequently their fitness increases once EPS production commences. However, at the lowest threshold (τ = 1 × 10 −7 kg/m 3 ), EPS production starts quickly and so the QS+ strain needs time (about 6 days, Figure 9A ) to dominate in the biofilm because it invests energy on both EPS matrix and QS+ cells, which is analogous to biofilm growth without quorum sensing regulation. There is an optimum QS threshold value for the control case at around τ = 5 × 10 −7 kg/m 3 ( P = 0.006) (Figure 9D ). However, for Case 1 and Case 2 (either strain initially deposited as an aggregate), the positive QS regulation of EPS is advantageous only at the beginning of biofilm growth (Figures 9B,C ). The long-term relative fitness of the QS+ strain decreases as the QS threshold increases ( P < 0.005) (Figure 9D ). Therefore, the relative fitness of the QS+ strain is negatively correlated with QS threshold for Case 1 and Case 2 (the respective Pearson's correlation coefficients are −0.9867 and −0.9644). Figure 9 The effect of positive quorum sensing regulation on relative fitness of QS+ strain at f = 0.5: (A) control case, both QS+ and EPS− strains are initially spread out on the substratum; (B) Case 1, QS+ cells are initially aggregated while EPS− strains are spread out; (C) Case 2, QS+ cells are initially spread out while EPS− strains are aggregated. At the highest threshold, the relative fitness rapidly increases to 6 and then decreases; (D) relative fitness of QS+ strain on day 12. The dotted lines in (D) are the relative fitness without QS for respective cases. The standard deviations of plots (A–C) are shown only at the end for clarity. The standard deviation increases over time. Generalized linear modeling was used to investigate the collective effects of aggregate type, quorum sensing threshold and the occurrence of positive or negative regulation on the relative fitness of EPS producers compared to non-producers as detailed in the Supporting information. The detrimental effect of the quorum sensing threshold ( t = −6.540, P = 3.83 × 10 −10 ) and the occurrence of positive vs. negative control of EPS production ( t = −10.248, P < 2 × 10 −16 ) on relative fitness is very significant. There is a significant correlation between the quorum sensing threshold and whether the EPS is positively or negatively regulated ( t = 13.868, P < 2 × 10 −16 ) indicating a synergy between the two variables in their effects on fitness. The relative fitness of EPS producers is also dependent on the nature of cell deposition, with aggregated EPS producers (Case 2) resulting in higher fitness than the other two deposition scenarios ( t = 6.660, P = 1.94 × 10 −10 ). Overall, we conclude that quorum sensing-regulated EPS production would enhance the fitness of EPS producers only marginally, or even reduces their competitive advantage, under the investigated conditions. This analysis shows that quorum sensing-mediated gene regulation in bacteria may be detrimental at times depending on the nature of the competition. Zhao and Wang ( 2017 ) argued that depending on the conditions there would be a “ right time ” and “ right place ” in which QS−regulated EPS production can favor biofilm growth; otherwise it would have unfavorable consequences for the EPS producers. Numerical experiments of Frederick et al. ( 2011 ) also show that QS−regulated EPS production rarely facilitates a biofilm to achieve a high cell population. However, maximizing offspring generation is not the only strategy bacteria may have, and sometimes production of EPS is beneficial if the objective is to produce a thick EPS protective layer. Therefore, further studies are needed to understand the role of QS− regulated EPS production for the cell deposition scenarios investigated here, taking into account the multiple functional roles of EPS in bacterial biofilms."
} | 11,705 |
31843766 | PMC6918757 | pmc | 928 | {
"abstract": "ABSTRACT Coral reefs are in global decline mainly due to increasing sea surface temperatures triggering coral bleaching. Recently, high salinity has been linked to increased thermotolerance and decreased bleaching in the sea anemone coral model Aiptasia. However, the underlying processes remain elusive. Using two Aiptasia host–endosymbiont pairings, we induced bleaching at different salinities and show reduced reactive oxygen species (ROS) release at high salinities, suggesting a role of osmoadaptation in increased thermotolerance. A subsequent screening of osmolytes revealed that this effect was only observed in algal endosymbionts that produce 2-O-glycerol-α-D-galactopyranoside (floridoside), an osmolyte capable of scavenging ROS. This result argues for a mechanistic link between osmoadaptation and thermotolerance, mediated by ROS-scavenging osmolytes (e.g., floridoside). This sheds new light on the putative mechanisms underlying the remarkable thermotolerance of corals from water bodies with high salinity such as the Red Sea or Persian/Arabian Gulf and holds implications for coral thermotolerance under climate change. This article has an associated First Person interview with the first author of the paper .",
"conclusion": "Conclusion Recent work showing reduced bleaching at high salinities and high levels of floridoside, an osmolyte with antioxidative capabilities, at high salinities, encouraged us to assess a link between osmoadaptation and thermotolerance in symbiotic cnidarians. Exposing the coral model Aiptasia to heat at different salinities confirms increased thermotolerance and reduced bleaching at high salinity, manifested by reduced ROS leakage. The decrease of ROS leakage followed increased levels of the ROS-scavenging osmolyte floridoside under increasing salinities, thus, arguing for a mechanistic link between osmoadaptation and thermotolerance in the cnidarian-dinoflagellate endosymbiosis. Our results may help to explain the extraordinarily high thermotolerance of corals from the Arabian Seas and may hold implications about the response of corals to rising sea surface temperatures considering salinity as a contributing factor. Future studies should assess whether salinity-conveyed thermotolerance is a mechanism that is present in corals and whether other osmolytes (and which ones) may be important and play a role in salinity-conveyed thermotolerance.",
"introduction": "INTRODUCTION Climate change leads to ocean warming and ocean acidification, which are threatening coral reefs at a global scale ( Hughes et al., 2017 a ). While ocean warming is identified as the main driver of coral bleaching ( Hughes et al., 2017b ), the effects of ocean acidification are less clear, but presumably affect coral calcification and reef growth ( Tambutté et al., 2015 ; Albright et al., 2016 ; Liew et al., 2018 ). Bleaching describes the loss of the coral-associated micro-algal photosynthetic endosymbionts in the family Symbiodiniaceae ( Hughes et al., 2017 b ; LaJeunesse et al., 2018 ). As such, corals lose their supply of photosynthates, which covers their main energy needs to build and maintain calcium carbonate skeletons that in turn provide the structural foundation of reef ecosystems ( Muscatine and Porter, 1977 ). Hence, it is becoming increasingly important to better understand the mechanisms and drivers of coral bleaching, as well as the factors that determine stress resilience and thermotolerance of corals ( Torda et al., 2017 ). As a rough estimate, corals start to bleach at about 1–2°C above their annual average summer temperatures ( Hoegh-Guldberg, 1999 ), suggesting that corals are adapted to their prevailing environmental conditions ( Hughes et al., 2017 b ), as supported by observed differences in thermotolerance across regions ( Osman et al., 2018 ). To date, we are missing a detailed understanding of the factors that contribute to such geographical differences of bleaching susceptibility. In addition, the cellular mechanisms of bleaching are not completely understood. While the production and accumulation of reactive oxygen species (ROS) as a consequence of increased temperatures, i.e. heat stress, certainly play a role ( Lesser, 1997 , 2011 ; Weis, 2008 ), recent studies have shown bleaching without heat stress ( Pogoreutz et al., 2017 ), bleaching without light ( Tolleter et al., 2013 ), and bleaching decoupled from oxidative stress ( Nielsen et al., 2018 ). Further, Gegner et al. (2017) showed increased thermotolerance and reduced bleaching at high salinities in the coral model Aiptasia, suggesting a possible role of osmoadaptation in stress resilience of symbiotic cnidarians ( Ochsenkühn et al., 2017 ; Osman et al., 2018 ). However, the underlying mechanism remained elusive. With regard to the putative importance of salinity in contributing to thermotolerance, it is important to note that some of the most thermotolerant corals can be found in the hottest and most saline water bodies, i.e. the Persian/Arabian Gulf ( Hume et al., 2013 , 2015 ; D'Angelo et al., 2015 ) and the northern Red Sea ( Bellworthy and Fine, 2017 ; Krueger et al., 2017 ; Osman et al., 2018 ). To gain insight into the potential mechanisms underlying salinity-conveyed thermotolerance of symbiotic cnidarians, we set up a series of bleaching experiments at different salinities. Using the coral model Aiptasia ( sensu Exaiptasia pallida ), we first assessed the thermotolerance of two host–endosymbiont pairings at different salinity and temperature conditions. Subsequent linking of the heat stress response to ROS and osmolyte levels allowed us to pinpoint potential processes that play a role in the increased thermotolerance and the decreased bleaching at high salinities in Aiptasia.",
"discussion": "DISCUSSION Prompted by previously published work that high salinity conveys thermotolerance in Aiptasia ( Gegner et al., 2017 ) and that the osmolyte floridoside is highly abundant in coral holobionts at high salinity ( Ochsenkühn et al., 2017 ), we here assessed the presence of a mechanistic link between osmoadaptation and thermotolerance in the cnidarian-algal endosymbiosis. Using different host–endosymbiont pairings of Aiptasia anemones, we could show that for the pairing H2-SSB01, increased thermotolerance and reduced bleaching under heat stress at high salinity resulted in reduced ROS leakage concomitant with increased floridoside levels. This lends further support to the proposed dual function of floridoside as an osmolyte and ROS scavenger in algal endosymbionts ( Ochsenkühn et al., 2017 ), due to its antioxidative capabilities ( Li et al., 2010 ; Pade et al., 2015 ). In particular, since floridoside is the only measured carbohydrate osmolyte that increased abundance levels with increasing salinity, it is a likely candidate (among other metabolites not measured here, e.g. glutathione) to reduce ROS levels, thereby conveying thermotolerance at high salinity. Nevertheless, a direct functional link remains to be established, e.g. via knockdown or overexpression of the underlying genes. Since corals as well as Aiptasia are lacking the genes to synthesize floridoside ( Pade et al., 2015 ; Ochsenkühn et al., 2017 ), such functional testing would need to be carried out in the algal endosymbiont, which seems particularly intractable to genetic manipulation ( Chen et al., 2019 ). Besides such limitations, our data suggest a link between osmoadaptation and thermotolerance in the cnidarian-algal symbiosis with implications for the importance of osmoadaptation to stress resilience. The osmolyte floridoside links osmoadaptation with thermotolerance as a putative broadly present mechanism Our results reveal a differential response of the two Aiptasia strains to heat stress under different salinities. Symbiotic H2-SSB01 anemones exhibited higher symbiont loss and reduction in photosynthetic efficiency during heat stress than symbiotic CC7-SSA01 Aiptasia. Further, reduced abundance of most of the carbohydrates points towards increased metabolism and energy consumption during heat stress in H2-SSB01. By comparison, carbohydrate levels in CC7-SSA01 were more stable under heat stress. Interestingly, however, H2-SSB01 became more thermotolerant with increasing salinities, whereas CC7-SSA01 did not seem to respond to an increase in salinity. This makes these Aiptasia strains a good model system to study differences in osmoadaptation-related thermotolerance. At this point, the underlying cause of the difference in salinity-conveyed thermotolerance remains unclear. In a recent study by Cziesielski et al. (2018) , the authors showed that different Aiptasia strains harbor similar antioxidant capacities, but that observed differences of ROS levels in hospite were endosymbiont-driven. Indeed, bleaching susceptibility in our experiments directly aligned with changes in ROS leakage of the endosymbionts during heat stress. Whereas CC7-SSA01 showed overall low ROS leakage across salinities, supporting a high inherent thermotolerance, H2-SSB01 showed reduced ROS leakage at increasing salinities. Thus, endosymbiont identity seems to, at least partially, determine thermotolerance of holobionts. Further, differences in symbiotic ROS leakage are likely not only underlying the differential bleaching susceptibility between host–endosymbiont pairings, but also play a role in the salinity-conveyed thermotolerance. In the case of H2-SSB01, the decrease of ROS leakage at increasing salinities during heat stress aligns with increased levels of the osmolyte floridoside, which was previously suggested to be an osmolyte of coral algal endosymbionts ( Ochsenkühn et al., 2017 ). Ochsenkühn et al. (2017) further hypothesized that floridoside might play an important role in countering ROS arising from salinity and heat stress, given that it is a potent antioxidant in many marine algae ( Li et al., 2010 ; Martinez-Garcia and van der Maarel, 2016 ). Our results corroborate this notion and mechanistically link osmoadaptation with thermotolerance via increased floridoside levels at high salinities, whereby floridoside plays a dual role as an osmolyte and ROS scavenger. Importantly, floridoside was only measured in H2-SSB01 and showed increased levels at increased salinities, in line with reduced ROS leakage of the algal endosymbionts. By contrast, floridoside was not detectable in CC7-SSA01, which neither exhibited reduced ROS leakage nor increased thermotolerance at high salinities. Future experiments could test for the presence of salinity-conveyed thermotolerance, reduced ROS leakage of the algal endosymbionts, and floridoside abundance levels in reversed host–endosymbiont pairings, i.e. H2-SSA01 and CC7-SSB01. This would also clarify the relative contribution of host and endosymbiont, respectively the importance of host and endosymbiont identity. While the Aiptasia system explicitly allows to test the same host with different endosymbionts and vice versa ( Voolstra, 2013 ), one must acknowledge that the performance of native host endosymbiont associations are optimized and often exceed non-native host endosymbiont associations ( Matthews et al., 2017 ; Rädecker et al., 2018 ). Therefore, results from such experiments may still be ambiguous. At this point, data obtained from Red Sea corals in a pilot study (data not shown) support the idea that salinity-conveyed thermotolerance might be a wider phenomenon, hinting towards a broadly present mechanism. Indeed, studies from plants grown at high salinities have shown an increased temperature tolerance and this was attributed to an increased production of osmolytes ( Lu et al., 2003 ; Rivero et al., 2014 ). It is important to note that floridoside is only one of many molecules that may link osmoadaptation with thermotolerance in symbiotic cnidarians. As such, we rather argue for the importance of the mechanistic link between osmoadaptation and thermotolerance, than for any particular osmolyte. There is a number of osmolytes that may contribute to the salinity-conveyed thermotolerance besides floridoside, such as dimethylsulphoniopropionate (DMSP) or amino acids ( Mayfield and Gates, 2007 ; Yancey et al., 2010 ; Ochsenkühn et al., 2017 ). A new perspective for corals in extreme environments The extraordinary thermotolerance of corals from the Red Sea and Persian/Arabian Gulf has been demonstrated in a number of studies ( Fine et al., 2013 ; Hume et al., 2016 ; Krueger et al., 2017 ; Osman et al., 2018 ). Importantly, D'Angelo et al. (2015) showed that superior heat tolerance is lost when corals from the Persian/Arabian Gulf are exposed to reduced salinity levels. While this may argue for strong local adaptation to high temperature and the exceptionally high salinity in the Persian/Arabian Gulf, it may, at least partially, relate to the here demonstrated link between salinity and thermotolerance. This is further supported by the higher heat tolerance of corals in the northern Red Sea ( Osman et al., 2018 ) and the Gulf of Aqaba ( Fine et al., 2013 ), in comparison to their central and southern Red Sea counterparts, in line with the northern Red Sea harboring much higher salinity levels (≥41) than the central and southern counterparts (36) ( Ngugi et al., 2012 ). As such, it remains to be determined whether salinity levels may affect the stress resilience of corals on a global scale. Notably, our results highlight the complexity of interactions underlying holobiont resilience. Besides algal endosymbionts, other microbiome members such as bacteria and archaea should also be taken into account, as they may rapidly respond to salinity ( Röthig et al., 2016 ) and may contribute to the thermotolerance of the coral holobiont ( Ziegler et al., 2017 ). Conclusion Recent work showing reduced bleaching at high salinities and high levels of floridoside, an osmolyte with antioxidative capabilities, at high salinities, encouraged us to assess a link between osmoadaptation and thermotolerance in symbiotic cnidarians. Exposing the coral model Aiptasia to heat at different salinities confirms increased thermotolerance and reduced bleaching at high salinity, manifested by reduced ROS leakage. The decrease of ROS leakage followed increased levels of the ROS-scavenging osmolyte floridoside under increasing salinities, thus, arguing for a mechanistic link between osmoadaptation and thermotolerance in the cnidarian-dinoflagellate endosymbiosis. Our results may help to explain the extraordinarily high thermotolerance of corals from the Arabian Seas and may hold implications about the response of corals to rising sea surface temperatures considering salinity as a contributing factor. Future studies should assess whether salinity-conveyed thermotolerance is a mechanism that is present in corals and whether other osmolytes (and which ones) may be important and play a role in salinity-conveyed thermotolerance."
} | 3,728 |
23284943 | PMC3527390 | pmc | 929 | {
"abstract": "Coral reefs are under considerable pressure from global stressors such as elevated sea surface temperature and ocean acidification, as well as local factors including eutrophication and poor water quality. Marine sponges are diverse, abundant and ecologically important components of coral reefs in both coastal and offshore environments. Due to their exceptionally high filtration rates, sponges also form a crucial coupling point between benthic and pelagic habitats. Sponges harbor extensive microbial communities, with many microbial phylotypes found exclusively in sponges and thought to contribute to the health and survival of their hosts. Manipulative experiments were undertaken to ascertain the impact of elevated nutrients and seawater temperature on health and microbial community dynamics in the Great Barrier Reef sponge Rhopaloeides odorabile . R. odorabile exposed to elevated nutrient levels including 10 µmol/L total nitrogen at 31°C appeared visually similar to those maintained under ambient seawater conditions after 7 days. The symbiotic microbial community, analyzed by 16S rRNA gene pyrotag sequencing, was highly conserved for the duration of the experiment at both phylum and operational taxonomic unit (OTU) (97% sequence similarity) levels with 19 bacterial phyla and 1743 OTUs identified across all samples. Additionally, elevated nutrients and temperatures did not alter the archaeal associations in R. odorabile , with sequencing of 16S rRNA gene libraries revealing similar Thaumarchaeota diversity and denaturing gradient gel electrophoresis (DGGE) revealing consistent amoA gene patterns, across all experimental treatments. A conserved eukaryotic community was also identified across all nutrient and temperature treatments by DGGE. The highly stable microbial associations indicate that R. odorabile symbionts are capable of withstanding short-term exposure to elevated nutrient concentrations and sub-lethal temperatures.",
"introduction": "Introduction The Great Barrier Reef (GBR) hosts high biodiversity and is the world’s largest coral reef ecosystem. At almost 2000 km long it was declared a World Heritage Area in 1981 [1] . Degradation of coastal marine ecosystems is occurring globally due to over-fishing, declining water quality and climate change [2] , [3] . Despite the GBR’s protected status it is still exposed to anthropogenic and environmental pressures, making degradation due to terrestrial run-off the focus of intense management efforts [4] . Twenty-six major catchments, in which a wide range of industrial and agricultural activities take place, border the GBR [5] . Fertilisers used in cattle grazing, sugarcane production and horticulture can flow into the marine environment [6] , with 80% of the total anthropogenic dissolved inorganic nitrogen (DIN) introduced into the GBR ecosystem thought to come from fertilisers ( Table 1 ) [2] , [7] . Moreover, nitrogen and phosphorus loads have increased by factors of approximately 6 and 9, respectively, since European settlement ca.1830 [8] . Catchment areas in the GBR are characterized by distinct wet/dry seasonal rainfall and are subject to intense cyclonic rainfall over periods of days to a few weeks [9] . River discharge of nutrients into the GBR therefore occurs almost entirely in large pulse events or flood plumes [6] , [10] which generally affect reefs within 20 km of the coast (∼27% of all reefs). The resulting elevated nutrient levels can be 2–100 times higher than ambient [6] , [10] , [11] but are relatively short-lived, detectable for only 3–14 days after flood plume events [12] . Cyclonic events are increasing in frequency and intensity, with the most recent on the GBR (in January 2011) delivering extremely high levels of nutrients from agricultural and urban catchments to the reef environment [13] – [15] . Nutrient levels in ambient (non-flood) conditions are presented in Table 2 . 10.1371/journal.pone.0052220.t001 Table 1 Total yearly inputs of anthropogenic nutrient loads into the GBR, 1 \n [2] , 2 \n [121] . Anthropogenic nutrient loads 1,2 \n Tonnes/year Total nitrogen (TN) 80000 Dissolved inorganic nitrogen (DIN) 11000 Dissolved organic nitrogen (DON) 6900 Particulate nitrogen (PN) 52000 Total phosphorus (TP) 16000 Dissolved inorganic phosphorus (DIP) 800 Dissolved organic phosphorus (DOP) 470 Particulate phosphorus (PP) 13000 10.1371/journal.pone.0052220.t002 Table 2 The level of nutrients in ambient (non flood) waters, reported from Pelorus Island from 2005–2011 1 \n [4] . Nutrient NH 4 \n NO 3 \n DON PN DOP PP DOC POC \n Ambient wet season 1 \n 0.13 0.06 5.64 1.08 0.14 0.1 70.66 9.44 \n Ambient dry season 1 \n 0.04 0.03 5.49 0.74 0.17 0.06 59.25 6.93 Parameters are in µM for dissolved inorganic nutrients (NH 4 , NO 3 ), dissolved organic nitrogen, phosphorus and carbon (DON, DOP, DOC) and particulate nitrogen, phosphorus and organic carbon (PN, PP, POC). Marine sponges are important components of coastal and offshore coral reefs, exhibiting high diversity, high biomass and the ability to influence both benthic and pelagic processes [16] . Sponges also harbour extensive microbial communities which can comprise up to 35% of sponge tissue volume and include bacteria, archaea and eukarya [17] , [18] . To date, 32 bacterial phyla and candidate phyla have been reported from sponges [19] , [20] , with some phylotypes appearing to occur exclusively in sponges and not in the surrounding environment i.e. so-called sponge-specific clusters (SCs) or sponge- and coral-specific clusters (SCCs) [17] , [21] – [23] . In areas such as coral reefs, where dissolved nutrients and particulate organic matter are scarce, sponges may experience nitrogen limitation and symbiotic microorganisms are thought to contribute to nitrogen cycling within the host. Both autotrophic (such as Cyanobacteria ) and heterotrophic symbionts may contribute to the nitrogen budget of sponges by fixing atmospheric nitrogen [24] – [26] . In low nutrient waters symbionts are thought to benefit by recycling nitrogenous waste excreted from the sponge host [18] . Ammonia-oxidising bacteria (AOB) of the genera Nitrosospira and Nitrosococcus \n [27] and the ammonia-oxidising archaea (AOA), such as “ Candidatus Cenarchaeum symbiosum” [28] , [29] , which convert ammonia to nitrite, have all been identified in sponges. Nitrite-oxidising bacteria (NOB) such as Nitrospina and Nitrospira have also been detected in many sponge species [21] , [30] – [33] , as have denitrification and anaerobic ammonia oxidation (anammox) processes [31] , [34] . Symbiosis between nitrogen-transforming microbes and sponges influences not only the ecology of the host but also the wider reef ecosystem (reviewed by [35] ). Elevated nutrient levels have been highlighted as a cause of coral reef decline, with some studies reporting an increase in the severity of coral diseases such as aspergillosis and yellow blotch [36] – [42] . Additionally, both resilience [43] , [44] and sensitivity of coral-microbial associations due to elevated nutrient levels have been reported [45] – [47] . Despite the importance of sponge nitrogen cycling to coral reef ecosystems [35] , very little research has addressed the sensitivity of sponge–microbial partnerships to nutrient enrichment, with the effects of eutrophication more widely reported for free-living microbial communities. In general, as the availability of nitrogen and phosphorus increases, phytoplankton and bacterial production increases which leads to a higher biological oxygen demand and increases the sedimentation rate of particulate matter [48] . In the early stages of nutrient loading within Chesapeake Bay [49] , bacterioplankton communities remain dominated by SAR11, SAR86 and picocyanobacteria, however as anoxic conditions develop the bacterial community shifts to anaerobic members of the Firmicutes, Bacteroidetes and sulphur-oxidising Gammaproteobacteria . Another effect commonly observed in nutrient-rich environments is an increased abundance of prokaryotic cells; in natural seawater amendments (e.g. the addition of nutrient-rich deep waters to nutrient-depleted surface waters) an increase in the abundance of taxa such as SAR11 and marine Actinobacteria was reported [50] – [52] . Several studies have also addressed the effect of nutrient addition on bacterial community structure in the marine environment [53] – [57] , however results have been variable due to different experimental methodologies and the high spatial and temporal variability of free-living marine communities [48] . Despite the variable microbial responses to experimental nutrient amendments, few microbial communities have been shown to be resistant to change after environmental disturbance. Here we analyzed how the microbial community of the Great Barrier Reef sponge Rhopaloeides odorabile responded to experimental nutrient exposures under ambient and elevated seawater temperature. While the interactive effects of multiple stressors have previously been explored in microbial biofilms [58] , coral larvae [59] , foraminifera [60] , coral pathogens [41] and adult corals [61] , the impact of combined anthropogenic stressors on marine sponges was unknown. Elevated seawater temperature has previously been shown to cause a shift in the dominant microbial community on marine sponges as well as a decline in sponge health [62] – [64] , with some mass mortality events concomitant with anomalies in sea surface temperature [65] , [66] . Previous experiments have demonstrated that adult R. odorabile exhibit necrosis and a loss of microbial symbionts within 72 h at 33°C [67] , with subsequent experiments confirming a narrow thermal threshold for the host and symbiont community between 31–32°C [68] , [69] . Here we investigated the combined effects of water quality and elevated seawater temperature by exposing sponges to a range of elevated nutrient levels under ambient (27°C) and sub-lethal (31°C) seawater temperatures.",
"discussion": "Discussion \n R. odorabile clones exposed to the combined effects of elevated nutrient levels and seawater temperature appeared visually similar to those maintained under ambient seawater conditions. The microbial communities of R. odorabile were not significantly affected by these environmental stressors, indicating that this sponge species is capable of withstanding short-term exposure to elevated nutrient concentrations and sub-lethal temperatures. R. odorabile is found throughout the GBR, occurring on inner-, mid- and outer-reef locations [91] , [92] . The broad distribution of R. odorabile throughout the GBR exposes this species to a well-defined water quality gradient, with inner reef sponges experiencing higher nutrient loads, particularly during flood events, compared to mid and outer reef sponges [6] , [93] . While reproductive output has been reported to decrease in female R. odorabile from inner reefs compared to outer reefs, these changes could not be directly linked to elevated nutrients or water turbidity [94] . Many cases have shown that microbial communities are sensitive to environmental perturbation [48] , [95] . The evidence presented here, however, suggests that microbial communities within R. odorabile can resist these nutrient perturbations, even at temperatures of 31°C. Eutrophication and poor water quality are major concerns for reef ecosystems globally. In addition to local factors, coral reefs are also faced with global stressors including elevated sea surface temperatures and ocean acidification [96] , [97] . Despite this, the interacting effects of multiple environmental stressors on marine invertebrates are seldom investigated. Ambient levels of nitrogen and phosphorus recorded over the duration of the experiment were higher than those for nearby reefs, which is potentially due to input from the Orpheus Island Research Station. Sponges were exposed to nutrient concentrations 9-fold, 7.5-fold, 7-fold and 2.1-fold (ammonium, phosphate, nitrite and nitrate concentrations, respectively) above ambient yet showed no adverse health effects or changes in symbiosis. These results are further supported by a recent study that found nutrient enrichment does not affect the sponge Aplysina cauliformis or its symbiont community [98] . Healthy and Aplysina Red Band Syndrome (ARBS)-affected A. cauliformis were exposed to nutrient-enriched conditions (up to 4.8-fold and 2.1-fold increases of nitrate and phosphate respectively, from ambient levels over 7 days). A combination of terminal restriction fragment length polymorphism (T-RFLP), histology and chlorophyll fluorescence measurements [98] revealed no change in the bacterial communities of healthy sponges, nor an enhanced rate of disease progression in ARBS-affected sponges. However, nutrient enrichment levels similar to those in this experiment have been shown to exacerbate the onset and severity of coral diseases, including Black Band Disease [42] , aspergillosis and Yellow Band Disease [36] . Although the mechanisms are unknown, this may be due to an enhancement of microbial growth rates [45] and/or increased pathogen virulence [36] , [37] . The microbial community of R. odorabile analyzed by 454 pyrotag sequencing was highly conserved for the duration of the experiment at both phylum and OTU levels. Consistent with other sponge amplicon pyrosequencing studies [19] , [20] , [87] , [99] , communities were dominated by Chloroflexi, Proteobacteria, “Poribacteria” and SAUL at both the phylum and OTU level. Whilst the abundance data and fold change data for the majority of OTUs showed little correlation with particular nutrient/temperature treatments, replicate C at time 7 days, seawater temperature 27°C and ambient concentrations of nutrients (727LC) was an exception. In 727LC the differences were attributed to two Gemmatimonadetes OTUs and one OTU within the Chloroflexi . Given the high similarity of all other replicates and that this clone appeared visibly healthy, it is possible that this anomaly was a consequence of associated infauna being inadvertently sequenced with the sponge tissue. Sponge-specific clusters (SC) and sponge-coral-specific clusters (SCC) [17] , [21] , [22] are monophyletic clusters of 16S rRNA sequences found only in sponges (or sponges and corals) and not the surrounding environment such as seawater or sediments. We saw neither a significant increase nor decrease in the proportion of reads assigned to SCs or SCCs, with approximately 70% of all reads assigned to these clusters per sample across all treatments. Even though the roles of SCs/SCCs are still largely uncharacterized, it is predicted that their loss would be detrimental to the health and survival of the host sponge [100] . Nutrient enrichment and sub-lethal temperature stress did not alter the symbiotic archaeal associations within R. odorabile . Archaeal sequences were affiliated with Thaumarchaeota , which is consistent with archaeal sequences previously found in this species and in other sponge studies [101] – [103] . To address changes in community composition and potential functionality of sponge-associated archaea due to elevated nutrients and temperature, we also screened samples for changes in the amoA gene. In AOA and AOB the AmoA enzyme catalyses aerobic oxidation of ammonia to nitrite (the first step of nitrification). Analysis by qPCR in both the marine and terrestrial environment has suggested that AOA outnumber AOB [76] , [104] , [105] and this is also the case in at least some cold water marine sponges [29] . In this study the highest level of ammonium that sponges were exposed to was 9-fold higher than ambient levels. However, no shifts in the composition of AOA could be correlated with nutrient or temperature treatment, indicating that the diversity of AOA is stable under multiple environmental stressors. Marine eukaryotic microbial communities mainly consist of algae, protozoa and marine fungi which play important roles in microbial food webs and in nutrient cycling [106] , [107] . Within marine sponges, diatoms, dinoflagellates and fungi are known to live symbiotically although their functional roles within the sponge remain unclear [17] , [108] . The effects of eutrophication on free-living eukaryotic microbial communities are widely reported: as the availability of nitrogen and phosphorus increases, primary production by algae is stimulated and the structures of phytoplankton communities change [109] – [112] . Ultimately these changes lead to increased turbidity in the water column coupled with oxygen depletion, which can greatly affect benthic communities [109] . Whilst the effect of nutrient amendment on host-associated microbial communities is less understood, nutrient enrichment studies with corals have shown increases in zooxanthellae abundance and indicate that these cells have preferential access to available CO 2 which is then used for photosynthesis. Within the sponge Cymbastela concentrica , nutrient enrichment had no effect on symbiotic micro-algal growth as detected via chlorophyll concentration and sponge growth rate [113] . In the current study, the conserved eukaryotic community in R. odorabile across all nutrient and temperature treatments further highlights the stability of microbial associations within this sponge species. Based on eutrophication studies with free-living systems, one possible scenario resulting from nutrient elevation is an increase in the relative abundance of some species (particularly photosynthetic organisms). However, further exploration is required to investigate this. Here we exposed sponges to seawater temperatures of 31°C which, based on Intergovernmental Panel on Climate Change (IPCC) scenarios, will occur before 2100 [114] . Previous studies assessing the response of R. odorabile to thermal stress have indicated sub-lethal effects at 31°C, including activation of the heat shock protein system [68] and a significant decrease in flow rate, filtration efficiency and choanocyte chamber density [115] . Whilst the bacterial community of R. odorabile is highly stable at 31°C, higher seawater temperatures cause a shift in the symbiont community which is concomitant with host tissue necrosis and mortality after four days at 32°C or three days at 33°C [67] , [69] . Anthropogenic stressors such as water pollution have been shown to negatively interact with elevated seawater temperature, reducing coral larval metamorphosis [59] and increasing the persistence of the coral pathogen Serratia marcescens \n [41] . Research on coral bleaching thresholds also identified higher temperature sensitivity after exposure to increased DIN concentrations [116] with a recent study confirming that increased DIN, combined with limited phosphate concentrations, increases the susceptibility of corals to temperature- and light-induced bleaching. This is thought to occur due to an imbalanced DIN supply causing phosphate starvation of the symbiotic zooxanthellae [61] . In contrast, sub-lethal thermal stress in the current study did not appear to increase the susceptibility of R. odorabile to elevated nutrients. Our study generated nutrient enrichment levels that are known to occur during major flood plume events and which have a destabilizing effect on the host-symbiont relationship in corals. The different response of sponges and corals to nutrient treatment may relate to the short timescale of exposure used in our study. However, the time scale of this study reflects real-time dispersal rates of elevated nutrients in the GBR environment. Many studies have reported the detrimental impacts of various water quality parameters on sponges and corals directly [36] , [40] , [42] , [89] , [94] , [117] – [120] . In contrast, we report for the first time the effects of multiple environmental stressors on the important partnerships that reef invertebrates form with symbiotic microbes. We detected no changes in the bacterial, eukaryotic or archaeal community from any of the nutrient and/or temperature treatments, indicating that R. odorabile will be able to withstand nutrient pulses associated with flood plume events that are becoming more frequent in occurrence and severity [2] , [13] , [14] , [93] . By assessing multiple stressors in combination, this study provides a first step for environmentally relevant sponge stress assessments which will enhance management strategies for GBR sponge populations."
} | 5,156 |
33945788 | PMC8162421 | pmc | 930 | {
"abstract": "Summary Stony corals are colonial cnidarians that sustain the most biodiverse marine ecosystems on Earth: coral reefs. Despite their ecological importance, little is known about the cell types and molecular pathways that underpin the biology of reef-building corals. Using single-cell RNA sequencing, we define over 40 cell types across the life cycle of Stylophora pistillata . We discover specialized immune cells, and we uncover the developmental gene expression dynamics of calcium-carbonate skeleton formation. By simultaneously measuring the transcriptomes of coral cells and the algae within them, we characterize the metabolic programs involved in symbiosis in both partners. We also trace the evolution of these coral cell specializations by phylogenetic integration of multiple cnidarian cell type atlases. Overall, this study reveals the molecular and cellular basis of stony coral biology.",
"introduction": "Introduction Scleractinian corals, also known as stony corals, are the main builders of the reefs that constitute the most diverse marine ecosystems, providing home to roughly a quarter of all marine species ( Reaka-Kudla, 1997 ). Stony corals belong to the Hexacorallia, a lineage within the Anthozoa class in the Cnidaria phylum. In addition to all of the stony corals, Hexacorallia includes sea anemones, such as the model cnidarians Nematostella vectensis and Exaiptasia pallida , and zoanthids ( Baumgarten et al., 2015 ; Genikhovich and Technau, 2009 ; Technau and Steele, 2011 ; Zapata et al., 2015 ). Anthozoan life cycles involve a swimming larva that disperses, settles, and metamorphoses into a sessile polyp, which in turn develops into the adult stage. In stony corals, larval settlement is followed by rapid accretion of a protein rich skeletal organic matrix and extracellular calcium carbonate crystals (in the form of aragonite) to form a stony skeleton ( Akiva et al., 2018 ; Vandermeulen and Watabe, 1973 ). Through this process of biomineralization, stony corals build the main mineral substrate of marine reefs ( Drake et al., 2020 ; Tambutté et al., 2011 ). Stony corals thrive in oligotrophic tropical and subtropical seas by forming a symbiotic consortium with photosynthetic dinoflagellate algae of the Symbiodiniaceae family ( Baker, 2003 ; LaJeunesse et al., 2018 ; van Oppen and Medina, 2020 ). The dinoflagellate cell resides within a lysosomal-like organelle inside the host cell and transfers diverse photosynthetic products to the coral, which in turn provides the symbiont with inorganic carbon ( Davy et al., 2012 ; Rosset et al., 2021 ). This photosymbiosis sustains the high-energy demands of coral growth and reproduction, including skeleton production by massive calcium carbonate deposition. In addition to diverse metabolic adaptations ( Morris et al., 2019 ), coral immunity is considered to be a major factor for coral endosymbiosis, as well as for modulating interactions with other microbial eukaryotes and prokaryotes ( Jacobovitz et al., 2019 ; Kwong et al., 2019 ). A worldwide decline in coral reefs has been reported in the past decades ( Hughes et al., 2017 , 2019 ; Sully et al., 2019 ). Global changes in ocean temperature and acidification directly impact coral symbiosis and skeleton formation, causing the release of symbionts (known as coral bleaching) and reducing calcification rates, respectively ( Hoegh-Guldberg et al., 2007 ; Putnam et al., 2017 ; Torda et al., 2017 ). The collapse in stony coral colonies has stirred investigations into the molecular basis of these unique coral specializations. To date, our understanding of the molecular biology of stony corals largely derives from genome sequencing efforts. These data have provided crucial information on coral population structure ( Dixon et al., 2015 ; Fuller et al., 2020 ; Shinzato et al., 2021 ) and on the evolution of coral gene repertoires, including genes potentially involved in symbiosis, skeleton-formation, and immunity ( Barshis et al., 2013 ; Bayer et al., 2012 ; Bhattacharya et al., 2016 ; Buitrago-López et al., 2020 ; Shinzato et al., 2011 ; Voolstra et al., 2017 ; Ying et al., 2018 ). However, fundamental aspects of stony coral biology are still to be clarified: the specific cellular context in which these genes are employed and the diversity of cell types encoded in scleractinian genomes. To address these questions, we used single-cell transcriptomics to systematically characterize cell type gene expression programs across the life cycle of Stylophora pistillata , a reef-building stony coral with a broad Indo-Pacific distribution ( Figures 1 A and 1B). From these whole-organism single-cell RNA sequencing (scRNA-seq) profiles, we derive detailed cell type maps for S. pistillata adult, primary polyp, and larval stages. Together with in situ hybridization validations, phylogeny-based gene annotation, and cross-species comparative analyses, these data offer insights into the molecular basis and evolution of stony coral cellular specializations, including symbiosis, calcification, and immunity. Figure 1 Stylophora pistillata multi-stage cell type atlas (A) S. pistillata adult colony. (B) S. pistillata phylogenetic position. (C) S. pistillata life cycle stages represented in this study. (D) 2D projection of S. pistillata adult single-cells (left) and normalized expression of selected variable genes across adult metacells (fold-change ≥2, allowing only a maximum of 10 genes per metacell). Broad cell types are indicated in the x axis and color bar in the y axis defines cell type in which the gene is specifically expressed. (E) Same as (D) for larva. (F) Same as (D) for primary polyp. See detailed single-cell expression maps in Figure S2 . (G) Schematic representation of S. pistillata anatomy and tissue architecture with the major sectioning planes used for ISH experiments. (H) RNA ISH on adult tissue sections showing the expression of selected marker genes for epidermis, gastrodermis, digestive filaments, neuron_3 and gland_1. Bar plots shows the expression across metacells of the selected markers (molecules/1,000 unique molecular identifiers [UMIs]). Scale bars, 50 μm. Ap, actinopharynx; Cal, calicodermis; DF, digestive filaments; Ep, epidermis; Ga, gastrodermis; GCn, gastrovascular canal; GCv, gastrovascular cavity; GHC, gastrodermal host cell; Me, mesoglea; MF, mesenterial filaments; Sk, skeleton; Ss, skeleton spine; Sp, spermary; Te, tentacle. See also Figures S1, S2, and S3 and Tables S1, S2, and S3.",
"discussion": "Discussion The whole-organism multi-stage S. pistillata cell atlas presented here uncovers the diversity of cell types in stony corals, advancing our understanding of the molecular pathways and regulatory programs involved in reef formation. Despite their ecological importance and decades of study, stony corals are still far from becoming stable model species, in part due to the difficulty of growing and reproducing them in laboratory conditions. A recent important development has been the first proof-of-concept application of CRISPR/Cas9 for gene knockout in a stony coral ( Cleves et al., 2018 ). This opens the window to functional studies, but the definition and prioritization of target study genes is still a major limitation ( Cleves et al., 2020 ). Our S. pistillata atlas constitutes a valuable resource in this direction, providing hundreds of marker genes for dozens of specific cell types. For example, we found that alga-hosting cells express genes associated with the lysosomal compartment and also genes involved in diverse lipid metabolic processes (from transport to fatty acid elongation and also a lipid metabolism-related TF- USF1 ), in protection against oxidative stress, in galactose metabolism, and in transport of other different metabolites (sugars, amino acids, ammonium, etc.). Similarly, we could identify genes overexpressed in the calicoblasts of skeleton-producing polyps, including diverse carbonic anhydrases, which catalyze the interconversion of CO 2 to carbonate and a proton (H + ) necessary to biomineralization process; as well as and acid-rich proteins, which are thought to be involved in calcium carbonate precipitation ( Mass et al., 2013 ). Finally, we identify two putative immune cell types in corals, characterized by expression of genes typically associated with immune signaling pathways (interleukin receptors, STING , and NOD-like receptors), antimicrobial responses (e.g., perforins, LPS binding proteins, and tyrosinase), and immune cell identity transcription factors ( NFAT and several interferon regulatory factors). The S. pistillata cell atlas also allowed us to initiate the comparative analysis of cnidarian cell types by phylogenetic integration of other cnidarian cell atlases. This multi-species single-cell comparison provides systematic evidence of the evolutionary conservation of major cnidarian cell type programs (e.g., cnidocytes, neurons, secretory/gland cells, digestive filament cells, and gastrodermal cells). Beyond that, we found evidence of deep conservation in sperm and immune cell type transcriptional programs between corals and other animal lineages, reflecting highly specialized and ancient effector gene repertoires. In contrast with these conserved expression programs, our study also traces the emergence of stony coral cell type novelties: calicoblasts and alga-hosting cells. The calicoblasts emerged from epidermal cells during scleractinian evolution and this was likely a pivotal event in the origin of coral reef ecosystems. Reconstructing the detailed molecular changes involved in the emergence of the calicoblast transcriptional program, however, will require the analysis of multiple additional cnidarian species, especially other scleractinians. In the case of S. pistillata alga-hosting cells, they transcriptionally resemble S. pistillata gastrodermal cells, and this pattern is paralleled in Xenia sp., a distantly related symbiotic cnidarian species. This suggests the independent evolution of dinoflagellate symbiosis in these two cnidarians, which is further supported by the apparent promiscuity of dinoflagellate symbioses across animal phyla ( Melo Clavijo et al., 2018 ). Interestingly, some common transcriptional signatures are shared between alga-hosting cells in both cnidarian species, particularly genes related to lipid metabolism. Some of these shared genes, like NPC sterol transporters, have also been found associated to symbiosis in the anthozoan anemone Exaiptasia pallida ( Hambleton et al., 2019 ). Future single-cell sampling efforts in cnidarian and non-cnidarian symbiotic species should elucidate the convergent evolution of dinoflagellate symbioses across animals. Single-cell transcriptomics has emerged as a powerful tool to explore the diversity of cell types in non-traditional model animal species ( Fincher et al., 2018 ; Musser et al., 2019 ; Sebé-Pedrós et al., 2018a , 2018b ; Siebert et al., 2019 ; Sladitschek et al., 2020 ). The increase in cell maps for phylogenetically diverse species should enable the comparative study of how animal cell type programs emerged and evolve ( Arendt et al., 2016 ). However, single-cell taxon sampling efforts are almost exclusively limited to species that can be grown in laboratory conditions. Consequently, the number of metazoan species and phyla with cell atlases available remains surprisingly small, especially if we compare it with genome data availability ( Dunn and Ryan, 2015 ). In this context, our study shows the power and feasibility of single-cell analysis in species sampled from the wild. As exemplified here, and together with improvements in sampling strategies ( García-Castro et al., 2020 ), we anticipate that a phylogenetically rich animal cell type tree of life should be within reach in the coming years. Overall, the S. pistillata cell atlas lays the foundations for a system-level molecular understanding of reef-building stony corals. This will empower the design and interpretation of studies on how environmental stressors linked to global change alter the normal function of stony coral cells. Beyond that, our cellular roadmap should enable the development of targeted strategies to improve coral resilience to global change, ultimately impacting the reef ecosystems that depend on stony coral health. Limitations of study Our study provides a reference molecular map of transcriptional cell states in Stylophora pistillata . There are several preliminary observations reported here that will deserve follow-up analyses to extend our understanding of stony coral cell biology. Among these, we want to highlight: (1) the involvement of Ntox44 homologs in polyp settlement, including the structure and function of this poorly characterized gene family that has been horizontally transferred from bacteria to scleractinian corals; (2) the role of the opsin gene ortholog expressed in algal-hosting cells, potentially coupling symbiont photosynthesis with host cell metabolic states; and (3) the tissue distribution and function of the immune cells identified here. Finally, we did not identify any adult progenitor or stem cell-like populations in our atlas. Future sampling efforts will be required to try to identify these cells in S. pistillata and other stony corals."
} | 3,333 |
26848568 | PMC4746064 | pmc | 931 | {
"abstract": "Background The overall metabolic/functional potential of any given environmental niche is a function of the sum total of genes/proteins/enzymes that are encoded and expressed by various interacting microbes residing in that niche. Consequently, prior (collated) information pertaining to genes, enzymes encoded by the resident microbes can aid in indirectly (re)constructing/ inferring the metabolic/ functional potential of a given microbial community (given its taxonomic abundance profile). In this study, we present Vikodak—a multi-modular package that is based on the above assumption and automates inferring and/ or comparing the functional characteristics of an environment using taxonomic abundance generated from one or more environmental sample datasets. With the underlying assumptions of co-metabolism and independent contributions of different microbes in a community, a concerted effort has been made to accommodate microbial co-existence patterns in various modules incorporated in Vikodak. Results Validation experiments on over 1400 metagenomic samples have confirmed the utility of Vikodak in (a) deciphering enzyme abundance profiles of any KEGG metabolic pathway, (b) functional resolution of distinct metagenomic environments, (c) inferring patterns of functional interaction between resident microbes, and (d) automating statistical comparison of functional features of studied microbiomes. Novel features incorporated in Vikodak also facilitate automatic removal of false positives and spurious functional predictions. Conclusions With novel provisions for comprehensive functional analysis, inclusion of microbial co-existence pattern based algorithms, automated inter-environment comparisons; in-depth analysis of individual metabolic pathways and greater flexibilities at the user end, Vikodak is expected to be an important value addition to the family of existing tools for 16S based function prediction. Availability and Implementation A web implementation of Vikodak can be publicly accessed at: http://metagenomics.atc.tcs.com/vikodak . This web service is freely available for all categories of users (academic as well as commercial).",
"introduction": "Introduction The field of metagenomics has significantly improved our overall understanding of the microbial world within and around us. Characterizing microbial communities using the metagenomics approach involves either (a) performing a whole genome (shotgun) sequencing (referred to as WGS) of the entire genomic content of a given environmental sample, or (b) performing amplicon sequencing of specific marker genes (e.g. 16S rRNA) from any microbial community. The focus/end-objectives of the research problem (as well as associated sequencing costs) typically drive the choice between adopting a WGS and a 16S amplicon sequencing approach. The 16S approach specifically helps in deciphering the taxonomic/microbial composition of a given environmental sample, thereby paving the way for performing specific diversity analysis and subsequent identification of environment-specific marker microbes. The WGS approach, on the other hand, offers a two-fold advantage. Besides helping in obtaining a taxonomic profile of the environmental sample under study, computational analysis of WGS data provides information at a functional level (i.e. types and relative abundances of genes encoded by various microbes in a given environment). Given the relatively higher costs associated with WGS sequencing (and the computational complexity of handling huge volumes of WGS data), a majority of metagenomic initiatives employ 16S sequencing for obtaining a quick comparative snap-shot of microbial diversity. Subsequently, a WGS follow-up experiment is sometimes performed (on a relatively smaller subset of samples), typically when 16S experiments indicate significant aberrations (in microbial composition) in the analysed sample classes (e.g. between healthy and diseased states). Metagenomic data analysis of microbial communities is typically aimed at (a) identifying the resident microbes and their relative proportions (i.e. taxonomic analysis), (b) profiling the functions encoded by these microbes (i.e. functional annotation), and (c) comparing/ co-relating the identified microbes and their functions with available sampling (phenotypic) metadata (i.e. comparative metagenomics). Several stand-alone tools and web-services are available for performing the above mentioned analyses [ 1 – 5 ]. Currently, the number of existing tools/ platforms for functional characterization of a metagenomic environment is relatively lower that that available for taxonomic analysis and comparative metagenomics. This scenario is however changing due to the emergence (and successful validation) of a new line of thought that proposes inferring/ (re)constructing the metabolic/ functional potential (of a given environmental sample) using its 16S taxonomic abundance profile. The rationale behind this line of thought is based on the fact that the overall metabolic/ functional potential of any given environmental niche is a function of the sum total of genes/ proteins/ enzymes that are encoded by various microbes residing in that niche. Consequently, a pre-computed database containing collated information pertaining to genes, enzymes encoded by various microbes (known till date) can potentially be employed for indirectly inferring/ (re)constructing the metabolic/ functional potential of a microbial community based on its 16S taxonomic abundance profile. The idea of inferring the functional potential using 16S-derived microbial abundance data has been attempted by a handful of research groups. For example, approaches like procrustes analysis [ 6 ] and ancestral-state reconstruction [ 2 ] have been successfully applied (and validated) for predicting functional potential of a microbial community from its 16S rRNA sequence data. Tools like METAGENEassist [ 4 ] have furthered this premise by extending the idea to prediction of phenotypic traits and physiological functions such as pH, oxygen requirement etc., of an environmental sample by analysing its microbial composition. Functional characterization of any metagenomic sample should ideally enable insights into (i) functional units (genes/ enzymes) expressed by resident microbes, (ii) relative abundance and expression levels of various metabolic pathways, (iii) relationships between various resident microbes based upon their individual functional characteristics, (iv) core functions of an environment, (v) differences between distinct environments, and (vi) genes/ enzymes associated with individual functions (viz. metabolic pathways). In addition, the functional profile should be devoid of functions specific to eukaryotes. While the first two objectives of an ideal functional characterisation are fulfilled by the conventional WGS-based functional annotation tools [ 3 , 5 ] and the recently developed 16S-based functional prediction algorithms [ 1 , 2 ], a majority of other objectives have not been addressed in the current state-of-art. In this study, we describe Vikodak ( decoder in Sanskrit)—a modular functional annotation tool that extends the functionality/ utility of the '16S-data-inferred-function-prediction' paradigm. A detailed description of (a) various modules of Vikodak (and the algorithmic work-flow they employ), and (b) additional features incorporated in Vikodak that address the unmet objectives of existing 16S-based functional prediction algorithms is provided. In addition to the validation of results obtained with greater than 1400 metagenomic samples, a case study (performed using publicly available Periodontitis metagenomics datasets) that highlights the functional utility of various modules incorporated in Vikodak is also presented.",
"discussion": "Discussion In addition to providing information about the presence/ abundance of various genes/ proteins/ metabolic-pathways encoded by various microbes inhabiting a given environment, an ideal function characterization tool should be able to provide insights pertaining to the functional interactions between the resident bacteria. A comparison of such information generated from two or more environments is expected to provide insights relevant from a biological standpoint. Vikodak—the tool presented in this study attempts to automate the overall process of obtaining and/ or comparing the functional profiles(s) of given environment(s). Fig 10 schematically depicts various modules and associated utilities of Vikodak. 10.1371/journal.pone.0148347.g010 Fig 10 Schematic summary of Vikodak. A schematic depiction of the overall workflow and application(s) of Vikodak. Three distinct functional modules, viz. Global Mapper, Inter Sample Feature Analyzer (ISFA), and Local Mapper, each catering to specific end-user requirements are depicted. Given a microbial abundance data profile of an environmental sample (e.g. 16S sequencing data classified using RDP classifier), the Global Mapper module enables (a) an in silico estimation of the relative abundance of various metabolic pathways in that sample, (b) quantifying the contribution of individual microbes (in that sample) to the predicted functions (at all three tiers of KEGG hierarchy), and (c) identification of the core set of metabolic functions defining a particular environment. The ISFA module in Vikodak is an extension of the Global Mapper module and is designed for performing a rigorous (pair-wise) comparative statistical analysis of the (inferred/predicted) functional profiles generated from two or more environments. The Local Mapper module further enables end-users to probe, in greater detail, the enzyme abundance profile(s) of individual metabolic pathway(s) identified as (a) the 'core' in one or more environments, or (b) differentially abundant between two or more environments. In comparison to existing functional annotation tools, the functional profiles obtained using the 'Global Mapper' module of Vikodak contain detailed information with respect to the core set of functions as well as the contribution of individual microbes to various functions identified in the studied environment(s). This facilitates end-users to decipher functional relationships between various resident microbes. Furthermore, the 'ISFA module' of Vikodak automates the functional comparison between a pair of environments. Given that the comparison work-flow employs a boot-strap approach (and also provides end-users the option to choose appropriate statistical thresholds), Vikodak represents a reliable one-stop package that infers as well as compares (in rigorous statistical terms) the functional profiles of the studied environment(s). The 'Local Mapper' module further complements the functionalities of the other two modules by enabling an in-depth (enzyme-level) analysis of specific metabolic pathways that are of interest to the end-user. In ideal circumstances, the process of inferring the presence of a pathway in an environmental sample should necessarily take into consideration the occurrence of a minimum quorum of genes/ enzymes encoding the pathway. For instance, it is inappropriate to report the 'presence' of a pathway (constituted of 20 enzymes) based on finding just 1 or 2 enzymes in the studied environment. The PEC value parameter in Vikodak duly takes care of this requirement, thereby enabling end-users to focus only on results which are bereft of possible false positive predictions. Comparison of microbial communities from two environments (using Vikodak's ISFA module) and identifying functions that are reported as significantly different at most PEC thresholds ( Fig 6 ) greatly increases the confidence in the set of functions identified as significantly different. On a conceptual note, Vikodak considers enzymes as the basic functional units. This in turn has enabled the development of a relatively simplistic (yet effective) modules catering to various aspects of function prediction. The back-end data in Vikodak is therefore based on a compilation of EC copy numbers (from greater than 37000 bacterial genomes) sourced from two major data repositories viz. PATRIC and IMG [ 8 , 9 ]. Given that current taxonomic classification algorithms [ 34 – 39 ] provide assignments at various levels of taxonomic hierarchy, the consensus mapping data in Vikodak has therefore been computed at all taxonomic levels. Vikodak is enabled to process taxonomic abundance data (generated using any of the currently available classification tools). With its additional provisions for comprehensive functional analysis/ comparison, deeper insights and more flexibility at the user end, Vikodak is expected to add value to the family of currently available tools for 16S based function prediction."
} | 3,196 |
36926689 | PMC10011134 | pmc | 932 | {
"abstract": "Biosorption of metal ions by phototrophic microorganisms is regarded as a sustainable and alternative method for bioremediation and metal recovery. In this study, 12 cyanobacterial strains, including 7 terrestrial and 5 aquatic cyanobacteria, covering a broad phylogenetic diversity were investigated for their potential application in the enrichment of rare earth elements through biosorption. A screening for the maximum adsorption capacity of cerium, neodymium, terbium, and lanthanum was conducted in which Nostoc sp. 20.02 showed the highest adsorption capacity with 84.2–91.5 mg g -1 . Additionally, Synechococcus elongatus UTEX 2973, Calothrix brevissima SAG 34.79, Desmonostoc muscorum 90.03, and Komarekiella sp. 89.12 were promising candidate strains, with maximum adsorption capacities of 69.5–83.4 mg g -1 , 68.6–83.5 mg g -1 , 44.7–70.6 mg g -1 , and 47.2–67.1 mg g -1 respectively. Experiments with cerium on adsorption properties of the five highest metal adsorbing strains displayed fast adsorption kinetics and a strong influence of the pH value on metal uptake, with an optimum at pH 5 to 6. Studies on binding specificity with mixed-metal solutions strongly indicated an ion-exchange mechanism in which Na + , K + , Mg 2+ , and Ca 2+ ions are replaced by other metal cations during the biosorption process. Depending on the cyanobacterial strain, FT-IR analysis indicated the involvement different functional groups like hydroxyl and carboxyl groups during the adsorption process. Overall, the application of cyanobacteria as biosorbent in bioremediation and recovery of rare earth elements is a promising method for the development of an industrial process and has to be further optimized and adjusted regarding metal-containing wastewater and adsorption efficiency by cyanobacterial biomass.",
"conclusion": "5 Conclusion In this study, a diverse group of 12 cyanobacteria was investigated for their potential in the enrichment of REE in a biosorption process. Metal uptake varied strongly among the tested strains, with Nostoc sp. 20.02 showing the highest maximum adsorption capacity of 84.2–91.5 mg g -1 . However, there was no apparent correlation between maximum adsorption capacity and phylogenetic relationship nor for the ecological habitat of the strains. This could be explained by variations in the composition of metal interacting functional groups located at the cell surface. Moreover, many cyanobacteria that showed high adsorption capacities for REE produce extracellular polymeric substances (EPS) that are known to facilitate metal adsorption ( Pagliaccia et al., 2022 ). The composition of these EPS and their influence on the adsorption of REE should be further investigated in future studies. The determination of relevant parameters for improving the metal uptake revealed a pH optimum at 5 to 6 and fast adsorption kinetics reaching adsorption equilibrium within an incubation time of a few minutes. In addition, metal analysis strongly indicated an ion-exchange mechanism during the biosorption process in which Na + , K + , Mg 2+ , and Ca 2+ ions are replaced by metal cations that bind to the surface of the biomass. These observations are in accordance with previous studies that were conducted on algal, bacterial, and other biomasses ( Acheampong et al., 2011 ; Sulaymon et al., 2013 ; Liang and Shen 2022 ). The isolation of single target elements in a technical biosorption process remains a challenging task due to the complex surface structure and the heterogeneity of functional groups. Nevertheless, based on the results of this study, the enrichment of metal elements from diluted solutions is possible. For the development of an industrial process, parameters need to be further optimized and adjusted depending on the metal composition in the wastewater and the biomass that is used as biosorbent.",
"introduction": "1 Introduction Rare Earth Elements (REE) consist of scandium, yttrium, and 15 elements of the lanthanide series. These elements have exceptional electromagnetic, catalytic, and optical properties making them crucial for the production and development of modern high-technology products. Due to their similar chemical properties, separating REE demands sophisticated industrial processes that are energy-intensive and use environmentally toxic chemicals ( Haque et al., 2014 ). Standard methods, for example, apply metal leaching with acids or bases and extraction methods to purify REE ( Opare et al., 2021 ). Moreover, REE production is focused on a few countries, resulting in an oligopoly that can dictate supply and price regimes. REE are crucial for technology transition towards a renewable energy-driven society. For instance, cerium or lanthanum have applications in catalysts for air purification or chemical processing. Other metals like neodymium or terbium are crucial for producing permanent magnets or modern LEDs ( Charalampides et al., 2015 ; Shan et al., 2020 ). Hence, industrialized countries increasingly focus on alternative supply routes and the development of cost and ecologically compatible recycling routes. In this context, REE recovered from dilute mining or industrial wastewater, as well as, electronic waste streams are opening new, regional supply routes. Establishing new biotechnologically based REE recovery methods therefore leads to enhanced market stability and supply chain independence for industrialized regions, such as the EU. Hence, there is a growing interest in the recovery and recycling of REE from industrial wastewater streams ( Li et al., 2013 ; Barros et al., 2019 ). Over the past decades, biosorption has been regarded as a relatively simple and cost-efficient method for wastewater treatment ( Volesky 2001 ). It is a physicochemical process that involves a solid phase (biosorbent) consisting of organic biomass and a liquid phase containing the dissolved or suspended chemical compounds to be sorbed (sorbate) ( Fomina and Gadd 2014 ). Biosorption has a wide range of potential applications in wastewater remediation, including the removal of organic substances like dyes, pharmaceuticals, or pesticides ( Bell and Tsezos 1987 ; Aksu 2005 ; Crini and Badot 2008 ; Menk et al., 2019 ). However, most research on biosorption in conjunction with the removal of pollutants has been conducted on metals, including heavy metals, actinides, and lanthanides ( Dhankhar and Hooda 2011 ; Abbas et al., 2014 ; Giese 2020 ; Mattocks and Cotruvo 2020 ). Yet, developed processes based on biosorption have not achieved a commercial breakthrough. For example, it has been shown that environmental factors, such as changes in the pH value, can alter the affinity of biomass towards different elements ( Zinicovscaia et al., 2019 ). A low technology readiness level, including a poor understanding of the underlying mechanisms, kinetics, and thermodynamics of the process are areas that require more research ( Fomina and Gadd 2014 ; Elgarahy et al., 2021 ). It is widely accepted that the chemical structure, in particular the composition of functional groups on the cell surface, profoundly influences the adsorption properties of biomass ( Eccles 1999 ; Volesky 2007 ). These active moieties may include hydroxyl-, carboxyl-, carbonyl-, phosphate-, sulfonate-, amine-, amide-, and imide-groups, among many others. Studies on biological, physical, or chemical modification of biomass by adding functional groups have shown that it is possible to improve binding specificity and capacity for target sorbates ( Wang and Chen 2006 ; Park et al., 2010 ; Abdolali et al., 2015 ; Ciopec et al., 2020 ). Especially the recovery of REE with chemically modified organic polymers has been the focus of recent studies ( Gabor et al., 2017 ; Negrea et al., 2018 ; Negrea et al., 2020 ). Nevertheless, these resulting biosorbents are still inferior in target selectivity to chemically synthesized ion-exchange resins with a defined structure and composition ( Gadd 2009 ). Due to the heterogeneity of functional groups on the cell surface of microbial biomass, binding specificity for elements remains a challenging factor for industrial applicability. The adsorption of heavy metals by eukaryotic algae and cyanobacteria is well documented ( Al-Amin et al., 2021 ; Ankit et al., 2022 ). At present, the screening of new species regarding biosorption and potential novel applications in metal recovery remains of great interest due to high variability in cell wall composition and structure, resulting in differences in adsorption properties [e.g. ( Micheletti et al., 2008 )]. Cyanobacteria have shown promising adsorption properties for heavy metals, which could be used in the sequestration of metals from water on a technical scale. If similar adsorption properties exist for the bioremediation of REE has not been studied extensively yet. Moreover, the adsorption properties of terrestrial cyanobacteria were seldom investigated. Therefore, we taxonomically and biotechnologically identified new and promising cyanobacterial strains and evaluated their properties for REE adsorption. In this context, we also aimed to correlate taxonomic identity and adsorption characteristics. In this study, 12 cyanobacterial strains with broad phylogenetic origin and inhabiting different ecological habitats such as terrestrial, freshwater, and saltwater habitats were investigated for their potential applicability in an adsorption process for the enrichment of REE. Their phylogenetic relationship was determined using 16S rRNA sequences. The screening for maximum adsorption capacity with four different REE (i.e. lanthanum, cerium neodymium, and terbium), as well as the effect of several parameters on biosorption, including initial pH value, incubation time, and metal concentration for cerium, were evaluated. Additionally, binding specificity for cerium in the presence of other metal cations was investigated.",
"discussion": "4 Discussion 4.1 Phylogenetic and taxonomical remarks In the broad context of biotechnology, cyanobacterial strains are often used without respecting their ecological niche. This is a problem, because some taxa e.g. from aquatic habitats, often cannot be used during biotechnological processes that involve heat or desiccation, while others, such as terrestrial strains, are better candidates and vice versa . In addition, it happens quite often that results are not linked to strain identifiers or to wrongly identified taxa what can lead to an incorrect comparison and interpretation of data—a mistake that can remain uncorrected over decades (e.g., Jung et al., 2021b ). For these reasons we respected the ecology of the strains used in this study and depicted the phylogenetic placement of the strains. This creates a transparent background for the cyanobacterial strains that we used and allows others to better compare their results. Besides publicly available cyanobacterial strains with a clarified identity, several new isolates were phylogenetically analyzed during this work based on their 16S rRNA gene region ( Figure 1 ). Among these were, for example, the heterocytous, false-branching strain S. hyalinum 02.01 that joined the large S. hyalinum cluster as outlined by Johansen et al., ( Johansen et al., 2017 ). In addition, the two true-branching, heterocytous strains Symphyonema bifilamentata 97.28 and Reptodigitus sp . 92.1 were included in the study in order to complement the setup of heterocytous, branching cyanobacteria. The strain 97.28 was treated as Fisherella ambigua for the last 50 years of biotechnological research on secondary metabolites but was recently re-assigned as the type strain of the genus Symphyonema ( Jung et al., 2021b ). This strain has great biotechnological potential, because it grows fast and produces a diverse set of secondary metabolites, such as various ambigols (summarized in ( Jung et al., 2021b )). The strain 92.1 was formerly treated as Nostochopsis lobatus, but doubts about this assignment arose because N. lobatus is only known from aquatic habitats. Recently, the new genus Reptodigitus was emerged, and the authors pointed out that strain 92.1 needs to be correctly described as a novel Reptodigitus species ( Casamatta et al., 2020 ) which the authors of this study will carry out in a follow up study. In contrast to the above named strains, which are low producers of EPS (extracellular polymeric substances), the genus Komarekiella and related genera are well known to produce cells and filaments covered by thick EPS sheaths ( Scotta Hentschke et al., 2017 ; Soares et al., 2021 ). EPS might play a role in metal adsorption (e.g. Al Amin et al., 2021 ). However, the two strains investigated here are the first strains of this genus described from a desert environment, while the other species of the genus have multiple origins, including lichen symbioses ( Jung et al., 2021a ; Soares et al., 2021 ; Panou and Gkelis 2022 ). All of them have a very complex life cycle in common that can hamper biotechnological applications due to different metabolic activity depending on the developmental stage of the culture. Also, the two strains 90.01 and 89.12 will be described as new species in the future. More challenging to interpret are the phylogenetic and taxonomical positions of Nostoc sp. 20.02 and C. brevissima SAG 34.79 ( Figure 1 ). The strain 20.02 was isolated as an epiphyte on a cyanolichens and can be considered as a Nostoc strain not involved in the symbiosis because most true Nostoc lichen photobionts usually join distinct Nostoc ‘photobiont clusters’ based on their 16S rRNA ( O'Brien et al., 2005 ). The overall taxonomic position of this strain remains unsure as it also does not cluster within the Nostoc sensu stricto clade. Similar uncertainties affect strain SAG 34.79 that could be assigned to C. brevissima based on its morphology and phylogenetic position, although there is no cohesive cluster formed and no type strain for the genus deposited. Closely related strains such as Tolypothrix tenuis SAG 94.79, Scytonema mirabile SAG 83.79, and T. tenuis J1 ( Figure 1 ) need further investigation to clarify the state of the genus. Calothrix exhibits a notorious morphological heterogeneity and extreme polyphyly, which is evident from various independent clades in the phylogenetic trees of past research [reviewed in ( Nowruzi and Shalygin 2021 )]. However, even if no phylogenetic or habitat correlation with adsorption capacity could be found, biotechnological studies of cyanobacterial strains should be more often accompanied with phylogenetical studies applying the current standard for taxonomical classification by the so called polyphasic approach ( Komárek et al., 2014 ) to identify taxonomic rearrangements and to avoid confusion regarding species names and strain names from culture collections for biotechnology. 4.2 Metal adsorption experiments For microalgae, the bioremediation, bioaccumulation, or biosorption of common heavy metals such as Pb, Cd, Cr, As, Hg, Ni, etc. is often studied [e.g. ( Ahuja et al., 1999 ; Ç etinkaya Dönmez et al., 1999 ; Mehta and Gaur 2005 )]. The mechanisms behind these adsorption processes vary with species and environmental conditions ( Kumar et al., 2015 ). However, different mechanisms are discussed, such as ion exchange, complexation, electrostatic attraction, and micro-precipitation ( Kumar et al., 2015 ; Yadav et al., 2021 ). In contrast, the biosorption of REE is studied less. For the adsorption process of REE, the results in this study indicate an ion-exchange mechanism in which cations of alkaline and alkaline earth metals (Na, K, Mg, and Ca) are replaced by other metal cations during the biosorption process with cyanobacterial biomass ( Figure 6 ). This is in agreement with previous experiments using biomass of different microorganisms ( Crist et al., 1994 ; Matheickal et al., 1997 ; Sulaymon et al., 2013 ; Liang and Shen 2022 ). Ion exchange has been proposed as a dominant mechanism during biosorption ( Chen et al., 2002 ; Iqbal et al., 2009 ). Apart from Synechocococcus elongates UTEX 2973 biomass , sodium was the predominant element during the ion exchange process. This differs from previous reports in which cations of earth alkaline metals were released in higher percentages ( Iqbal et al., 2009 ; Sulaymon et al., 2013 ). Additionally, studies reported the replacement of protons with metal cations leading to a decrease in pH during the sorption process ( Mashitah et al., 1999 ; Vasudevan et al., 2002 ). However, this aspect was not focused on in the experimental setup of this study. The strong influence of pH value on metal uptake shown in this study further emphasizes the correlation between charges on the surface of the biosorbent and the adsorbed metal ions. In previous studies, the effect of pH value on biosorption has been confirmed ( García-Rosales et al., 2012 ; Abdel-Aty et al., 2013 ). At low pH values, functional groups on the cell surface are either neutral or positively charged. Carboxyl groups for instance are protonated at pH values below 3, whereas amino groups are protonated at pH 4.1 ( Eccles 1999 ). As similar charges create a repulsive force, positive charges on the biomass surface repel metal cations, leading to poor metal uptake at low pH values. Previous studies described a strong influence of hydroxyl and carboxyl groups on the adsorption process for different biomasses ( Gupta and Rastogi 2008 ; Luo et al., 2010 ; Utomo et al., 2016 ). Experiments on adsorption kinetics showed a quick metal uptake for all tested biomasses, reaching equilibrium within only a few minutes. In general, the process of metal cations attaching to adsorbents with a mesoporous surface involves two stages ( Zinicovscaia et al., 2021 ). Specifically, the steps involve the migration of ions from the main solution to the boundary layer surrounding the intermediate-pore matrices, and the attachment of the metal ions to the active sites of the adsorbent material via adsorption. Previous studies have reported fast kinetics for the adsorption of metals on biomass of other green algae and cyanobacteria ( Klimmek et al., 2001 ). On the other hand, experiments in other studies resulted in incubation times of up to 60 min and more before reaching the maximum adsorption equilibrium ( Ahuja et al., 1999 ; Zinicovscaia et al., 2017 ). Fast metal uptake is a beneficial factor for the process development beyond laboratory scale as long incubation periods can be avoided, and higher flow rates can be achieved. Adsorption experiments with equimolar mixed-metal solutions were carried out, revealing a preference for certain elements influenced by the total metal concentration. The tested biomasses showed the highest overall adsorption capacity for Ce 3+ at low metal concentrations. However, cations of these elements were replaced by Pb and Al at higher metal concentrations (2–4 mM) in this experimental setup. Zn and Ni showed to lowest affinity to the tested biomasses. Similar results have been reported for biomass of other microorganisms ( Klimmek 2003 ; Wilke et al., 2006 ; Huang et al., 2018 ). At present, our ability to make predictions on binding specificity based on single-element adsorption experiments is limited ( Wilke et al., 2006 ). Regarding a potential industrial application for the recovery of REE, these are promising results, as metal concentrations usually are lower than the highest concentrations in the experimental setup of this study. Furthermore, it should be considered that this study predominately focused on the adsorption of the element cerium. Due to high chemical similarities between REE, it is likely that the adsorption properties of the tested biomasses will be similar for other elements of this group. Nevertheless, additional experiments with other REE are advisable. Target elements could be extracted from the resulting metal-loaded biomass in follow-up processes. The destructive recovery by combustion, resulting in metal-enriched ash, is a simple method with the drawback of losing the initial biomass. An economically more desirable approach is the targeted desorption of elements from loaded biomass, enabling the recycling of the biosorbent. Previous studies have tested various approaches using different acids or complexing agents ( Gong et al., 2005 ; Abdolali et al., 2015 ). Unfortunately, the adsorption properties of biosorbents are impaired over the curse of a few cycles ( Hammaini et al., 2007 ). Future studies should address the binding specificity and durability of biosorbents to implement biosorption in industrial processes successfully. In competitive systems, the adsorption of different metal cations on biomass is influenced by functional groups on the cell surface. The interaction between metal cations and functional groups still requires more research. According to the current state of knowledge, various ionic properties of metal cations, such as electronegativity, redox potential, and ionic radius can influence the adsorption on biomass ( Naja et al., 2010 ). Depending on the biomass and physico-chemical conditions, multiple mechanisms may be involved in metal sorption simultaneously ( Gadd 2009 ). With respect to different cyanobacterial strains, FT-IR analysis indicated the involvement of various functional groups during like hydroxyl or carboxyl groups during metal adsorption. However, at present, there is no discrete chemical entity that has been identified as dominant cell wall feature that governs metal binding. In a previous study, for instance, it was shown that complex polymeric sugars are involved in the adsorption of terbium by C. brevissima ( Jurkowski et al., 2022 ). Cell wall-derived binding entities most likely vary for every organism and metal presented."
} | 5,465 |
33893309 | PMC8065059 | pmc | 933 | {
"abstract": "Geothermal environments, such as hot springs and hydrothermal vents, are hotspots for carbon cycling and contain many poorly described microbial taxa. Here, we reconstructed 15 archaeal metagenome-assembled genomes (MAGs) from terrestrial hot spring sediments in China and deep-sea hydrothermal vent sediments in Guaymas Basin, Gulf of California. Phylogenetic analyses of these MAGs indicate that they form a distinct group within the TACK superphylum, and thus we propose their classification as a new phylum, ‘Brockarchaeota’, named after Thomas Brock for his seminal research in hot springs. Based on the MAG sequence information, we infer that some Brockarchaeota are uniquely capable of mediating non-methanogenic anaerobic methylotrophy, via the tetrahydrofolate methyl branch of the Wood-Ljungdahl pathway and reductive glycine pathway. The hydrothermal vent genotypes appear to be obligate fermenters of plant-derived polysaccharides that rely mostly on substrate-level phosphorylation, as they seem to lack most respiratory complexes. In contrast, hot spring lineages have alternate pathways to increase their ATP yield, including anaerobic methylotrophy of methanol and trimethylamine, and potentially use geothermally derived mercury, arsenic, or hydrogen. Their broad distribution and their apparent anaerobic metabolic versatility indicate that Brockarchaeota may occupy previously overlooked roles in anaerobic carbon cycling.",
"introduction": "Introduction Advances in DNA sequencing and computational approaches have accelerated the reconstruction of metagenome assembled genomes (MAGs) from natural communities 1 . This approach has revealed many novel lineages on the tree of life and is advancing our understanding the ecological roles of uncultured microbes 1 – 3 . For example, many new archaeal phyla have been described from hot springs including Geoarchaeota 4 , Marsarchaeota 5 , Aigarchaeota 6 , and several Asgard phyla from deep-sea hydrothermal vents 7 – 12 . However, diversity surveys have demonstrated there are many novel taxa left to be explored 13 . Moreover, there are several gaps between our knowledge of active biogeochemical processes and the metabolic mechanisms and taxa mediating them. For example, the description of microbes mediating anaerobic methylotrophy is still limited, and it is unclear which non-methanogenic heterotrophs utilize methylated compounds on the anoxic seafloor 14 . Little is known about the microorganisms or pathways mediating this process 15 . Methylotrophs are organisms that are capable of using simple organics including single-carbon (C1 e.g., methanol) and methylated (e.g., trimethylamine) compounds as a source of energy and carbon 16 , 17 . In nature, the most prevalent are compounds such as methanol and methylamines, which are derived from a variety of sources such as phytoplankton, plants, and the decay of organic matter 15 , 18 , 19 . As a result, they are ubiquitous in oceans and atmosphere and are important components of the global carbon and nitrogen cycles 15 . In oxic environments, methanol is converted to formaldehyde by the classical pyrroloquinoline quinone (PQQ)-linked methanol dehydrogenase pathway found in aerobic methylotrophs 15 , 18 . In anoxic settings, these compounds are used as substrate for methylotrophic-methanogenesis 20 – 23 and sulfate reduction 24 . Anaerobic methylotrophs utilize the methyltransferase system (MT) to break and transfer the methyl residue to coenzyme M (in the case of methanogens) or tetrahydrofolate (H 4 F) (in acetogens and sulfate reducers) 20 – 24 and conserved energy via the Wood–Ljungdahl pathway (WLP). Methylotrophic archaea include methanogenic orders (in Euryarchaeota): Methanosarcinales , Methanobacteriales , Methanomassiliicoccales , and the recently discovered uncultured methylotrophic phylum, Verstraetearchaeota 20 . Methylotrophy has not been described in archaeal lineages outside of these methanogenic groups. Here we describe a new archaeal phylum, the Brockarchaeota, whose members are metabolically versatile and can be found in geothermal environments around the world. The Brockarchaeota appear to possess diverse pathways for carbon cycling including fermentation of complex organic carbon compounds, anaerobic methylotrophy, and chemolithotrophy.",
"discussion": "Discussion Brockarchaeota gene content indicates they are facultative or obligate anaerobic fermentative organisms that produce acetate, CO 2 , and H 2 as byproducts (see Supplementary Information for details). Some Brockarchaeota have unique pathways for non-methanogenic methylotrophy. This puts them a unique ecological position in nature, where they degrade abundant methylamines in anoxic environments without the production of methane (Fig. 6 ). Brockarchaeota are also able to degrade complex carbon compounds such as xylan. Xylans are a major structural polysaccharide in plant cells, and is the second most abundant polysaccharide in nature, accounting for approximately one-third of all renewable organic carbon on Earth after cellulose 54 , 55 . This suggests that Brockarchaeota are players in organic matter degradation in geothermally active environments. Interestingly, detrital proteins can be used as a substrate by Brockarchaeota, indicating potential role in protein remineralization in geothermally active environments. Fig. 6 A model of the biogeochemical roles of Brockarchaeota in the anaerobic carbon cycle. C1 and methylated compounds, such as methanol or methylamines, are utilized biologically as carbon and energy sources in the ocean and deep-sea sediments resulting in a considerable carbon reservoir. The biodegradation of organic carbon in the water column and subsurface is a source of these compounds. The utilization of methyl compounds as precursors in methane synthesis is confined to a small group of methylotrophic methanogens (i.e., Verstraetearchaeota). The only described anaerobic methylotrophs include members of methanogenic archaea, acetogenic bacteria, and sulfate-reducing bacteria. These organisms compete for these compounds geochemically produced in anoxic settings. Brockarchaeota may recycle methanol and methylamines in anoxic environments without methane formation and may be sequestered in deep sea sediments and hot springs. Orange and purple arrows represent sources and sinks, respectively. Organic Matter (OM) includes dissolved and particulate organic matter feeding the microbial loop (adapted from Evans et al., 2019 and Zhuang et al., 2018). The protein repertoire of GB and hot spring genomes have some important distinctions that reflect different anaerobic metabolisms. GB genomes appear to be obligately fermenting organisms that rely mostly on substrate-level phosphorylation since they lack all the complexes for the respiratory chain with exception of the ATPase. In contrast, hot spring genomes appear to have mechanisms to increase their ATP yield including the use of geothermally derived inorganic substrates as possible terminal electron acceptors such as mercury (Hg), arsenic (As), and hydrogen (H 2 ). Deep-sea hydrothermal vents, hot springs, and fumaroles are natural sources of Hg 42 , H 2 52 , arsenic 56 , and sulfur 57 . The discovery of Brockarchaeota genomes from sediments around the world, overlooked by conventional rRNA gene diversity approaches, highlights the need for further exploration of subsurface microbial communities. Although they are relatively low in abundance in the communities described here, the addition of these genomes to public databases, will enhance their detection in future environmental studies, like other recently described novel archaeal lineages 1 , 58 . A lack of recognition of their existence prior to this limited our ability to fully describe sediment community structure and function. Given their broad distribution, and versatile carbon metabolism, they are likely key players in global carbon cycling. However, this first description is limited to genomic characterization, thus culturing or in activity measurements are needed to confirm their physiological activities 59 . Overall, the description of this new phylum enhances our understanding of biodiversity of archaea and suggests they are mediating unique roles in anoxic carbon cycling."
} | 2,067 |
36132822 | PMC9418118 | pmc | 934 | {
"abstract": "Random networks of nanoparticle-based memristive switches enable pathways for emulating highly complex and self-organized synaptic connectivity together with their emergent functional behavior known from biological neuronal networks. They therefore embody a distinct class of neuromorphic hardware architectures and provide an alternative to highly regular arrays of memristors. Especially, networks of memristive nanoparticles (NPs) poised at the percolation threshold are promising due to their capabilities of showing brain-like activity such as critical dynamics or long-range temporal correlation (LRTC), which are closely connected to the computational capabilities in biological neuronal networks. Here, we adapt this concept to networks of Ag-NPs poised at the electrical percolation threshold, where the memristive properties are governed by electro-chemical metallization. We show that critical dynamics and LRTC are preserved although the nature of individual memristive gaps throughout the network is fundamentally changed by filling the gaps with an insulating matrix. The results in this work generate important contributions towards the practical applicability of critical dynamics and LRTC in percolating NP networks by elucidating the consequences of NP network encapsulation, which is considered as an important step towards device integration.",
"conclusion": "Conclusion and outlook In conclusion, we expand the concept of implementing brain-like critical dynamics and LRTCs via percolating NP networks for neuromorphic applications 26 towards Ag-NP based systems, where the memristive gap dynamics is governed by electrochemical metallization. The network dynamics were characterized via ACFs and DFA (indicating the existence of LRTCs) and analysis of scale-invariant dynamical features and avalanches (indicating critical dynamics). More importantly, it was shown that these functionalities are preserved, when insulating SiO x N y matrices were added onto the percolating NP networks, which fundamentally changes the nature of memristive gaps from air-type to solid-state type. This was supported by the absence of qualitative differences in the critical dynamics and LRTCs of networks with and without matrix. Both systems exhibited long-range temporal correlations in the sequence of network transition events, scale-invariance of dynamical features like magnitudes of network transition events and interevent intervals and presence of scale-invariant avalanches. These findings strengthen the prospects regarding to practical applicability of percolating NP networks in neuromorphic systems, because embedding the system (which must be carefully poised at the percolation transition) without functionality loss is inevitable for a potential device integration procedure. For future progress in this field, we suggest to combine the understanding of nanoscale Ag-based electrochemical memristive switching dynamics (which has been extensively studies in a broad range of materials systems) with network science approaches, to model the collective network behavior. This may allow for understanding of the emergent features observed here for the percolating Ag-NP networks, and especially may elucidate further details on the impact of an insulating matrix.",
"introduction": "Introduction Conventional computer technology is facing fundamental limitations, which are related to hardware architecture (von-Neumann bottleneck), the integration density of transistors (envisioned end of Moore's law) and a tremendous increase in estimated power consumption. These limitations have greatly stimulated the research into novel and unconventional computation concepts. 1 The field of neuromorphic engineering aims to solve these challenges by designing novel types of computational hardware, which draw inspiration from biological principles like signal thresholding, synaptic plasticity, parallelism and hierarchy or in-memory computing. 2 In the past decade, memristive devices played a key role as fundamental building units in the design of neuromorphic hardware and significant effort was focused on mass integration of memristive devices on wafer scales. 3–6 The key characteristic of a memristive device is its reconfigurable resistance state. Among different types of memristive devices, filamentary switching devices based on electrochemical metallization (ECM) principles 7 are of special interest. The major working principle of this type of memristive devices is the reconfiguration of metallic filaments in nanoscale switching gaps in response to the application of external voltage or current stimuli. Accordingly, the conductance of the nanogaps is determined by the state of the metallic filament. Diverse switching dynamics such as non-volatile filamentary switching, 7 diffusive switching 8 or highly dynamic spiking behavior 9,10 have been reported. However, the creation of networks of memristive devices approaching brain-like complexity via traditional top-down fabrication technologies poses several challenges as the performance of each memristive device and the enormous degree of connectivity within the network has to be under precise control. 1 In view of these challenges, fabrication of neuromorphic devices based on self-assembly approaches appears to be a promising and feasible alternative route. 11,12 Such devices are typically implemented via the formation of a complex network of memristive gaps with a stochastic distribution. In these networks, the emerging collective dynamical is exploited. Following such approaches neuromorphic functionalities can be implemented, circumventing the necessity of a precise wiring, spatial assembly, and tailoring of switching characteristics of individual memristive units. Such approaches turned out to be feasible for reservoir computing, where complex dynamical systems showing short-term memory and spatio-temporal correlations are required. 13–16 Recently several reports have shown the existence of different potential building units from which complex emergent networks can be formed. Networks assembled from Au-NPs above the percolation threshold exhibit complex memristive switching patterns, which are caused by atomic rearrangements between adjacent Au-NPs induced by electrical currents. 17 Networks of polymer-coated Ag-nanowires formed by random self-assembly also show emergent dynamics applicable for the design of neuromorphic systems. 18,19 Recently, the technical implementation of critical dynamics in neuromorphic systems via self-assembled networks of memristive switches gained considerable interest. This is motivated by findings from neuroscience indicating that biological neuronal networks operate in a regime of critical dynamics, which is seen as beneficial for solving computational tasks efficiently. 20 The presence of critical dynamics in biological neuronal systems was supported by the fact that spontaneous neural activity in cortical tissues takes place via brief bursts separated by periods of highly reduced activity, so-called “ neural avalanches ”. 20,21 Experimental observations of this avalanche activity in cortical neuron tissues from rats suggest that avalanches exhibit scale-invariant dynamics where the occurrence of different avalanche sizes can be described by a probability distribution in the form of a power law. 21 The origin of such behavior is frequently related to self-organized criticality. Self-organized criticality can be seen as a generating mechanism for avalanches and associated scale-invariant dynamics. 22 In general, critical dynamics universally occur in systems which are poised at the transition between two phases, which are composed of a multitude of dynamical units that influence each other. 20 There, a macroscopically observed avalanche may be triggered by a microscopic local change in the system that is collectively passed along the system due to the mutual interaction between single units. Several computational benefits have been described for this state, including the maximization of dynamic range, signal transmission and information capacity. 23 Based on this, it was hypothesized that the brain also operates at the edge of the transition between complete ordering and disordering. 24 With regard to memristive systems, hints on critical dynamics were found in networks of metal-insulator-metal switches (Ag–Ag 2 S–Ag). This network showed spatially distributed switching throughout the network and power-law scaling of persistent metastable network states. 25 Recently, networks of Sn-NPs poised at the electrical percolation threshold (a second order phase transition) were shown to exhibit critical dynamics with corresponding avalanche patterns similar to those observed in neural tissues. 26 Critical dynamics in random networks of memristive switches are expressed by scale-invariance in dynamic features of the network. These dynamic features include the fluctuations in magnitude and temporal structure of network conductance that originate from the underlying memristive activity. Moreover, fluctuations of the network conductance are organized in avalanche patterns, which indicates that the memristive activity in the network is correlated. Characteristic for critical dynamics are scale-invariant distributions of avalanche sizes and durations. 26 Further, those networks implement long-range temporal correlations (LRTC), which is commonly a concomitant feature of critical dynamics. LRTC is a dynamical feature of a system, describing that the past activity of the system determines the future activity at any given time, which indicates capabilities to implement a dynamical memory. Such a dynamical memory is beneficial for mapping of temporal information into a system, a property that is important in the context of reservoir computing. 15,16 One common procedure for proving the presence of LRTC in a system is to demonstrate a power-law decay of the autocorrelation functions in the time domain. 27 Another indication for LRTC can be found by scale-invariant fluctuations in the network activity, also in the time domain. In this connection, detrended fluctuation analysis (DFA) is a frequently used method to characterize such scale-free fluctuation and to demonstrate LRTC. 28 Because of the dynamical memory properties, LRTC is seen as beneficial for neuromorphic computation approaches. 16 To implement critical dynamics and LRTC, both features that suggest brain-like degree of complexity, into neuromorphic systems tailoring the connectivity (by poising at the percolation threshold) within networks of NP-based memristive switches appear to be a feasible strategy. However, the practical applicability of this approach can be still debated and also the understanding of the origin of the emergent phenomena has to be improved. Particularly, an elaboration of this concept towards composite systems is still missing so far and would be beneficial to support practical applicability. In this work, we extend the concept of implementing critical dynamics and LRTC for neuromorphic systems via Ag-NPs networks poised at the electrical percolation threshold (in the following named as “percolating NP networks”) and address the question how filling the memristive gaps in the network with an insulating matrix influences the network dynamics and applicability. The use of Ag-NPs in this work is motivated by the fact, that the nanoscale behavior of Ag-NP-based memristive gaps is already well-understood, 8,29 which provides suitable complementary knowledge for future development, aiming to understand the emergence of collective phenomena in random memristive networks. A Haberland-type gas aggregation source (GAS) 30 was used for the vapor phase synthesis of Ag-NPs. Generally, a GAS offers a broad choice between elemental and alloy NPs and good control on the properties, 31 which allows to extend the engineering of memristive gaps towards Ag-based alloy NP systems with enhanced stability and degree of freedom. 29 For the practical application of percolating NP networks in neuromorphic systems maintaining their functionality within a composite system is highly relevant. This is because encapsulation of the network into an insulating medium is in most cases an inevitable step of electronic device integration. However, there is only insufficient knowledge about the consequences on the overall network dynamics, when the character of the underlying memristive units are fundamentally altered by filling the gaps with an insulating material. This is because critical dynamics of percolating NP networks reported so far were in an exposed scenario, i.e. without encapsulation of the network. Although several reports on the electronic properties of composite systems comprising percolating NP networks exists, 32,33 a connection between the brain-like dynamics ( i.e. critical dynamics and LRTC) and integration of percolating NP networks into insulating matrices has not yet been made. In particular, this poses questions regarding to the consequences for the network functionality (and therefore practical applicability), when the nature of the memristive gaps throughout the network is fundamentally altered by filling the gaps with an insulating matrix. Therefore, we fabricated exposed percolating NP networks and compared them with similar networks, which were covered with ceramic layers of SiO x N y . Comparing both network types, the nature of the memristive gaps is changed from air-type (for exposed NPs) to solid-state-type (for embedded NPs). Characterization of percolating NP networks is done by evaluating their temporal patterns of memristive switching activity in response to a constant stimulus (voltage input) towards hallmarks of brain-like behavior, such as critical dynamics 21 and LRTC. 27 For the quantification of LRTC, autocorrelation functions 34 and detrended fluctuation analysis 28 were applied. Critical dynamics in the network activity is demonstrated by the emergence of scale-invariance and the according power laws in dynamical network features. Particularly, this requires an avalanche analysis analogous to approaches used in neuroscience. 21 The main focus of this work is to demonstrate, that the network behavior applicable for neuromorphic systems ( i.e. critical dynamics and LRTC) is preserved when an insulating matrix is added to the percolating NP network.",
"discussion": "Discussion The results in this work demonstrate that the implementation of brain-like behavior, such as critical dynamics or LRTC, via NP networks poised at the percolation threshold is feasible for a broader range of material systems. In contrast to similar NP-based memristive networks in the literature, 26,41 the memristive activity Ag-NP based networks presented here is expected to be significantly dominated by electrochemical metallization (ECM). 8,35,42 More importantly, the comparison between percolating Ag-NP networks with and without a SiO x N y matrix shows, that the switching dynamics is preserved after the addition of an insulating matrix. Similar network behavior, in terms of temporal correlations in the sequence of network transition events, scale-invariance of dynamical features like magnitudes of network transition events and interevent intervals and presence of scale-invariant avalanches, was observed in both systems. This indicates that the collective behavior of memristive gaps is preserved upon addition of SiO x N y . These observations lead to important implications regarding the practical application of percolating NP networks. The results imply that tailoring the network connectivity (which is done here by the electrical percolation threshold) and insulating matrix integration (through addition of a ceramic layer) can be treated as independent from each other in the fabrication procedure. This is highly beneficial, because the establishment of network functionality ( i.e. critical dynamics and LRTC) still solely requires tailoring of the NP filling factor and a deliberate insulating matrix integration does not considerably affect these functionalities. From a general point of view, the dynamics in a system of highly interconnected dynamical units crucially depends on the underlying network connectivity 23,43 and dynamical properties of the individual units, 44,45 which makes both features important for the engineering of critical dynamics and LRTC in memristive material systems. Regarding the properties of individual memristive gaps, it can be expected that their dynamics are governed by a volatile character. This can be argued from the fact that rather low currents are flowing through individual gaps during formation of filamentary structures, which commonly leads to formation of thin (and therefore volatile) filaments. 10,29 Comparable volatile dynamics were, for instance, also observed for Ag/PVP/Ag cross-points in Ag-nanowire networks. 46 Moreover, it is expected that the memristive gaps do not behave uniformly, but that the degree of volatility underlies a certain variance. This is reasoned by the fact, that the time for spontaneous decay of a filament strongly depends on parameters like the filament thickness or curvature on the filamentary structures, which was comprehensively described for the dynamics in Ag-nanowire/silk/Ag-nanowire cross-point structures. 47 The observed similarity in the critical dynamics and LRTC of both systems is to a certain degree surprising, because the underlying ECM-based physical mechanisms, by which filaments within the memristive gaps are formed and disintegrated, 8,29,48 are presumably altered upon addition of an insulating layer like SiO x N y . Consequently, the dynamical properties of individual memristive gaps are also expected to be altered. Major consequences of adding a matrix would include for instance, that the migration of Ag + -species is now enabled within a volume instead of migration on a surface (when a matrix is missing). Furthermore, the interfacial energies of filaments and diffusivity of Ag-species, which affects filament morphology, are altered. 8,47 Although the conditions, that are responsible for the memristive gap dynamics, are different for percolating NP networks with and without matrix, our results suggests that the underlying dynamics behave similarly. A possible explanation for this can be provided from a kinetic point of view, under the assumption that a common rate determining mechanism influences the dynamics of both systems similarly. According to literature, 35,42 the ECM-based memristive dynamics of filamentary Ag-structures can be kinetically limited by one of three different mechanisms, which contribute to the formation of filaments: Nucleation at cathodic sites (which initiates filamentary growth), migration of Ag + -species across the gap or electron-transfer at the Ag-gap interface during electrochemical oxidation. If one of these three mechanisms is rate limiting for the percolating NP networks with and without matrix, from a kinetic viewpoint, this mechanism could determine the dynamics equally in both systems. We expect that rate limitation by nucleation does not play a role for percolating Ag-NP networks, because a growing filament and a cathodic site will be the same metal, which consequently excludes any significant nucleation barrier. Further, the migration of Ag + -species across the gaps is not considered as a common rate-determining step, because surface migration (networks without matrix) and volume migration (with matrix) are expected to behave kinetically different. Only the electron-transfer rate at the Ag-gap interface, mainly depending on the kind of active metal, could contribute similar kinetics to both systems. Reasoning from a kinetic point of view, a common rate determining mechanism will most likely result in a similar behaviour of both systems. We thus propose that a similarity in the electron-transfer rate of both systems (with and without insulating matrix) could be an explanation for the similarity in the observed dynamics."
} | 5,026 |
37461420 | PMC10350088 | pmc | 935 | {
"abstract": "Microalgae are key players in the global carbon cycle and emerging producers of biofuels. Algal growth is critically regulated by its complex microenvironment, including nitrogen and phosphorous levels, light intensity, and temperature. Mechanistic understanding of algal growth is important for maintaining a balanced ecosystem at a time of climate change and population expansion, as well as providing essential formulations for optimizing biofuel production. Current mathematical models for algal growth in complex environmental conditions are still in their infancy, due in part to the lack of experimental tools necessary to generate data amenable to theoretical modeling. Here, we present a high throughput microfluidic platform that allows for algal growth with precise control over light intensity and nutrient gradients, while also performing real-time microscopic imaging. We propose a general mathematical model that describes algal growth under multiple physical and chemical environments, which we have validated experimentally. We showed that light and nitrogen colimited the growth of the model alga Chlamydomonas reinhardtii following a multiplicative Monod kinetic model. The microfluidic platform presented here can be easily adapted to studies of other photosynthetic micro-organisms, and the algal growth model will be essential for future bioreactor designs and ecological predictions.",
"conclusion": "Conclusions and future perspectives Modeling algal growth under the influence of various physical and chemical factors in their microenvironment is important for understanding the ecology and evolution of phytoplankton-related aquatic microbial communities, controlling harmful algal blooms, and improving biofuel production. Here, we implemented a microfluidic platform with precisely controlled light and chemical gradients to study the light and nitrogen-controlled growth of C. reinhardtii , then used the experimentally measured data to propose and validate a general growth kinetics model. We found that algal growth response to environmental parameters was described by an independent multiplicative Monod kinetic model. Interestingly, we found that the contributing term to the multiplicative model for both a physical and a chemical parameter were similar, because they can both be described by Monod kinetics. In our experiment, the multiplicative model consists of the Monod growth kinetics due to light multiplied by the Monod growth kinetics due to nitrogen. We hypothesize that more physical and chemical parameters can be included in this general growth model by simply multiplying the Monod kinetics term of each single source together, assuming the resources are independently colimiting. Future experiments will be needed to develop a true general growth model in a complex environment. In this work, we demonstrated that a microfluidic platform enabled studies of algal growth under a well-defined physical and chemical environment, and that the results were ideal for data driven theoretical modeling. In addition, dynamic information can be obtained through this platform’s real time imaging capabilities. This platform can be easily adapted to studies of photosynthetic microbes under two chemical gradients (e.g., nitrogen and phosphorous) alone or with a light gradient. We note that the generated light and ammonium gradients were kept at low levels for the investigation of colimitation effects of the two factors. The range can be easily extended by varying the lamp power output and the concentration of nitrogen in the medium perfused from the side channels. Our platform is also amenable to studies of other cellular behavior such as competition of microbiomes or cell motility. While this small-scale microfluidic device provided a way to precisely define a complex microenvironment for cells, the downside is that there was a small number of cells in each habitat. This was reflected in the variability of the measurements. A future modeling improvement can include cell number variability into the formulation. A multiplicative growth model of light and nitrogen was used to describe the independent colimitation effect of light and nitrogen on algal growth. The model predictions were used to estimate which environmental driver, nutrient or light, would be the limiting growth factor in two Finger Lakes with different clarities (thus light condition) and different total nitrogen concentrations. The model treated the presence of acetate as contributing to an equivalent baseline light intensity. For independent environmental parameters, the multiplicative model could be further simplified as follows:\n \n (2) \n μ = μ m a x ∏ i = 1 n ( E i + E i o K E i + E i + E i o ) . \n \nHere, E i stands for the i-th factor of the n number of independent colimiting factors such as light, nitrogen, phosphorous, and CO 2 ( 46 , 65 , 66 ). K E i is the half saturation constant for the i-th environmental factor, and E i o is the storage term. μ max is the maximum growth rate at saturation. Beyond freshwater bodies, mathematical models of phytoplankton growth are being incorporated in global models of ocean circulation and biogeochemistry ( 67 ), to better predict the magnitude and spatial intensity of the carbon pump concentration and elemental ratios in the ocean. We anticipate that a better understanding of phytoplankton growth in nutrients and light gradients will provide the foundation for more accurate predictions of the global biogeochemical processes to which phytoplankton contributes. In response to nitrogen stress, along with the reduction in photosynthesis, algae such as C. reinhardtii modify their metabolism to accumulate high amounts of storage molecules including triacylglycerol (TAG), which makes them a promising candidate for biofuel production ( 7 , 9 , 49 , 68 – 71 ). While nitrogen deprivation studies have mainly compared milli molar ammonium to the nitrogen-deprived state, here, we found that an increase of nitrogen concentration in the micro molar range could clearly affect cells utilization of light for growth. This information could possibly be used to search for a nitrogen condition with high TAG accumulation as well as a reasonable growth rate to optimize biofuel production. In addition, optimizing biofuel production requires further investigation of the effect of acetate, where experiments could be performed to figure out L 0 proposed in Eqn. 1 as a function of acetate concentration.",
"introduction": "Introduction Phytoplankton, including microalgae and cyanobacteria, are essential in maintaining the balance of ecosystems contributing to primary production and the carbon cycle. By producing and releasing oxygen as a byproduct of photosynthesis, cyanobacteria are known to be the first oxygen producers on Earth, dating back billions of years ago, and are essential for life on Earth. Disruption of phytoplankton growth can lead to environmental problems such as Harmful Algal Blooms (HABs) ( 1 – 3 ), which deplete threatened water resources. The problem of HABs has been recently exacerbated by climate change, including the effect of warming temperatures and increasing frequency of storms and flooding events ( 4 – 6 ). In contrary to HABs, lipid production from controlled growth of microalgae is a promising avenue for clean alternative bioenergy ( 7 – 9 ). A number of microalgae species have been found to be excellent nutritional food sources for animals as well as humans ( 10 – 12 ). As such, a mechanistic understanding of and the ability to control algal growth is an essential step towards finding solutions for sustainable living. The biophysical (e.g., light and temperature) and biochemical (e.g., nitrogen and phosphorous) environment critically impacts the growth of microalgae. Traditionally, the environment has been considered as the major selection force acting on genes. Recent studies have revealed important roles played by the environment in ecology, evolution, and development of biological organisms, within the so-called “eco-evo-devo” field ( 13 – 17 ). Phenotypic plasticity of cells can emerge under time-varying environments, which help cells respond quickly to environmental fluctuations ( 18 – 20 ). Also, the passing down of extra-cellular environments could serve as another mechanism for non-genetic inheritance, which could not only affect single cells, but also a community of cells and their interactions ( 21 – 25 ). For example, microorganisms and the interactions among them can shape their environment, a phenomenon known as niche construction, leading to the co-evolution of organisms and their environment ( 26 – 29 ). Despite the importance of cell-environment interactions in algal growth, mechanistic understanding of how complex environments regulate algal growth is limited. This is due to the lack of tools where environmental conditions can be directly controlled while cell growth is simultaneously monitored in real time at a quantitative level. Microfluidics has emerged to provide well controlled environments for algal cell growth. Algal growth has been studied extensively under single environmental gradients, including nutrients ( 30 – 33 ) and light ( 34 , 35 ). Due to the complexity of the environment in nature, recent work has started to explore roles of multiple environmental parameters in cell growth. Dual microfluidic chemical gradients have been reported to rapidly screen cell response to multiple stressors or characterize cell responses to stressors under complex environments ( 36 – 39 ). Specific to algal cells, a microfluidic dual chemical gradient generator device revealed the synergistic effect of nutrients (nitrogen and phosphorous) on cell growth ( 40 ). Although photosynthesis is an important part of algal cell growth, the quantitative understanding of the effect of light exposure on algal growth is limited, especially under a controlled nutrient environment ( 41 , 42 ). Nguyen et al. recently presented an important millimeter-scale platform that examined the optimal light intensity and nutrient condition for lipid production of algal cells under controlled light and nutrient condition for cells immobilized in hydrogels ( 42 ). Mathematical modeling of algal growth kinetics, especially in complex environmental conditions, has lacked precise experimental validation. Current algal growth models have been derived from first principles such as mass action kinetics and growth optimality. These models were used to describe observations of natural waters and experiments with large-scale static and continuous cultures for the dependence of algal growth on multiple resources ( 43 – 47 ). However, limitations in the range of substrate concentrations and temporal resolution prevented the validation of these models, limiting our ability to identify the fundamental principles controlling the dependence of algal growth parameters on the concentration of multiple resources. In this spirit, microfluidics could serve as a useful tool to provide well-defined microenvironments and quantitative measurements for the development of accurate growth models for microalgae. Here, we propose a microfluidic platform that allows us to establish gradients of both light intensity and nutrient concentration directly on the array microhabitats placed on a microscope stage. Key innovations of our platform include that growing C. reinhardtii cells can swim freely within microhabitats, closely mimicking its natural habitat, and that the growth dynamics can be continuously monitored in real time instead of end point measurements. We propose a general mathematical model developed alongside experimental data demonstrating that light intensity and nutrient concentration colimit the growth of algal cells following a general multiplicative Monod kinetic model.",
"discussion": "Results and Discussion A microfluidic platform for growing photosynthetic microbes under a dual light and nitrogen gradient A hydrogel-based array microhabitat device in conjunction with a light gradient was used to study algal growth under well-defined physical and chemical conditions (See Fig. 1A ). To create a nitrogen gradient, we used a previously developed hydrogel-based chemical gradient generator ( 40 ). The device consists of an 8 × 8 array of microhabitats (each with size of 100μm × 100μm × 100μm) flanked by two side channels (400μm W × 200μm H), patterned on a 1 inch × 3 inch size and 1 mm thick agarose gel membrane ( Fig. 1A and Fig. S1 ). Nitrogen-starved cells were seeded in the microhabitats. The nitrogen gradient was established by flowing medium with known nitrogen concentrations and blank buffer through the top and bottom side channels respectively, creating a nutrient gradient at the location of the array microhabitats via molecular diffusion. The time to reach a steady state nutrient gradient was measured to be about 90min ( 40 ). The light intensity gradient was generated by modifying the bright field illumination light path of a commercial microscope ( Fig. 1A and ref. (48) ). Briefly, the light intensity gradient at the location of the array microhabitat was created by placing a half-moon shaped mask directly below the field iris of an Olympus IX81 microscope. Fig. 1B shows an image of cells growing in microhabitats under light and nitrogen gradients, which were characterized in the following ways. The nitrogen concentration gradient in the dual gradient experiment is shown in Fig. 1C . To characterize the chemical gradient in the array microhabitat, we flowed a solution of a known concentration of a fluorescent dye in the source channel and blank buffer in the sink channel as described previously ( 40 ). The steady linear gradient computed from the fluorescence image is shown in Fig. 1C . Here, we assumed the nitrogen concentration in the sink and source channels was 5.3 and 35.3μM respectively, and the concentration of the fluorescence solution was linearly related to the nitrogen concentration. Nitrogen concentrations at the middle line of each row of microhabitats were taken as the concentration of that row. The light intensity was characterized using a light meter and readouts from a CCD camera. Fig. 1D shows an almost linear light intensity profile across the array microhabitat. The maximum PAR value was approximately 45 μmol·m −2 ·s −1 and the minimum was 0.1 μmol·m −2 ·s −1 ( Fig. 1D ). PAR values were converted from the grayscale values measured from bright field images of the light intensity gradient captured with a CCD camera as described previously ( 48 ). The light intensity of each column was calculated as the average across the width of the microhabitats in that column, as cells were observed to swim freely within single habitats. To monitor cell growth dynamics, time lapse fluorescence images were taken every four hours during the 7-day experimental period ( Fig. 1E ). Cell numbers were measured using the fluorescence intensities in each microhabitat, and were used to calculate the growth rates. Our platform advances the current technology to study photosynthetic microorganisms under well-controlled nutrient and light gradients. The unique feature is the integration of a microfluidic chemical gradient generator together with a microscope-based light gradient generator. To our knowledge, this is the first time that cell growth can be monitored in real time while the cells are subjected to well defined nutrient and light gradient conditions. This platform is particularly suitable for the creation of data driven mathematical growth models for photosynthetic cells that are sensitive to complex microenvironments. Algal growth sensitivity to light was enhanced at high nitrogen concentration C. reinhardtii cells grew very differently under the same light intensity gradient with low versus high nitrogen concentrations ( Fig. 2A – B ). Cells were starved in medium with 5.3μM nitrogen before loading them into the microhabitats ( Materials and Methods ). When the cells were provided with a low nitrogen concentration of 5.3μM in the device, no clear response to light was observed ( Fig. 2A ). In contrast, when 30.8μM of nitrogen was provided uniformly across all the habitats, cells showed a clear increase in growth rate as light intensities increased ( Fig. 2B ). Fig. 2C – D shows the corresponding growth curves, normalized by the initial cell number that was usually 1–6 cells per habitat due to the random seeding process. Microhabitats without cells were excluded from the analysis. The normalization with respect to the initial cell number revealed a slight growth response to light intensity at 5.3μM nitrogen ( Fig. 2C ), which was not apparent from the fluorescence images ( Fig. 2A ). In the presence of 30.8μM nitrogen, the growth curves at different light intensities were clearly spread out ( Fig. 2D ), indicating that cell growth had higher sensitivity to light at the higher nitrogen concentration. Light is an important energy source for photosynthetic microbes like microalgae. Energy harnessed from light can be used for the biosynthesis of structural and functional components of the cell, contributing to cell proliferation. Upon the removal of nitrogen, proteins and pigments related to photosynthesis decrease in abundance, including RuBisCO, a key enzyme in the Calvin cycle, and chlorophyll, the pigment for capturing photons, which leads to reduced efficiency of light utilization ( 49 – 52 ). Our results in the microhabitats indeed showed that cell growth was less sensitive to light at low nitrogen concentration, indicating a suppressed photosynthetic capacity. Increased sensitivity to light in the presence of nitrogen as compared to no nitrogen has been shown in previous studies either by -omics response or by photosynthetic functional measurements ( 49 , 50 ). Here, we showed that growth rate, which results from all cellular processes combined, was more sensitive to light under the high nitrogen concentration. Algal growth sensitivity to nitrogen was enhanced under high light intensity Higher light intensity, in turn, increased the growth sensitivity of algal cells to nitrogen ( Fig. 3 ). Under lower light intensity (0.1PAR), the cell growth did not vary distinctly across different nitrogen concentrations ( Fig. 3A , C ). In contrast, at higher light intensity (41.1PAR), cells at lower nitrogen concentrations grew significantly slower than those at higher nitrogen concentrations ( Fig. 3B , D ). This phenomenon was clearly visible in the fluorescence images ( Fig. 3A – B ) and was reflected in the growth curves ( Fig. 3C – D ). Nitrogen is one of the most important nutrients for algae, contributing to the synthesis of amino acids and proteins. Various forms of nitrogen can be utilized by algal cells including nitrate, nitrite, ammonium, and some organic forms, among which ammonium is the preferred form. Ammonium is transported into cells via the ammonium transporter proteins (AMT family) and is assimilated through the glutamine synthetase-glutamate synthase pathway. Synthesized glutamine and glutamate can be used for the biosynthesis of macromolecules for cell function and proliferation. Increase in light intensity was found to induce responses in various metabolic processes including photosynthesis, as well as amino acid, fatty acid, and nucleotide biosynthesis. Nitrogen metabolism is also known to be affected by light ( 53 – 55 ). Specifically in the case of ammonium utilization, the presence of light contributes to higher activity of glutamate synthases ( 56 ). The experimental observation presented here begins to reveal the interactions between nitrogen metabolism and photosynthesis machineries. Light intensity and nitrogen concentration synergistically influence the growth rate of algal cells Light and nitrogen were found to synergistically promote algal growth when the algal cells were grown in the array microhabitats in the presence of light and nitrogen gradients. Here, the nitrogen gradient was generated by flowing medium with 35.3μM nitrogen in the source channel and medium with 5.3μM nitrogen in the sink channel. The light intensity was provided by the bright field light source of the microscope, ranging from 0 to 45 μmol·m −2 ·s −1 . The synergistic effect can be seen clearly in Fig. 4A , which shows fluorescence images of cells in the array microhabitats. Therein, the upper right corner represents the habitat with the highest light intensity and highest nitrogen concentration. To understand quantitatively how cell growth depends on light intensity and nitrogen concentration gradients, we calculated growth rates of each microhabitat and displayed them in Fig. 4B . Results from three replicated experiments were shown in Fig. 4B . In all three replicates, the data was consistent, and shows that growth rate was highest at high nitrogen concentration and light intensity. We noticed an interesting parallel between the cell growth response to dual light and nitrogen gradients (shown in Fig. 4 ) and our previous work where cell growth response was studied under dual nitrogen and phosphorous gradients ( 40 ). Under a single nutrient or light gradient, we used a microfluidic platform to reveal that algal cell growth followed a Monod growth kinetic model ( 31 , 48 ). Under a dual nitrogen and phosphorous gradient, a multiplicative model fitted to the growth data showed the colimitation by two nutrients ( Supplementary Information ). The experimental data presented here inspired us to ask whether there is a general growth kinetic model describing algal cell growth under complex physical and nutrient environments. Colimitation of algal growth by light and nitrogen: theoretical modeling We propose a general colimitation growth model for algal cells subjected to both physical and chemical parameters. We hypothesized that algal growth response to light or nitrogen was described by Monod kinetics, and the colimitation of algal growth by nitrogen and light was described by a multiplicative model:\n \n (1) \n μ = μ m a x ( L + L 0 K L + L + L 0 ) ( [ N ] + N 0 K N + [ N ] + N 0 ) . \n \nHere, μ max is the maximum growth rate, K L is the half-saturation constant of light intensity, and K N is the half-saturation constant of nitrogen concentration. We note that the response term to nitrogen is a Monod kinetic model, where N 0 represents stored nitrogen. The response term to light is also a Monod kinetic model, with a storage term L 0 . This term is required since we observed residual growth in the absence of light, but no residual growth in the absence of both light and acetate ( Fig. S4 ). Detailed discussion on the effect of acetate on algal growth can be found in the Supplementary Information . We fitted Eqn. 1 to experimental data under the dual light and nitrogen gradients. The fit included growth rate data obtained with and without acetate. During fitting, K L and K N were kept as free fitting parameters. μ max was fixed to be 2.4day −1 , which was the maximum growth rate obtained in the microhabitats previously ( 31 ). N 0 was set to 0, as the cells were starved in low N concentration media prior to experiments (See Supplementary Information ). In the case with acetate, L 0 was left as a free parameter. In the case without acetate, L 0 was set to 0 because ( 1 ) we observed no growth in the absence of both light and acetate and ( 2 ) fitting Eqn. 1 to growth data obtained in the absence of acetate gave a best-fit value of L 0 compatible with 0 (p-value = 0.18, Supplementary Information ). The fitted surface to Eqn. 1 is plotted together with experimental data in Fig. 5A . The best fit parameters were L 0 = 50.8μmol·m −2 ·s −1 , K L = 57.2μmol·m −2 ·s −1 , and K N = 2.8μM, with standard deviations of 7.9μmol·m −2 ·s −1 , 8.8μmol·m 2 ·s −1 , and 0.4μM, respectively. Distributions and correlations of the fitted parameters are shown in Fig. S3 . Colimitation of algal growth by light and nitrogen was revealed by the microhabitat dual gradient experiments and subsequent fitting to a general multiplicative growth model. While growth response to a single resource has often been described via Monod kinetics, here, we showed that growth response to a physical and a chemical parameter could be described by the multiplication of the two Monod kinetics terms with respect to each single resource, known as independent colimitation. Multi-resource colimitation on algal growth has been observed in natural waters, and has been categorized into 1) independent colimitation, where two resources were both at low levels and potentially limiting, and 2) dependent colimitation, where two resources were either biochemical substitutions or biochemically dependent ( 43 – 46 ). Different types of multi-resource-controlled algal growth kinetic models were discussed in the comprehensive reviews by Lee et al. and Bekirogullari et al. ( 57 , 58 ). As for light and extracellular nitrogen, previous studies that used field data and large-scale culture experiments have described them as independent growth-limiting resources, whose influence could be expressed in a multiplicative form ( 46 , 59 – 64 ). However, growth data corresponding to the precise light and nitrogen conditions in the algal microenvironment was unavailable. Our microhabitat platform provides results that fill this gap and demonstrate the independent colimitation of light and nitrogen on the growth of C. reinhardtii . Previous estimates of K L and K N in macro-scale systems were 81.4–215 5μmol·m −2 ·s −1 and 2.2–17mM respectively ( 57 , 58 ). These values were larger as compared to the half-saturation constants obtained from our microhabitat platform, and varied in a wide range. The differences could come from the self-shading effect and varying nutrient concentrations in large scale experiments. We also note that the half-saturation constants here depended on the history of the cell (starved or not), the provided light spectrum, and the form of nitrogen. Previous studies using non-starved cells gave K L and K N based on single light and single nitrogen gradients in the microhabitat platform to be 1.9μmol·m −2 ·s −1 and 1.2μM, respectively ( 31 , 40 ), which is smaller than those found in this work. In addition, we also compared different forms of models, including a multiplicative form with only μ 0 , K L , and K N as fitted parameters and a law of minimum form of the growth kinetic model (see Supplemental Information ). It was seen that the proposed model in Eqn. 1 had better goodness of fit as compared to the other forms. Predicting growth rate under various light intensities and N concentrations using the colimitation model The general co-limiting model can be used to predict and understand how light and nutrient synergistically control the growth of algal cells. Using the fitted model, a look up map was generated that predicts algal growth rate given the light intensity and the nitrogen concentration ( Fig. 5B ). In general, growth is suppressed when either N is under 20 uM, or light intensity is under 20 PAR. This look up map could be used to make predictions on algal growth trends under various light and nitrogen conditions. For example, one could predict the effect of environmental drivers on algal growth in freshwater bodies, where light intensity and nitrogen conditions could vary seasonally, as well as geographically. Take two of the Finger Lakes in upstate New York, USA as an example. In the 2020 sampling season, the light intensity at 6m depth and nitrogen concentration at the lake surface in Cayuga Lake (South Shelf Site) were 0.1PAR and 93μM respectively, while the values for Hemlock Lake (Mid Site) were 100PAR and 17μM (data was taken from the Citizens Statewide Lake Assessment Program reports, conversion from clarity measurements to light intensities could be found in SI ). Comparing the growth response to nitrogen concentration at the two different light intensity levels in Fig. 5D , it was shown that nitrogen concentration would be a more important driver for algal growth in Hemlock Lake (100PAR) than in Cayuga Lake (0.1PAR). In addition, comparing the growth response to light intensity at the two different nitrogen concentration levels in Fig. 5E , it was seen that light intensity would be a slightly more important driver in Cayuga Lake (93μM) than in Hemlock Lake (17μM). We note that in natural lakes, the spectrum of light and its attenuation down the water column, the forms of nitrogen that TN involves, and the blooming species are not exactly accounted for in our experiments that led to the model. However, this example shows the potential use of the experimental method and the model to make relevant predictions to some level of generalization."
} | 7,220 |
31259241 | PMC6598769 | pmc | 936 | {
"abstract": "Leidenfrost drops, known to levitate on very hot solids, exhibit a “cold” regime on superhydrophobic solids.",
"introduction": "INTRODUCTION A volatile liquid on a hot solid levitates above its vapor if the substrate temperature T exceeds the so-called Leidenfrost point ( 1 ). This temperature, often denoted as T L , is on the order of 200°C for water on smooth metals ( 2 ), a value that remains to be understood ( 3 , 4 ). Above T L , levitation provides nonadhesiveness, and it makes liquids spectacularly mobile ( 4 ): Drops in the Leidenfrost state move under the action of tiny forces, which was exploited to generate self-propulsion on asymmetric textures ( 5 ). In addition, vapor insulates the liquid from its substrate, which triggers a strong reduction of thermal exchanges ( 2 ). In contrast, if the solid temperature lies between the boiling point T b and the Leidenfrost point T L , then the liquid experiences nucleate boiling, with marked consequences on both heat transfer and liquid persistence ( 3 ). The thermal properties of both liquid ( 2 , 3 , 6 ) and solid ( 6 , 7 ) affect the Leidenfrost temperature. However, the combination of liquid and solid is often imposed by applications, which requires ingenious strategies to control T L and, thus, the conditions where boiling or insulation happens. Roughness at the solid surface was found to deeply affect the Leidenfrost point. On the one hand, hydrophilic texture can increase T L up to about 450°C, a way to enhance thermal fluxes and evaporative cooling ( 8 , 9 ). On the other hand, experiments by Vakarelski et al. ( 10 ) recently suggested that hydrophobic texture may stabilize the vapor layer down to the boiling point T b of water. The latter situation thus generates a “cold Leidenfrost regime” in water where levitation and its thermal and hydrodynamic consequences are extended by about 100°C compared to usual cases ( 11 , 12 ). This finding is of obvious practical interest, considering the gain in thermal energy needed to trigger levitation, drag reduction of hot solids ( 13 , 14 ), or augmented drop lifetime ( 10 ). By scanning T between room temperature and T L , we explore here the characteristics of the cold Leidenfrost regime. The Leidenfrost transition In Fig. 1A , we first compare the conformation of water drops (volume, Ω = 4 μl) placed on hydrophilic (blue frame) or superhydrophobic (red frame) materials brought to temperature T . The hydrophilic solid is a bare silicon wafer that water meets with advancing and receding angles θ a = 42 ± 2° and θ r = 16 ± 2°. The repellent material is a wafer coated with hydrophobic nanobeads (Glaco coating; see Materials and Methods), which provides θ a = 165 ± 2° and θ r = 160 ± 2°. The wettability contrast between both solids is obvious at T = 20°C, and it persists up to T b = 100°C. Beyond T b , nucleate boiling occurs on the hydrophilic substrate, as expected, while neither boiling nor apparent change in drop shape is seen on the superhydrophobic material. It is only above the Leidenfrost point on the hydrophilic material ( T L ≈ 210°C) that both drops become undistinguishable, a consequence of a similar levitation on vapor. Hence, the Leidenfrost transition on repellent materials cannot be evidenced by direct visualization since water switches from a poorly wetting state at ambient temperature to a vapor-levitating state at high temperature with little change in contact angle. In addition, nucleate boiling does not act as an indicator of temperature when the substrate temperature crosses 100°C, which can be seen as a hallmark of hot repellent materials, an effect that can be further exploited to reduce thermal exchanges and avoid massive gas production. Fig. 1 Water drops on hot hydrophilic and superhydrophobic materials. ( A ) Water drops (Ω = 4 μl) on a hydrophilic silicon wafer (blue frame) or on a superhydrophobic Glaco-treated wafer (red frame) brought to temperature T . Scale bar, 1 mm. While boiling occurs above 100°C in the hydrophilic case, neither boiling nor apparent change in shape is observed on the repellent solid. Both drops only become similar above 210°C, in a common Leidenfrost state. ( B ) Lifetime τ of water drops (Ω = 20 μl) as a function of the substrate temperature T on bare aluminum (blue data) and Glaco-treated aluminum (red data). Each point is an average over at least five measurements, and error bars represent standard deviations. The Leidenfrost transition is observed at T L ≈ 210°C on the hydrophilic substrate, whereas τ( T ) monotonically decreases in the repellent situation. Beyond T L , both curves superimpose. The sharp contrast between the two materials is also obvious when plotting the lifetime τ of a given volume of water (Ω = 20 μl) as a function of the substrate temperature T ( Fig. 1B ). Drops are trapped in shallow cavities machined in aluminum blocks, the metal being either bare (hydrophilic, blue) or Glaco coated (superhydrophobic, red). As seen in the figure, the lifetime on the hydrophilic solid sharply decreases from about 2 min at T = 85°C to a fraction of a second above T b = 100°C (boiling regime). At larger T , τ markedly increases up to a maximum that defines the Leidenfrost temperature ( T L ≈ 210°C). Above T L , τ slowly decreases with T , a classical observation in the Leidenfrost regime: Vapor insulates water from its substrate, which prevents boiling and impedes evaporation. On the superhydrophobic material, the behavior is very different below T L . The lifetime is always much larger than the former, and its decay with temperature is slower, both facts arising from the repellency-induced reduction of solid-liquid contact area. Beyond T b , τ remains high (a few minutes) and it smoothly decreases with T so that the Leidenfrost transition seems to be continuous instead of abrupt. Last, both lifetimes above T L become comparable, showing that the Leidenfrost regime at high T does not depend on the solid wettability anymore. The plot in Fig. 1B raises a number of questions. We know that water (weakly) contacts superhydrophobic materials at room temperature, while it levitates at high temperature, so that we still expect a Leidenfrost transition. The absence of nucleate boiling makes us anticipate a Leidenfrost point T* < T L , but the continuity in the data does not allow us to detect this point although it makes us suspect that the nature of the transition is modified. Our aim here is to describe what happens on water-repellent substrates below T L , which we do by characterizing the water adhesion and by visualizing the solid-liquid interface. Adhesion measurements Adhesion is classically quantified by the roll-off angle of millimeter-sized drops. We placed a given volume Ω of water on a Glaco-coated substrate brought to a temperature T and tilted until it reaches the value α at which water departs. As sketched in Fig. 2A , we assume that the drop with apparent contact radius r meets the substrate with respective angles θ a and θ r at its leading and trailing edges so that the contact angle hysteresis Δcosθ = cosθ r – cosθ a can be deduced from the force balance at departure, as first discussed by Furmidge ( 15 ). This balance states π r γ(cosθ r – cosθ a ) ≈ ρΩ g sinα, denoting γ and ρ as the surface tension and density of water and g as the acceleration of gravity. Contact angle hysteresis is a dimensionless measurement of adhesion possibly varying between 0 (no adhesion) and 2 (maximum adhesion). Fig. 2 Adhesion of water on hot repellent materials. ( A ) Sketch of the experiment: A water drop with volume Ω and contact radius r is placed on a substrate brought to a temperature T and tilted until the drop departs. At departure (tilting angle α), contact angles at the drop edges are the receding and advancing angles θ r and θ a , respectively. ( B ) Roll-off angle α as a function of temperature T for Ω = 3.9 μl (blue data), Ω = 5.4 μl (red data), and Ω = 9.2 μl (green data). Error bars show the standard deviation for a minimum of five measurements. ( C ) Contact angle hysteresis Δcosθ = cosθ r – cosθ a deduced from Furmidge’s equation: Water adhesion is nonmonotonic as a function of T , and it becomes nonmeasurable above ~130°C. The critical tilt α is plotted in Fig. 2B as a function of the substrate temperature T for three volumes Ω. Its value logically decreases when drops are larger. The graphs are not monotonic in T . The low value of α at T = 24°C gradually increases with temperature, and it is multiplied by a factor 3 as the temperature reaches ~70°C, but this regime of enhanced sticking is followed by a decrease in adhesion, up to T ≈ 130°C, where the critical tilt even becomes nonmeasurable (α ≈ 0). Hence, Fig. 2B allows us to unambiguously distinguish the small pinning on a water-repellent material (seen at room temperature) from the zero adhesion characterizing a Leidenfrost state. Specifically, the Leidenfrost point is found to be around T* ≈ 130°C, a temperature both much smaller than T L ≈ 210°C and substantially higher than T b ≈ 100°C. By measuring the contact radius r in each experiment (fig. S1) and using Furmidge’s equation (where both γ and ρ are taken at the substrate temperature T ), we deduce the contact angle hysteresis Δcosθ. As we plot Δcosθ as a function of T , we observe that the data fairly converge ( Fig. 2C ): Being a local quantity, Δcosθ is not expected to depend on the drop volume. Similar results are obtained if the initial drop temperature is the same as that of the substrate (fig. S2) or if experiments are performed with other hydrophobic textures, either colloidal or regularly etched (figs. S3 and S4): Adhesion of water on a warm superhydrophobic solid generally follows three successive regimes when increasing temperature, which we now discuss. The different regimes of adhesion 1) As a substrate gets warmer, water evaporation is favored. The repellent Glaco coating consists of random aggregates of nanobeads forming a porous structure with submicrometric depth, as seen in the image displayed in Fig. 3A . Vapor produced by evaporation can condense inside the pores, which eventually creates liquid bridges between the substrate and the drop ( 16 , 17 ) and enhances adhesion, as reported in Fig. 2C . We can assess this interpretation by testing a substrate where the formation of these bridges was shown to be negligible. Water condensing on dense arrays of hydrophobic nanocones ( Fig. 3B ) does not stick on them, a consequence of the geometrical expulsion of water nuclei from conical structures ( 17 ). Performing the experiment sketched in Fig. 2A allows us to compare adhesion on Glaco coating to that on nanocones with similar adhesion at 20°C ( Fig. 3C ). Instead of the nonmonotonic behavior reported earlier (blue data), we observe a continuous decay of Δcosθ from its low value at room temperature to zero above T* ≈ 130°C (red data). This experiment thus validates our scenario of condensation-induced adhesion on common water-repellent materials between 20° and 60°/70°C. Fig. 3 Adhesion of water on two kinds of hot hydrophobic nanotexture. ( A ) SEM (scanning electron microscopy) image of a Glaco-coated brass substrate. Hydrophobic nanobeads deposited on the substrate provide a submicrometric roughness. Scale bar, 500 nm. ( B ) SEM picture of a dense array of nanocones (height, 115 nm; spacing, 52 nm) textured in silicon and coated by fluorosilanes. Scale bar, 200 nm. The picture is adapted from the work of Checco et al. ( 25 ). ( C ) Contact angle hysteresis Δcosθ of a water drop (Ω = 3.9 μl) on Glaco coating (a, blue data) and on nanocones (b, red data) as a function of the substrate temperature T . 2) As seen in Fig. 2C , adhesion decreases from its maximum at ~70°C to its vanishing at ~130°C. Increasing temperature and approaching the boiling point oppose the formation of water nuclei in the texture, which contributes to lower adhesion. We can go further by imaging the bottom interface of the drop. To that end, we use sapphire as a substrate, which combines high thermal conductivity with transparency, the latter property being conserved after Glaco coating owing to the nanosize texture. An inverted microscope (see Materials and Methods) provides an image of the interface at the drop base ( Fig. 4A ). As shown by Mahadevan and Pomeau ( 18 ), the radius r of the contact area is expressed by the relationship r ≈ R 2 κ, denoting κ −1 = (γ/ρ g ) 1/2 as the capillary length. κ −1 varies from 2.7 mm at room temperature to 2.5 mm at the boiling point so that the contact radius of a millimeter-size drop is typically 400 μm, significantly smaller than R . At moderate temperature ( T < 60°C), the contact zone is gray with white dots (for T = 51°C; Fig. 4B ), a heterogeneous appearance arising from air trapped in the texture. This picture is deeply modified above 60° to 70°C. Then, we observe the formation of gray patches with well-defined contours (highlighted in red in the figure for T = 75°C). These patches grow as a function of temperature until they fully invade the contact zone where they generate interferences, as seen in Fig. 4B for T = 150°C. Fig. 4 Focus on the base of water drops placed on hot repellent substrates. ( A ) Setup: A drop is deposited on a Glaco-treated sapphire brought to a temperature T . The contact zone is observed from below with an inverted microscope. ( B ) Visualization of the contact area whose radius r is 0.4 mm for a drop with radius R = 1.0 mm at T = 51°C (left), T = 75°C (center), and T = 150°C (right). Vapor patches appear around 70°C, and we highlight their contour in red. ( C ) Close-up on the fringes seen in the central region of the vapor patches seen in (B) at T = 75°C. ( D ) Fraction ϕ v occupied by the vapor patches as a function of temperature T and measured 0 to 10 s after drop deposition. Each data point is an average over at least three drops, and error bars represent standard deviations. The bars are large in the critical regime of vapor formation and get smaller at larger T where they even become negligible when reaching the vapor patch stationary state at large time t . The patches are vapor bubbles, as better seen in Fig. 4C , where we display close-up views of their central region. We observe fringes, from which we can deduce that these spherical vapor/liquid interfaces meet the substrate with an advancing contact angle θ v as low as 2° (fig. S5): Vapor is close to “wet” the material whose superhydrophobicity implies superaerophilicity. At the same time, the contours of the bubbles are found to be distorted: Vapor bubbles are pinned in the texture and just grow from their nucleation site. Low θ v also implies that even a small volume of vapor induces a significant coverage of the solid: A bubble with radius r v = 100 μm encloses a volume π r v 3 θ v /4 of typically 30 pl, which would cover a surface area about 20 times smaller on a smooth hydrophobic surface (θ v = 90°). The total coverage ϕ v of the surface by vapor can be determined through image analysis. Defined as the ratio of the patch area over the whole contact area π r 2 , ϕ v is, for instance, ~0.4 at 75°C ( Fig. 4B ). At much larger T (for instance, 150°C in the same figure), ϕ v has reached its maximum ϕ v = 1, and the image is fully covered by the fringes arising from the presence of a thin continuous vapor film, as reported on regular (hydrophilic) solids above the Leidenfrost point ( 19 , 20 ). We report in Fig. 4D how the vapor coverage ϕ v increases with temperature T . We obtained each ensemble of data after depositing a water drop with volume Ω = 4 μl and following the evolution of ϕ v during the first 10 s after deposition (a shorter time compared to the lifetime τ). At fixed temperature T , we observed that ϕ v quickly reaches a stationary value that corresponds to the balance between vapor leakage inside the porous texture and vapor injection from the evaporating drop. This stationary value of ϕ v rapidly increases with T around 70°C, a critical behavior that explains the large error bars observed in this regime. Then, it gradually tends toward unity, a behavior accompanied by a decrease of the error bars. The invasion of vapor above 70°C tends to depin water from the solid substrate, which explains the decay of adhesion constituting the second regime in Fig. 2C . The Leidenfrost transition on a superhydrophobic material eventually appears to be a continuous phenomenon, instead of a discontinuous one on regular solids, in agreement with the qualitative observations in Fig. 1 . 3) Adhesion becomes nonmeasurable when the Leidenfrost film fully occupies the contact zone, which consistently occurs around 130°C in both Figs. 2C and 4C . Interferences in Fig. 4B show a buoyancy-driven blister, as observed on conventional materials above T L ( 19 – 21 ). Apart from an increase in the film thickness, this situation does not evolve when increasing the temperature. Hence, the third regime is an extended Leidenfrost regime, which confirms the observations in Fig. 2C . The Leidenfrost point is lowered by about 80°C compared to flat hydrophilic solids. This strong reduction is made possible by the invasion and coalescence of wetting vapor patches on the highly hydrophobic material, which happens around 130°C. The Leidenfrost transition might naturally occur at the boiling point of water, but this value is slightly shifted in our experiments. The fact that evaporative cooling lowers the solid temperature in the contact zone and that the vapor film insulates the liquid qualitatively explain that the Leidenfrost temperature is larger than 100°C, although the use of hydrophobic texture allows us to approach this limit.",
"discussion": "DISCUSSION Our interpretation was based on a quasi-static representation of water drops. However, motion is expected in liquids contacting hot solids and thus subjected to temperature differences of a few degrees between their base and their top ( 22 ). Convection was reported in water evaporating on repellent materials and attributed to both Marangoni and buoyancy effects ( 22 , 23 ). In fig. S6, we report particle image velocimetry measurements performed in millimetric drops placed on hot substrates. In all cases, we observed a rolling motion at the scale of the drop, with typical velocities V in the range of millimeters per second and increasing with the substrate temperature T . The viscous force exerted by the drop on the substrate scales as (η V / R ) r 2 , and it becomes comparable to the adhesion force γ r Δcosθ when the flow velocity is on the order of γ R Δcosθ/η r , a speed that can fall to ~10 cm/s for our less adhesive substrates (Δcosθ ≈ 10 −3 ). This velocity, however, remains large compared to that measured in the liquid, which justifies why we could neglect the role of these flows in our analysis. Also in the context of dynamics, another case of interest is that of impacting drops. Then, the Leidenfrost transition is known to shift to a (much) higher temperature ( 24 ) owing to the enhancement of liquid/solid contact brought by inertia. It would be interesting to see how this effect is modified when using repellent materials, a situation where we should observe a weaker Leidenfrost shift than that found with hydrophilic solids."
} | 4,873 |
28415180 | null | s2 | 937 | {
"abstract": "Swarming groups of bacteria coordinate their behavior by self-organizing as a population to move over surfaces in search of nutrients and optimal niches for colonization. Many open questions remain about the cues used by swarming bacteria to achieve this self-organization. While chemical cue signaling known as quorum sensing is well-described, swarming bacteria often act and coordinate on time scales that could not be achieved via these extracellular quorum sensing cues. Here, cell-cell contact-dependent protein exchange is explored as a mechanism of intercellular signaling for the bacterium Myxococcus xanthus. A detailed biologically calibrated computational model is used to study how M. xanthus optimizes the connection rate between cells and maximizes the spread of an extracellular protein within the population. The maximum rate of protein spreading is observed for cells that reverse direction optimally for swarming. Cells that reverse too slowly or too fast fail to spread extracellular protein efficiently. In particular, a specific range of cell reversal frequencies was observed to maximize the cell-cell connection rate and minimize the time of protein spreading. Furthermore, our findings suggest that predesigned motion reversal can be employed to enhance the collective behavior of biological synthetic active systems."
} | 335 |
31371705 | PMC6672015 | pmc | 938 | {
"abstract": "Owing to their attractive application potentials in both non-volatile memory and unconventional computing, memristive devices have drawn substantial research attention in the last decade. However, major roadblocks still remain in device performance, especially concerning relatively large parameter variability and limited cycling endurance. The response of the active region in the device within and between switching cycles plays the dominating role, yet the microscopic details remain elusive. This Review summarizes recent progress in scientific understanding of the physical origins of the non-idealities and propose a synergistic approach based on in situ characterization and device modeling to investigate switching mechanism. At last, the Review offers an outlook for commercialization viability of memristive technology.",
"introduction": "Introduction Memristive devices have been studied intensively since the link between memristor theory and physical resistive switching devices was established in 2008 1 , which was initially driven by the need for high-performance non-volatile memory and has more recently been fueled by energy-efficient unconventional computing 2 . Postulated as the fourth fundamental passive circuit element in addition to resistor, capacitor, and inductor, the memristor can store information in a form of resistance, which can be modulated by the history of its external stimuli 3 . The typical structure of memristors is a two-terminal three-layered stack, consisting of a switching layer sandwiched between two metallic electrodes. The switching layer ranges from semiconducting to insulating inorganic or organic materials. Through materials engineering, memristive devices can be tailored to provide non-volatile or volatile memory. Non-volatile memristor maintains its resistance state after the removal of the applied switching voltage or current. The stable resistance state is used to represent stored information, making memristive devices suitable for data storage applications. Integrated memristive crossbar arrays are considered promising candidates for future application in mainstream non-volatile memory. This is because memristive devices can store information at the sub-2-nm scale 4 and possess many other desired properties, including high speed 5 , low energy consumption, three-dimensional integration capability, and compatibility with complementary metal oxide semiconductor (CMOS) technologies 6 . Furthermore, in-memory analog computing is being developed to process information where it is stored 7 . Such in-memory computing is expected to offer an efficient and reconfigurable solution to process analog information in artificial intelligence (AI) applications 8 . On the other hand, the programmed resistance state of a volatile memristor gradually relaxes toward a thermodynamically stable state upon the removal of the programming signal, offering desirable dynamics for emulating biological synapses and neurons 9 . Memristors at the individual device level have shown short- and long-term plasticity similar to that of biological synapses 9 , 10 . Neural networks with memristors at the array level have been used to demonstrate brain-inspired functions 11 . Consequently, memristive devices have attracted significant attention in the past decade as a key enabler of new computing paradigms to overcome the limitations of the conventional von Neumann computing architecture. However, although significant research efforts have been directed toward memristive devices, a large-scale commercialization of these devices has not yet been achieved. In addition to challenges at the circuit, algorithm and architecture levels, issues at the device level are likely still the primary reason. Two of the major remaining challenges at the device level are relatively large parameter variability and poor cycling endurance (Fig. 1a ). To improve the cycling endurance and reduce the parameter variability to a level sufficient for large-scale commercialization, it is necessary to acquire an in-depth understanding of ion migration and its coupling with electron transport—the dominating dynamics in memristive mechanism—during the switching process. To reveal the dynamic process of switching, in situ characterization techniques are necessary; then, device modeling is needed to thoroughly explain the phenomena observed in situ. As such, in situ characterization combined with device modeling is the most efficient approach to achieve a complete and in-depth understanding of the switching mechanism. In this Review, we summarize state-of-the-art understanding of memristive switching mechanism and discuss future research directions. We primarily focus on in situ characterization techniques and device modeling methods, aiming to monitor and analyze the switching behavior of memristors at the single-atom level. Fig. 1 Synergistic approaches for mechanistic research of memristive devices for improving the device performance. a Parameter variability of the set and reset voltages (left), and typical endurance failure behaviors in memristors (right). HRS represents the high resistance state, and LRS represents the low resistance state. b Possible microscopic origins responsible for the device behaviors: conduction filaments in different forms contribute to the parameter variability (left of panel a ); after cycling operation, the formation of an interfacial layer with high series resistance (orange disc), or the expansion of migration area for mobile ions (migrated blue balls marked by black arrows), leading to the cycling endurance failure (right of panel a ). c The in situ characterizations and device modeling can complement and complete each other and together provide a holistic picture of the resistive switching, as shown in the middle where a device model is schematically presented with switching I – V loops of simulated and experimental data. Reproduced from ref. 12 , Macmillan Publishers Ltd ( c ), Ag/SiO 2 /Pt device; ref. 22 , Macmillan Publishers Ltd ( c ), Ta 2 O 5-x /SiO 2 /Pt device"
} | 1,519 |
37549275 | PMC10437433 | pmc | 939 | {
"abstract": "Significance Changes in an animal’s behavioral state, such as arousal and movements, induce complex modulations of the baseline input currents to sensory areas, eliciting sensory modality-specific effects. A simple computational principle explaining the effects of baseline modulations on recurrent cortical circuits is lacking. We investigate the benefits of baseline modulations using a reservoir computing approach in recurrent neural networks with random couplings. Baseline modulations unlock a set of network phases and phenomena, including chaos enhancement, neural hysteresis, and ergodicity breaking. Strikingly, baseline modulations enable reservoir networks to perform multiple tasks, without any optimization of the network couplings. Baseline control of network dynamics opens directions for brain-inspired artificial intelligence and sheds light on behavioral modulations of cortical activity.",
"discussion": "Discussion We presented a brain-inspired framework for reservoir computing where we controlled the dynamical phase of a recurrent neural network by modulating the mean and quenched variance of its baseline inputs. Baseline modulations revealed a host of phenomena. First, we found that they can set the operating point of the network activity by controlling whether synaptic inputs overlap with the high gain region of the transfer function. A manifestation of this effect is a noise-induced enhancement of chaos. Second, baseline modulations unlocked access to a large repertoire of network phases. On top of the known fixed point and chaotic ones, we uncovered three bistable phases, where the network activity breaks ergodicity and exhibits the simultaneous coexistence of a fixed point and chaos, of two different fixed points, and weak and strong chaos. By driving the network with adiabatic changes in the baseline statistics one can toggle between the different phases, charting a trajectory in phase space. These trajectories exhibited a manifestation of the phenomenon of neural hysteresis, whereby adiabatic transitions across a phase boundary retain the memory of the adiabatic trajectory. Moreover, we showed that baseline control can achieve optimal performance in a memory task at a second-order phase boundary without any fine-tuning of the network recurrent couplings. In the bistable phases, we showed that the reservoir can perform different decision-making tasks, leveraging neural hysteresis and ergodicity breaking. Strikingly, we found that by simply varying the network baseline the reservoir can perform multiple tasks without any weight optimization. Our work provides a conceptual framework to achieve flexible performance and multitasking via the simple neural mechanism of baseline control, paving the way for an approach to reservoir computing. Noise-Induced Enhancement of Chaos. Previous theoretical work found a noise-induced suppression of chaos in random neural networks driven by time-varying inputs both in discrete time ( 29 ) and continuous time ( 22 , 24 , 28 , 30 , 34 ). In previous cases, featuring a mean synaptic input centered in the middle of the high-gain region of the transfer function, suppression of chaos occurs because an increase in the variance drives the network away from the chaotic regime. In contrast, we found that, when the baseline statistics sets the mean synaptic input away from the center of the high gain region, one can induce a transition from fixed point to chaos at intermediate values of the variance ( Fig. 3 ). Larger values of the variance eventually suppress chaos, such that a nonmonotonic dependence of the Lyapunov exponent on the baseline variance or mean can be realized. This is an example of noise-induced chaos in recurrent neural networks with additive interactions, although a similar phenomenon was recently found in networks with gated recurrent units ( 35 ) (for the logistic map see ref. 36 ). We believe that noise-induced modulation of chaos in discrete time networks is similar for both quenched and dynamical noise ( 24 ) since the LLE and the edge of chaos are the same for both cases. We speculate that introducing a leak term and generalizing our results to a continuous time system may induce a dynamical suppression of chaos on general grounds, based on the memory effect. Another interesting direction is to drive the network with dynamical noise at different values of the baseline input and investigate its effect on the different monostable and bistable phases we uncovered via baseline modulation. Optimal Sequential Memory. Previous studies showed that optimal performance in random networks can be achieved by either tuning the recurrent couplings at the edge of chaos ( 23 ) or by driving the network with noisy input tuned to a particular amplitude ( 24 ). Both those methods requires simple tuning of two hyperparameters [mean and variance of the random couplings ( 23 ) or noise ( 24 ) distribution], as in our model. It would be interesting to compare these alternative methods, test whether any of them is realized in cortical circuits and develop optimization algorithms to learn their parameters. Comparison with Other Multitasking Frameworks. Humans learn to perform new cognitive tasks by directly following instructions, without any training at all ( 37 ). On the other hand, brain-inspired RNNs can be trained to perform multiple tasks by optimizing their recurrent weights via gradient descent ( 33 , 38 ). This optimization procedure is costly, scaling as the square of the network size, and typically requires thousands or millions of training epochs to achieve good task performance; moreover, their maintenance is biologically implausible, as it requires a mechanism to fine-tune the value of the recurrent weights. Recent work showed that RNNs trained to perform a library of tasks via gradient descent can then quickly learn a new task by reutilizing learned computational motifs, such as learned fixed points or line attractors ( 38 , 39 ). Here, we took a different approach to multitasking by interpreting the reservoir’s own dynamical phases as a library of ’innate’ computational motifs. Each of the multiple bistable phases already present with random recurrent couplings was shown to implement a different binary choice, relying on the combination of their ergodicity breaking and neural hysteresis property. Task rules were implemented as values of the baseline input mean and variance ( Fig. 5 ). Unlike previous studies, our approach does not require any training of recurrent weights, thus avoiding the issues listed above. A limitation of our approach is that only a small number of bistable phases are available, and therefore, the expressivity of the reservoir is not large as the one achieved by trained RNNs ( 33 ). It is tantalizing to speculate that by combining our reservoir approach with some limited weight optimization one could learn a larger variety of computational motifs and lead to a more biologically plausible theory of multitasking RNNs. Information Processing Capabilities and Bistability. Bistable phases with coexistence of fixed points and chaos were previously reported in recurrent networks with random couplings ( 40 ) and with gated recurrent units ( 35 ). We generalized this to a set of bistable phases featuring the coexistence of two fixed points and, remarkably, two chaotic attractors with slow and fast chaos, respectively. This is a report of a doubly chaotic phase in recurrent neural networks. Are there any information processing benefits of the double chaos phase? Neural activity unfolding within the weakly chaotic branch of this bistable phase has large sequential memory capacity, as the Fisher information diverges at the edge of chaos. On the other hand, the strongly chaotic branch erases memory fast. In this doubly chaotic phase, the network’s information processing ability can be changed drastically by switching between the two branches, for example, via an external pulse. It would be tantalizing to explore the computational capabilities of these bistable phases unlocked by baseline modulation. Here, we only considered homogeneous inputs where the baseline statistics is the same for all network neurons. Although, one may consider a more general setup with heterogeneous inputs, where different neural populations receive baseline modulations with different statistics. The simplest such possibility would be the ability to perform different tasks by gating in and out specific subpopulations, driving them with negative input. This is a promising direction for multitasking, and we leave it for future work. Evidence for Baseline Modulations in Brain Circuits. In biologically plausible models of cortical circuits based on spiking networks, it was previously shown that increasing the baseline quenched variance leads to improved performance. This mechanism was shown to explain the improvement of sensory processing observed in the visual cortex during locomotion ( 11 ) and in the gustatory cortex with general expectation ( 2 ). In these studies, the effect of locomotion or expectation was modeled as a change in the constant baseline input to each neuron realizing an increase in the input quenched variance. This model was consistent with the physiological observation of the heterogeneous neuronal responses to changes in behavioral state, comprising a mix of enhanced and suppressed firing rate responses [during locomotion ( 3 , 11 , 25 ), movements ( 4 – 6 ), or expectation ( 14 , 41 )]. Intracellular recordings showed that these modulations are mediated by a change of baseline synaptic currents, likely originating from subcortical areas ( 8 , 9 ). Because the effects of these changes in behavioral state on neural activity unfolded over a slower timescale (a few seconds) compared to the typical information processing speed in neural circuits (subsecond), we modeled them as constant baseline changes, captured by changes in the mean and variance of the distribution of input currents. Our results provide an interpretation of these phenomena, leading to the hypothesis that they could enable cortical circuits to adapt their operating regimes to changing demands. Baseline Modulations and Gain Modulation. The effect of the baseline modulations on network dynamics highlighted in this study can be understood in terms of changes in the network effective transfer function Φ eff ( x ) = ∫ D z ϕ ( C z + μ + x ) , where Dz is a standard Gaussian measure, and C is the self-consistent variance of the activity, giving the self-consistent equation for the mean rate M = Φ eff ( M ) ( Materials and Methods ). Baseline modulations lead to changes in the slope of the effective transfer function, a relationship previously derived in spiking networks ( 2 ). This is consistent with experimental observations that changes in behavioral states are mediated by gain modulation, as observed at the level of single cells ( 1 ) as well as populations ( 11 ). Alternative mechanisms for gain modulation include changes in the background synaptic currents controlling the single-cell conductances ( 42 ), which are not captured by our rate-based model. Ergodicity Breaking. We found ergodicity breaking in network dynamics occurring in a series of bistable phases, which include phases with two fixed points, with a fixed point and chaos, and with weak/strong realizations of chaos. Ergodicity breaking was recently reported independently in a dynamically balanced neural network of inhibitory units in ref. 43 . The origin of the ergodicity breaking in these two models is different. While in our case it is driven by heterogeneity, or disorder, in the input baseline, in ref. 43 , it is caused by an overrepresentation of symmetric connections, leading to non-Gaussian inputs for each neuron as a consequence. Moreover, while we relied on DMFT to prove the existence of bistability, Berlemont and Mongillo ( 43 ) applied the cavity method to reveal a large number of metastable states. Neural Hysteresis. A prediction of our model is that baseline modulations may induce neural hysteresis when crossing a bistable phase boundary. Hysteresis is a universal phenomenon observed in many domains of physics. Hysteresis in neural networks was first observed in the presence of recurrent inhibition ( 44 , 45 ) and later confirmed in visual areas in vitro ( 46 ). In the Wilson–Cowan model ( 47 ), hysteresis was observed in the transitions between fixed points. In our case, hysteresis occurs in the transition between different network phases including chaotic and fixed-point regimes. Our results suggest a potential way to examine the existence of hysteresis in brain circuits, within the assumption that increasing baseline variance represents increasing values of a continuous behavioral modulation such as arousal [e.g., measured by pupil size ( 48 )]. A potential signature of hysteresis could be detected whether the autocorrelation time of neural activity at a specific arousal level exhibited a strong dependence on whether arousal levels decreased from very high levels or increased from very low levels. We leave this interesting direction for future work."
} | 3,297 |
39025960 | PMC11258262 | pmc | 941 | {
"abstract": "Circuit implementations of neuronal networks so far have been focusing on synaptic weight changes as network growth principles. Besides these weight changes, however, it is also useful to incorporate additional network growth principles such as guided axon growth and pruning. These allow for dynamical signal delays and a higher degree of self-organization, and can thus lead to novel circuit design principles. In this work we develop an ideal, bio-inspired electrical circuit mimicking growth and pruning controlled by guidance cues. The circuit is based on memristively coupled neuronal oscillators. As coupling element, we use memsensors consisting of a general sensor, two gradient sensors, and two memristors. The oscillators and memsensors are arranged in a grid structure, where oscillators and memsensors realize nodes and edges, respectively. This allows for arbitrary 2D growth scenarios with axon growth controlled by guidance cues. Simulation results show that the circuit successfully mimics a biological example in which two neurons initially grow towards two target neurons, where undesired connections are pruned later on.",
"conclusion": "Conclusion In this work, we have proposed a bio-inspired electrical circuit mimicking axon growth controlled by growth and pruning cues. Our circuit is based on a Morris-Lecar circuit as a node of Ranvier or axon hillock, and a memsensor as axon segment. The memsensor consists of one general sensor, two gradient sensors, and two memristors. The latter enable a permanent growth of axon segments when decreasing their resistance based on an STDP-like mechanism. As an overall and very general architecture, we have used a grid structure, where each node is implemented by the Morris-Lecar circuit and each edge by the memsensor. This way, once the target points are fixed, the setup allows for arbitrary growth paths, because starting points are chosen by just applying a current stimulus to the corresponding Morris-Lecar circuits. We like to stress that the STDP-like mechanism of the growth concept makes it possible to use other neuronal oscillators as well. Note that this requires e.g. the memsensor parameters or the input current signal to be redesigned. As a bio-inspired application example, we have considered two neurons both growing towards to two target neurons, although two of the resulting connections should be pruned later on. Simulations have verified that our circuit can mimic this behavior in terms of dynamically growing structures that affect the delay of the signal transmission. This underlines the bio-inspired and functional character of the self-organized network topology formation implemented by our circuit.",
"introduction": "Introduction The way a nerve network grows fundamentally affects its functionality. Understanding the underlying principles is not only relevant from a biological point of view, but also of great interest for technical implementations of neuronal networks. Up to now, most technical approaches focus on the growth and change of synapses. Based on biological findings, e.g. learning rules such as spike-timing-dependent plasticity (STDP) have been derived and used for neuromorphic circuits 1 , 2 . From a biological point of view, however, varying synaptic coupling strengths are only one aspect among many others that shape the structure of a neuronal network. Especially relevant, although often neglected, is axon growth, since it is actually this aspect that forms the connectome. A further aspect of axon growth is the emergence of signal transmission delays. For instance, transmitting an action potential of 5 ms duration across a one-meter-long axon takes 1.6–400 longer than the action potential itself 3 , 4 . Thus, axons directly influence synaptic learning rules such as STDP, and should not be neglected for neuronal functionality. Considering axon growth principles for technical implementations therefore offers both a functional advantage and a higher degree of self-organization for network growth. Growth of electrical circuits has for example been considered in the context of a circuit tile assembly model 5 – 7 . This circuit model, however, does only account for axon growth in a structural sense, as it neglects oscillatory signal (re)generation and propagation. A biologically abstract circuit implementation has been presented in 3 and is based on memristive Jaumann-structures and delay lines. Here, memristors are nonlinear resistors whose resistance depends on past signals and is maintained when the supplying voltage or current is turned off. A more bio-inspired circuit approach has been proposed in 8 , where neuronal oscillators are coupled via memristors to implement axon growth and pruning. All these circuit models account for axon growth, but neglect guided growth. How axons are guided during their growth is an active field of research in biology, see e.g. 9 – 11 . Since this aspect contributes to the self-organizing ability of neuronal networks, considering it for hardware realizations can improve the self-structuring of circuits. Mathematical models for guided axon growth have been discussed in 12 , 13 . Electrical circuit implementations of this aspect are extremely rare. Recently, an approach making use of memristive sensors in combination with delay lines and Jaumann structures has been reported in 14 . This approach is, however, a very specific implementation of one growth principle and as such not well usable for extending it towards more detailed growth mechanisms such as pruning. Our aim in this work is to design an electrical circuit that can mimic guided axon growth and pruning in a bio-inspired way. Unlike existing circuit implementations of unguided axon growth, this approach allows for controlled growth that affects both the functionality and the structure of the circuit. In contrast to 14 , the bio-inspired nature of our circuit also allows additional growth aspects to be incorporated more easily. To achieve these goals, we use memristively coupled neuronal oscillators. This is inspired by biologically reasonable axon models relying on Hodgkin-Huxley models 15 coupled via resistors, see for example the McNeal 16 , 17 or the spatially extended nonlinear node (SENN) model 17 , 18 . The use of memristors is especially popular for neuromorphic applications, where memristors are, for instance, utilized as synapses 19 , 20 . Moreover, they have also been considered for neuron models such as the Hodgkin-Huxley model 21 , the Morris-Lecar model 22 , and the Hindmarsh-Rose model 23 , 24 . In this work, memristors enable us to achieve a permanent growth of axon paths, similar to 8 . To verify the proposed electrical circuit, we use a wave digital simulation 25 . The wave digital method has, for example, successfully been applied to simulate both neuronal oscillators 26 , 27 and neuronal oscillator networks 8 , 28 . The remainder of this manuscript is structured as follows: In Section “ Circuit model ”, circuit models for neuron building blocks are presented and used to design an overall circuit for guided axon growth. In Section “ Simulation results and discussion ”, simulation results for an axon growth and pruning example taken from the neocortex of mice are discussed. Concluding remarks are given in Section “ Conclusion ”.",
"discussion": "Simulation results and discussion Minimal example \n Figure 3 Circuit setup for guided axon growth. ( a ) Complete circuit structure for the minimal example, with synapse memristors in orange and the growth cue concentration as background. ( b ) Morris-Lecar circuit as oscillator model for axon hillocks and nodes of Ranvier. \n To demonstrate the general functionality of the circuit concept, let us consider the minimal example shown in Fig. 3 a that displays the growth cue concentration as background for the circuit setup. The corresponding circuit consists of four oscillators, one for each of the four possible locations on the \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$2\\times 2$$\\end{document} 2 × 2 grid. Oscillators are connected to each other by two memsensors and two synapse memristors highlighted in black and orange, respectively. In the example, only the oscillator related to the position (1, 2) where growth starts is supplied with an external current signal j . The latter consists of rectangular shaped pulses, with an amplitude of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$2.4$$\\end{document} 2.4 μA, pulse width \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$T_\\mathrm{pw} = 2\\,\\textrm{ms}$$\\end{document} T pw = 2 ms , and a pause between the pulses of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$T_\\mathrm{p} = 4\\,\\textrm{ms}$$\\end{document} T p = 4 ms . The oscillator of the target position at (2, 1) is connected to a pruning cue generation circuit that should dissolve the oscillatory connection via the Memsensor \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 . Both oscillators at start and target position are considered as axon hillocks, since they act as the first segment of a neuron. Results for the setup are depicted in Fig. 4 and are obtained through wave digital simulations. Note that we set \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\eta _2=0.5$$\\end{document} η 2 = 0.5 to narrow the sensor range in this minimal example. As can be seen from the top of Fig. 4 b, both total memristances \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},2}$$\\end{document} M a , 2 decrease below \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$100\\,\\mathrm {k\\Omega }$$\\end{document} 100 k Ω within less than \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$10\\,\\textrm{ms}$$\\end{document} 10 ms . This is due the memristors in positive direction switching to the low resistance states, while memristors in negative direction remain in the high resistance states. This can be inferred from the center and bottom of Fig. 4 b. \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$100\\,\\mathrm {k\\Omega }$$\\end{document} 100 k Ω is set as threshold value for determining grown connections, since this value permits signal transmission. We have highlighted the corresponding network structure in Fig. 4 a for specific time points. The left side shows the growth cue concentration and the network structure, while the right side displays the pruning cue concentration. This shows that shortly after \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$10\\,\\textrm{ms}$$\\end{document} 10 ms , at which both connections from position (1,2) to position (2,1) are established, pruning cues are mainly generated in the direction of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 . \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$R_{\\textrm{p},1}$$\\end{document} R p , 1 senses this generated pruning cue concentration and increases, due to which \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 rises as well. This in turn reduces the voltage at the memristor being in the low resistance state, making it slowly switch back to its high resistance state, cf. Fig. 4 b. The sharp increase of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 in the beginning is mainly due to the sensor resistor \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$R_{\\textrm{p},1}$$\\end{document} R p , 1 , since the pruning cue concentration is slowly building up. After the concentration and hence the sensor resistance reach saturation, the weaker increase of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 stems from the memristor that switches back. The connection via \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},1}$$\\end{document} M a , 1 is completely dissolved after \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$4.3\\,\\textrm{s}$$\\end{document} 4.3 s , while the desired connection via \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_{\\textrm{a},2}$$\\end{document} M a , 2 is still below \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$100\\,\\mathrm {k\\Omega }$$\\end{document} 100 k Ω . This enables a working signal transmission from position (1,2) to position (2,1), as can be seen from Fig. 4 c displaying the corresponding oscillator voltages. The example thus shows the successful growth and subsequent pruning of desired and undesired connections. Figure 4 Simulation results for the minimal example. ( a ) The resulting network structure and the growth cue concentration at a specific time point are shown on the left side, with grown axon segments and active axon hillocks marked in yellow, and synaptic connections in orange. The right side contains the pruning cue concentration. ( b ) Total memristances of the memsensor (top), memristor states for positive direction (center), and memristor states for negative direction (bottom). ( c ) Axon hillock voltages. Biological example In the following, we consider a circuit setup inspired by axon growth and pruning observed from the neocortex of mice 34 . In particular, axons from both a visual cortex neuron \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_1$$\\end{document} N 1 and a motor cortex neuron \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_2$$\\end{document} N 2 grow towards a spinal cord neuron \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_2'$$\\end{document} N 2 ′ and a superior colliculus neuron \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_1'$$\\end{document} N 1 ′ . Note that the superior colliculus is a visual computation center. However, only the connections from the visual cortex to the superior colliculus and from the motor cortex to the spinal cord are desired. As a result, the false connections between neurons \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_1$$\\end{document} N 1 and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_2'$$\\end{document} N 2 ′ and neurons \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_2$$\\end{document} N 2 and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_1'$$\\end{document} N 1 ′ are pruned afterwards. As an environment for the growth and pruning, we consider a \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$15\\times 15$$\\end{document} 15 × 15 grid, containing 225 Morris-Lecar circuits, 416 memsensors, four synapse memristors, and two RC circuits for generating pruning cues. In this grid, see Fig. 5 a, we place the axon hillocks of the growing neurons \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1}$$\\end{document} N 1 and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2}$$\\end{document} N 2 at (5, 15) and (15, 15), respectively. Corresponding Morris-Lecar circuits are excited by the external current signal j that is chosen equal to the minimal example. The axon hillocks of the target neurons \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1'}$$\\end{document} N 1 ′ and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2'}$$\\end{document} N 2 ′ are located at (5, 3) and (10, 8), respectively. Synapses are placed above and right of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1'}$$\\end{document} N 1 ′ , as well as left and right of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2'}$$\\end{document} N 2 ′ . Figure 5 Simulation results for the biologically inspired example. ( a ) The resulting network structure and the growth cue concentration at a specific time point are shown on the left side, while the right side displays the pruning cue concentration. ( b ) Total memristances of the memsensor (top), memristor states for positive direction (center), and memristor states for negative direction (bottom). ( c ) Axon hillock voltages. Simulation results are shown in Fig. 5 . Similar to the minimal example, Fig. 5 b displays the total memristances \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_\\mathrm{a}$$\\end{document} M a and both internal memristor states \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$z_\\mathrm{a,p}$$\\end{document} z a , p and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$z_\\mathrm{a,n}$$\\end{document} z a , n for the positive and negative direction in the top row, center row, and bottom row, respectively. Blue curves indicate desired connections, red curves indicate connections that should later on be pruned, and black curves are connections that are not part of the directs paths between start and target neurons. Memsensors that sense a sufficiently large growth cue gradient start at memristance values of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$800\\,\\mathrm {k\\Omega }$$\\end{document} 800 k Ω , while all other memsensors start at \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$1.8\\,\\mathrm {M\\Omega }$$\\end{document} 1.8 M Ω . Of the memsensors sensing a large enough growth cue gradient, those connected by a continuous pathway to the oscillators modeling active axon hillocks switch to the low resistance state. This can be inferred from the internal memristor states and indicates that corresponding connections have grown. Overall growth is finished after \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\approx 0.22\\,\\textrm{s}$$\\end{document} ≈ 0.22 s . Pruning starts earlier, because connections from the start neurons to the target neurons \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1'}$$\\end{document} N 1 ′ and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2'}$$\\end{document} N 2 ′ are established after \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$0.105\\,\\textrm{ms}$$\\end{document} 0.105 ms and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$0.141\\,\\textrm{ms}$$\\end{document} 0.141 ms , respectively. These time points can be taken from the axon hillock voltages illustrated in Fig. 5 c that shows when the target neurons become active. As the pruning cue concentration rises, corresponding sensor resistances increase, resulting in rising total memristances \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$M_\\mathrm{a}$$\\end{document} M a of those connections to be pruned. As in the minimal example, this reduces the voltage of the internal memristors, such that they slowly drop back to the high resistance state. As can be seen from the center and bottom row of Fig. 5 b, this high resistance state is reached after \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$~4.5\\,\\textrm{s}$$\\end{document} 4.5 s for almost all pruned connections. This demonstrates that self-organized growth and pruning is also possible for larger setups. To highlight the delayed signal transmission that comes with this growth and pruning process, we have excited the target neurons with a separate current pulse at \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$4.99\\,\\textrm{s}$$\\end{document} 4.99 s , following a short pause. This can be seen from the right side of Fig. 5 c and illustrates that signal transmission from \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1}$$\\end{document} N 1 to \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{1'}$$\\end{document} N 1 ′ and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2}$$\\end{document} N 2 to \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$${\\mathcal {N}}_{2}'$$\\end{document} N 2 ′ takes \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$8\\,\\textrm{ms}$$\\end{document} 8 ms and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$6\\,\\textrm{ms}$$\\end{document} 6 ms , respectively. The biologically inspired example also highlights the advantages of the proposed circuit approach compared to the existing one for guided axon growth 14 . First, our approach provides a pruning concept that is missing in the previous work. This pruning improves the self-organized growth of connections and allows for dynamically reconfigurable setups. Second, we deploy a more elaborated memsensor model based on separated memristors and sensors instead of a memristor with sensing abilities. Since the latter are advantageous in functionality, but hardly available as they are very idealized, this makes our approach more flexible for possible implementations. Third, the concept of 14 relies on circulators that relay voltages from ones axon segment to another. This limits its applicability to branching axons, since each branch halves the transmitted voltage. In contrast to this, our concept makes use of coupled neuronal oscillators that are able to regenerate the transmitted voltage. This way, our concept is well suited for branching axons, as has been demonstrated. Robustness against parameter variations \n Figure 6 Simulation results for the biologically inspired example with memsensor parameters varied by ± 10%. Results for parameters with \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$+\\,10\\%$$\\end{document} + 10 % deviation are shown in ( a ), ( b ), while ( c ), ( d ) display results for parameters with − 10% deviation. Resulting network structures, growth cue concentrations, and pruning cue concentrations are shown in ( a ), ( c ). Total memristances and internal state variables for positive and negative direction are illustrated in ( b ), ( d ). \n To observe the influence of parameters variations on the biologically inspired example, we have investigated two worst-case scenarios where the memsensor parameters are varied by either + 10% or − 10%. Corresponding results are shown in Fig. 6 a, b (+ 10%) and Fig. 6 c, d (− 10%). As can be seen, the emerging structures with respect to the threshold memristance \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$100\\,\\mathrm {k\\Omega }$$\\end{document} 100 k Ω remain equal except for one element less to be pruned in the case of \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$+10\\%$$\\end{document} + 10 % . Little effect can also be observed for a full pruning of memristors, meaning they returned to their high resistance state. This can again be inferred from the state variables \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$z_\\mathrm{a,p}$$\\end{document} z a , p and \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$z_\\mathrm{a,n}$$\\end{document} z a , n . These state variables show that compared to 0% parameter variation, see Fig. 5 , a similar amount of memristors returns to the high resistance state. Differences in the results are due to variations of pruning sensor sensitivity, high resistance states of pruning sensor and memristors, and the threshold voltage of the memristors, as these parameters mainly determine which and how effectively connections are pruned."
} | 9,089 |
38834541 | PMC11150455 | pmc | 942 | {
"abstract": "Biological systems interact directly with the environment and learn by receiving multimodal feedback via sensory stimuli that shape the formation of internal neuronal representations. Drawing inspiration from biological concepts such as exploration and sensory processing that eventually lead to behavioral conditioning, we present a robotic system handling objects through multimodal learning. A small-scale organic neuromorphic circuit locally integrates and adaptively processes multimodal sensory stimuli, enabling the robot to interact intelligently with its surroundings. The real-time handling of sensory stimuli via low-voltage organic neuromorphic devices with synaptic functionality forms multimodal associative connections that lead to behavioral conditioning, and thus the robot learns to avoid potentially dangerous objects. This work demonstrates that adaptive neuro-inspired circuitry with multifunctional organic materials, can accommodate locally efficient bio-inspired learning for advancing intelligent robotics.",
"introduction": "Introduction Advancements in the field of robotics have witnessed a notable shift towards bio-inspiration, motivated by the remarkable capabilities of biological nervous systems 1 – 3 . Bio-inspired robotics introduces novel ways for robots to interact with and be integrated into the physical world. Achieving this goal often necessitates the use of functional materials chosen for their ability to provide the desired flexibility, deformability, or adaptability 4 , 5 . At the same time, artificial intelligence (AI) is already demonstrating its proficiency for highly complex tasks in various domains such as data analysis, decision making and computer vision 6 . AI systems mostly utilize large-scale (deep) neural networks for learning, pattern recognition, classification, and language processing inside a static environment 7 , 8 . These systems are based on gradient-based algorithms that require high computing power and memory storage as well as a large amount of labeled training data. Although these systems are highly effective, their biological plausibility is limited 9 , and they can be power-hungry 10 . Hence, there is a desire to explore alternative bio-inspired algorithms, such as spiking neural networks, genetic and evolutionary algorithms, and swarm strategies, and to further enhance the development of specialized neuromorphic hardware platforms 11 – 13 . Such innovations in algorithms and hardware have proven to be powerful tools for simulating neural processes, accelerating the training of artificial neural networks, and leading to increasingly sophisticated hardware for artificial neural systems. However, essential adaptive neuronal processes, including associative learning and behavioral conditioning, exist in primitive organisms like the box jellyfish which even lack centralized nervous systems 14 . This raises the question of whether complexity in algorithms and architectures is always imperative for achieving cognitive functions and intelligent behavior. The relatively simple neural circuits of primitive species still exhibit significant capabilities, suggesting that emulating fundamental biological learning principles locally with functional materials and devices could be equally important as complexity while gaining efficiency 4 , 15 . Primitive biological organisms employ fundamental strategies for learning, such as exploration, multimodal processing, and behavioral conditioning. From early developmental stages, living beings instinctively start to learn from experience and through trial and error by interacting with their surroundings 16 . During this initial exploration phase, behaviors tend to be somewhat random and lack a specific goal while the organism is engaged with the environment via a wide range of sensory modalities (touch, vision, olfaction, etc.). The randomness of certain behaviors, such as bumping into an object, leads to the discovery of new sensations and, consequently, learning opportunities. Through this physical interaction of organisms with their surroundings, behavioral randomness develops gradually into consistency 17 . In this context, multimodal sensing enables the collection of various sensations describing the same event. These concurrent multimodal observations are synchronized in time and, as a result, become correlated, establishing autonomic connections across different sensory modalities and enabling behaviors such as respondent (Pavlovian) conditioning and associative learning. Indeed, a recent study of the complete connectome of a Drosophila brain reveals that the majority of neurons process multimodal signals 18 . Adaptivity and plasticity in function and behavior - essentials for biological development - are especially effective if previous experiences and memory are taken into account as well 19 . For instance, behaviors are associated with consequences through affirmative (rewards or reinforcement) and adverse (punishment) stimuli to strengthen or weaken a specific behavior (operant conditioning). By providing diverse sensory feedback and abundant opportunities to learn from the environment, explorative behavior and multimodal processing allow for instruction-free processes that converge into optimal behavioral conditions via adaptivity. Emerging functional materials and devices can offer unique properties that go beyond what conventional systems and electronics could achieve 20 . Organic mixed ionic-electronic (semi)conductors have recently experienced a notable upswing in neuromorphic engineering 21 – 23 . They are able to replicate bio-inspired functionalities such as synaptic plasticity 24 , 25 , neural processing 26 , high connectivity and recurrence 27 , 28 and even forgetfulness 29 just by material-inherent mechanisms. Key features of organic synaptic devices are their adaptivity through linear, symmetric, and analog tuning of electrical conductance and their operation at low voltage with high energy efficiency 30 . The compatibility of organics with solution-based processes and large-area integration into flexible or stretchable substrates can enable the merging of organic neuromorphic electronics in unconventional form factors (body, robotics, buildings, etc.) 31 . Indeed, significant steps have been made using conductive polymers regarding localized handling of data via on-chip training 32 , 33 , real-time operation with online learning 34 and spiking circuits for bio-integration 35 , 36 . Despite these significant demonstrations, applications are often limited to abstract and conceptual demonstrations in well-defined laboratory settings or mock environments, enabled by simple binary decisions. Robotic setups offer a realistic platform for interaction-rich, real-life setups 37 . Robotic manipulators, for example, are crucial for a variety of applications serving in versatile and dynamic environments, ranging from industrial assembly lines to neural prostheses. Highly adaptive and localized control close to the sensory nodes can drastically improve performance and can also warrant operational safety which is essential for human-oriented purposes such as neuroprosthetics 38 , 39 . In this work, we present a robotic system that uses multimodal sensory stimuli to explore and interact with a real-world environment in real time while adapting to it using bio-inspired mechanisms. At the core of adaptivity and learning of the robotic system is an organic neuromorphic circuit that consists of organic electrochemical transistors (OECTs) and organic neuromorphic devices (also called electrochemical random-access memories, ECRAMs). This bio-inspired approach enables the robotic agent to incrementally learn and perform a complex behavioral task, showcasing its adaptability and distributed intelligence in responding locally to dynamic and multimodal environmental cues. More specifically, the robotic system gains the ability to distinguish between safe and potentially harmful objects through local adaptation of neuromorphic circuitry. This work demonstrates that highly functional organic materials can reform neuromorphic hardware, rethinking adaptive intelligent systems as small(er)-scale local circuitry that interacts with the environment with bio-inspired learning mechanisms.",
"discussion": "Discussion Taking inspiration from the versatile capabilities of biological systems, we combine bio-inspired processing, learning, and control paradigms with the development of organic neuromorphic circuitry, and we demonstrate a standalone robotic system that interacts intelligently with a non-static environment. Through the integration of an organic neuromorphic circuit, the system adapts its behavior based on multimodal sensory feedback from environmental cues. The synaptic devices in the circuit enable associative learning, leading to both respondent (Pavlovian) conditioning and more complex operant conditioning. The robotic agent learns to associate positive and negative consequences with multimodal stimuli, showcasing adaptability and the ability to distinguish between safe and potentially harmful objects. The use of functional materials, such as organic (semi-)conducting polymers, in the neuromorphic circuit is elemental to the system’s capabilities, replicating bio-inspired functionalities like synaptic plasticity, dendritic summation, and neural processing. This is possible using small-scale, locally integrated, and low-voltage monolithic polymer electronics. Moreover, due to the modular-like structure of the neuromorphic circuit, the concept can be extended into multiple branches in order to handle sensory signals of arbitrary complexity and multimodality. The presented robotic system serves as a tangible example of how combining bio-inspired principles with localized organic neuromorphic circuitry can lead to the development of highly adaptive, intelligent, and efficient systems for real-world applications."
} | 2,484 |
24708438 | PMC3997834 | pmc | 943 | {
"abstract": "Background The currently accepted thesis on nitrogenous fertilizer additions on methane oxidation activity assumes niche partitioning among methanotrophic species, with activity responses to changes in nitrogen content being dependent on the in situ methanotrophic community structure Unfortunately, widely applied tools for microbial community assessment only have a limited phylogenetic resolution mostly restricted to genus level diversity, and not to species level as often mistakenly assumed. As a consequence, intragenus or intraspecies metabolic versatility in nitrogen metabolism was never evaluated nor considered among methanotrophic bacteria as a source of differential responses of methane oxidation to nitrogen amendments. Results We demonstrated that fourteen genotypically different Methylomonas strains, thus distinct below the level at which most techniques assign operational taxonomic units (OTU), show a versatile physiology in their nitrogen metabolism. Differential responses, even among strains with identical 16S rRNA or pmoA gene sequences, were observed for production of nitrite and nitrous oxide from nitrate or ammonium, nitrogen fixation and tolerance to high levels of ammonium, nitrate, and hydroxylamine. Overall, reduction of nitrate to nitrite, nitrogen fixation, higher tolerance to ammonium than nitrate and tolerance and assimilation of nitrite were general features. Conclusions Differential responses among closely related methanotrophic strains to overcome inhibition and toxicity from high nitrogen loads and assimilation of various nitrogen sources yield competitive fitness advantages to individual methane-oxidizing bacteria. Our observations proved that community structure at the deepest phylogenetic resolution potentially influences in situ functioning.",
"conclusion": "Conclusions Microbial ecologists still struggle to link microbial community structure to ecosystem functioning, notwithstanding that in-depth analyses of microbial community structure has never been easier than now, with the affordability of deep sequencing of amplified genes as well as whole communities. But to make sense out of sequence data, basic knowledge about the biology of the microorganisms involved in the biogeochemical functions under study is pivotal. Unfortunately, current insights are often retrieved from sporadic pure culture studies including only few distantly related strains. Here, investigation of closely related, genotypically different methanotrophic strains revealed differential responses to overcome inhibition and toxicity from high nitrogen loads and assimilation of various nitrogen sources, yielding competitive fitness advantages to individual methane-oxidizing bacteria. We have not assessed the specific effect of nitrogen on methanotrophic activity rate, but rather demonstrated the surprising versatility below the commonly used cut-off for operational taxonomic units, i.e. genus level. Our results proved that community structure at the deepest phylogenetic resolution potentially influences in situ functioning. Until molecular tools become available to allow much finer analyses of microbial diversity, metabolic variability at those unmeasurable levels should be taken into account.",
"discussion": "Discussion The ability of Methylomonas strains to produce nitrite and nitrous oxide from moderate levels (10 mM) of nitrate and ammonium, indicative of detoxification, demonstrated metabolic variability on both genus- and species level (Table 2 ; Figure 1 ). Ammonium amendments introduce a fraction of ammonia to the culture depending on the pH, which can be oxidized to hydroxylamine by methane monooxygenase. The conversion of hydroxylamine directly to nitric oxide or indirectly via nitrite is well studied and limited to the activity of hydroxylamine oxidoreductase [ 8 - 10 ], and to a much lesser extent cytochrome P460 [ 6 ]. In contrast, nitrate metabolism is underexplored in methanotrophs and little is known, except that nitrite from nitrate can be produced by both assimilatory and dissimilatory nitrate reductases [ 6 , 28 ]. Nevertheless, high levels of nitrite produced from nitrate should also be converted to nitric oxide for nitrite detoxification. Upon production, the cytotoxic nitric oxide needs to be immediately detoxified to the, at least for the cell, harmless nitrous oxide by one of the many known nitric oxide reductase enzymes [ 29 - 33 ]. In the present study, most strains were able to produce nitrous oxide from ammonium without preceding measurable levels of nitrite, either due to an immediate conversion of hydroxylamine to nitric oxide by hydroxylamine oxidoreductase or to small transient nitrite peaks below detection levels. The latter hypothesis is plausible when considering the nearly identical nitrous oxide profiles observed for M. methanica NCIMB 11130 T and Methylosinus sp. R-45379, except the nitrite peak (Figure 1 : phenotypes I and VIII), suggesting the same detoxification mechanism. Nitrate as sole nitrogen source was converted to nitrite in all strains once oxygen concentrations became low, except for M. methanica R-45363 that probably lacks an enzyme for nitrate reduction. Strains of both M. methanica and M. koyamae showed a drop in nitrite levels with a corresponding rise in nitrous oxide levels (Table 2 ; Figure 1 ). These observations would fit nicely with activities of an oxygen-sensitive nitrate reductase producing nitrite in actively growing cells, which is subsequently detoxified to nitrous oxide (via nitric oxide, not measured in this study) during stationary phase, similar to previous observations for non-denitrifiers [ 34 , 35 ], but unreported in methanotrophs. M. methanica NCIMB 11130 T and R-45364 produced both nitrite and nitrous oxide from nitrate, but did not show a subsequent drop in nitrite and corresponding rise in nitrous oxide (Table 2 ; Figure 1 : phenotypes I, IV and VIII). M. koyamae R-45383 was the only strain not able to produce nitrous oxide from either nitrate or ammonium, probably because it lacks a nitric oxide reductase. On the other hand, M. koyamae strain R-49807 did produce nitrous oxide from nitrate but not from ammonium. This indicates that the nitrous oxide produced was truly derived directly from nitrate via nitrite, and not via an indirect process via ammonium, i.e. assimilatory reduction of nitrate to ammonium. It is clear that, although variable phenotypes for nitrite and nitrous oxide production from nitrate and ammonium can be tentatively explained using the currently described genomic inventories of other methanotrophs [ 6 ] or a combination thereof, various novel features on specifically nitrate reduction should be the subject of further genome and expression studies. Within a single species, strain-dependent differences were also observed. This was most obvious for strains within M. koyamae regarding their tolerance to high ammonium levels. Noteworthy are the identical 16S rRNA gene and (almost) identical functional gene sequences between the highly ammonium-tolerant M. koyamae strains NCIMB 14606 T and R-49799 (up to 100 mM) and the low ammonium-tolerant strain R-49807 (up to 40 mM) (Additional file 2 : Figure S4 and Additional file 4 : Figure S5; Additional file 3 : Table S1 and Additional file 5 : Table S2). This differentiation among M. koyamae strains is further demonstrated in their tolerance to hydroxylamine (up to 1 mM). In addition, R-49807 unexpectedly did not exhibit N 2 fixation, as sole M. koyamae strain, despite a >99% nifH gene sequence similarity with strains NCIMB 14606 T and R-49799, both positive for N 2 fixation under low oxygen tension. This might be explained by a higher oxygen sensitivity of the nitrogenase of this strain [ 14 , 36 ], requiring oxygen concentrations lower than the 2% used here. In addition, something worth considering as well is the applicability of M. koyamae R-49799 for methane mitigation in high-ammonium sites, since this strain grows at 100 mM ammonium levels, utilizes 2 mM nitrite as sole nitrogen source, tolerates hydroxylamine levels up to 1 mM and possesses pathways to detoxify ammonia and nitrite. Some features appeared to be general within Methylomonas with some strain specific exceptions. The observed higher tolerance to ammonium than nitrate confirmed previous reports of methanotrophic growth inhibition above 40 mM nitrate [ 37 , 38 ]. Nevertheless, two M. lenta strains (R-45370 and R-45377) tolerated nitrate concentrations up to 100 mM. This could not be linked to a possible higher nitrite tolerance, as all strains could grow with up to 2 mM nitrite as sole nitrogen source. This nitrite tolerance and assimilation was in contrast to earlier findings that nitrite utilization was rare for M. methanica members [ 7 ]. The presence of the nifH gene and the ability to fix nitrogen with an oxygen-sensitive nitrogenase was also found in most strains, although both traits appear to be strain-dependent when combining the results of this study with other reports [ 12 ]. Although initially thought to be limited to mostly Type II and Type Ib methanotrophs [ 11 ], our findings are in agreement with more recent reports suggesting that nitrogen fixation is a common feature of many methanotrophs, including Type Ia and verrucomicrobial methanotrophs [ 12 - 14 ]."
} | 2,343 |
25866914 | null | s2 | 944 | {
"abstract": "Biofilms are surface-attached microbial communities that have complex structures and produce significant spatial heterogeneities. Biofilm development is strongly regulated by the surrounding flow and nutritional environment. Biofilm growth also increases the heterogeneity of the local microenvironment by generating complex flow fields and solute transport patterns. To investigate the development of heterogeneity in biofilms and interactions between biofilms and their local micro-habitat, we grew mono-species biofilms of Pseudomonas aeruginosa and dual-species biofilms of P. aeruginosa and Escherichia coli under nutritional gradients in a microfluidic flow cell. We provide detailed protocols for creating nutrient gradients within the flow cell and for growing and visualizing biofilm development under these conditions. We also present protocols for a series of optical methods to quantify spatial patterns in biofilm structure, flow distributions over biofilms, and mass transport around and within biofilm colonies. These methods support comprehensive investigations of the co-development of biofilm and habitat heterogeneity."
} | 284 |
36679373 | PMC9866600 | pmc | 945 | {
"abstract": "Triboelectric nanogenerators (TENGs) stand out as an attractive form of technology for the efficient harvest of mechanical energy and the powering of wearable devices due to their light weight, simplicity, high power density, and efficient vibration energy scavenging capabilities. However, the requirement for micro/nanostructures and/or complex and expensive instruments hinders their cheap mass production, thus limiting their practical applications. By using a simple, cost-effective, fast spray-coating process, we develop high-performance UV-curable triboelectric coatings for large-scale energy harvesting. The effect of different formulations and coating compositions on the triboelectric output is investigated to design triboelectric coatings with high output performance. The TENG based on a hybrid coating exhibits high output performance of 54.5 μA current, 1228.9 V voltage, 163.6 nC transferred charge and 3.51 mW output power. Moreover, the hybrid coatings show good long-term output stability. All the results indicate that the designed triboelectric coatings show great potential for large-scale energy harvesting with the advantages of cost-effectiveness, fast fabrication, easy mass production and long-term stability.",
"conclusion": "4. Conclusions In summary, we produced high-performance triboelectric coatings by using a simple, cost-effective, fast spray-coating process. In addition, the effect of different formulations and coating compositions on the triboelectric output was fully investigated. The hybrid composition coatings showed the best output performance as well as good long-term output stability. The hybrid composition-based TENG exhibited an output power approximately 87 times higher than that of the commercial PTFE film-based TENG, which opens up an exciting opportunity to improve the output performance of coating-based TENGs towards large-scale energy harvesting.",
"introduction": "1. Introduction The rapid development of flexible functional materials and advanced fabrication technologies has led to wearable devices being widely used in our daily lives, allowing us to monitor our health status and achieve individual eHealth [ 1 , 2 , 3 ]. Wearable devices require electricity to perform various functions, so the power source, typically conventional batteries, is an essential component. A significant increase in demand for sustainable and independent operation, lightweight, and flexibility has been seen with the development of wearable devices for health applications [ 4 , 5 , 6 ]. Conventional batteries, which are bulky and rigid, do not satisfy these requirements and will cause additional environmental burdens. Mechanical energy, the most widely distributed form of energy in the body, is the best source of energy when wearing wearable devices. A large number of body movements (e.g., running, walking, heart beating, breathing, talking, blinking, and swallowing) are performed every moment of every day, containing a large amount of biomechanical energy, which can be collected to power the wearable devices [ 7 , 8 , 9 ]. There are different types of energy harvesting devices that can convert mechanical energy into electricity, including the mechanisms of electromagnetic induction [ 10 , 11 ], piezoelectric effect [ 12 , 13 ], and triboelectric effect [ 14 , 15 ]. Taking the advantages of light weight, simple structure, high power density and efficient low-frequency vibration energy scavenging, triboelectric nanogenerators (TENGs) stand out as an attractive technology for efficient mechanical energy harvesting [ 16 , 17 , 18 ]. Based on the coupling effect of triboelectrification and electrostatic induction, TENGs can efficiently collect electricity from random, irregular, and/or low-frequency energy, such as mechanical vibration [ 19 , 20 ], wind [ 21 , 22 ], body motion [ 23 , 24 ], and ocean waves [ 25 , 26 , 27 ]. To push the TENGs into practical applications, many research advances have been made to improve the output performance of the devices, including surface modification [ 28 , 29 ], structure optimization [ 30 , 31 ], ion injection [ 32 , 33 ], and intermediate layer implantation [ 34 , 35 ], expanding the fields of application to self-powered sensing, smart wearables, and implantable electronics [ 36 , 37 , 38 ]. Despite these advances, most devices need micro/nanostructures and/or complex and expensive instrumentation, making inexpensive and large-scale mass production difficult, which has ultimately limited their practical applications. Developing new materials compatible with existing mass production techniques is urgent and meaningful to solve this problem. As a well-established industrial process, painting (mainly spraying, rolling, and brushing) is a versatile method for the mass production of films, showing great potential for the mass production of TENGs. In this respect, Chung and co-workers reported a superhydrophobic water-solid TENG, which was prepared by a commercial aerosol hydrophobic spray [ 39 ]. Later on, Yun and co-workers developed a commercial spray paint-based solid–solid TENG for smart traffic systems and security applications [ 40 ]. In addition to commercial spray, Liu and co-workers fabricated silk-fibroin based TENG by using a spray-coating process, which exhibits a maximum voltage of 213.9 V and power density of 68.0 mW/m 2 [ 41 ]. Saqib and co-workers proposed a natural seagrass-based material for spray-coatable TENG [ 42 ]. Wang and co-workers fabricated new hydrophobic organic coatings for water-solid TENG and hydropower harvesting [ 43 ]. Kong and co-workers developed solid–solid coating TENGs with antiwear and healing properties [ 44 ]. By adding mesoporous silica and perfluorooctylethanol, this coating TENG reached the short-circuit current of 10 μA and the output voltage of 220 V. Although significant advancements have been achieved in paintable TENGs, few studies have been conducted to develop painting materials specially designed for high output TENGs. In order to benefit the most from this energy-harvesting technology for wearable devices, we developed an organic coating that can be used for high-output TENGs via a simple painting process. Different formulations and coating compositions were evaluated to design high-performance triboelectric coatings. The TENG with 1:1 mixture of DFHMA and BA showed the highest output performance of 54.5 μA current, 1228.9 V voltage, 163.6 nC transferred charge, and 3.51 mW output power, as well as good long-term stability, and it can be used for large-scale energy harvesting through a fast and cost-effective spray-coating and a UV-curing process.",
"discussion": "3. Results and Discussion Among the polymers used in coatings, acrylic resins based on acrylate and methacrylate monomers dominate due to their excellent durability, weather resistance, gloss retention, adhesion, abrasion, and thermal resistance. So, in this work we chose four different acrylic monomers to develop an organic coating that can be used for triboelectricity harvesting. A commercial Fluororesin and PTFE film were also used as comparison. As mentioned in Section 2 , UV-curable triboelectric coating solutions with different formulations were prepared and spray-coated onto the substrates, followed by UV curing for 3–5 min, as shown in Figure 1 a. Spray coating is a simple, cost-effective, fast, and versatile process that can be applied to a wide range of surfaces, including flexible substrates. Further, this fabrication method can be easily extended to larger surfaces for mass production ( Figure 1 b). The inset in Figure 1 b shows that our triboelectric coating is flexible after being sprayed onto PET film. As shown in Figure 1 c, the triboelectric coating with copper foil tape exhibits stable mechanical flexibility during repeated bending and the resistance changed by less than 2 ohms. It is important to note that the coating solutions can be used not only for spray-coating, but also for other painting techniques, like brushing, rolling, and spin-coating, making it more practical. To obtain high output performance triboelectric coatings, we fabricated TENGs using triboelectric coatings with different formulations, including HFBMA, DFHMA, Fluororesin, MMA, BA, and DFHMA + BA. As a comparison, a commercial PTFE film based TENG was also fabricated. Nylon-11 spin-coated on Kapton film was used as another friction layer to construct TENGs. When the TENG was periodically pressed and released, Nylon-11 rubbed against the coating and generated positive charges on the Nylon-11 surface and negative charges on the surface of coating, according to the Triboelectric Series. As shown in Figure 2 , when the TENG is periodically pressed and released, there is an alternating current generated between the top and bottom electrodes because of the coupling effect of triboelectrification and electrostatic induction. In this paper, the output performances of all the TENGs were tested under the same conditions of the working frequency of 5 Hz. Figure 3 a–c shows the short circuit current, output voltage and transferred charge of the TENGs based on triboelectric coatings with different formulations and the commercial PTFE film. With a current of 43.1 μA, voltage of 1148.5 V, and transferred charge of 148.3 nC, the TENG based on DFHMA clearly has the highest output performance. The TENG based on BA has the second highest output performance (34.1 μA, 698.4 V and 88.3 nC) and the lowest comes from the TENG based on the commercial PTFE film (3.0 μA, 207.2 V, and 13.6 nC). To further investigate the output performance, we calculated the current density and charge density for all the TENGs ( Figure 3 d). According to the results, the TENG based on DFHMA achieves a current density and charge density of 26.9 mA/m 2 and 92.0 μC/m 2 , showing the same trend as short circuit current, output voltage and transferred charge. In order to determine the TENGs’ effective electric power, resistors are connected as external loads. The instantaneous current drops as the load resistance increases ( Figure 3 e) and the instantaneous output power reached its maximum value at a load resistance of 10 MΩ ( Figure 3 f). The output power of the TENG based on DFHMA, BA and commercial PTFE film is 2.76, 1.94, and 0.04 mW, respectively. Together these results indicate that the TENG based on DFHMA has the best output performance, which offers 69 times more output power than the TENG based on the commercial PTFE film. Along with coating formulations, we also examined the coating composition in relation to the triboelectric output. Specifically, the mass ratio between DFHMA oligomers and UV monomers was modulated as 4:1 to 2:1, 1:1, 1:2, and 1:4. To simplify the discussion, the composition of triboelectric coating is defined as Φ DFHMA oligomers + (1 − Φ ) UV monomers, where Φ is the mass ratio of DFHMA oligomers to UV monomers. TENG output performance with different Φ was evaluated and shown in Figure 4 a,b. The current, voltage, and transferred charge all showed the same trend and increased with Φ . This is primarily because DFHMA oligomers contain a large number of fluorine groups, while fluorine is the most electronegative element due to its strong ability to attract electrons. Increasing means adding more fluorine, which is beneficial to the output performance. This is also confirmed by the results of TENGs’ effective electric power ( Figure 4 c,d). It can be seen that the output power of the TENG with Φ = 80% is the highest, which is consistent with previous results. As part of our efforts to design a high-output triboelectric coating, we investigated how mixing two components would affect its performance. A TENG with a 1:1 mixture of DFHMA and BA was fabricated and examined. As shown in Figure 5 a–d, the TENG based on hybrid compositions has very high output performance of 54.5 μA current, 1228.9 V voltage, 163.6 nC transferred charge and 3.51 mW output power. It is very interesting that hybrid composition coatings have higher output performance than single composition coatings ( Figure 5 e,f). It is particularly noteworthy that the TENG based on hybrid compositions has an output power approximately 87 times higher than the TENG based on a commercial PTFE film. Table 1 shows the comparison of output performances of TENGs based on a spray-coating process with our acrylic resin-based hybrid coating TENG. This finding provides an exciting avenue to optimize the output performance of coating based TENGs. As a triboelectric coating, robustness plays a crucial role in practical applications besides outputting high performance. To fully evaluate our triboelectric coating, we carried out a long-term output stability experiment for the TENG with hybrid composition coatings. As shown in Figure 6 a, the TENG with hybrid composition coatings exhibits very good mechanical stability even after more than 16,000 working cycles. The current does not decrease between the starting and ending points ( Figure 6 b,c), showing that our triboelectric coating has great potential for long-term large-scale energy harvesting."
} | 3,281 |
26490957 | PMC4702795 | pmc | 946 | {
"abstract": "Quorum sensing is a widespread phenomenon in prokaryotes that helps them to communicate among themselves and with eukaryotes. It is driven through quorum sensing signaling molecules (QSSMs) in a density dependent manner that assists in numerous biological functions like biofilm formation, virulence factors secretion, swarming motility, bioluminescence, etc. Despite immense implications, dedicated resources of QSSMs are lacking. Therefore, we have developed SigMol ( http://bioinfo.imtech.res.in/manojk/sigmol ), a specialized repository of these molecules in prokaryotes. SigMol harbors information on QSSMs pertaining to different quorum sensing signaling systems namely acylated homoserine lactones (AHLs), diketopiperazines (DKPs), 4-hydroxy-2-alkylquinolines (HAQs), diffusible signal factors (DSFs), autoinducer-2 (AI-2) and others. Database contains 1382 entries of 182 unique signaling molecules from 215 organisms. It encompasses biological as well as chemical aspects of signaling molecules. Biological information includes genes, preliminary bioassays, identification assays and applications, while chemical detail comprises of IUPAC name, SMILES and structure. We have provided user-friendly browsing and searching facilities for easy data retrieval and comparison. We have gleaned information of diverse QSSMs reported in literature at a single platform ‘SigMol’. This comprehensive resource will assist the scientific community in understanding intraspecies, interspecies or interkingdom networking and further help to unfold different facets of quorum sensing and related therapeutics.",
"introduction": "INTRODUCTION Quorum sensing (QS) is a signaling mechanism by which bacteria communicate among themselves and with other organisms ( 1 – 3 ). Through QS, they sense and respond to environmental changes via signal transduction events using signaling molecules in a density dependent manner ( 4 ). In QS, signaling molecules (autoinducers) are secreted out of the cell and on attaining a certain threshold these are sensed by other cells present in their vicinity. It further activates cascade of signaling events resulting in the activation of QS genes ( 5 ). This phenomenon was firstly reported by Nealson et al., in two bioluminescent marine bacterial species i.e. Vibrio fischeri and Vibrio harveyi ( 6 ). However, the term QS was coined by Greenberg et al . ( 5 ). Later on, this mechanism was also discovered in Natronococcus occultus , an archaeal species ( 7 ). Quorums sensing signaling molecules (QSSMs) are broadly distributed into different signaling systems namely acyl homoserine lactones (AHLs), quorum sensing peptides (QSPs) diketopiperazines (DKPs), diffusible signal factors (DSFs), 4-hydroxy-2-alkylquinolines (HAQs), autoinducer-2 (AI-2), autoinducer-3 (AI-3) and others ( 8 ). Amongst all the known QS signaling systems, AHLs are the most prevalent molecules predominantly found in Gram-negative bacteria ( 9 ). AHLs have acyl side chain that varies from C 4 -C 18 of homoserine lactone moiety which are usually straight and in some cases may have branched configuration ( 9 – 11 ). AHLs are synthesized by LuxI or its homologues utilizing S-adenosylmethionine (SAM) and acyl–acyl carrier protein (acyl–ACP) as substrates ( 1 ). These signals, in turn sensed by LuxR or its homologues proteins leads to the activation of various physiological functions ( 9 ). Of the other QS signaling systems DKPs ( 12 ), DSFs ( 13 ), HAQs ( 14 ), AI-3 ( 3 ) are reported in Gram-negative bacteria, while QSPs are majorly found in Gram-positive bacteria ( 15 ). Moreover, AI-2 system is reported in both ( 16 , 17 ). QSSMs help prokaryotes to adapt in diverse environment through various biological processes. One such important aspect is the development and dispersion of biofilms to cope up with harsh conditions. Biofilm formation is widely reported in numerous bacterial species, e.g. Streptococcus mutans ( 18 ), Pseudomonas aeruginosa ( 19 ), Vibrio cholerae ( 20 ), etc., whereas biofilm dispersion in Staphylococcus aureus, Vibrio cholerae, Xanthomonas campestris and so forth ( 21 ). Similarly other process like release of virulence factors causes extensive damage to the host. This helps bacteria to escape from host immune response as described in Staphylococcal spp. ( 22 ), Streptococcal spp. ( 23 ), Burkholderia cepacia complex ( 24 ) and many more. Likewise other QS mediated processes viz. swarming motility, genetic competence, bioluminescence, etc. also assist bacteria in multifarious ways ( 1 , 15 , 25 ). Apart from signaling functions, some QSSMs are also involved in non-signaling events like iron chelation, and membrane modification with the help of 2-heptyl-3-hydroxy-4-quinolone (PQS) ( 26 , 27 ). Additionally, they also possess antimicrobial properties as mediated by autoinducer (lantibiotics) like nisin and subtilin from Lactobacillus lactis and Bacillus subtilis , respectively ( 28 ). Despite inevitable importance of QSSMs, this field is still computationally under explored. Only one depository ‘Quorumpeps’ is available for QSPs system with 231 QS peptides entries from 51 bacteria ( 15 ). Another is an algorithm ‘QSPpred’ for analyzing and predicting QSPs ( 29 ). Therefore, there is an exigent need of a resource for majority of QS signaling systems. To fill this void, we have developed a comprehensive database ‘SigMol’ of QS signaling molecules in prokaryotes.",
"discussion": "DISCUSSION Presence of QS phenomenon among prokaryotes in regulating numerous physiological processes and aiding in crosstalk with eukaryotes further highlights its importance ( 4 ). In this study, we are providing a compendium ‘SigMol’, which integrates QSSMs of various QS signaling systems reported in prokaryotes since 1970. Inference from the data statistics revealed that many organisms have more than one QS signaling system like in E. coli (AI-2 and AI-3), Vibrio spp. (AHLs and AI-2), Dickeya spp. (AHLs and AI-2), etc. Amongst various bacteria provided in the repository, only Burkholderia spp. and Pseudomonas spp. showed presence of four QS systems namely AHLs, DKPs, DSFs and HAQs as depicted through heatmap at http://bioinfo.imtech.res.in/manojk/sigmol/summary.php . Concurrently, within the same QS signaling system, a specific bacterium also generates diverse signaling molecules of that class. For example, 17 diverse AHLs have been reported for Sinorhizobium melliloti . Similarly, Burkholderia phytofirmans , Burkholderia xenovorans , Roseovarious tolerans , Pseudomonas aeruginosa are also known to produce 15, 11, 11, 10 QSSMs, respectively (Supplementary Figure S3). It seems that existence of so many AHLs within the same bacteria may help it to respond in different environments, however, this observation requires experimental validation. AHLs signaling system is the most abundant and important among prokaryotes. We have categorized AHLs into five groups according to acyl chain modifications namely saturated, unsaturated, carbonyl, hydroxyl and alanine methyl ester. Out of these, majority of AHLs belong to saturated and carbonyl followed by hydroxyl group. C6-HSL and C8-HSL are preferred among saturated AHL molecules, while OC6-HSL and OC8-HSL are for carbonyl group. Similarly, for hydroxyl group OHC8-HSL and OHC10-HSL are favored. Unlike bacteria, archaea have uncommon AHLs i.e. carboxylated-HSLs. Formation of biofilm is the representative outcome of intricate patterns of communication to enhance pathogenicity of bacteria. In a classical example, two bacterial species Pseudomonas aeruginosa and Burkholderia cepacia are known to reside together in a biofilm in lungs of cystic fibrosis patients reflecting intraspecies, interspecies and interkingdom networking ( 37 , 38 ). Likewise, multispecies biofilms ( 39 ) formed by various bacterial species involved in chronic wounds, dental plaque, etc. exhibit complex networking among different organisms. This QS based group-behavior of microbes is termed as ‘sociomicrobiology’ ( 40 ). SigMol is a comprehensive resource of signaling molecules providing their biological and chemical information. It integrates all the facilities to explore QSSMs for searching signaling molecule of particular bacteria, browsing or comparing capability for specific systems and signaling molecules, structure based search and summary of all the QS systems present till date in the form of heatmaps. Here, all the prokaryotic QSSMs are integrated on one platform that can accelerate the research in field of quorum quenching therapeutics, mechanistics and sociomicrobiology. Researchers can explore the role of signaling molecules to understand complex pattern of communication networking."
} | 2,180 |
23209709 | PMC3510208 | pmc | 947 | {
"abstract": "In plants, transpiration draws the water upward from the roots to the leaves. However, this flow can be blocked by air bubbles in the xylem conduits, which is called xylem embolism. In this research, we present the design of a biomimetic microfluidic pump/valve based on water transpiration and xylem embolism. This micropump/valve is mainly composed of three parts: the first is a silicon sheet with an array of slit-like micropores to mimic the stomata in a plant leaf; the second is a piece of agarose gel to mimic the mesophyll cells in the sub-cavities of a stoma; the third is a micro-heater which is used to mimic the xylem embolism and its self-repairing. The solution in the microchannels of a microfluidic chip can be driven by the biomimetic “leaf” composed of the silicon sheet and the agarose gel. The halting and flowing of the solution is controlled by the micro-heater. Results have shown that a steady flow rate of 1.12 µl/min can be obtained by using this micropump/valve. The time interval between the turning on/off of the micro-heater and the halt (or flow) of the fluid is only 2∼3 s. This micropump/valve can be used as a “plug and play” fluid-driven unit. It has the potential to be used in many application fields.",
"conclusion": "Conclusions A biomimetic micropump/valve based on transpiration and xylem embolism has been demonstrated in this paper. Results have shown that water potential generated by the micropump/valve is 72.5 KPa which can lift the water upward 7 m. The water potential is not generated by single agarose gel. The diffusion of water through the slit-like micropores can obviously increase the potential due to the edge effects of micropore transpiration. The halt and flow of the fluid has been controlled by mimicking the xylem embolism and its self-repairing behavior. The time interval between the turning on/off of the micro-heater and the halt (or flow) of the fluid is only 2∼3 s. The micropump/valve can work well at normal temperature and humidity conditons with a steady flow rate of 1.0∼1.2 µl/min.",
"introduction": "Introduction The micropump/valve is the “beating heart” of a microfluidic system [1] , [2] . The development of a miniaturized, portable, low cost and easy operation micropump/valve is important [3] , [4] . However, present micropumps/valves, have some disadvantages [5] , [6] , such as requiring a continuous connection with external large equipments, expensive fabrication procedure and unsteady flow rate, which results in the difficulty in integrating these micropumps/valves onto a microfluidic device to obtain a true micro total analysis system (μTAS). Transpiration is the loss of water through the slit-like stomata on the leaves, which may generate a water potential gradient in the stem vessels of a plant [7] , [8] . The water potential gradient lifts the water upward from the roots, via the xylem vessels and the mesophyll cells, eventually diffusing into the sub-cavities of the stomata ( Fig. 1a ). Transpiration is a powerful method to drive the fluid [9] , [10] . The water can be lifted up to a height of 100 meters with a steady and adjustable flow rate. Because the driving is a passive process, it costs little metabolizable energy of the plant cells [11] , [12] . Xylem embolism ( Fig. 1b ) is mainly caused by cavitation [13] . It readily occurs at scorching heat and drought conditions in which the tension of water generated by the transpiration becomes great enough to separate the air from the water [14] , [15] . Embolism can completely block the water transport path in a plant. Xylem embolism can be self-repaired as the atmospheric temperature decreases (especially at night). Solutes can be imported into the xylem conduits via the ray cells or via the bordered pits to redissolve the air-bubbles [16] , [17] . 10.1371/journal.pone.0050320.g001 Figure 1 Water transport and xylem embolism in plants and their inspirations in the developing of a biomimetic microfluidic pump. (a) Water transport in plants induced by transpiration through the stomata; (b) xylem embolism induced by cavitation; (c) the structure of the micropump/valve based on water transpiration and xylem embolism. In this paper, a biomimetic microfluidic pump is designed based on water transpiration and xylem embolism. Fig. 1c shows the structure of the micropump/valve. We use a silicon sheet with an array of slit-like micropores to mimic the stomata in a leaf. And the agarose gel is used to mimic the mesophyll cells. The silicon sheet and the agarose gel form an artificial “leaf” to drive the fluid. A micro-heater is placed into the agarose gel. Several button cells are used to give the electric power supply to the micro-heater. As the temperature in the agarose is increased by the micro-heater, the air in the agarose gel will expand to block the water transport path, which mimics the formation process of the xylem embolism at scorching heat condition. As the micro-heater is turned off, the temperature in the agarose gel will decrease to open the water transport path, which mimics the self-repairing process of the xylem embolism at night. Compared with our previous works [18] , [19] , the micropump/valve presented here mimics the xylem embolism in plants to control the fluid flow. Its structure and bimimetic mechanism are both different. The slit-like micropores which are used to mimics the stomata have high reproducibility in size due to the use of UV-LIGA method in their fabrication. The study of edge effect in a slit-like micropore has been studied by recording the change of the fluorescence density.",
"discussion": "Results and Discussion Water Potential Generated by the Micropump/valve This research uses the method reported in our former work to measure the water potential generated by the micropump/valve [19] . A plant water potential meter (TEN-15, Zhejiang TOP Instrument Co., Ltd.) is inserted into a measurement chamber. The chamber is filled with the agarose gel. The water potential generated by the agarose gel will be directly read through the gauge of the water potential meter. It is found that the water potential has reached 72.5 KPa which is large enough to lift water upward to a height of 7 m. We think that the water potential generated by the micropump/valve is composed of two parts. One is the potential generated by the agarose gel. The other is the potential caused by the diffusion of water through the slit-like micropores. The water potential generated by single agarose gel has been measured. The porous ceramic cup is covered with agarose gel. As the water diffuses from the agarose gel into the air, a vacuum generates within the plastic body tube, which equilibrates with the water potential generated by the agarose gel. Results have shown that the water potential generated by single agarose gel is only about 30∼35 KPa. It can be found that this water potential is only half of that generated by the micropump/valve reported here. Halt and Flow of Fluid Controlled by Mimicking Xylem Embolism It has been mentioned above that a micro-heater which is placed into the agarose gel is used to mimic xylem embolism to controll the flow of fluid in a microchannel. As the temperature of the agarose gel is increased by the micro-heater, the air in the gel will expand to block the water transport channels, which results in the halting of the fluid flow. As the micro-heater is turned off, the air in the gel shrinks to open the channels, which has mimic the self-repairing process of xylem embolism in a plant. Fig. 3a to e shows a series of photographs which exhibit the halt and flow of the fluid in a microchannel controlled by using this bio-inspired method. Fig. 3a shows the fluid flow before the micro-heater is turned on. Fig. 3b shows the fluid as the micro-heater is turned on for 2 s. The fluid flow has stopped. Fig. 3c shows the meniscus of the water head as the micro-heater is turned on for 55 s. Fig. 3d shows the fluid flow after the micro-heater is turned off for 2 s. The fluid flow has been restarted. Fig. 3e shows the fluid flow after the micro-heater is turned off for 60 s. 10.1371/journal.pone.0050320.g003 Figure 3 Fluid control and flow rate. (a) to (e) A series of photographs which exhibit the halt and flow of the fluid in a microchannel controlled by mimicking xylem embolism and its self-repairing; (f) the relationship between temperature increase and time; (g) relationship between flow rate and time. To evaluate the performance of this control method, we define closing time and opening time. The closing time is the length of time from the micro-heater being turned on to the fluid flow being stopped. The opening time is the length of time from the micro-heater being turned off to the meniscus of the water head changing obviously which means the transpiration has restarted. It has been found that the closing time and the opening time of this control method are both within 2∼3 s. We have calculated the temperature change of the agarose gel within 2∼3 s. It has been found that the temperature only can be increased or reduced 0.2∼0.3°C within this time. Hence, the response time of this controlling method is very short. As the turning on state of the micro-heater is retained, the temperature of the agarose gel will increase. However, it has been found that the temperature will keep steady after being increased to 1.0∼1.2°C. Fig. 3f shows the temperature-time profile. The data below 0.5°C are obtained by calculation (by using joule law and the second law of thermodynamics). The data above 0.5°C are measured by a temperature sensor (DS18820, DALLAS). We think that the steady state of the temperature exhibit the balance between heating and dispersing. As the temperature of the agarose gel rises 1.0∼1.2°C, the curvature of the meniscus of water head decreases. But backflow of the fluid is not observed. Flow Rate The research uses the method reported in our former work to measure the flow rate [18] . Firstly, a photo mask with many reference lines (the interval between two adjacent lines is 1 mm) is placed near the microchannel. Secondly, the time taken by the water head to travel between two reference lines is recorded to calculate the flow velocity of the water. Finally, the flow rate is calculated by using a equation of , where represents the transpiration volume, represents the cross-section area of a microchannel, represents the flow velocity. An experiment is performed to test the flow rate of this micropump/valve at 25°C and 50% humidity conditions. We record ten flow rates at different positions of the microchannel. Fig. 3g shows the profile. It can be seen that variation of the flow rate is slight. The maximum value of flow rate is 1.17 µl/min. The minimum value is 1.07 µl/min. The average value is about 1.12 µl/min. Effect of Temperature and humidity We use an air-conditioner to change the ambient temperature and humidity to evaluate the effects of temperature and humidity on flow rate. The temperature is increased from 25°C to 30°C. To simplify the study, the ambient humidity is maintained at 50%. At each temperature degree, we measure three flow rates. It is found that the increase of flow rate is only 0.2 µl/min as the temperature rises from 25°C to 28°C. As the temperature increases to 30°C, the increase of flow rate is 0.3∼0.4 µl/min. It can be found that the micropump/valve can work well at normal temperature, even the temperature has a variation of 2∼3°C. As we study the effects of humidity, the ambient temperature is retained at 26°C. The humidity is increased from 50% to 90% to test the effects of humidity on flow rate. Three flow rates are measured at each humidity level. There are not obvious variations in flow rate as the humidity increases from 50% to 70%. However, an obvious decrease in flow rate is observed as the humidity increases above 80%. As the humidity increases to 90%, the flow rate decreases to about 0.81 µl/min. According to the experimental results, the micropump/valve can work well at a normal humidity condition (below 70%)."
} | 3,021 |
35149724 | PMC8837658 | pmc | 948 | {
"abstract": "Social insects, such as ants, use various pheromones as their social signal. In addition, they use the presence of other ants for decision-making. In this study, we attempted to evaluate if individual decision-making is influenced by the complementary use of pheromones and presence of other ants. Ants were induced to form a one-way flow system. We found that when ants entered such a system at a right angle, they tended to move in the opposite direction of the one-way flow system. Interestingly, the target ants moved randomly in the experiments in which no ant and/or no pheromone trails were present. We also developed simulation algorithms and found that artificial ant foragers could reach a certain goal more often if they adopted the reverse run (similar mechanism found in ant experiments) over the forward run (moving in the same direction as their nestmates).",
"introduction": "Introduction Chemical communications of ant foragers are well known as one of the self-organization phenomena. Ant foragers usually deposit pheromones when returning to their nest after finding a food item. Pheromones act as signals and recruit other ants to food resource 1 – 6 . These other ants also deposit pheromones, resulting in the establishment of a pheromone trail 6 , 7 . According to previous research, argentine ant foragers modify their travel direction by reacting to the gradient of pheromone concentrations, suggesting that ants continuously update their travel direction by scanning pheromone concentrations around them 5 . In fact, they may consider other factors in addition to the pheromone trail, such as geometrical information, route memory, and the presence of other ants, which may result in the selection of a path among multiple paths 8 – 11 In fact, ant workers modulate trail following, which may depend on factors such as visual information 12 . In summary, ants do not always obey pheromones blindly. Several studies have reported that the presence of nestmates on the pheromone trail coordinates the actions of ant foragers 8 , 13 , 14 . Black garden ants Lasius niger decrease the rates of pheromone depositions when they encounter their nestmates on the trail and at the food source, suggesting that ant foragers may modulate the overpopulation on a foraging route based on physical contact with the nestmates 13 , 15 . However, it is also reported that foraging ant groups tend to select a pheromone trail occupied by their nestmates 8 . Thus, it is possible for ant foragers to respond differently to the information, which may depend on the presence of ant nestmates. Moreover, physical contact with the nestmates may also determine the foraging action of ant workers 9 , 16 , 17 . Considering the abovementioned information, the movement of the garden ant L. niger , particularly naïve ants, on the trail may be influenced by the encounter with ant nestmates. This is because experienced ants will confidently move forward on a trail owing to their own experiences 18 . Contrarily, naïve ants may have to use additional information to update their position on a trail. A similar mechanism was observed in laden Pheidologeton diversus workers on a foraging trail. They used the flow of other laden workers to navigate themselves toward the nest 17 , suggesting that ant workers tended to follow a stream of ant traffic. Few studies have focused on the orientation cue on a pheromone trail at individual-level and its relation with the presence of other ants although several researches have reported that L. niger workers could modify their actions on a pheromone trail, as already mentioned. Here we tackled this issue and investigated a possibility that contacting other ants oriented L. niger workers on a pheromone trail toward a certain direction. Consequently, we investigated the following question in this study: which mechanism do L. niger naïve workers rely on? To be more precise, we assessed whether they move in the same direction with ant nestmates or in the opposite direction from ant nestmates when coming in contact with them on the trail. To investigate this, we developed a maze apparatus with the following two features: (1) A set-up to establish one-way ant trails. By doing so, we could separate the outbound trips of ant foragers from their inbound trips 19 . (2) Individual naïve ants were allowed to join the ant traffic at right angles to a pheromone trail to easily judge whether they choose both sides (right and left side for those ants) equally if they do not modify their movements when coming in contact with ant nestmates on the trail 20 . Herein, we propose that the garden ant L. niger uses two information sources—pheromones and presence of other ants—to navigate through the established ant trail when entering the trail at a right angle. Using this maze apparatus, we hypothesized that individual ants, which entered the established ant trail at a right angle, followed the same direction of other nestmates using two complementary information sources—pheromones and presence of other ants. Contrary to our expectation, however we found that ants tended to move in the opposite direction of the one-way flow system. We also developed a multiagent-based model to evaluate the mechanistic understanding of the action of individual ants.",
"discussion": "Discussion In this study, we used an apparatus with two paths to separate outward flows from inward flows. Target ants that had been isolated from the nest in advance were allowed to join the established ant trails. The results of these experiments suggested that the direction of movement of the target ants may be determined by the presence of other ants when they reach the trails. Considering the results of this study, we concluded that the target ant may make decisions and move in the opposite direction in ant trails in the presence of both pheromones as well as other ants. In previous studies, complementary use of several information seemed to determine the actions of L. niger foragers 8 , 12 , 13 , 25 . Thus, we confirm that pheromones and other ants served as information for the target ants in determining the direction to move forward. We also propose that the direct contact immediately after entering the main bridge is not important for the target ants, suggesting which position the direct contact occurs is not important for target ants. When returning to the nest, foraging ants deposit pheromones on their inward path. On the other hand, experienced ants deposit foraging pheromones on both the inward and outward paths 21 . Therefore, it is possible for the target ants to deduce that there is a feeding site by moving in the opposite direction of ant flows and update their direction. Further investigations such as forcing target ants to join a trail only of inexperienced ants will clarify these possibilities. Our findings were contrary to our expectations as we expected that the target ants would move in the same direction as their nestmates 17 . In some collective behaviors of animals, the alignment rule can be a key point in explaining the emergent property of those living systems 26 , 27 . For example, birds get closer to each other and progressively align their headings. Notably, we found that target ants chose the goal side randomly when we used a single path in the main apparatus. Then, why do the target ants move in the opposite direction from the direction of other ants rather than moving in the same direction as them? This could perhaps be related to the continuous updating of their actions 5 , 28 . We think that contact with other foraging ants may provide reliable directional cues to the target ants considering that pheromones do not offer them any directional information when they join a trail. By doing so, the target ants will not get lost and trace a pheromone trail more effectively, thereby helping them reach the desired place. Our findings do not suggest that ants conduct information transfer regarding food positions when contacting each other 16 . Rather, our findings imply that the target ants are allowed to reinforce forward movement when contacting their nestmates. In fact, we also developed a multiagent model and found that the target agents reached a goal more often when moving against an artificial ant flow than when moving along an artificial ant flow. These findings suggest that naïve ants may adopt adaptive decision-making when they encounter an ant on a pheromone trail as they do not know in which direction the destination, such as food location or their nest, is located 8 , 13 , 14 . Although desynchronizing the flow of outward and inward ants has been observed in laboratory experiments on the garden ant L. niger 29 , in natural conditions, the outward path is not separated from the inward path of ant trails. Having said that, our findings indicate an interesting notion—ants may maintain their travel direction along a pheromone trail as long as they continuously encounter nestmates, thereby leading them to a certain destination. This mechanism will help a naïve forager to reliably reach a certain location when it cannot detect any pheromone gradients. At a glance, physical contact between foraging ants is likely to interfere with the forward motion of ant traffic. Considering our results however, it will have an adverse effect on ant flow. This flow property of ant traffic implies that masses of foraging ants may be beyond an aggregation of solid particles. Notably, masses of some ant species do not always behave like a solid, which is dependent on the situation 30 , 31 . Individual ants can behave adaptively according to the environment condition and modify their actions via an interaction with nestmates. Thus, individuals of living particles influence the macro-pattern of the ensemble 32 , 33 . To this end, our findings imply and predict a behavioral significance for foraging ants: they may determine the direction to move forward by continuously encountering nestmates on a pheromone trail after leaving the nest. This implication will be particularly true for naïve ants and will be linked to a self-organizing process that contribute to rapid flow and efficient transportation on ant traffic. Further experiments are warranted to evaluate whether it is possible for naïve ants to present reverse runs when they join a pheromone trail not at a right angle, join a pheromone trail not alone but with ant nestmates, or join a pheromone trail not on the one-way flow system."
} | 2,622 |
32973923 | PMC7507662 | pmc | 949 | {
"abstract": "Background Extracellular electron transfer (EET) is essential in improving the power generation performance of electrochemically active bacteria (EAB) in microbial fuel cells (MFCs). Currently, the EET mechanisms of dissimilatory metal-reducing (DMR) model bacteria Shewanella oneidensis and Geobacter sulfurreducens have been thoroughly studied. Klebsiella has also been proved to be an EAB capable of EET, but the EET mechanism has not been perfected. This study investigated the effects of biofilm transfer and electron mediators transfer on Klebsiella quasipneumoniae sp. 203 electricity generation performance in MFCs. Results Herein, we covered the anode of MFC with a layer of microfiltration membrane to block the effect of the biofilm mechanism, and then explore the EET of the electron mediator mechanism of K. quasipneumoniae sp. 203 and electricity generation performance. In the absence of short-range electron transfer, we found that K. quasipneumoniae sp. 203 can still produce a certain power generation performance, and coated-MFC reached 40.26 mW/m 2 at a current density of 770.9 mA/m 2, whereas the uncoated-MFC reached 90.69 mW/m 2 at a current density of 1224.49 mA/m 2 . The difference in the electricity generation performance between coated-MFC and uncoated-MFC was probably due to the microfiltration membrane covered in anode, which inhibited the growth of EAB on the anode. Therefore, we speculated that K. quasipneumoniae sp. 203 can also perform EET through the biofilm mechanism. The protein content, the integrity of biofilm and the biofilm activity all proved that the difference in the electricity generation performance between coated-MFC and uncoated-MFC was due to the extremely little biomass of the anode biofilm. To further verify the effect of electron mediators on electricity generation performance of MFCs, 10 µM 2,6-DTBBQ, 2,6-DTBHQ and DHNA were added to coated-MFC and uncoated-MFC. Combining the time–voltage curve and CV curve, we found that 2,6-DTBBQ and 2,6-DTBHQ had high electrocatalytic activity toward the redox reaction of K. quasipneumoniae sp. 203-inoculated MFCs. It was also speculated that K. quasipneumoniae sp. 203 produced 2,6-DTBHQ and 2,6-DTBBQ. Conclusions To the best of our knowledge, the three modes of EET did not exist separately. K. quasipneumoniae sp.203 will adopt the corresponding electron transfer mode or multiple ways to realize EET according to the living environment to improve electricity generation performance.",
"conclusion": "Conclusions In this study, the coated-MFC anode was covered with a microfiltration membrane to investigate whether K. quasipneumoniae sp. 203 can conduct EET through the electron mediator mechanism. In the absence of short-range electron transfer, we found that K. quasipneumoniae sp. 203 can still produce certain electricity generation efficiency and redox activity. The protein content, the integrity of biofilm and the biofilm activity all proved that the existence of microfiltration membrane prevented EAB from attaching and growing on the anode, resulting in the difference in the electricity generation performance between coated-MFC and uncoated-MFC. Finally, to further prove the effect of electron mediators on the performance of electricity generation, combining the time–voltage curve and CV curve of MFCs after adding electron mediators, we found that 2,6-DTBBQ and 2,6-DTBHQ had high electrocatalytic activity toward the redox reaction of K. quasipneumoniae sp. 203-inoculated MFCs. Our work demonstrated that K. quasipneumoniae sp. 203 can be coupled to realize EET in a variety of ways in MFCs. Through understanding the EET mechanism of K. quasipneumoniae sp. 203 in MFCs, it provides a theoretical basis for improving its power generation performance.",
"discussion": "Results and discussion Preliminary verification of the electricity generation performance in coated-MFC and uncoated-MFC Figure 1 a shows that 5 cycles of K. quasipneumoniae sp. 203-inoculated MFCs were observed, a total of 5 cycles, each cycle lasted approx 120 h. Due to the consumption of the substrate sodium citrate in the anolyte, the output voltage will start to decline, and it is necessary to replace the anolyte. The average maximum output voltage in coated-MFC and uncoated-MFC were detected with 621 mV and 327 mV, respectively. Although the average maximum output voltage of coated-MFC was 300 mV lower than that of uncoated-MFC, it can be seen from Fig. 1 b that coated-MFC still showed a certain electrochemical performance, coated-MFC reached 40.26 mW/m 2 at a current density of 770.9 mA/m 2 and the uncoated-MFC reached 90.69 mW/m 2 at a current density of 1224.49 mA/m 2 . The difference in the electricity generation performance between coated-MFC and uncoated-MFC was probably due to the microfiltration membrane covered in anode, which inhibited the growth of EAB on the anode. Therefore, we speculated that K. quasipneumoniae sp. 203 can also perform EET through the biofilm mechanism. C. Yuvraj et al. also indicated that K. quasipneumoniae can directly transfer electrons to the anode without any external mediator, and the increase in electrochemical performance is directly proportional to the electroactive biofilm formed on the electrode surface [ 18 ]. Shewanella and Geobacter are considered model exogenous electrons, and are known to be able to perform direct extracellular electron transitions through outer membrane redox proteins [ 6 ]. According to previous studies, the common feature of EAB with this EET ability is the presence of many polyheme c-type cytochromes (MH-cytC) in their genome [ 19 ]. Fig. 1 a MFC-coated and MFC-uncoated were incubated for 600 h of output voltage, and the anode medium was replaced when the output voltage droped. b Polarization curve and power density curve when the output voltage of the reaches its maximum value. c Nyquist plots from electrochemical impedance spectroscopy measurements of MFC-coated (black empty circle) and MFC-uncoated (red empty circle) scanned at 0.1 ~ 100 kHz at open-circuit potential (the inset on the right is the charge transfer resistance in the high frequency region of MFC-uncoated). d Cyclic voltammograms curve of coated-MFC and uncoated-MFC Figure 1 c presents that the internal resistance of the MFCs, the semicircles in the high frequency region and the straight lines in the low frequency region represent the charge transfer resistance (Rct) and diffusion resistance (W 1 ) of the MFCs, respectively. The R ct and W 1 of coated-MFC were 78.20 Ω and 55.10 Ω, the R ct and W 1 of uncoated-MFC were 6.84 Ω and 30.20 Ω, the R ct and W 1 of coated-MFC were both higher than uncoated-MFC. After the operation of MFCs, the anode biofilm continued to grew and generate electrons until a complete biofilm was formed on the anode surface, and the electricity generation performance of the system reached a stable state. For EAB, the phospholipid bilayer structure of the cell membrane acts as a capacitor, and the electron shuttle (or redox mediator) generated endogenously on the cell membrane acts as an electrochemical active site. The surface of the coated-MFC anode was covered with a microfiltration membrane. It was difficult for microbial cells to adhere to the smooth surface of the microfiltration membrane, and the electrons generated by it cannot be transmitted to the anode surface over a long distance without electron mediators [ 7 ]. It is generally believed that, within a certain period of time, the thickness of the anode biofilm and the content of the redox mediator are negatively related to the internal resistance [ 20 ]. MFCs performance metrics were summarized in Table 1 and the data indicated that the output power of MFCs is inversely proportional to the internal resistance. Table 1 Comparison of electricity generation performance and internal resistance in coated-MFC and uncoated-MFC Power density (mW/m 2 ) Current density (mA/m 2 ) Internal resistance (Ω) Rohm Rct W 1 CPE1 CPE2 Coated-MFC 40.26 770.97 23.54 78.20 55.10 1.76 2.74 Uncoated-MFC 90.69 1224.49 7.41 6.84 30.20 7.20 8.61 Figure 1 d shows the electron transfer mechanism and catalytic efficiency during the stable stage of 1st and 3rd operation of the MFCs. The CV curve revealed that no significant redox peaks were observed in 1st cycle of coated-MFC and uncoated-MFC. After 3rd cycle of operation, the existence of a reversible redox process in both MFCs, but the peak of coated-MFC was significantly lower than that of uncoated-MFC, and more than one pair of redox peaks of uncoated-MFC can be observed. In 3rd cycle of coated-MFC, a redox peak that was not detected in the 1st operation was observed. Therefore, it can be inferred that the electrochemical activity of coated-MFC due to the self-excreted electron mediators lead to the mechanism of electron transfer from bacterium to the anode. A similar result was also obtained from the analysis by Deng et al. [ 8 ]. Interestingly, uncoated-MFC showed more intense redox activity than coated-MFC, indicating that uncoated-MFC can also transfer electrons through other EET mechanisms in addition to generating electron mediators. Islam et al. also observed a similar redox peak from the MFCs inoculated with Klebsiella variicola [ 7 ]. Consequently, the appearance of more intense redox peaks indicated that the mature and effective biofilm was formed on the anode surface, which shortened the diffusion distance of extracellular electron transfer between EAB and the anode. Effect of microfiltration membrane on the growth of cells We were interested in whether the microfiltration membrane affects cell growth, so we measured the biomass of the anolyte and anode. The biomass of the microbial population can be indirectly calculated by measuring the protein content [ 21 ]. During the electricity generation process of MFCs, the amounts of EAB in anode biofilm and anolyte suspension in coated-MFC and uncoated-MFC as assessed by the protein contents were compared (Fig. 2 ). We found that the protein content is basically the same in coated-MFC and uncoated-MFC, and the presence of microfiltration membrane has little effect on the anolyte suspension biomass (Fig. 2 a). Moreover, we found that the biomass of uncoated-MFC anode biofilm was much higher than that of the coated-MFC, and the average protein content by more than 5 times (Fig. 2 b). The existence of the microfiltration membrane only affected the biomass of the anode biofilm, and had little effect on the biomass in the anolyte. Therefore, it can be considered that the difference in the electricity generation performance between coated-MFC and uncoated-MFC was due to the extremely little biomass of the anode biofilm. Fig. 2 Protein contents of coated-MFC and uncoated-MFC. a Anolyte suspension; b Anode biofilm Effect of anode biofilm on electricity generation performance of MFCs Figure 3 reveals that the formation of the biofilm on anode electrode surface by SEM. Results clear showed that almost no EAB were attached to the carbon paper surface in coated-MFC (Fig. 3 a–c). This further verified the existing results, the extremely low protein content of coated-MFC anode biofilm (Fig. 2 b). However, the adsorption capacity of rod-shaped EAB at the anode increased with time in uncoated-MFC (Fig. 3 d–f). It can be seen that an incomplete biofilm is formed on the electrode surface due to the adsorption of EAB in the 1st operation of uncoated-MFC (Fig. 3 d). It was not until the third operation of uncoated-MFC that we observed that the microorganisms attached to the surface of the carbon paper began to secrete an enveloping matrix consisting mainly of polysaccharides and proteins, known as EPS. According to the research by Wu et al., EPS can accumulate group-sensing effect signaling molecules, extracellular enzymes, and bacterial secondary metabolites, providing a place for microbial to exchange information [ 22 ]. Kim et al. tested the impedance of the biofilm formation process in the early adhesion stage of P. aeruginosa PAO1 and found that early adhesion of microorganisms on the anode would lead to a reduction in electrical resistance [ 23 ]. In combination with the above electrochemical performance results, we considered that microorganisms clusters form complete biofilms with metabolic activity, thereby exhibiting higher power generation performance (Figs. 1 a, 3 e). However, in the 5th cycle, several layers of biofilm were adsorbed on the anode of uncoated-MFC which decreased the electrochemical performance (Fig. 3 f). This result was consistent with the result of protein content of coated-MFC anode biofilm (Fig. 2 b). In the 3rd cycle of uncoated-MFC, the output voltage (621 mV) and the protein content anode biofilm (2.00 g prot /L) reached their peak values at the same time (Fig. 1 a). And then the protein content decreased with the output voltage - time. This may be related to the EAB proliferation rate of anode biofilm. Fig. 3 SEM images of anode biofilm of coated-MFC and uncoated-MFC, a coated-MFC: 1st cycle, b coated-MFC: 3rd cycle, c coated-MFC: 5th cycle d uncoated-MFC: 1st cycle, e uncoated-MFC: 3rd cycle, f uncoated-MFC: 5th cycle In addition, the changes in biofilm viability were examined over time using fluorescent staining to distinguish live versus dead cells (Fig. 4 and Additional file 1 : Fig. S1). As previously described, the presence of microfiltration membrane maked it almost impossible to observe the presence of living and dead cells (Additional file 1 : Fig. S1a–c). The fluorescence intensity of dead cells was much less than that of living cells (Fig. 4 ). On the contrary, as the biofilm grew, cells existed in two states (live and dead) in uncoated-MFC (Additional file 1 : Fig. S1d–f). When the maximum output voltage reached 621 mV in the 3rd cycle, we observed that the entire biofilm of K. quasipneumoniae sp. 203 was alive, with very few dead cells (Fig. 1 a and Fig. S1e). As the biofilm grew, the fluorescence intensity of red dead cells increased from 30.49 to 64.95, the increased fluorescence intensity of dead cells can help explain that the output voltage gradually decreases after reaching the peak (Fig. 4 , Additional file 1 : Fig. S1e, f). Although we did not analyze the location and spatial structure of live and dead cells in biofilm, previous studies suggested that living cells can only exist in the outer layer of thick biofilm, which may be due to the availability of anode electrolyte substrates [ 24 , 25 ]. Furthermore, the accumulation of dead cells (more than active cells) at the bottom of the biofilm over time will not exert redox activity. Thus this increased the intrinsic resistance of MFCs. Fig. 4 Fluorescence intensity of live and dead cells of coated-MFC and uncoated-MFC Effect of electron mediators on electricity generation performance of MFCs The model strains for EET mechanism were Geobacter sulfurreducens and Shewanella oneidensis . G. sulfurreducens can also synthesize a small amount of flavin compound, but it can only binds to the outer membrane protein and cannot be released as an electron mediator, like S. oneidensis . Therefore, it is generally believed that G. sulfurreducens undergoes short-range electron transfer by direct contact with extracellular electron acceptors [ 5 ]. Combined with this study, it was found that when a new anode medium was replaced at the end of each cycle, the output voltage decreased significantly and rose slowly, even over 24 h, which was in sharp contrast to G. sulfurreducens [ 26 ]. Therefore, compared with the CV curve of coated-MFC, we found that a pair of redox peak that appeared in the range of − 0.4 ~ 0 V in the CV curve of the supernatant did not appear in the MFCs (Figs. 1 d, 5 a). According to the report by Freguia et al., it may be due to EAB actively removing mediators in an oxidized state [ 27 ]. To better detect the electron mediators, we conduct CV detection on the anode supernatant. As shown in Fig. 5 a, we found that more than one pair of redox peaks appeared in the anode supernatant. By comparing the standard library and consulting related references, it is inferred that the possible quinone electron mediators secreted by K. quasipneumoniae sp.203 were 2,6-di-tert-butyl-p-benzoquinone (2,6-DTBBQ), 2,6-Di-tert-butylphenol (2,6-DTBHQ), 1,4-dihydroxy-2-naphthoic acid (DHNA) and 2-amino-3carboxy-1,4-naphthoquinone(ACNQ) (Fig. 5 b, Additional file 1 : Tables S1 and S2) [ 14 , 28 , 29 ]. Fig. 5 a CV of the supernatant of the anode on the 1–3 cycle, b Quinones electron mediators secreted by K. quasipneumoniae sp. 20 Herein, combined with the existing research ideas of this experiment, MFCs constructed by K. quasipneumoniae sp. 203 was used as the research object, upon the addition of 10 µM 2,6-DTBBQ, 2,6-DTBHQ and DHNA to the reactor, respectively, and the results showed that the voltage of coated-MFC and uncoated-MFC increased immediately, except for the addition DHNA of MFCs (Fig. 6 a). Although the voltage of the coated-MFC was lower than that of the uncoated-MFC, the voltage reached the highest output voltage within 40 h. The sharp rise in the voltage proved that the 2,6-DTBBQ and 2,6-DTBHQ can function as electron mediators in MFCs, thus facilitating the electron transfer from EAB to electrode. This speculation was similar to the results of Deng et al. [ 30 ]. HPLC-MS detected a low content of ACNQ, so we did not add ACNQ to the reactors. Fig. 6 Effect of three quinone electron mediators secreted by K. quasipneumoniae sp. 203 on electricity generation performance of MFCs. a The output voltage–time; b Cyclic voltammogram curve We also performed CV to examine the redox state of MFCs after the addition of electron mediators (Fig. 6 b). The redox peaks appeared in coated-MFC and uncoated-MFC after adding 2,6-DTBBQ and 2,6-DTBHQ. Notably, the pair of redox peak at 0 ~ 0.8 V were attributed to 2,6-DTBBQ, and the pair of redox peak at − 0.4 ~ 0 V were assigned to 2,6-DTBHQ. This result almost corresponded to the CV curve of the supernatant (Fig. 5 a). In contrast, no redox peaks were found in the addition DHNA of MFCs. Combined with the time–voltage curve, we also found that after adding DHNA, the voltage of the MFC did not show a significant upward trend. Therefore, we speculate that DHNA was not a redox metabolite of K. quasipneumoniae sp. 203. Combining the time–voltage curve and CV curve, we found that 2,6-DTBBQ and 2,6-DTBHQ had high electrocatalytic activity toward the redox reaction of K. quasipneumoniae sp. 203-inoculated MFCs. Moreover, the oxidation and reduction peaks at the range of − 0.4 ~ 0 V correspond to 2,6-DTBBQ, 2,6-DTBHQ, this result was similar to that of Zeng. et al. in detecting the electrocatalytic activity of Klebsiella pneumoniae on 2,6-DTBBQ [ 31 ]. This contributes to the electron transfer between the EAB and the MFCs anode and also contributes to the power density of the MFCs. According to previous studies, the electron mediators-mediated extracellular electron-transport mechanism is a circulation mechanism. The oxidized mediators are converted into the reduced mediators after being coupled intracellularly with the reduction products on the respiratory chain [ 9 ]. Then reduced mediators are discharged out of the extracellular, and the electrons are transferred to the electrode to be oxidized. It is speculated that the electron mediator secreted by K. quasipneumoniae sp. 203 can be reused. In the stable growth stage of anode EAB in MFCs, the addition of electron mediators can effectively improve the electricity generation performance. Quinones can be used as electron mediators, mainly because of their quinone group with electron transfer function. Based on the existing research of this experiment, it is considered that the transfer mechanism of the electron mediators in coated-MFC and uncoated-MFC is: the quinone compounds secreted by EAB were obtained electrons and reduced into hydroquinones, and then electrons were transferred to the anode, and hydroquinones were oxidized to benzoquinones. The main role of quinones as electron mediators is due to their quinone group with electron transfer function. The group is circulated intracellularly in three states: oxidized, semiquinone radical, reduced or hydroquinone. The transition of the electron mediator from the oxidized state to the reduced state is accomplished by quinone reductases under the action of flavin adenine dinucleotide (FAD) [ 28 ]. And the studies by Ramos et al. has shown that the redox mediators secreted by EAB can also promote the formation of biofilms [ 32 ]."
} | 5,193 |
38914624 | PMC11196684 | pmc | 950 | {
"abstract": "The application of beneficial microorganisms for corals (BMC) decreases the bleaching susceptibility and mortality rate of corals. BMC selection is typically performed via molecular and biochemical assays, followed by genomic screening for BMC traits. Herein, we present a comprehensive in silico framework to explore a set of six putative BMC strains. We extracted high-quality DNA from coral samples collected from the Red Sea and performed PacBio sequencing. We identified BMC traits and mechanisms associated with each strain as well as proposed new traits and mechanisms, such as chemotaxis and the presence of phages and bioactive secondary metabolites. The presence of prophages in two of the six studied BMC strains suggests their possible distribution within beneficial bacteria. We also detected various secondary metabolites, such as terpenes, ectoines, lanthipeptides, and lasso peptides. These metabolites possess antimicrobial, antifungal, antiviral, anti-inflammatory, and antioxidant activities and play key roles in coral health by reducing the effects of heat stress, high salinity, reactive oxygen species, and radiation. Corals are currently facing unprecedented challenges, and our revised framework can help select more efficient BMC for use in studies on coral microbiome rehabilitation, coral resilience, and coral restoration.",
"introduction": "Introduction Coral reefs are one of the most diverse and productive ecosystems on Earth. However, they are currently facing unprecedented challenges due to overfishing, pollution, and climate change 1 – 4 . Increased greenhouse gas concentrations leading to global warming 5 , pollution and ocean acidification have emerged as key stressors, threatening the survival of coral reefs worldwide 3 , 6 – 8 . Some of these stressors cause periods of high seawater temperatures, leading to the loss of Symbiodiniaceae, the algal symbionts that reside within the holobiont and fulfill up to 90% of the corals’ nutritional requirements 9 except when they become nutrient competitors and can be expelled 10 – 12 . This loss, especially over a long-term, can lead to devastating bleaching events worldwide, in which corals are deprived of vital nutrients, leading to increased disease susceptibility and reduced reproductive success and skeletal growth 2 , 13 – 16 . This disruption extends beyond the loss of algal symbionts because it is followed by shifts in the composition and function of the entire coral-associated microbiome 17 . This causes post-heat stress disorder 18 that compromises the coral’s ability to grow, resist pathogens, modulate nutrient cycling, and maintain holobiont homeostasis 18 – 27 . The symbiotic relationship between corals and their bacterial symbionts 28 , 29 plays a pivotal role in maintaining the health of these organisms and consequently the functioning of reef ecosystems. Although some corals and their microbiomes have shown the capacity to recover from acute stress events 30 , 31 , the coral microbiome may not fully restore its original composition and functionality following a severe disturbance 32 . Rehabilitation of the coral microbiome 33 has recently emerged as a promising strategy to enhance coral health and resilience 18 , 24 , 26 , 33 – 42 . Coral-associated bacteria possess potential beneficial traits that enhance the fitness and resilience of their coral host by maintaining homeostasis through their involvement in various essential processes, such as nutrient cycling, production of antibiotics and antimicrobial compounds, and mitigation of toxic compounds 33 , 43 . Harnessing the symbiotic interactions between corals and their associated bacteria 28 allows researchers to selectively introduce these beneficial microorganisms for corals (BMC) 33 into the coral holobiont for use as customized probiotics. This can help maintain or even enrich the proportion of native beneficial microorganisms to enhance coral health and resilience under stress 33 , 35 . In this study, we performed an in silico investigation of the probiotic potential of six putative BMC (pBMC) strains isolated from corals in the Red Sea with the aim of developing the first Red Sea BMC consortium. This was achieved by analyzing the complete genomes of all six strains and screening for the presence of potential beneficial mechanisms for corals. Our research provides a framework for the selection of novel, customized BMC consortia based on the presence of particular BMC characteristics that support host health and survival under stress conditions.",
"discussion": "Discussion Halomonas sp. pBMC5 and Sutcliffiella sp. pBMC6 as candidate novel species Our results, particularly those based on the ANI and DDH values, suggest that both pBMC5 and pBMC6 are novel species, and further characterization is planned. The systematic classification of the family Bacillaceae has undergone numerous modifications in recent years due to the implementation of new taxonomic polyphasic techniques 80 , 81 . This has resulted in the creation of new genera from the previously classified Bacillus genus, such as the new genus Sutcliffiella 81 , and justifies the presence of B. horikoshii strains in the phylogenomic tree of pBMC6—likely an older classification of the current S. horikoshii —and the dominance of Bacillus sp. genomes in this tree. This is further corroborated by the presence of two putative prophages in pBMC6 that were predicted to infect S. horikoshii hosts. Genome screening reveals previously proposed beneficial traits of pBMC Following the classification of each pBMC for identification and phylogenomic analysis, we screened the pBMC genomes for genes encoding proteins that are potentially beneficial for corals. We screened for genes related to catalase, urease, and siderophore production; phosphate assimilation; and nitrogen cycle and DMSP degradation through biochemical tests and PCR assays 44 , which are typically employed for BMC selection. We also detected genes involved in other potential beneficial traits (Table S3 ), including those related to oxidative stress, such as superoxide dismutases (all pBMC genomes), which exert an antioxidant effect by catalyzing the dismutation of superoxide (an ROS molecule that causes cell damage) 84 ; catalase KatE (all pBMC genomes) and catalase-peroxidase KatG (pBMC1, pBMC2, pBMC5, and pBMC6 genomes), both of which protect cells from the toxic effects of H 2 O 2 and aerated growth conditions 85 – 89 ; manganese catalase (pBMC5 and pBMC6 genomes), which is also involved in the protection of cells from H 2 O 2 90 , 91 ; and glutathione synthetase (all pBMC genomes except for pBMC6), which produces glutathione that can subsequently be used by glutathione peroxidase (all pBMC genomes except for pBMC3 and pBMC4) to scavenge ROS, such as H 2 O 2 92 , 93 . When seawater temperatures rise, the coral holobiont produce ROS, resulting in cell damage in both host and its symbionts 94 – 96 . A direct correlation between bleaching and ROS production has been previously reported 3 , and ROS-scavenging pBMC strains were hypothesized to mitigate coral bleaching 33 , making this a crucial trait when selecting pBMC. Several genes involved in vitamin B-complex biosynthesis were found in our pBMC genomes, such as riboflavin synthase (all pBMC), pyridoxine 5’-phosphate synthase (all pBMC except for pBMC6), biotin synthase (all pBMC), dihydrofolate synthase and thymidylate synthase (all pBMC), and cobalamin synthase (all pBMC except for pBMC3 and pBMC4) for the biosynthesis of vitamins B2, B6, B7, B9, and B12, respectively. Vitamin B2 is necessary for glutathione reductase activity, which is involved in stress reduction by increasing antioxidant potential, and B2 deficiency increases lipid peroxidation 97 . Vitamin B6 catalyzes approximately 2% of all prokaryotic functions 98 , but it has not been widely studied in the marine environment 99 , 100 . It acts as an antioxidant during light exposure and against oxidative stress 101 , 102 . Vitamin B7 is a cofactor in several metabolic pathways, such as fatty acid biosynthesis, amino acid metabolism, and gluconeogenesis 103 . Vitamin B12 is involved in several metabolic pathways, including the production of the antioxidants glutathione and DMSP 104 , which are important for neutralizing high concentrations of ROS generated from heat stress events 33 , 43 . Bacteria that exist in association with corals possess genes encoding for proteins related to the biosynthesis of essential vitamins, such as B1, B2, and B7, whereas their coral host does not have the capacity to produce them 105 . This suggests that the coral holobiont can only take up these essential vitamins through heterotrophic feeding and/or from its bacterial symbionts. Furthermore, coral symbionts from the family Symbiodiniaceae are auxotrophs for various B-complex vitamins, which they acquire from exogenous sources such as bacteria 83 , 104 , 106 – 108 . This highlights the important role of bacterial symbionts in ensuring coral health. For these reasons and because of the close interaction between several B-complex vitamins, the presence of genes encoding proteins related to B-complex vitamin biosynthesis is suggested as a BMC trait. We also screened for other genes related to metabolism. Siderophore synthase was present in pBMC6. This enzyme produces siderophores that can scavenge iron from the environment, a trait that is beneficial for corals 33 , 82 and other organisms 109 . In general, the bioavailability of iron in oceans is extremely low; consequently, the growth and survival of organisms that use iron for essential physiological processes, such as photosynthesis and nitrogen fixation, cannot be guaranteed 110 , 111 . Thus, bacteria that exist in association with other marine organisms, such as corals and microalgae, produce siderophores to capture and concentrate iron into a bioavailable form that can be used by other organisms 112 , 113 Apart from siderosphere production, we also found that some of our pBMC produced ectoine (pBMC3, pBMC4, and pBMC5) and betaine (pBMC1, pBMC2, pBMC3, pBMC4, and pBMC5), which have been previously described in BMC genomes and proposed as compounds that plays a role in beneficial mechanisms in corals 82 . Ectoines and betaines are important for osmoregulation and act as protective agents under thermal stress and high irradiance 114 – 116 . They also contribute to the nitrogen biomass of corals in reefs 117 . In marine microalgae, the ectoine content was found to increase in the presence of bacteria, highlighting the crucial role of these microorganisms in host health 118 . Betaine and ectoine production significantly improves environmental stress tolerance, including pH stress 119 and heat stress 120 in aquatic organisms 121 , such as corals and their symbionts 116 . Ectoines can help mitigate the harmful effects of heat stress, high salinity, ROS, and radiation 122 . Pei et al. identified betaine lipids as leading metabolite drivers for differentiating heat-bleached corals from healthy ones, revealing new tools to screen for heat-resistant corals and their symbionts, such as BMC 116 , 123 . We also found genes involved in the nitrogen cycle, including nitrate reductase (pBMC3, pBMC4, and pBMC5) and cyanate hydratase (pBMC5). The presence of these genes was previously proposed as a BMC trait 33 , 43 because balancing this nutrient’s availability contributes to maintaining desirable levels of bioavailable nitrogen, limiting algal growth and leading to an accumulation of photosynthates in algal cells that, when released, feed the coral host and promote its growth. Additionally, increased coral catabolic activity due to an environmental stressor leads to host starvation and increased nitrogen availability to Symbiodiniaceae members of the holobiont, potentially causing destabilization of the host’s nutrient cycle and of the Symbiodiniaceae–coral interaction 10 , 124 . Screening for genes related to DMSP degradation and sulfur metabolism revealed the presence of DMSP CoA transferase/lyase DddD (pBMC3 and pBMC4) and acryloyl-CoA reductase AcuI/YhdH (BMC1 and pBMC2). DMSP is found in several marine organisms, including Symbiodiniaceae 107 , 125 , and is a ROS scavenger in marine algae 126 , an attribute that has been proposed as a BMC trait due to its antioxidant activity 33 . Mechanisms of DMSP breakdown have also been hypothesized as BMC traits because a high DMSP concentration can lead to dysbiosis and signal the location of more vulnerable coral to pathogens through chemotaxis 18 , 33 , 127 . Discovery of the presence of BMC-associated prophages Prophages are DNA from bacteriophages (or bacteria-infecting viruses) that are integrated into the genomes of their bacterial host upon infection. They are mainly studied as contributors to the virulence of pathogenic reef bacteria, such as V. coralliilyticus , because they encode virulence factors 128 . The detection of putative prophages in two of our six pBMC strains revealed the potentially presence of prophages across marine bacteria taxa, and also implies a positive role of prophages associated with beneficial bacteria. As prophages also protect the bacterial host against virulent phage infections via superinfection exclusion 129 , pBMC containing prophages may have a competitive advantage against other bacteria, including pathogens. Thus, prophages might both expand the metabolic capabilities and protection of beneficial bacteria, and further research is warranted. Pangenomes exhibit novel pBMC beneficial traits and other applicable functions We conducted pangenomic analyses that included our pBMC strains and strains closely related to them in the phylogenomic trees to screen for genes unique to our pBMC strains. In the Pseudoalteromonas pangenome, 110 genes with known functions were only present in the pBMC genomes (i.e., they were present in one or both of our P. galatheae pBMC strains). This included a protein involved in chemotaxis and some proteins involved in iron/sulfur metabolism. Chemotaxis may be an important BMC trait because it was previously suggested to play a crucial role in defining patterns of microbial diversity, coral metabolism, coral infection dynamics, and chemical cycling processes, thereby influencing coral holobiont health 130 . Despite the absence of DMSP-related genes in the genomes of the Pseudoalteromonas sp. pBMC strains compared with the other Pseudoalteromonas sp. strains, the metabolism and production of sulfur compounds has been proposed as a BMC trait because they inhibit the growth of coral pathogens and also play a role in the structure of bacterial communities of the coral holobiont 33 . In the Cobetia pangenome, we found five genes with known functions that were only present in the pBMC genomes, but none of them showed a clear benefit to the host. In the Halomonas pangenome, we found 204 genes with known functions that were only present in the pBMC5 genome, inclunding genes involved in carotenoid biosynthesis, antibiotic biosynthesis, and sulfur metabolism. Although not directly related to host health but more to the environment as a whole, the presence of mercury resistance-related genes in the pBMC5 genome is of crucial interest. Heavy metals are known to be environmental toxins due to their bioaccumulation in the food chain, becoming increasingly hazardous for the higher trophic levels 131 , 132 The use of bioremediation for the removal of toxic metals has been studied, but only a few studies were performed in a marine environment 133 . The trend is similar for mercury-resistant marine bacteria, and the use of mercury-resistant marine bacteria for bioremediation of mercury contamination has received little attention 134 . However, their use is associated with certain advantages, such as simple process, lower amount of secondary metabolites, and lower cost than commonly used chemical technologies as well as better adaptability and higher resistance to adverse environmental conditions compared with terrestrial bacteria 134 , 135 . Some studies have reported that mercury-resistant marine bacteria show a higher capability for mercury bioremediation and can reduce the toxic effects of mercury in contaminated environments 134 , 136 . We found that the Sutcliffiella genome had the highest amount of singleton genes of all the pangenomes generated in this study, with 1333 genes with known functions present only in the pBMC6 genome, including genes encoding proteins related to the biosynthesis of siderophores, carotenoids, antibiotics, vitamins B1, B2 and B12; nitrogen, iron, and sulfur metabolism; chemotaxis; and oxidative stress resistance. Other than the abovementioned genes, we identified a gene that was involved in the detoxification of reactive aldehydes, which are highly reactive organic chemical compounds that mostly arise due to oxidative stress 137 . We also identified proteins involved in the degradation of aromatic compounds, such as phenols, cresols, catechols, and diphenyl phosphate (DPHP). Phenols and cresols are harmful to the environment 138 , and catechol bioaccumulation negatively affects the entire ecosystem 139 . DPHP is used as a chemical additive in numerous industrial products, and because it does not bind with other chemicals, it is easily spread to the environment, where it has been widely detected 140 – 142 . When in the environment, DPHP has a long half-life and is immunotoxic and neurotoxic to other organisms 143 . Despite not being comprehensively studied in marine organisms, recent studies have reported its negative effects on fish growth, energy metabolism, and reproduction 144 – 147 . Further investigations are necessary to understand the toxic and negative effects of DPHP on corals, and it may be beneficial to possess the necessary genetic machinery to degrade such harmful compounds. Lastly, we identified a regulatory system that functions under specific stress conditions, such as hypoxia and starvation, and appears to be beneficial to host health and resilience. pBMC and the importance of their secondary metabolites Secondary metabolites are important for defense against microorganisms, toxic compounds, and UV radiation as well as essential for symbiotic relationships 148 ; they are abundantly found within the coral holobiont 149 . Accordingly, we assessed the BGCs and Pfams of our six pBMC strains using the antiSMASH platform, revealing 11 secondary metabolite production clusters: ectoine, beta-lactone, terpene, aryl polyene, lasso peptide, NRPS, nonribosomal peptide metallophores, NI siderophores, opine-like zincophores, class I and IV lanthipeptides, and type I and III PKS. Ectoines are important because they reduce the effects of heat stress, high salinity, ROS, and radiation. Beta-lactone derivatives are extremely diverse, comprising 30 distinct families, many of which have antimicrobial activity 150 , while others are important elements for antibiotic production 148 . Terpenes are diverse organic compounds that play a role in defense mechanisms in plants and fungi 151 , 152 , serving as antioxidants and protecting cells from oxidative stress 153 , and their presence indicates production of diverse bioactive compounds 154 . Some terpenes and carotenoids, such as squalenes, are pigments involved in photosynthesis and signaling. However, they also possess antioxidant activity and can neutralize oxidative stress 155 – 157 . Aryl polyenes are pigments that are structurally and functionally related to carotenoids and confer protection against photo-oxidative damage and lipid peroxidation 158 , 159 . Lasso peptides have antimicrobial and antiviral properties and also show thermal and chemical resistance 148 , 160 . NRPSs are sources of newly discovered antibacterial agents that have also been widely studied for their antiviral and anti-inflammatory properties 148 , 161 . Microorganisms scavenge metal ions from the environment via metallophores 162 . Some of these metallophores, such as siderophores like desferrioxamine E, play important roles in biocontrol and bioremediation 163 – 165 , and are considered beneficial when selecting for BMC 33 . Zinc is also an essential nutrient for several cellular processes and is taken from the surrounding environment by bacteria 166 . Similar to iron availability, zinc is also present at low concentrations in the environment and is captured by zincophores (also called opine-like zincophores in bacteria), such as bacillopaline, produced by microorganisms 167 . Lanthipeptides that possess antimicrobial activity are known as lantibiotics 168 , but their functions are not limited to this; they may also possess antifungal and antiviral properties 169 , 170 . The presence of at least one PKS BGC in each pBMC genome suggests that these bacterial strains have the necessary genomic tools to synthesize various polyketides that are likely to have beneficial properties 161 . Red Sea pBMC and their Indo-Pacific Ocean counterparts Rosado and colleagues 26 successfully manipulated the coral microbiome of P. damicornis by adding BMC to coral fragments in a mesocosm setting, and the genomes of these BMC were screened to identify potential beneficial mechanisms and/or traits based on previous literature 82 . Some of the BMC from the study by Rosado and colleagues 26 , 82 share several gene functions related to putative beneficial traits for corals with some of our pBMC (e.g., superoxide dismutase, glutathione synthetase, catalase-peroxidase, adenosylcobinamide-phosphate synthase, adenosylcobinamide kinase, betaine aldehyde dehydrogenase, choline dehydrogenase, ectoine hydroxylase, L-ectoine synthase, nitrite reductase, and CoA transferase) (Table S3 ) 82 . However, when examining the pangenomes of each genus, we noted that our pBMC had a unique set of genes that were absent in the genomes of the BMC from the study by Rosado and colleagues 82 . Moreover, some of these genes represent potential beneficial traits or mechanisms (Figs. 6 , 7 , 8 and Table 1 ). Our results revealed potential insights into how bacteria help corals during periods of stress. Although some of the abovementioned beneficial traits are hypothetical, others have already been validated in previous studies on the differential expression of genes during heat stress experiments in corals 18 . Moreover, the identification and classification of these beneficial features depend on and is limited to existing databases. Additionally, relying on cultivation-dependent methods to obtain candidate pBMC introduces an element of chance in discovering and identifying potential candidates. This can be mitigated by obtaining a large number of isolates and meticulously screening and testing them for beneficial traits. Notably, all pBMC examined in this study, even those from the same genus, were distinct and demonstrated potential to contribute to the health and resilience of corals, indicating the need for continued efforts to isolate 149 , 171 , 172 , explore 82 , 173 – 175 and test 176 , 177 novel pBMC. We also highlight the discovery of prophages associated with two of the BMCs and their potential role in providing competitive advantage to coral probiotics against other bacteria."
} | 5,825 |
38371992 | PMC10873675 | pmc | 954 | {
"abstract": "Microbial fuel cells (MFCs) are promising for generating renewable energy from organic matter and efficient wastewater treatment. Ensuring their practical viability requires meticulous optimization and precise design. Among the critical components of MFCs, the membrane separator plays a pivotal role in segregating the anode and cathode chambers. Recent investigations have shed light on the potential benefits of membrane-less MFCs in enhancing power generation. However, it is crucial to recognize that such configurations can adversely impact the electrocatalytic activity of anode microorganisms due to increased substrate and oxygen penetration, leading to decreased coulombic efficiency. Therefore, when selecting a membrane for MFCs, it is essential to consider key factors such as internal resistance, substrate loss, biofouling, and oxygen diffusion. Addressing these considerations carefully allows researchers to advance the performance and efficiency of MFCs, facilitating their practical application in sustainable energy production and wastewater treatment. Accelerated substrate penetration could also lead to cathode clogging and bacterial inactivation, reducing the MFC's efficiency. Overall, the design and optimization of MFCs, including the selection and use of membranes, are vital for their practical application in renewable energy generation and wastewater treatment. Further research is necessary to overcome the challenges of MFCs without a membrane and to develop improved membrane materials for MFCs. This review article aims to compile comprehensive information about all constituents of the microbial fuel cell, providing practical insights for researchers examining various variables in microbial fuel cell research.",
"conclusion": "9 Concluding remarks and future perspectives ترجمه بیش از حد طولای است و د. This comprehensive review article thoroughly examines the latest developments in materials, methodologies, structural innovations, and microbial components pertinent to Microbial Fuel Cell (MFC) technology. Despite the significant promise of MFCs in sustainable electricity generation and wastewater treatment, their practical application on a larger scale poses substantial challenges. In this context, this review identifies key areas for future research and outlines specific recommendations to overcome these challenges. These areas encompass materials, methodologies, structural considerations, and the utilization of microorganisms in MFC technology. By addressing these aspects, we aim to facilitate the progression of MFCs towards becoming viable and sustainable green energy solutions. • Material innovations: one critical avenue for improving MFC technology involves the development of innovative materials. This review underscores the need to: a. Engineer low-cost, high-performance anode materials, including metal oxides and conductive polymers, to replace expensive metals and enhance MFC efficiency. b. Investigate cathode materials, such as heteroatom-doped carbons and transition metal complexes, as alternatives to costly noble metal catalysts. c. Design proton exchange membranes with high conductivity are cost-effective and exhibit antifouling properties. • Methodological improvements and the refinement of methodologies are pivotal in furthering MFC technology. Our recommendations include: a. Optimization of reactor configurations to mitigate internal resistance and enhance mass transfer. Notably, stacked and cascaded MFCs exhibit potential for scalability. b. Integration of MFCs into existing wastewater treatment infrastructure for simultaneous electricity generation and bioremediation. c. Development of standardized protocols to facilitate comparing results across different research studies. • Structural innovations enhance the practicality of MFCs for real-world applications; we emphasize the importance of structural innovation. This entails: a. The design of compact and portable single-chamber MFCs featuring air-cathodes. b. Utilizing 3D printing techniques to fabricate miniature MFCs suitable for remote sensing and robotics deployment. c. The engineering of modular and stackable MFC systems tailored for large-scale electricity generation. • Harnessing Microorganisms and microbial components plays a crucial role in MFCs, and exploring their potential is vital. Recommendations in this domain include: a. Exploration of cooperative microbial communities and metabolic engineering techniques to augment electricity generation. b. Investigation into extremophiles and their robust bioelectrochemical systems, particularly in high-temperature MFCs. c. Elucidating direct interspecies electron transfer mechanisms among electroactive microbes can significantly enhance MFC efficiency. In conclusion, overcoming the challenges associated with scaling up MFC technology while simultaneously improving performance and reducing costs is paramount. The suggestions delineated in this review aim to provide clear guidance for future researchers, ultimately advancing MFCs toward becoming sustainable and practical green energy solutions.",
"introduction": "1 Introduction Microbial fuel cells (MFCs) are devices harnessing the catabolic activity of microorganisms to generate electricity from organic materials. Safeguarding freshwater resources is essential to meet the growing needs of communities, especially amid the water crisis and resource scarcity. Both industrial and domestic wastes harbor a concealed reservoir of water and energy. Domestic wastewater and wastewater from food processing industries in the United States alone are estimated to contain approximately 17 GW of energy [ 1 ]. The increasing global population and industrialization have resulted in challenges concerning worldwide access to water and energy resources. [ 2 ], Kober et al. (2019) highlighted that the increasing population and industrialization are causing challenges in ensuring global access to water and energy resources [ 3 ]. The increasing demand for energy and decreased fossil fuel supply could result in a global energy crisis with significant environmental and human health impacts [ 4 , 5 ]. Fossil fuels, nuclear energy, and renewable energy are the main energy sources [ 6 ]. Lee et al. (2014) noted a recent upswing in interest regarding renewable energy systems. They underscored the urgency of addressing environmental pollution, focusing on water pollution and the associated challenges and pollutants in discharged wastewater [ 7 ]. Gupta et al. (2021) highlighted that freshwater availability could decrease by up to 40% in the next decade [ 8 ]. Nie et al. (2020) emphasized that the advancement and integration of new technologies for environmentally friendly bioenergy production face substantial challenges, including rising levels of greenhouse gases, societal and economic instability, and various other obstacles [ 9 ]. In summary, MFCs present a promising solution for generating electricity from organic materials. They should be considered part of a larger effort to address the water and energy crisis while reducing environmental pollution (see Table 3 , Table 4 , Table 5 )."
} | 1,784 |
35928942 | PMC9343942 | pmc | 955 | {
"abstract": "With the rapid development of synthetic biology, a variety of biopolymers can be obtained by recombinant microorganisms. Polyhydroxyalkanoates (PHA) is one of the most popular one with promising material properties, such as biodegradability and biocompatibility against the petrol-based plastics. This study reviews the recent studies focusing on the microbial synthesis of PHA, including chassis engineering, pathways engineering for various substrates utilization and PHA monomer synthesis, and PHA synthase modification. In particular, advances in metabolic engineering of dominant workhorses, for example Halomonas, Ralstonia eutropha, Escherichia coli and Pseudomonas, with outstanding PHA accumulation capability, were summarized and discussed, providing a full landscape of diverse PHA biosynthesis. Meanwhile, we also introduced the recent efforts focusing on structural analysis and mutagenesis of PHA synthase, which significantly determines the polymerization activity of varied monomer structures and PHA molecular weight. Besides, perspectives and solutions were thus proposed for achieving scale-up PHA of low cost with customized material property in the coming future.",
"conclusion": "Conclusion and perspective In this study, we highlighted the global trends of industrial PHA productions reported by different companies and start-up teams, and briefly summarized and discussed the advances of different building blocks focusing on PHA synthase, biosynthesis pathways of SCL-, MCL- and LCL-PHA, dominant PHA workhorses of industrial potential and optimization strategies for effective PHA synthesis. This study provides an overview of PHA biosynthesis from enzyme engineering, cell factory design, towards scale-up bio-manufacturing. However, more attempts are still required to achieve further cost-reduction and improved material properties of tailor-made PHAs against the petrol-based plastics.",
"introduction": "Introduction Polyhydroxyalkanoates (PHAs) is a series of polyesters synthesized by different microbes ( Steinbüchel, 2001 ), which have been widely used as bio-plastics for replacing petrol-based plastic due to their outstanding biodegradability and biocompatibility. Accordingly, PHA can be divided into three categories ( Sudesh et al., 2000 ) including short-, medium- and long- chain-length PHAs, namely SCL-, MCL- and LCL-PHA, respectively. Of which, the monomers of SCL-, MCL- and LCL-PHA generally contain 2–5, 6–14 and over 15 carbon atoms, respectively. Because of the competitive material properties, PHA has attracted growing attentions of commercial interests in different application areas, such as medical implant ( Chen and Wu, 2005 ), cosmetic beads ( Choi et al., 2020 ), packaging ( Chen and Patel, 2012 ), agricultural film ( Chen, 2009 ), textile ( Chen, 2009 ), feeding additives ( Chen, 2009 ) and so on. In the past decades, intensive efforts have been made to generate various PHA productions consisting of diverse polymerized units with different carbon-chain-length and structures by genetically modified bacterial ( Chen and Jiang, 2017 ), such as Halomonas spp. ( Tan et al., 2011 ; Fu X. Z. et al., 2014 ) , Ralstonia eutropha ( Antonio et al., 2000 ; Raberg et al., 2018 ; Xiong et al., 2018 ) , Escherichia coli ( Park et al., 2001 ; Linares-Pastén et al., 2015 ; Sudo et al., 2020 ) , Pseudomonas spp ( Chanasit et al., 2016 ; Liang et al., 2020 ; Li M. et al., 2021 ) and so on ( Hyakutake et al., 2014 ; Tariq et al., 2015 ). Therefore, over 150 types of PHAs have been obtained including homopolymers (PHB, poly-3-hydroxybutyrate) ( Tan et al., 2011 ), random- and/or block- copolymers such as poly(3-hydroxybutyrate- co -4-hydroxybutyrate) (P34HB), poly(3-hydroxybutyrate- co -3-hydroxyvalerate) (PHBV) ( Fu X. Z. et al., 2014 ), poly(3-hydroxybutyrate- co -3-hydroxyhexonate) (PHBHHx) ( Park et al., 2001 ), etc. ( Li M. et al., 2021 ). To date, many building blocks, including rational designed enzymes ( Chek et al., 2019 ; Lim et al., 2021 ), fine-tuned metabolic pathways towards monomer synthesis ( Pacholak et al., 2021 ) and genetically engineered chassis of predominant PHA accumulation performance ( Liang et al., 2020 ; Ye and Chen, 2021 ), have been developed for sufficient PHA synthesis using a variate of substrates. In particular, scale-up industrial production lines for various PHA manufacturing have been recently launched or established by several companies, for example, MedPHA (operating production line of 1,000 ton/year PHB and/or P34HB, China) ( Obruča et al., 2022 ), PhaBuilder (10,000 ton/year, under construction, China) ( Yang et al., 2010 ), Tianan (3,000 ton/year PHBV, China) ( Modi et al., 2011 ), Tepha (P4HB for medical uses, United States) ( Martin and Williams, 2003 ), Danimer Scientific (6,000 ton/year PHBHHx, United State) ( Mehrpouya et al., 2021 ), Keneka (5,000 ton/year PHBHHx, Japan) ( Tanaka et al., 2021 ). However, the production cost of PHA still challenges for wide range commercial uses. Therefore, many solutions have been proposed and developed to reduce the industrial cost of PHA, including high cell density fermentation based on optimized feeding solution ( Silva et al., 2017 ), non-sterile open fermentation process based on recombinant halophiles ( Tan et al., 2011 ), cell factory engineering for effective utilization of low-cost carbon sources ( Murugan et al., 2017 ; Panich et al., 2021 ), carbon fixation engineering for the improved conversion rate from glucose to PHA ( Salehizadeh et al., 2020 ), co-production of PHA and value-added chemicals ( Lan et al., 2016 ; Li et al., 2016 ) and so on. Therefore, this study summarized recent advances of various PHA production and industrial trends thereof. Additionally, major building blocks, including representative workhorses, metabolic pathways and critical enzymes, for PHA synthesis have been reviewed and discussed. This study provides an entire landscape of PHA productions powered by synthetic biology, as well as perspectives focusing on cost-effective PHA manufacturing in the coming future."
} | 1,517 |
19636465 | null | s2 | 957 | {
"abstract": "Polymer dynamics play an important role in a diversity of fields including materials science, physics, biology and medicine. The spatiotemporal responses of individual molecules such as biopolymers have been critical to the development of new materials, the expanded understanding of cell structures including cytoskeletal dynamics, and DNA replication. The ability to probe single molecule dynamics however is often limited by the availability of small-scale technologies that can manipulate these systems to uncover highly intricate behaviors. Advances in micro- and nano-scale technologies have simultaneously provided us with valuable tools that can interface with these systems including methods such as microfluidics. Here, we report on the creation of micro-curvilinear flow through a small-scale fluidic approach, which we have been used to impose a flow-based high radial acceleration ( approximately 10(3) g) on individual flexible polymers. We were able to employ this microfluidic-based approach to adjust and control flow velocity and acceleration to observe real-time dynamics of fluorescently labeled lambda-phage DNA molecules in our device. This allowed us to impose mechanical stimulation including stretching and bending on single molecules in localized regimes through a simple and straightforward technology-based method. We found that the flexible DNA molecules exhibited multimodal responses including distinct conformations and controllable curvatures; these characteristics were directly related to both the elongation and bending dynamics dictated by their locations within the curvilinear flow. We analyzed the dynamics of these individual molecules to determine their elongation strain rates and curvatures ( approximately 0.09 microm(-1)) at different locations in this system to probe the individual polymer structural response. These results demonstrate our ability to create high radial acceleration flow and observe real-time dynamic responses applied directly to individual DNA molecules. This approach may also be useful for studying other biologically based polymers including additional nucleic acids, actin filaments, and microtubules and provide a platform to understand the material properties of flexible polymers at a small scale."
} | 568 |
39943669 | PMC11867021 | pmc | 958 | {
"abstract": "Domain-wall electronics based on the tunable transport\nin reconfigurable\nferroic domain interfaces offer a promising platform for in-memory\ncomputing approaches and reprogrammable neuromorphic circuits. While\nconductive domain walls have been discovered in many materials, progress\nin the field is hindered by high-voltage operations, stability of\nthe resistive states and limited control over the domain wall dynamics.\nHere, we show nonvolatile memristive functionalities based on precisely\ncontrollable conductive domain walls in tetragonal Pb(Zr,Ti)O 3 thin films within a two-terminal parallel-plate capacitor\ngeometry. Individual submicron domains can be manipulated selectively\nby position-sensitive low-voltage operations to address distinct resistive\nstates with nanoampere-range conduction readout. Quantitative phase-field\nsimulations reveal a complex pattern of interpenetrating a- and c-domain\nassociated with the formation of 2D conducting layers at the intertwined\nregions and the emergence of 3D percolation channels of extraordinary\nhigh conductivity. Subnanometer resolution polarization mapping experimentally\nproves the existence of such extensive segments of charged tail-to-tail\ndomain walls with unconventional structure at the ferroelastic-ferroelectric\ndomain boundaries.",
"conclusion": "Conclusions To conclude, the integration of PZT films\nwith highly conductive\nDWs and focused ion beam (FIB)-deposited electrodes presents a unique\ncombination of properties, making them highly promising for multistate\nmemristor applications. These films exhibit high DW conduction, achieving\ncurrents in the range of 1–10 nA/μm 2 with\nthe application of remarkably low voltages of just 2 V. The system\nsupports multiple reproducible and randomly accessible states, as\nwell as consistent, position-sensitive switching of individual domains.\nThis innovative concept opens new pathways for research and development,\nparticularly in integrating the memristor with access circuitry for\nindividual domain read/write operations without relying on an AFM\nprobe. Such circuitry could involve a network of high-conductivity\nleads connected to the high-resistance layer at specific points where\ndomains are to be nucleated. With potential domain sizes as small\nas 10–20 nm, the design holds significant promise for scaling.\nHowever, challenges such as cross-talk between adjacent domains and\nachieving precise domain addressability require further investigation. Compared to previously presented LNO based DW-memristors 33 in which a high-density of electrically controllable\ndomains allowed for the reproducible access to numerous conduction\nstates, the memristor introduced in this work functions by manipulation\nof single DWs. It operates at much lower read/write voltages and may\nsupport more aggressive scaling by requiring fewer domains through\na higher current over single domain wall ratio. However, these advantages\ncome with trade-offs, including the need for more complex access circuitry\nand a reduced number of conduction states compared to the LNO-memristor. The mechanism underlying the high domain wall conduction in tetragonal\nstrained PZT films reveals a more intricate domain structure than\npreviously thought. Phase-field simulations, supported by STEM atomic-scale\nanalysis and polarization mapping, indicate that the interaction between\nDWs between the entangled a- and c-domains is central to the observed\nmemristive properties. The simulations show the emergence of the complex\nnetwork of bound charges, leading to the formation of an extended\nsystem of conductive channels percolating throughout the bulk of the\nfilm, and demonstrating that the enhanced conductivity is inherently\na 3D volumetric effect. While conventional 2D analytical methods,\nsuch as STEM, have provided valuable insights and are able to support\nthe proposed model, fully capturing the nature of the current transport\nin these systems necessitates a 3D reconstruction of the conductive\nchannels. Thus, this study demonstrates that highly conductive\ncharged DWs\ncan be effectively harnessed for information processing within well-established\ntetragonal perovskite ferroelectrics, which exhibit very stable and\npredictable domain structures.",
"discussion": "Results and Discussion Confining Conductive 180°-DWs inside a Capacitor PZT (Zr/Ti = 10:90) films of 60 nm thickness were epitaxially grown\nby pulsed laser deposition (PLD) onto a (110) DyScO 3 (DSO)\nsubstrate together with a 20 nm thick SrRuO 3 (SRO) bottom\nelectrode (details in Methods). Through PFM imaging the pristine downward\npolarized c-domains, which are intersected by ferroelastic a-domains,\ncan be visualized ( Figure 1 a). Phase loops acquired with the AFM-tip on the surface yield\na switching window of ∼±2 V (see Supporting Information).\nIn these samples the 90°-DWs as well as the 180°-DWs, which\nare created by poling, exhibit a conductive response, which can be\nrevealed by cAFM imaging ( Figure 1 a, bottom row). Figure 1 Ferroelectric domain wall properties and\nhigh-R Pt device characteristics.\n(a) From top to bottom: topography, PFM-amplitude, PFM-phase and cAFM\nimages of the PZT bare surface. In the left column the pristine film\nis shown with uniformly downward polarized c-domains (purple) and\nweak 90°-DW conduction along the a-domain pattern (black lines\nof the cross-hatch pattern). In the right column, a square in a square\npoling was done indicated by the red dotted line. The poled segments\nshown in yellow, are bordered by conductive 180°-DWs. (b) PFM\nimages of a high-R Pt electrode before and after a 5 V/1 ms poling\npulse. A round polarized domain is observed after the poling together\nwith a ring-like 180°-DW. (c) IV characteristics recorded by\nthe AFM-tip at the same position before and after the poling (green\ndot). Without any 180°-DW, the conductance of the device is below\nthe noise level of 5 pA (HRS). After the injection of the 180°-DW\nthe conduction is increased to around 3 nA (LRS). (d) Retention test\nof the LRS. A circular domain is poled first and imaged by PFM, then\n1100 1.5 V/2 ms pulses are applied and the current response is recorded.\nA PFM image is taken after the test to confirm the stability of the\ncircular poled domain. (e) Endurance test by an alternating sequence\nof injecting and erasing of a polarized domain. After each operation\nthe conductance state of the device is recorded by 1.5 V/2 ms pulses.\nPFM images before and after are taken to monitor the reproducibility\nof the poling process. In previous works 20 , 39 , 40 the conduction properties of the 90°- and 180°-DWs\nwere\nstudied and basic memristive operations were demonstrated. However,\nby using Cr/Au evaporated electrodes as described in ref., 20 it was not possible to confine written 180°-DWs\nunder the top electrode area. The demonstrated configurations used\na DW-electrode connection which was obtained by pushing the 180°-DWs\nfrom outside of the device area to the boundary of the electrode to\nform binary on/off-switching devices. Device states in which the 180°-DW\nwas positioned directly under the electrode were unstable and tended\nto collapse once electrical readout or consecutive PFM imaging was\nperformed (see Supporting Information ),\npreventing any precise DW-conduction tuning. To enable a better\nand flexible control over the domain wall position\ninside the device area, electron beam induced deposited (EBID) platinum\n(Pt) top electrodes were employed (details in Methods). These electrodes\npossess a relatively high sheet resistance – up to 3 orders\nof magnitude higher than that of conventional electrodes of the same\nthickness. 41 The elevated resistance limits\nthe speed of charge movement, resulting in a voltage gradient across\nthe high resistive (high-R) electrode when subject to sufficiently\nshort voltage pulses. 42 , 43 Therefore, by tuning the amplitude\nand dwell time of the poling pulses, the propagation of the DWs can\nbe controlled enabling position-sensitive poling/reading operations.\nThis approach allows to confine the polarization domains to submicron\nareas beneath the conductive probe inside the capacitor device. In\nthe pulse mode, each individual domain can be selectively written\nand erased without impacting adjacent domains. Additionally, the high-R\nPt electrodes facilitate the separation of nondestructive readout\nand write operations, allowing for domain-wall current sensing that\ndoes not significantly change the shape of the written domains. In Figure 1 b (left\ncolumn) PFM images of a pristine 5 × 5 μm 2 high-R\nPt electrode with a thickness of 12 nm are shown. PFM amplitude and\nphase images show that the domain configuration and polarization state\nof the capacitor can be monitored through the top electrode. For manipulating\nthe domain configuration, to inject or erase DWs and to probe the\nconductance state of the device, the AFM-tip is used as a nanometric\ncontact. By placing the AFM-tip in the center of the electrode and\nafter the application of a short voltage pulse (5 V/1 ms), a circular\ndomain is formed under the tip’s location (right column of Figure 1 b). The resulting\nconductance change of the device from its original high-resistive\nstate (HRS) to its low-resistive state (LRS) can be observed in Figure 1 c, in which IV curves\nfrom before (red) and after (blue) the poling pulse are recorded.\nIn the pristine HRS without any 180°-domain wall, a current lower\nthan the noise level of around 5 pA for 1.5 V is measured. After the\npoling pulse, the injected domain wall mediates a current transport\nproviding a readout of ∼3 nA which results in an on/off ratio\nof at least 3 orders of magnitude. The on/off switching as well as\nthe current readout are performed from the same location in the center\nof the electrode (green dot). In Figure 1 d,e,\nretention and endurance cycling tests are performed, respectively.\nFor the retention test, a single circular domain is poled (insert\nPFM image, left side) and successively probed by rectangular 2 V/2\nms readout pulses. A total of 1100 pulses are recorded consecutively\nwithout degradation of the LRS. Subsequent PFM imaging (insert PFM\nimage, right side) confirmed the stable domain structure. In the endurance\ntest, a circular domain is alternately created and erased with readout\npulses before and after each operation. The test performed over 100\ncycles shows excellent current stability. PFM images before and after\ncycling (see PFM inserts) show nearly the same domain configuration\nconfirming the stability and control of the injected and erased DWs.\nSubsequent tests showed that nondegrading cycling for more than 700\ncycles can be performed with only slight variations of the LRS (see\nSupporting Information). From Binary to Multistate Memristor These characteristics\npave the way for more advanced multiresistive level device concepts,\nrelying on repeatedly creating and erasing of multiple domains and\na resulting step-like conductance change. In Figure 2 a, the working principle of the device is\nillustrated. By varying the tip position on the high-R Pt electrode\nand the application of short (5 V/5 ms) voltage pulses an independent\nset of DWs can be injected in the device. In Figure 2 b PFM images taken after each poling operation\nare shown, revealing the step-by-step creation of new domains (top\nrow). Notably, the application of subsequent poling pulses does not\nlead to any visible change of the shape or position of the already\nexisting domains due to the well confined electric field. By simultaneously\nrecording IV curves after each operation (see Supporting Information ), a gradual increase of the conductivity\ncan be observed (red squares in Figure 2 c). The IV curves are taken from the center of the\ndevice (green dot) and show a DW-mediated current up to a maximum\nof ∼6 nA for 8 created domain wall rings. Even more importantly\nit is possible to selectively turn off the switched domains again\nby adjusting the width and amplitude of the backpoling pulses (−3\nV/80 μs). By placing the tip at the previously poled domain\nlocations, a sequential deactivation of single domains is realized\n(bottom row of Figure 2 b). Consequently, by erasing the DWs associated with the poled domains\nthe conductivity of the device is gradually decreased (blue squares\nin Figure 2 c). Moreover,\nthe conduction profile (”potentiation/depression curve”)\nplotted over the number of created/erased domains is found to be nearly\nlinear and fully symmetric, compared to similar pulse modulation schemes\nin memristive FeFET devices, which suffer from high nonlinearity and\nasymmetry. 44 , 45 Figure 2 Multistate domain wall memristor. (a)\nDevice schematic and poling\noperations. (b) PFM-images of each device state after consecutive\npoling operations. Top row shows the one-by-one addition of poled\ndomains (yellow) by application of positive pulses. Bottom row shows\nthe step-by-step turning off of individual domains by negative pulses.\n(c) Simultaneously recorded change of conduction for each of the 9\nlevels (8 poled domains + off state) from the experiment in (b). The\nred curve indicates the increase in conduction when more domains are\npoled and the blue curve corresponds to the case when domains are\nswitched off and the conductance decreases. (d) Plot of the conduction\nvalues over the total domain wall length (DW-length acquired by calculating\nthe accumulated domain circumference of each state). A linear behavior\nis apparent with a slope value of 410 pA/μm. A high degree of\nsymmetry between the potentiation and depression curves is obvious\nand supported by the root-mean-square error (RMSE) values of 0.189\nnA and 0.187 nA respectively. In Figure 2 d, the\npotentiation and depression curves are plotted over the total DW-length\nof each device state (DW-length is measured as the accumulated circumference\nof the polarized domains), confirming the linear dependency of conduction\nover DW-length with a slope of around 410 pA/μm. The high symmetry\nbetween the potentiation and depression curve results from the high\ndegree of control over the size of domains and their excellent reversible\ncharacter. The high linearity is explained by the fact that each poling\noperation at a different location creates another low-resistive path\nfor the current to pass the capacitor similar to a parallel resistor\ncircuit. Moreover, by tuning of the poling pulses and reducing the\nthickness of the high-R Pt electrode it was possible to inject 5 domains\nin a sub 0.25 μm 2 area without a collapse of the\ndomain configuration and a single domain size of 50 nm (see Supporting\nInformation), demonstrating the scalability of the approach. Repeatable Addressing of Multiple Resistive States To demonstrate the potential and robustness of this approach for\na multistate memristor, a sequence is performed in which different\nconduction levels are addressed independently of the previous state\nand in a repeatable manner. For the data in Figure 3 a device similar to the one from Figure 2 is used with a 7\n× 7 μm 2 surface area. In Figure 3 a, starting from the off-state with monodomain\nconfiguration and a current readout lower than the noise level of 5 pA, the device is cycled through\ndifferent domain configurations ranging from 0 to 4 poled domains.\nThe domains are arranged in a square (see PFM-phase inserts at each\nconduction level) and are located around the readout point (center\nof the device). An irregular sequence is performed in which each conduction\nlevel is addressed 3 times. After applying the poling (5 V/0.5 ms)\nand depoling (−2.5 V/80 μs) pulses needed to reach the\nnew state, the conduction is read from the center with 3 consecutive\npulses (1.5 V/100 ms). Each resistive state could repeatably be addressed,\nmeaning that after cycling through different configurations, it was\npossible to restore the conduction values of that level within a margin\nof 200–300 pA. Figure 3 Repeatable multilevel addressing of nonvolatile conduction\nstates.\n(a) Independent addressing of 4 conduction levels in a random sequence.\nEach conduction level is addressed 3 times and the parameters of the\npoling/depoling pulses were 5 V/0.5 ms and −2.5 V/80 μs,\nrespectively. After each operation 3 readout pulses (1.5 V/100 ms)\nare recorded before switching to the next level and PFM images are\ntaken after each step to confirm the correlation between conduction\nstate and domain configuration. (b) Multilevel endurance test by cycling\nbetween the fourth and second conduction level of the same device\nas in (a). PFM images before and after the test (inserts) confirm\nthe stability of the written domain patterns. (c) Retention characteristics\nof the 4 conduction levels, probed by consecutive application of around\n70 pulses (1.5 V/100 ms). After a stabilization phase for around 5\npulses (∼1 s) the conduction remains stable and each level\nis well separated by at least 500 pA. PFM inserts show the domain\nconfiguration of each state. In Figure 3 b, the\nmultilevel endurance is tested by cycling between the conduction level\ncorresponding to 2 and 4 poled domains. Between each readout pulse\ntwo poling/depoling pulses are applied, respectively. For up to 40\ntimes, the domains were cycled and showed reasonable stability. For\nthe second level an increase of around 100 pA after 28 cycles is observed,\nwith the overall conduction being confined within a margin of 280\npA. During this fatigue test, some domains exhibit a tendency to grow\nafter cycling due to the alternating poling pulses with higher amplitude\nthan the readout pulses. Importantly, this domain expansion has minimal\nimpact on the multilevel functionality. The current variation in Figure 3 b remains within\n10% without any consistent trend of increase across cycles. A plausible\nexplanation is that domain wall conduction does not scale linearly\nwith the domain circumference, and additional sections formed during\ndomain expansion seem to contribute very weakly to conduction. Further\nexplanations of this behavior are presented in the next section. The\nmultistate fatigue test illustrates that repetitive cycling and measuring\nis possible not only for complete on/off-switching as shown in Figure 1 , but also between\ndifferent groups of domains. It is worth noting that the pulses used\nfor cycling the system in the multibit storage mode ( Figure 3 b) are different compared to\nthe single-bit switching in Figure 1 e. Cycling between the 4-domain state and 2-domain\nstate requires precise voltage pulse adjustments to selectively erase\ntwo domains while leaving the other two virtually unaffected. Pulses\nof −3 V/80 μs were determined to be sufficiently short\nto reproducibly remove the domain under the tip with the minimal impact\nto the adjacent domains. These pulses could also be applicable for\nthe endurance test in Figure 1 e, however in that case much longer −3 V/50 ms pulses\nwere used. These long pulses ensure complete polarization reversal,\neffectively erasing small domain nuclei near the bottom interface,\nthereby subjecting the fatigue test to more rigorous conditions. To complement these experiments, Figure 3 c shows the retention of the 4 resistive\nlevels used in Figure 3 . For each level around 70 readout pulses (1.5 V/100 ms) are recorded.\nBetween each 5 pulses a waiting time of several seconds is introduced\nto simultaneously test for a medium time-stability. Each level was\nprobed for a total time of 2–3 min and the total experiment\nlasted for around 30 min (taking into account the time to switch levels\nand to acquire the PFM images for each state). A slight decrease for\nthe first 3–5 pulses (∼1 s) of around 500 pA is observed\nwhich readily stabilizes. The stabilized discrete levels show a separation\nof ≥500 pA between each other with a scattering within one\nlevel of ≤300 pA. These observations further strengthen the\nnonvolatile character of each addressable conduction state with high\nrobustness over repetitive readouts and longer time scales. The operation\nspeed is another important characteristic of the memory element, which\ncan be estimated from the data in Figure 3 . The demonstrated writing/erasing speeds\nof 500 μs/80 μs ( Figure 3 b) are comparable to those of flash memories, albeit\nwith lower voltage requirements. In the presented experiments, the\nswitching speed is primarily constrained by extrinsic factors such\nas capacitance and electrode resistance. Further iterations of this\nconcept with electrical contacts replacing the AFM tip and with a\nsmaller capacitor size, are expected to yield significantly faster\ncurrent responses. Modeling To gain deeper insights into the switching\nmechanism between high- and low-resistivity states, we conducted phase-field\nmodeling of a PZT thin film region with a thickness of 60 nm and periodically\nconstrained lateral dimensions of 250 × 250 nm 2 (details\nsee Methods). In our simulation, we identified the bound charges,\ncharacterized by the divergence of the polarization field and concentrated\nwith a density of ρ = −div P . These bound\ncharges are screened by semiconducting free charges, typically originating\nfrom lattice imperfections and impurities such as oxygen vacancies. 39 The free charges, attracted by the bound charges,\nform memristive channels. We demonstrate that the complex topological\nnetworking of the 180° and 90° DWs results in an intricate,\npercolating configuration of these channels. Notably, this is an inherently\n3D effect of domain wall interlacement, which makes it challenging\nto identify on the 2D slices of the structure commonly used for the\nanalysis of polarization patterns. We investigate the dynamics\nof polarization domains and bound charges that form the memristive\nchannels within the film during the poling process. A cylindrical\nvolume with a radius of R = 100 nm is poled by applying\na bias −6 V at the surface. Panels a–c of Figure 4 illustrate the distribution\nof domains and charges before poling, while panels d-i show the distribution\nafter poling when the voltage is removed and the system is relaxed. Figure 4 Domains,\nDWs and bound charges in PZT films: a phase-field simulation.\n(a) Polarization distribution at the surface of the pristine film\nwith a- and c-type domains. The color map illustrating the polarization\norientation is displayed at the top-center of the figure. (b) Bound\ncharges at the termination of 90° DWs at the surface. (c) 3D\nphase-field tomography of the domain wall structure and bound charge\ndistribution inside the pristine film. Gray inclined surfaces represent\nthe 90° DWs. The positive and negative bound charges are shown\nin red and blue, respectively. (d) Polarization distribution at the\nsurface of the film after poling and relaxation. (e) Distribution\nof the bound charge at the surface after poling and relaxation. (f)\n3D phase-field tomography of the DW structure and bound charge distribution\ninside the film after poling and relaxation. The 180° DW is shown\nin yellow. (g) Top view of the 3D phase-field tomography image of\npanel (f) showing the extended areas of negative bound charges (blue)\naround the 180° DWs (yellow). (h) 2D cross-cut slice along the\nline AB at panel (d) showing the distribution of polarization. (i)\n2D cross-cut slice along the line AB at panel (e) showing the distribution\nof bound charge. Before poling, the sample is predominantly polarized\ndownward along\nthe c-direction (dark red color). The network of narrow a-type domains\n(blue and green color) emerges to compensate for the strain introduced\nby the interface with the DSO substrate. Two crossing a-type domains\npiercing the film are shown in Figure 4 a. Figure 4 b illustrates the surface emergence of bound charges, associated\nwith 90° DWs of these domains. The origin of the charges is related\nto a slight deviation of the 90° walls from the charge-neutral\n45° orientation. This configuration optimizes the elastic energy\nassociated with matching a-type and c-type domains. 39 These charges exhibit alternating, positive (red) and negative\n(blue) signs on opposite sides of the a-type domains, depending on\nthe direction of polarization turn at the DWs. The 3D phase-field\ntomography of the DWs and bound charges distribution\nbeneath the surface, shown in Figure 4 c, gives more information about the structure of the\nsystem. We observe that the regions with the highest concentration\nof bound charges are located at the 90° DWs near the interface\nwith the substrate. This phenomenon is attributed to the additional\ndeviation of the DWs from their 45° orientation to align with\nthe substrate. Another notable feature is the emergence of a-type\ndomains at the interface that do not extend to the upper surface,\ncollapsing, instead, within the bulk of the c-phase. Due to their\nsubstantial curvature at the collapse points, these walls also host\nrelatively large bound charges, that are mostly positive. We\nnow focus on the situation after poling, which reverses the\npolarization within the poled cylindrical volume. As illustrated in\nthe surface view at Figure 4 d, the majority of the poled area consists of upward-directed\nc-type domains (light red color). Most of the a-type domains have\nbeen displaced from the surface of the poled region. Only a small\nsegment of an a-type domain, shown in green, remains within the poled\narea, near the border. Importantly, the surface-bound charges are\nnow located not only along the DWs of this residual 90° domain,\nbut also around the perimeter of the poled area, where the 180°\ndomain wall emerges, marking the change in polarization of the out-of-plane\ndirection (see Figure 4 e). Analysis of the full 3D phase-field tomography of the DWs\nand bound\ncharges gives a comprehensive understanding of the bound charge distribution.\nThe side and top views of the tomography are shown in Figure 4 f,g, respectively. These images\ndemonstrate the interlacement of the 180° (yellow color) and\n90° DWs (gray color), resulting in the intricate distribution\nof bound charges which are heavily concentrated at the intersection\npoints of the a- and c-type domain boundaries. Even more notable is\nthe interaction between the 180° and 90° DWs at the sides\nof the poled region, where the 180° DWs are pierced by intersecting,\nmutually perpendicular 90° DWs, separating two variants of a-type\ndomains. Figure 4 h,i, illustrates\nthe domain and charge distributions beneath the surface along the\nvertical cross-section A-B, referenced in panels (d) and (e). These\nimages provide useful information on the location of the bound charges\nwithin the bulk of the film. A spot of bound charges is observed in Figure 4 i at the junction\nof the 90° and 180° DWs in the upper left corner of the\npoled area, just below the place of the surface emergence of the 180°\ndomain wall, as shown in Figure 4 h. This observation aligns with findings from our previous\n2D simulations, 20 which concluded that\nthe observed conductivity of the 180° DWs are due to their networking\nwith conducting 90° DWs. Another important observation is that\nthe 180° DWs that bound the central c-domain exhibit a slight\ndeviation from their equilibrium vertical orientation. This deviation\nbecomes stronger as it approaches the interface with the substrate.\nFurthermore, a small portion of a-type domains is nucleated at the\nregion where the domain wall meets the substrate. This effect, observed\nin both the left and right 180° DW is associated with the polarization\nbending near the substrate to accommodate the lattice matching between\nthe ferroelectric material and the substrate. These deviations of\nthe 180° DWs, which are typically charge-neutral in equilibrium,\nlead to the emergence of bound charges at the walls, and thereby providing\ntheir conductivity. Another observation in Figure 4 i is the nearly horizontal\nlocus of bound charges at\nthe central part of the cross-sectional area, which may also host\na memristive channel. These charges originate from the intersection\nof the [01̅0] a-type domain (blue area in Figure 4 h) and the [001] c-type domain (light red\narea in Figure 4 h).\nHowever, the 2D slice does not provide complete information regarding\nthe origin of the bound charge spots or their interconnectivity, which\nis needed to understand the formation of the memristive channels. The complex arrangement of entangled a-type and c-type domains,\nwhich is fully revealed in the 3D view, leads to the formation of\nextended zones with predominantly negative bound charges (see Figure 4 f,g), which is consistent\nwith the formation of a 2D hole gas that supports domain wall conduction.\nSignificantly, the conductive channels exhibit an intricate, continuous\npercolating distribution throughout the volume from bottom to top,\na characteristic that can not be fully captured in 2D cross sections.\nThese channels likely serve as current pathways within the memristor.\nOverall, the system reveals a complex network of interconnected conductive\nspots. The issue of conductivity becomes a percolation problem, where\nfree charges navigate through these spots from the top to the bottom\nof the film. By tuning and switching the domain wall network with\nan electric field, the conductive channels within the volume can be\nrearranged, imparting the system with distinct memristive properties\nthat are ideal for neuromorphic application. STEM-Cross Section Analysis of 180° Domain Walls To confirm the emergence of the charged domain wall segments calculated\nby the phase-field modeling, we prepared a scanning transmission electron\nmicroscopy (STEM) specimen from one of the poled areas of a similarly\nproduced PZT film (details in Methods). More specifically, ferroelectric\ndomain switching was implemented in areas of 5 × 10 μm 2 by applying a positive bias of +5 V at the SrRuO 3 and scanning the grounded tip in successive rectangular regions\noriented along the <100> pc direction of the DyScO 3 substrate. This switching created consecutive downward poled\nregions between the pristine upward polarized areas separated by the\nformed 180° DWs. A cross-sectional lamella oriented along <010> pc was extracted from the poled region by a focused ion beam,\nto determine the location of the 180° DWs and analyze their polarization,\nwhich could reveal the presence of charged segments. The 180°\nDWs locations within the film were located and imaged using an annular\nbright field (ABF) imaging. The 180° DWs, which form at the boundaries\nbetween the upward-poled domain and the two downward-poled domains,\nare shown in Figure 5 a as dark-shaded regions. This shading stems from diffuse scattering\ncaused by the shear strain that emerges at the defect following the\npoling process. In addition, the film is populated by a-type domains\nwhich are inclined along the (011) pc and (01̅1) pc planes of the tetragonal PZT. Figure 5 STEM images and polarization\nanalysis of the PZT film and the 180°\nDWs. (a) Low-magnification ABF image of the film. The three a-domains\nare highlighted with dashed white dashed lines and the 180° DWs\nare identified by a black contrast. (b–d) High-resolution ABF\nimages of the 180° domain wall at different locations collected\nfrom the yellow, blue and red square areas in (a). The images are\noverlaid with a color map of the analyzed polarization vector orientation\nhighlighting the differently polarized domains and the domain boundaries.\nThe polarization discontinuity at the domain boundary is highlighted\nwith a yellow dotted line. The polarization of the domains was calculated\nby analyzing the\npositions of the Ti/Zr and O atoms. Specifically, the analysis showed\nthat the c-domains of the as-grown film have an out-of-plane polarization\npointing upward and the switched c-domains downward. The a-domain\npolarization is adapted to minimize the energy cost corresponding\nto the bound charges that emerge at the DWs between the c- and a-domains.\nTo gain a deeper insight into the formation of the 180° DWs,\nwe performed high-resolution ABF-STEM at the three designated areas\nshown in Figure 5 a.\nThe first image (highlighted in yellow, Figure 5 b) is focused on the interconnection of the\ntwo a-domains. The image shows that the c-domain above the left a-domain\nand the c-domain beneath the two a-domains are switched and polarized\ndownward. On the other hand, the c-domain above the right a-domain\nis polarized upward and therefore unswitched. The analysis of polar\ndisplacements reveals that both a-domains are polarized to the right,\nforming a polarization discontinuity at the boundary between the bottom\ndomain wall of the right a-domain and the downward polarized c-domain\nwhich runs parallel to the (110) pc plane. This discontinuity\nforms a tail-to-tail domain wall configuration which induces negative\nbound charges. Further down the film, the high-resolution ABF image\nnear the bottom interface (highlighted in blue, Figure 5 c) displays a 45° inclination of the\n180° domain wall along the (110) pc orientation before\nit reaches the substrate creating another charged segment with a tail-to-tail\nconfiguration. An explanation for this unexpected deviation from the\ncharge neutral configuration could be that an a-domain previously\nexisted in the region where the domain wall twist occurs but was eliminated\ndue to the strong electric fields applied during poling. The residual\nstress left at the location of the former a-domain causes the 180°\ndomain wall to twist and aligning it parallel to the (110) pc plane. Finally, the 180° domain wall in the third image (highlighted\nin red, Figure 5 d),\nwhich runs parallel to the (001) pc plane and separates\nthe two differently polarized c-domains, shows a slight bending confirming\nits weakly charged nature even in the absence of a-domain interactions. Both inclined charged segments as well as the slightly bent 180°\ndomain wall generate bound charges which attract screening charges\nfrom within the film and contribute to the conductivity of the formed\ndomain wall channels. Interestingly, both predicted mechanisms (slight\nbending of the 180° DW and strongly charged tail-to-tail segments\nat the a-c domain boundary interaction points) from the phase field\nsimulations are found in the 2D cross-section polarization STEM analysis,\nwhich support the proposed concept of a percolation mediated current\ntransport through the 3D domain wall channels."
} | 8,588 |
33305000 | null | s2 | 959 | {
"abstract": "At the biointerface where materials and microorganisms meet, the organic and synthetic worlds merge into a new science that directs the design and safe use of synthetic materials for biological applications. Vapor deposition techniques provide an effective way to control the material properties of these biointerfaces with molecular-level precision that is important for biomaterials to interface with bacteria. In recent years, biointerface research that focuses on bacteria-surface interactions has been primarily driven by the goals of killing bacteria (antimicrobial) and fouling prevention (antifouling). Nevertheless, vapor deposition techniques have the potential to create biointerfaces with features that can manipulate and dictate the behavior of bacteria rather than killing or deterring them. In this review, we focus on recent advances in antimicrobial and antifouling biointerfaces produced through vapor deposition and provide an outlook on opportunities to capitalize on the features of these techniques to find unexplored connections between surface features and microbial behavior."
} | 275 |
36620452 | PMC9817103 | pmc | 961 | {
"abstract": "Spiking neural networks (SNNs), as one of the algorithmic models in neuromorphic computing, have gained a great deal of research attention owing to temporal information processing capability, low power consumption, and high biological plausibility. The potential to efficiently extract spatio-temporal features makes it suitable for processing event streams. However, existing synaptic structures in SNNs are almost full-connections or spatial 2D convolution, neither of which can extract temporal dependencies adequately. In this work, we take inspiration from biological synapses and propose a Spatio-Temporal Synaptic Connection SNN (STSC-SNN) model to enhance the spatio-temporal receptive fields of synaptic connections, thereby establishing temporal dependencies across layers. Specifically, we incorporate temporal convolution and attention mechanisms to implement synaptic filtering and gating functions. We show that endowing synaptic models with temporal dependencies can improve the performance of SNNs on classification tasks. In addition, we investigate the impact of performance via varied spatial-temporal receptive fields and reevaluate the temporal modules in SNNs. Our approach is tested on neuromorphic datasets, including DVS128 Gesture (gesture recognition), N-MNIST, CIFAR10-DVS (image classification), and SHD (speech digit recognition). The results show that the proposed model outperforms the state-of-the-art accuracy on nearly all datasets.",
"conclusion": "5. Conclusion In conclusion, this work proposes to endow synaptic structures with spatio-temporal receptive fields and additional temporal dependencies in an effort to enhance the temporal information processing capabilities of SNNs. We propose the STSC module from the standpoints of both computational models and biological realities, which consists of TRF and FLI, implemented with temporal convolution and attention mechanisms. We verified the method's reliability on neuromorphic datasets of SHD, N-MNIST, CIFAR10-DVS, and DVS-Gesture. Notably, the STSC supports SNNs in reaching the SOTA result (92.36%) on the SHD dataset, which is comparable to ANNs' methods (89 and 92.4%), validating the potential of SNNs in the spatio-temporal data processing.",
"introduction": "1. Introduction Spiking neural networks (SNNs) are regarded as the third generation of neural networks (Maass, 1997 ), with the purpose of addressing the fundamental mysteries of intelligence and the brain by emulating biological neurons and incorporating more biological mechanisms (Roy et al., 2019 ). The two fundamental components of SNNs are spiking neurons and synapses, which create a hierarchical structure (layers) and subsequently construct a network. SNNs have attracted a significant deal of academic interest in recent years due to their prospective properties, such as the ability to process temporal information (Petro et al., 2019 ), low power consumption (Roy et al., 2019 ), and biological interpretability (Gerstner et al., 2014 ). Currently, SNNs are capable of processing event stream data with low latency and low power (Pei et al., 2019 ; Gallego et al., 2020 ). However, there is still a performance gap between SNNs and traditional Artificial Neural Networks (ANNs). Recent SNN training techniques based on surrogate gradients and back-propagation have significantly enhanced the performance of SNNs (Wu et al., 2018 ; Fang et al., 2021c ), while also promoting the further integration of ANNs' modules into SNNs (Hu et al., 2021 ; Yao et al., 2021 ; Zheng et al., 2021 ), greatly accelerating the development of SNNs. However, it remains challenging to connect these computational techniques with the biological properties of SNNs. Due to the time-dependent correlation of neuron dynamics, it is believed that SNNs naturally process information in both temporal and spatial dimensions (Petro et al., 2019 ; Roy et al., 2019 ). Further researches are necessary to harness the spatio-temporal information processing capabilities of SNNs. Combining ANNs' modules has significantly increased the performance of SNNs in several research studies. In terms of spatial information processing, CSNN (Xu et al., 2018 ) was the first to validate the application of convolution structure on SNNs, followed by the proposal of NeuNorm to improve SNNs' usage of convolution through auxiliary neurons (Wu et al., 2019 ). In the time dimension, Zheng et al. ( 2021 ) implements the time-dependent batch normalization (tdBN) module to tackle the issue of gradient vanishing and threshold balancing, and Yao et al. ( 2021 ) uses the Squeeze-and-Excitation (SE) block (Hu et al., 2018 ) to realize the attention distribution of the temporal dimension in order to improve the temporal feature extraction. Notably, Zhu et al. ( 2022 ) proposes Temporal-Channel Joint Attention (TCJA) to concurrently process input in both temporal and spatial dimensions, which is a significant effort for SNNs' spatio-temporal feature extraction. These studies effectively improve the performance of SNNs by transplanting established ANNs' modules and methodologies. However, applying these computational modules to SNNs from the standpoint of deep learning dilutes the fundamental biological interpretability, bringing SNNs closer to a mix of existing concepts in machine learning, such as recurrent neural networks (RNNs), binary neural networks (BNNs), and quantization networks. From a biological standpoint, some works focus on the synapse models, investigating the potential of SNNs in respect of connection modes and information transmission. Shrestha and Orchard ( 2018 ), Fang et al. ( 2020a ), and Yu et al. ( 2022 ) integrate impulse response models with synaptic dynamics, hence enhancing the temporal information representation of SNNs; Cheng et al. ( 2020 ) implements intra-layer lateral inhibitory connections to improve the noise tolerance of SNNs; from the standpoint of synaptic plasticity, Zhang and Li ( 2019 ) and Bellec et al. ( 2020 ) introduce bio-plausible training algorithms as an alternative to back-propagation (BP), allowing for lower-power training. Experiments revealed that the synaptic models of SNNs have a great deal of space for modification and refinement in order to handle spatio-temporal data better (Fang et al., 2020a ). We propose a Spatio-Temporal Synaptic Connection (STSC) module for this reason. Based on the notion of spatio-temporal receptive fields, the structural features of dendritic branches (Letellier et al., 2019 ) and feedforward lateral inhibition (Luo, 2021 ) motivate this study. By merging the ANNs' computation modules (temporal convolutions and attention mechanisms) with SNNs, we propose the STSC module, consisting of Temporal Response Filter (TRF) module and Feedforward Lateral Inhibition (FLI) module. As shown in Figure 1 , the STSC can be attached to spatial operations to expand the spatio-temporal receptive fields of synaptic connections, hence facilitating the extraction of spatio-temporal features. Figure 1 Illustration of receptive fields in synaptic connections. (A) The receptive fields of typical spatial operations used in SNNs, e.g., fully-connected layers (full) and 2D convolutional layers (sparse); (B) The STSC modules proposed to extend spatial operations with spatio-temporal receptive fields. The main contributions of this work are summarized as follows: We propose STSC-SNN to implement synaptic connections with extra temporal dependencies and enhance the SNNs' capacity to handle temporal information. To the best of our knowledge, this study is the first to propose the idea of synaptic connections with spatio-temporal receptive fields in SNNs and to investigate the influence of synaptic temporal dependencies in SNNs. Inspired by biological synapses, we propose two plug-and-play blocks: Temporal Response Filter (TRF) and Feedforward Lateral Inhibition (FLI), which perform temporal convolution and attention operations and can be simply implemented into deep learning frameworks for performance improvements. On neuromorphic datasets, DVS128 Gesture, SHD, N-MNIST, and CIFAR10-DVS, we have produced positive results. Specifically, we acquire 92.36% accuracy on SHD with a simple fully-connected structure, which is a great improvement above the 91.08% results obtained with recurrent structure and reaches performance comparable to ANNs.",
"discussion": "4. Discussion The incorporation of temporal operations inevitably increases the model's complexities and the analysis of trade-off value. Here, we explore the time and space complexity induced by the TRF and FLI modules in convolutional layers for 3D cases. Assuming the STSC modules are inserted prior to a spatial 2D convolution, the input and output tensor dimensions are R T × C × H × W and R T × C o u t × H × W , and the size of the convolution kernel is O ( K c × K c ). Temporal convolution (Equation 6) needs just O ( K F ) time complexity per element for a TRF module with a K F receptive field, and the total time complexity is O ( T × C × H × W × K F ). For the FLI module with a K G receptive field, each time slot requires a computational complexity of O ( C × C r × K G + C × C r ) = O ( C × C r × ( K G + 1 ) ) , and overall time complexity is O ( T × C × C r × ( K G + 1 ) ) . In contrast to the O ( T × C × C out × H × W ) time complexity required for spatial 2D convolution operations, the O ( H × W × K c × K c ) and O ( T × C × C r × ( K G + 1 ) ) time complexity of TRF and FLI are acceptable. In addition, spatial 2D convolution needs O ( C × C out × K c × K c ) parameters, whereas TRF requires merely O ( C × K F ) parameters and FLI requires O ( C × C r × ( K G + 1 ) ) parameters. In general, the space complexity of TRF is substantially less than that of FLI, and its additional parameters are negligible when compared to 2D convolution; the time complexity of TRF and FLI is relatively efficient in comparison to 2D convolution. Notably, both the TRF and FLI modules are based on the sliding of time windows, and the computations for distinct time frame inputs are identical; thus, there is potential parallelism in the time dimension, and hardware implementation and optimization are possible. In the meanwhile, the computation of STSC-SNN depends on floating-point multiplication, which may reduce the energy efficiency of hardware based on the binary nature of spike transmission. Nevertheless, there is a good reason to believe that binary signals are not a strict constraint for the development of neuromorphic computing, as the carrier (electrical signal or neurotransmitter) used to transmit the spike signal in the biological synapse is not a binary information representing just presence or absence of spike activities (Rothman, 2013 ); in fact, a substantial amount of research has moderately loosened the binary constraint (Shrestha and Orchard, 2018 ; Fang et al., 2020a ; Wu et al., 2021 ; Yao et al., 2021 ; Yu et al., 2022 ; Zhu et al., 2022 ). We believe that with the development of neuromorphic chips, spiking neural networks based on analog circuits and in-memory computing will be capable of surpassing the binary constraints and reconcile the biological plausibility and computational complexity of synaptic operations (Roy et al., 2019 ; Fang et al., 2021a ; Tao et al., 2021 )."
} | 2,832 |
37124285 | PMC10131214 | pmc | 963 | {
"abstract": "Poly(ethylene terephthalate)\npolyester represents the most common\nclass of thermoplastic polymers widely used in the textile, bottling,\nand packaging industries. Terephthalic acid and ethylene glycol, both\nof petrochemical origin, are polymerized to yield the polyester. However,\nan earlier report suggests that polymerization of methoxyterephthalic\nacid with ethylene glycol provides a methoxy-polyester with similar\nproperties. Currently, there are no established biobased synthetic\nroutes toward the methoxyterephthalic acid monomer. Here, we show\na viable route to the dicarboxylic acid from various tree species\ninvolving three catalytic steps. We demonstrate that sawdust can be\nconverted to valuable aryl nitrile intermediates through hydrogenolysis,\nfollowed by an efficient fluorosulfation–catalytic cyanation\nsequence (>90%) and then converted to methoxyterephthalic acid\nby\nhydrolysis and oxidation. A preliminary polymerization result indicates\na methoxy-polyester with acceptable thermal properties.",
"conclusion": "Conclusions In\nsummary, we have shown the first biobased synthesis of methoxyterephthalic\nacid directly from lignocellulose. Our developed multistep procedure\ndisplays high-yielding reaction steps, such as for fluorosulfation,\ncyanation, and hydrolysis, while also a good yield for the final oxidation.\nThe developed conditions for the palladium-catalyzed cyanation, consisting\nof a Pd–XPhos system, display an unusual stability toward cyanide-mediated\ndeactivation, and investigations on the mechanism indicate a favorable\nPd II catalyst resting state and a thermodynamically unfavored\noxidative addition of HCN to Pd 0 –XPhos, which are\npotential reasons for the observed stability. A preliminary polymerization\nreaction showed the possible synthesis of MPET from MTA and ethylene\nglycol, and future work should further optimize the polymerization\nand also focus on a more in-depth characterization of the polymer,\nto evaluate whether it has industrial potential as a new polyester\nmaterial.",
"introduction": "Introduction The global dependency on crude oil for\nthe production of poly(ethylene\nterephthalate) (PET) plastics and fibers was estimated to reach 87.2\nmillion metric tons of PET in 2022. 1 PET\nplastics play a major role in the production of beverage bottles,\nfood packaging, and textiles 2 , 3 because of their highly\ndesirable properties. 4 , 5 As crude oil is a finite and nonrenewable\nresource, the continued manufacturing of PET products will not be\ncompatible with the needs of a rising global population. 6 Furthermore, the end-of-life PET plastic waste\nrepresents a global problem as recycling of this polymer only occurs\nto a low degree, 7 with the majority ending\nup in landfills or incinerated. For example, in the U.S. alone, 57%\nof the PET plastic bottles and containers produced in 2018 were landfilled,\nwhile 14% were incinerated, and only 29% were recycled (see Scheme 1 a). 8 Gasification and pyrolysis represent alternative solutions,\ntransforming the waste into smaller hydrocarbon compounds suitable\nfor heating or as fuels. 9 However, common\nfor these solutions and the direct incineration of plastics, are their\nultimate generation of CO 2 from crude oil, thus exacerbating\nglobal warming. A renewable and carbon-neutral alternative is to replace\ncrude oil with biomass as the starting material in PET production. 10 Some solutions already exist as ethylene glycol\ncan be produced from bio-ethylene via an epoxidation-hydrolysis approach\nor from sugars by a two-step pyrolysis-hydrogenation. 11 For the other monomer, terephthalic acid (TA), only less\nelaborated biobased routes exist. These are multistep procedures that\ncommonly rely on sugar fermentation, leading to a range of small-molecule\ncompounds such as ethylene, 12 isoprene, 13 iso -butanol, 14 and 2,5-dimethylfurfural. 15 Subsequently,\nthese compounds can be transformed to p -xylene, 16 representing the industrial precursor for TA\nin the AMOCO process. 17 Another fermentation\nproduct, trans , trans -muconic acid,\nis also a substrate for TA synthesis, namely, via a Diels–Alder\nreaction followed by dehydrogenation. 18 Scheme 1 Synthesis and Numbers on PET and a Lignin-Based Route to MTA There is currently a high demand for biobased\nPET production, exemplified\nby the company goals of Coca-Cola and Virent, diverting toward biobased\nor recycled PET in the future. 19 Synthetic\nprocedures toward TA also exist, not relying on sugars as the main\nfeedstock, including via limonene, which can be extracted from orange\npeels. 20 , 21 However, a more desirable resource for TA\nsynthesis is to rely on the omnipresent lignocellulose. This natural\npolymer is the most abundant renewable carbon resource, being composed\nof cellulose, hemicellulose, and lignin. 22 While many applications exist for cellulose, there are only few\nuses of lignin, even though it has the potential to be a feedstock\nof the future for aromatic compounds. 22 The main problem stems from the traditional pulping techniques that\neffectively separate the cellulose from lignocellulose, but in the\nprocess delivers a less reactive lignin polymer byproduct. 23 More modern techniques such as oxidative and\nreductive catalytic fractionation (OCF and RCF) are on the contrary\nable to generate a useful lignin product stream, which is composed\nof methoxy-substituted phenols (see Scheme 1 b). 23 − 26 The highest yields are achieved with RCF, which depolymerizes\nlignocellulose under catalytic hydrogenative conditions applying a\nheterogeneous catalyst and a polar protic solvent, such as MeOH or\nH 2 O. 23 , 25 , 26 Furthermore, applying hydrogen as a reagent aligns amiably with\nmodern power-to-hydrogen strategies. 27 The\nRCF method is believed to become a major industrial process in the\nfuture; 25 thus, a synthetic route toward\nTA from lignin RCF products could be highly desirable. Previously,\ntwo multistep routes from lignin to TA have been disclosed\nby the groups of Zhu and Yan. 28 , 29 In the first, TA was\nprepared from vanillic and syringic acid from the oxidative degradation\nof lignin. However, these two products only account for 5 wt % of\nall of the oxidation products. 16 Furthermore,\nharsh reaction conditions were necessary to overcome the challenging\ndemethoxylation and carboxylation steps, which are both conducted\nat a minimum of 400 °C and a pressure of 40 bar CO 2 /H 2 using noncommercial catalysts. 28 In the subsequent work by Yan et al., a reductive catalytic\nfractionation was developed, providing significantly higher yields\nof lignin monomer products. 29 The challenges\nof this multistep protocol also lie in the harsh conditions necessary\nfor a productive demethoxylation step, operating at 320 °C with\na pressure of 30 bar H 2 that results in a 71% yield, while\nthe remainder consists of undesired side-products. 29 Furthermore, the subsequent triflation of the RCF phenol\nproducts suffers from the use of reactive and expensive triflic anhydride,\nwhich is not compatible with a crude lignin product mixture that may\nhold aliphatic alcohol functionalities and water. 30 In a patent from 1959, the General Electric Company\n(GE) reported\nthe synthesis of polyesters including methoxy poly(ethylene terephthalate)\n(MPET) from methoxyterephthalic acid (MTA) and ethylene glycol. 31 However, these materials have been largely overshadowed\nby conventional PET. This is explained by the conventional and direct\nproduction of TA from crude-oil-derived p -xylene. 4 , 17 With a future potentially relying on renewable lignocellulose as\none resource for creating polyester materials, the methoxylated feedstock\ncould represent an advantageous starting point for preparing a biobased\nMTA, which could be useful for an eventual MPET production. Here,\nwe envisage a route from the methoxy-phenol products of RCF to MTA\napplying a two-step fluorosulfation, catalytic cyanation strategy\nwith industrially viable reagents. MTA is then reached from the aryl\nnitrile cyanation products by two well-described industrial processes,\nafter which a polymerization with ethylene glycol is able to yield\nMPET (see Scheme 1 c).",
"discussion": "Results\nand Discussion To explore the efficacy of the proposed route\nfor synthesizing\nMTA from lignin, we collected lignocellulose samples (sawdust) from\neight different tree species and subjected them to the reported RCF\nconditions by Sels and co-workers, 32 which\nwas previously shown to effectively depolymerize various lignocelluloses\nvia the Ru/C-catalyzed hydrogenolysis reaction under a hydrogen atmosphere\n(30 bar) in MeOH at a temperature of 250 °C. The resulting methoxy-phenol\nproducts were obtained in up to 75 mg/g of sawdust (for maple sawdust,\nsee p. S14 of the SI), which corresponds\nto 27 wt % with respect to the lignin weight. This is acceptable with\nrespect to the results obtained by Sels and co-workers 32 and other main contributors in the RCF field,\nsuch as Beckham, Román-Leshkov, and co-workers (regarding results\nwith Ru/C and various conditions and tree species, 21–32 wt\n% monomer yields). 33 − 35 However, it should also be noted that the monomer\nyields depend to a great extent on the specific lignocellulose composition\nof the wood sample, which can vary according to the age, growth period,\nand home region. 36 , 37 The obtained phenol products\nwere subsequently converted to aryl nitriles via sequential fluorosulfation\nand catalytic cyanation (see Scheme 2 a). Sulfuryl fluoride is the ideal reagent for the\nactivation of the RCF phenol fraction because of its industrial-scale\nproduction of 3000 tons/year and its use for phenol activation while\nstable toward hydrolysis (see Scheme 2 b). 38 , 39 Recently, related C–O\nactivation strategies have been reported for lignin phenol monomers\nin the context of electrochemical reductive coupling and phenol deoxygenation. 40 , 41 The specific RCF phenol products, 1 and 2 , have not before been subjected to fluorosulfation; however, gratifyingly,\nwe observed quantitative yields of the corresponding aryl fluorosulfates\nwhen the reaction was conducted with ex situ generated sulfuryl fluoride\nin a two-chamber reactor system (see Scheme 2 b). 42 The selection\nof K 2 CO 3 instead of the widely used NEt 3 as base for fluorosulfation 39 , 42 does not only\ngive quantitative yields of the reaction but is also more viable from\nan industrial viewpoint, representing a less expensive and greener\noption. 43 − 45 Furthermore, the use of a heterogeneous base facilitates\npurification through simple filtration. Scheme 2 Three-Step Route\nto Aryl Nitriles via Reductive Catalytic Fractionation,\nFluorosulfation, and Cyanation Fluorosulfation reactions\nwere\nconducted on a scale of 5–10 mmol in a two-chamber reactor\nsystem using ex situ generated sulfuryl fluoride\n(see pp. S4–S5 of the SI). The palladium-catalyzed\ncyanation of aryl fluorosulfates was optimized on a 0.2–0.4\nmmol scale (see the full optimization in Tables S1–S3 in the SI). All of the presented DFT results were\ncalculated at the PBE-D3/Def2-TZVPP//PBE-D3/Def2-SVP level of theory\nwith SMD (acetonitrile). The ensuing catalytic\ncyanation step takes advantage of the excellent\nreactivity of aryl fluorosulfates in transition-metal-catalyzed cross-coupling\nreactions, being generally viewed as aryl triflate surrogates. 39 , 42 Initially, we explored the industrial source of cyanide, HCN, with\na two-chamber setup, 46 in a palladium-catalyzed\ncyanation of aryl fluorosulfates 3 and 4 relying on XPhos as the ligand. Quantitative yields of both aryl\nnitriles 5 and 6 were achieved (see Supporting Tables S1 and S2 ). As deactivation\nof palladium catalysts has been reported from the presence of minute\nquantities of HCN for the cyanation of aryl electrophiles, 47 , 48 we were surprised by the efficiency of this protocol, which suggests\nan unusual stability for the Pd–XPhos complexes. Subsequently,\nit was discovered that both KCN and NaCN in lieu of HCN rendered equally\nquantitative transformations (see Scheme 2 c and Supporting Table S3 ), being more desirable from a safety perspective. Through\noptimization, we found that the successful transformation of aryl\nfluorosulfates 3 and 4 necessitates 0.5–2\nmol % of a Pd–XPhos catalyst generated from the Buchwald G4-precatalyst 49 using industrially viable solvents such as MeCN\nand biomass-derived 2-MeTHF. 50 Furthermore,\nfor the previously reported cyanation reactions with aryl fluorosulfates,\nlower yields are observed with sterically hindered ortho -methoxy-substituted substrates. 51 , 52 However, the\nPd–XPhos catalyst employed in this transformation reacts amiably\nwith even the dimethoxy-substituted substrate, 4 (see Scheme 2 c). Another strong\npoint of this reaction is clearly illustrated by its ability to tolerate\nsmall amounts of water (see Scheme 2 c), which can be detrimental to palladium-catalyzed\ncyanation reactions as illustrated by the seminal work from the groups\nof Beller, Grushin, and Macgregor. 47 , 48 We postulate\nthat the addition of K 2 CO 3 is\ncrucial to the reaction, as it serves to deprotonate and remove HCN\nformed in situ from the reaction of KCN and water. Trace HCN has been\nreported to deactivate palladium catalysts by oxidative addition,\nyielding off-cycle H–Pd II –CN and eventually\ninactive palladium(II) cyanide complexes. 47 , 48 In the presence of K 2 CO 3 , reductive elimination\nof HCN from the H–Pd II –CN complex becomes\nfavored by an ensuing deprotonation of the HCN. Furthermore, we were\nable to support our claim on the effect of K 2 CO 3 being more than just a dehydration agent of the solvent mixture,\nas the same reaction with an equal amount of the drying agent, Na 2 SO 4 instead of carbonate resulted in no conversion\nwhen conducted in the presence of 5 equiv of H 2 O (see Scheme 2 c). Furthermore,\nwe hypothesize that the success of 2-MeTHF and MeCN as a solvent mixture\ncould be due to an ideal solubility of cyanide, which has been suggested\nbefore for the Pd-catalyzed cyanation of aryl bromides using a similar\nTHF/MeCN solvent mixture. 53 A mechanism\nfor the catalytic cyanation reaction is depicted in Scheme 2 d, which follows\nthe generally accepted catalytic cycle for oxidative addition ( MC4 – MC5 ), cyanation ( MC5 – MC2 ), and reductive elimination ( MC2 – MC3 ). 47 , 48 Additionally, it is likely that\na solvent-coordinated complex S is formed after the dissociation\nof fluorosulfate counterion from MC5 , which is supported\nby the obtained crystal structure for S (see Supporting Figure S21 ). Energies of intermediates\nand barriers in the catalytic cycle were determined with DFT calculations\n(PBE-D3,SMD, see Supporting Figures S11 and S13 ). The calculations identify MC2 – MC3 (reductive elimination) as the rate-limiting step for the reaction\nwith 3 (Δ G ‡ =\n10.5 kcal mol –1 ), and MC4 – MC5 (oxidative addition) for 4 (Δ G ‡ = 15.4 kcal mol –1 ). The overall reaction free energy for the catalytic cyanation with 3 was found to be −58.7 kcal mol –1 (343.15 K), while for 4 it was −60.1 kcal mol –1 (343.15 K). To better elucidate the underlining reasons\nfor the stability of the catalytic system against HCN deactivation,\nfurther DFT studies were undertaken. From the Pd 0 -intermediate MC4 , oxidative addition of HCN leads to MC6 of\nthe deactivation pathway. From the calculations, it is evident that\nthe conversion of MC6 – MC7 with HCN\nis plausible due to an energy barrier of Δ G ‡ = 9.3 kcal mol –1 . However,\nthe reverse reaction ( MC7 – MC6 ) has\na lower barrier (Δ G ‡ = 7.5\nkcal mol –1 ) and is thus likely to occur as well.\nFor comparison, we calculated the energies for the same deactivation\nmechanism with the Pd–P( t Bu) 3 system\nreported by Grushin and co-workers (see Scheme 2 d). 53 This system\nis known to be effective for the cyanation of aryl bromides with KCN\nbut requires dry conditions to circumvent HCN formation from trace\nwater. Our calculations for the Pd–P( t Bu) 3 system revealed a higher vulnerability toward the HCN deactivation\npathway as the back reaction from MC7″ – MC6″ has an energy barrier, being twice the height\nof our system (Δ G ‡ = 15.1\nvs Δ G ‡ = 7.5 kcal mol –1 ). Furthermore, MC7″ is −3.2\nkcal mol –1 lower in energy compared to MC6″ , thereby providing a thermodynamic driving force for the formation\nof MC7″ . Another postulation as to why our reported\nsystem is less susceptible to HCN deactivation is that the catalyst\nresting state is likely to resemble the Pd II –complex S or MC5 , as deduced by monitoring the progression\nof the reaction with time-resolved 31 P–NMR spectroscopy\n(see Supporting Figure S8 ). With\nthe optimized fluorosulfation and cyanation conditions at\nhand, we explored the sequential transformation of sawdust originating\nfrom different tree species to the corresponding aryl nitriles. By\nqualitative observations made with GCMS, we assessed that the fluorosulfation\nstep required 3 equiv of sulfuryl fluoride and K 2 CO 3 each for a full conversion in all cases. In the case of the\ncyanation reaction, a slightly higher catalyst loading of 5 mol %\nand 3 equiv of K 2 CO 3 were required for high\nconversion with all tree species (see Scheme 3 ), although eventual industrialization of\nthe process would require a significantly lower catalyst loading.\nFrom exploring the scope of sawdust samples, best yields were obtained\nwith hardwoods, such as birch, willow, oak, hazel, and maple, compared\nto softwood species including spruce, grandis, and noble fir, which\nis in line with the general observations made in earlier studies on\nRCF. 25 , 32 However, interestingly, with softwood in\ncontrast to the hardwood samples, the 2-methoxy-substituted phenol\nproduct was exclusively isolated, being precisely the precursor required\nfor MTA synthesis. 31 This product distribution\nis accounted for by the different composition of H-, G-, and S-lignin\nsubunits present in softwoods versus hardwoods. 22 , 23 As such, these observations obviate eventual separation of nitriles 5 and 6 when relying on softwoods, and as this\ntree sort is generally faster growing than the corresponding hardwoods, 54 this could be advantageous from an industrial\nviewpoint. A scale-up experiment (10 g) was also conducted with noble\nfir sawdust, and we observed an acceptable 6.1 wt % yield of the aryl\nnitrile 5 , which is, however, a lower yield in comparison\nwith the small-scale yield (9.1 wt %) due to not achieving full conversion\nin the cyanation step. Instead, if 5 was purified before\nthe cyanation reaction, we could observe quantitative isolated yields\non 2–10 mmol scales with a low catalyst loading of 0.5–2\nmol % (see Table S3 in the SI). With the\nisolated 2-methoxy-substituted aryl nitrile from the various sawdust\nsamples studied, the corresponding carboxylic acid product was finally\nproduced by an initial hydrolysis with NaOH ( Scheme 4 a). Subsequently, the conditions of the AMOCO\nprocess 17 , 29 promoted the synthesis of MTA by oxidation,\nand the product as a colorless solid was isolated in high purity as\ndeduced from NMR spectroscopic analysis. Besides the AMOCO process,\nthere are many other excellent strategies toward C(sp 3 )–H\noxidation that have been recently reviewed. 55 − 57 Furthermore,\nthe research group of Maes also studied the benzylic oxidation of\nlignin-derived propylguaiacol and syringol by using Na 2 S 2 O 8 in order to obtain aryl ketones. 58 An alternatively interesting pathway for reaching\nMTA involves a carboxylation strategy, which we showcased from spruce\nsawdust using a nickel-catalyzed reductive carboxylation system inspired\nby the work of Mei and co-workers. 59 Nonetheless,\na major limitation to this strategy is the necessity of a stoichiometric\nreductant as exemplified by the need for manganese metal. Scheme 3 Scope of\nthe Three-Step Synthesis of Aryl Nitriles Directly from\nSawdusts Lignin weight percentages\nwere\ndetermined by the Klason lignin method and are based on triplicate\nresults (see Supporting Figure S3 and Table S4 ). Mass and wt % yields for aryl nitrile products 5 and 6 are given as averages of duplicate reactions and are reported\nrelative to the mass of lignin (see the Supporting Information pp. S12–S15 ). SO 2 F 2 was generated ex situ from SDI (1,1′-sulfonylbis(1 H -imidazole)) and potassium fluoride in TFA. The large-scale\nreaction with noble fir required Pd–XPhos–G4 (6.5 mol\n%), K 2 CO 3 (3.9 equiv), and KCN (2.0 equiv).\nSee the General Procedures section pp. S11–S12 of the SI. Qualitative assessment by 1 H- and 13 C-NMR indicates a high purity of the respective compounds 5 and 6 after purification by flash column chromatography,\nhowever with the co-elution of minute amounts of ethyl-substituted\naryl nitriles (see the SI page S12 ). Purification\nmethods used: Step 1 (filtration through celite, then filtration through\nsilica), Step 2 (filtration through celite), Step 3 (flash column\nchromatography); see the General Procedures A and B, pp. S11–S12 . Scheme 4 Synthesis and Polymerization\nof MTA Experimental procedure\nfor nitrile\nhydrolysis (see the SI, pp. S16–S17 ), oxidation (see the SI, pp. S17–S18 ), carboxylation (see the SI, p. S17 ),\nand polymerization (see the SI, p. S18 ).\nThe 2.4 mmol scale hydrolysis of 5 to 7 was\nconducted with 5 obtained from noble fir, and the following\noxidation step was conducted with the noble fir-derived 7 . The MTA used for the polymerization was obtained from a reported\nprocedure. 61 Qualitative assessment by 1 H- and 13 C-NMR analysis of the hydrolysis, oxidation,\nand carboxylation steps indicate a high purity of the respective compounds 7 and 8 . Purification methods used: hydrolysis\nstep (acid–base wash), oxidation step (decantation with AcOH,\nH 2 O, and pentane), carboxylation step (acid–base\nwash), polymerization (no purification). The glass-transition temperature\nwas evaluated with differential scanning calorimetry (DSC), while\nthe decomposition temperature was analyzed with thermogravimetric\nanalysis (TGA) (see the SI, Figures S6 and S7 ). The error was measured to be ±1.0 °C for DSC analysis\nand ±1.7 °C for TGA. Lastly, we\nattempted a preliminary un-optimized synthesis of MPET\nvia the depolymerization of MTA with ethylene glycol. The reaction\nconditions were inspired from that described in the patent, 31 and they deviate slightly from standard PET\nsynthesis, in which catalysts such as antimony oxides or titanium\nalkoxides are frequently applied. 60 MTA\nwas first reacted with 3 equiv of ethylene glycol at 220 °C for\n18 h in a closed system under an argon atmosphere. The subsequent\npolymerization step was conducted under vacuum, heating at 220 °C\nfor 4 h, then at 270 °C for 3.5 h, and finally at 310 °C\nfor 18 h, to yield MPET as a dark polymer, which was characterized\nby infrared spectroscopy (IR), thermogravimetric analysis (TGA), and\ndifferential scanning calorimetry (DSC) (see Supporting Figures S5–S7 ). The polymer product displayed a glass-transition\ntemperature ( T g ) of 59 °C and a decomposition\ntemperature ( T d ) of 407 °C, indicating\nacceptable thermal properties. Future work should include further\noptimization of MPET polymerization, while also expanding the characterization\nto include mechanical properties, crystallinity, gas permeability,\nchemical resistance, and rheological parameters."
} | 5,856 |
26781463 | PMC4717592 | pmc | 964 | {
"abstract": "Background Metagenomics allows us to acquire the potential resources from both cultivatable and uncultivable microorganisms in the environment. Here, shotgun metagenome sequencing was used to investigate microbial communities from the surface layer of low grade copper tailings that were industrially bioleached at the Dexing Copper Mine, China. A bioinformatics analysis was further performed to elucidate structural and functional properties of the microbial communities in a copper bioleaching heap. Results Taxonomic analysis revealed unexpectedly high microbial biodiversity of this extremely acidic environment, as most sequences were phylogenetically assigned to Proteobacteria , while Euryarchaeota -related sequences occupied little proportion in this system, assuming that Archaea probably played little role in the bioleaching systems. At the genus level, the microbial community in mineral surface-layer was dominated by the sulfur- and iron-oxidizing acidophiles such as Acidithiobacillus -like populations, most of which were A. ferrivorans -like and A. ferrooxidans -like groups. In addition, Caudovirales were the dominant viral type observed in this extremely environment. Functional analysis illustrated that the principal participants related to the key metabolic pathways (carbon fixation, nitrogen metabolism, Fe(II) oxidation and sulfur metabolism) were mainly identified to be Acidithiobacillus -like, Thiobacillus -like and Leptospirillum -like microorganisms, indicating their vital roles. Also, microbial community harbored certain adaptive mechanisms (heavy metal resistance, low pH adaption, organic solvents tolerance and detoxification of hydroxyl radicals) as they performed their functions in the bioleaching system. Conclusion Our study provides several valuable datasets for understanding the microbial community composition and function in the surface-layer of copper bioleaching heap. Electronic supplementary material The online version of this article (doi:10.1186/s12863-016-0330-4) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusions The properties of environments, especially extremely acidic, oligotrophic and heavy metals containing bioleaching heap discussed in this study, shape the microbial community composition and function. Metabolic activities occur in the microbial community, conferring the role as a recycler of substance circulation in the bioleaching system and even in nature. Besides, whether environmental microorganisms harbor a suit of genes involving the response mechanisms is probably as a determinant factor to adapt the particular environmental conditions. Microorganisms in the extremely acidic environments have to cope with environmental stresses to survive and proliferate, before they can perform their functions in the bioleaching system.",
"discussion": "Results and discussion Sequencing, de novo assembly, gene prediction and functional annotation Metagenomic DNA was subjected to Illumina MiSeq sequencing, and approximately 3.4 million short DNA sequences were then used for bioinformatics analysis. After quality control using NGS QC Toolkit, 2,941,297 (87.80 %) reads with high-quality were obtained (Additional file 1 ). Subsequently, all high-quality reads aforementioned were assembled, and a self-writing script was used to filter the assembled sequences under 300 bp, resulting in a total of 301,907,459 bases, with an N50 of 641 bp (481,688 contigs range from 301 bp to 49,868 bp, and the mean length was 626 bp). For gene prediction, 660,572 coding sequences (CDS) were identified using the program MetaGeneAnnotator. All putative protein coding sequences were searched against the databases including NCBI-nr, the extended COG [ 28 ] and KEGG, and we obtained a total of 535,887 (81.12 %), 517,948 (78.41 %) and 494,721 (74.89 %) significant BLAST hits respectively. Moreover, 497,601 (75.33 %) and 261,595 (39.60 %) sequences were assigned to the COG categories and KEGG Orthology respectively (Additional file 1 ). Among the 25 COG categories, metagenome sequences were assigned to 23 of them (Fig. 1 ). A large proportion of sequences were assigned to COG category [S] (function unknown) (80,561 CDSs; 16.19 %) and COG category [R] (General function prediction only) (39,507 CDSs; 7.94 %), indicating large pools of potential unknown functional genes in copper bioleaching operations. Furthermore, the large amount of genes associated with basal metabolisms such as amino acid transport and metabolism (COG category [E]) and energy metabolism (COG category [C]) indicated the ubiquitous substance and energy metabolism in the extremely environments, maintaining the basic microbial activities. Fig. 1 The COGs categories of metagenome data from mine tailings Taxonomic assignment of metagenome datasets To reveal metagenome sequence classification of microbial communities in tailings sample, taxonomic analyses at the genus level were performed. Taxonomic assignment using the program MEGAN revealed unexpectedly abundant microbial biodiversity (over 100 genera) of this extreme environments (surface-layer of copper mine tailings), to some extent, which hindered the sequence assembly due to the low sequencing depth. Copper mine tailings in this study harbored diverse microbial populations possibly because of various niches related to gradients of physico-chemical conditions, which was discussed previously in AMD environments [ 2 , 29 – 31 ]. MEGAN analysis showed that the microbial community in mineral surface-layer was dominated by the sulfur- and iron- oxidizing acidophiles Acidithiobacillus -related and Leptospirillum -related groups (Fig. 2 ). In these Acidithiobacillus -related sequences, most of them were assigned to Acidithiobacillus ferrivorans , followed closely by A. ferrooxidans . In the extremely acidic tailings, approximately 93.47 % of the total Acidithiobacillus -related sequences were affiliated with A. ferrivorans and A. ferrooxidans (Additional file 2 ). As a major participant of iron- and sulfur-oxidizing acidophilic bacteria, A. ferrivorans has been widely found in metal mine-impacted environments [ 32 ]. Likewise, A. ferrooxidans , which utilized energy from the oxidation of sulfur- and iron-containing minerals, was a principal member in consortia of microorganisms associated with the bioleaching or biomining (industrial recovery of copper) [ 33 ]. The numerical dominance of Acidithiobacillus -related sequence indicated its importance in surface-layer of copper mine tailings during the industrial bioleaching operations. Moreover, Rhodanobacter (7.34 %), Thiobacillus (6.03 %), Leptospirillum (5.57 %), and Acidiphilium (4.51 %) were also found in the surface-layer mine tailings. In addition, 82 CDSs were assigned to virus, most of which were affiliated with the dsDNA viruses with no RNA stage. Of these sequences, the majority of taxonomic hits (74 %) shared sequence identity with sequences in the order Caudovirales , based on the taxonomy of viral genomes provided by GenBank database (Additional file 3 ). This was consistent with the viruses previously described from the desert [ 34 , 35 ] and other metaviromes from other environments such as marine environment [ 36 ]. Fig. 2 Taxonomic composition analysis at the genus level based on contigs sequences (≥ 300 bp) in the metagenome dataset. Only those genera with the specified percentage abundance (≥ 1 %) are shown Depend on the automated analysis pipeline implemented in the MG-RAST platform, the microbial populations at the phylum level were phylogenetically assigned to the Proteobacteria , Actinobacteria , Nitrospirae , Bacteroidetes , Gemmatimonadetes , Acidobacteria , Firmicutes , Deinococcus - Thermus , Euryarchaeota , and several other phyla mainly belonged to the domain Bacteria (Additional file 4 ). In more detail, Proteobacteria -related sequences with the most abundance were composed of the class Gammaproteobacteria , Betaproteobacteria , Alphaproteobacteria , Deltaproteobacteria , Epsilonproteobacteria and Zetaproteobacteria in an order from the highest to the lowest. Similarly, community diversity analysis based on a PCR-based cloning approach showed that the majority of sequenced clones were affiliate with the Gammaproteobacteria [ 37 ]. As the most abundant microbes were similar to Acidithiobacillus -like genus, it was proposed to belong to the new class Acidithiobacillia (a sister group of class Gammaproteobacteria ) [ 38 ]. Thus, the most abundant sequences at the class level in this extreme environment could be assigned to the Acidithiobacillia -related microorganisms. The phylum Euryarchaeota occupied the largest proportion in domain Archaea . However, it was relatively low in the whole metagenome dataset, suggesting that the Archaea might play little role in the surface-layer of copper mine tailings. Key genes coding for enzymes associated with principal metabolisms The vital activities of chemolithotrophy-based microbial community present in the mine tailings mainly rely on metabolic capabilities to metabolize carbon, nitrogen, iron and sulfur. Thus, it is necessary to investigate the general metabolisms of microbial processes, aiming to understand the sub-cycling of those elements within a copper bioleaching heap. Autotrophic carbon fixation Cellular carbon acquired from inorganic carbon is essential for life, suggesting the transition of carbon from inorganic to organic world. Recent research revealed that six different pathways for carbon fixation existed in microorganisms, including Calvin–Benson–Bassham (CBB) cycle, reductive citric acid (rTCA) cycle, reductive acetyl-coenzyme A (acetyl-CoA) pathway, 3-hydroxypropionate bicycle, 3-hydroxypropionate/4-hydroxybutyrate cycle (hydroxypropionate–hydroxybutyrate cycle) and dicarboxylate–4-hydroxybutyrate cycle (shortened to the dicarboxylate–hydroxybutyrate cycle) [ 2 , 39 ]. Functional annotation against databases, i.e., NCBI-nr, the extended COG and KEGG, showed that approximately all genes encoding for CBB cycle and rTCA cycle were identified, whilst no gene involved in the other four pathways for carbon fixation was identified. This finding was largely consistent with previous studies [ 2 ]. In CBB cycle, there are a series of enzymatic reactions, one of which is involved in CO 2 fixation catalyzed by ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) and the others are responsible for the regeneration of ribulose 1,5-bisphosphate (RuBP) [ 40 ]. In addition, the key enzymes for CBB cycle are Rubisco and phosphoribulokinase (or ribulose-5-phosphate kinase) [ 39 , 41 ]. Rubisco exists in various forms in diverse organisms from all domains of life [ 42 ]. Our results indicated that genes coding for Rubisco were enriched in the metagenome, most of which were affiliated with Acidithiobacillus -like populations (Additional file 5 ). Indeed, genomic as well as proteomic evidence showed that Acidithiobacillus occupied an important position in carbon fixation in AMD [ 43 ]. In addition, the majority of phosphoribulokinase (PRK) genes were from Thiobacillus -like microorganisms, indicating its importance in the bioleaching system. However, no gene encoding the sedoheptulose-1,7-bisphosphatase (SBPase), which catalyzed sedoheptulose 1,7-bisphosphate to generate sedoheptulose 7-phosphate, was found in the metagenome (Additional file 5 ). The lack of such genes was likely due to the low sequencing depth. Besides, the candidate genes that probably presented in this metagenome dataset were not yet identified because of our limited knowledge of SBPase genes within the acidophiles. Autotrophic CO 2 fixation via rTCA cycle was considered to be an important pathway in microbial communities [ 44 ]. There were two key enzymes (2-oxoglutarate ferredoxin oxidoreductase and ATP citrate lyase) related to rTCA cycle [ 39 ]. In some species, the citrate cleavage pathway could be catalyzed by other two enzymes, citryl-CoA synthetase (EC6.2.1.18) and citryl-CoA lyase (EC4.1.3.34), instead of ATP citrate lyase (ACL; EC2.3.3.8) [ 45 ]. These two enzymes, i.e., citryl-CoA synthetase and citryl-CoA lyase, however, were phylogenetically related to ACL [ 46 – 48 ]. Furthermore, research indicated that ACL was formed by a gene fusion of citryl-CoA lyase and citryl-CoA synthetase [ 48 ]. These findings could explain why no genes encoding ACL were identified in previous studies about lead/zinc mine tailings [ 49 ]. In Leptospirillum ferriphilum , a complete set of enzymes involved in rTCA cycle were found [ 40 ]. Based on the aforementioned research, our results in this study showed that gene homologs for many steps in the rTCA cycle were assigned to the L. ferriphilum -like populations (Additional file 5 ). As for citryl-CoA synthetase, however, the small subunit gene ( ccsB ) was not detected. Nitrogen metabolism As the main nitrogen sources, generally speaking, atmospheric nitrogen, nitrate, nitrite, ammonium and glutamine are used wildly by microbes in natural environments. Six subsystems related to the nitrogen metabolisms, including nitrogen fixation, dissimilatory nitrate reduction, assimilatory nitrate reduction, denitrification, nitrification and anammox, were discussed in previous papers [ 50 , 51 ]. In this metagenome, a large number of genes involved in nitrogen fixation including nifD , nifK , nifH , nifE, nifN and nifX , which formed a nif operon encoding the Mo-Fe nitrogenase enzyme complex [ 52 ] , were detected (Additional file 5 ). The relevant genes were largely found in Leptospirillum -like and had the highest similarity to those from L. ferrooxidans -like species, suggesting the vital position of Leptospirillum -like organisms in bioleaching systems with limited fixed nitrogen from external sources. Nitrogen in the form of ammonium could be either directly assimilated into biomass or be oxidized by two key enzymes involved in the nitrification (ammonium monooxygenase and hydroxylamine oxidoreductase, which were encoded by amoCAB operon and hao gene, respectively) [ 52 ]. In this metagenome, however, the lacked hao gene suggested that ammonium could not be utilized via the nitrification. Mine tailings might contain elevated concentrations of nitrate caused by the nitrogen-based explosives. As for dissimilatory nitrate reduction and assimilatory nitrate reduction, most genes encoding metabolic enzymes were related to those from Thiobacillus -like, Acidithiobacillus -like and Leptospirillum -like populations (Fig. 3 ). Given that there was limited fixation nitrogen from atmosphere and relatively low content of available nitrogen in mine tailings, it was speculated that these microorganisms probably played an important role in maintaining the nitrogen sources of microbial communities via utilizing nitrate and nitrite. Fig. 3 Schematic diagram of nitrogen metabolism in the surface-layer mine tailings representing one important part of bioleaching system. On the basis of metagenomic data, solid lines indicate the presence of protein-coding genes associated with six major pathways, whereas dashed lines show that no gene was found in this metagenome. Different colored lines depict various metabolic pathways. Most main implicated taxa/groups governing the enzymatic reactions are displayed respectively, and the percentages of CDSs related to each group are also showed. Abbreviations: nif , nitrogenase (various subunits); amo : ammonia monooxygenase; hao : hydroxylamine dehydrogenase; norAB : Nitrate reductase (A and B represent alpha subunit and beta subunit, respectively); narGHIJ : nitrate reductase (dissimilartory); nirK : nitrite reductase (NO-forming); norBC : nitric oxide reductase; nosZ : nitrous oxide reductase; nasAB : nitrate reductase (assimilatory); nirA : ferredoxin-nitrite reductase; nirBD : nitrite reductase (NADH); hzo : hydrazine oxidoreductase Four enzymes are reported to be involved in denitrification i.e., nitrate reductase, nitrite reductase (NO-forming), nitric oxide reductase (cytochrome c) and nitrous oxide reductase, which are encoded by nar operon, nirK , norBC and nosZ , respectively [ 2 , 53 ]. In this process, nitrate or nitrite ions could be used as the terminal electron acceptors under anoxic or low-oxygen conditions [ 52 ]. In addition, the incomplete ammonification in this system, which lacked the relevant genes encoding hydrazine oxidoreductase (Fig. 3 ), presumably indicated that organic nitrogen compounds in the environments could not be degraded by microbes in the bioleaching system. Ferrous iron oxidation Under acidic conditions, ferrous iron in a stable status (no matter whether atmospheric oxygen exists or not) could be utilized by microorganisms [ 11 ]. Widespread distribution of ferrous iron oxidation capabilities has been observed in acidophilic Bacteria and Archaea [ 54 – 56 ]. The ability to oxidize iron has been widely distributed in microbes from neutral pH environments. So far, details of ferrous iron oxidation and electron transport pathways were available for the Gram-negative bacterium A. ferrooxidans that was identified until recently to be affiliated with the class Acidithiobacillia [ 38 ]. Besides, iron oxidation pathways in other known acidophilic prokaryotes (e.g., A. ferrivorans, L. ferrooxidans and Thiobacillus prosperus , which was reclassified as Acidihalobacter prosperus recently [ 57 ]) have been also studied [ 11 ]. The results showed that there were no relevant genes for the pio operon ( pioA , pioB , and pioC ) or the fox operon ( foxE , foxY and foxZ ) in the metagenome, which were reported to be involved in the phototrophic iron oxidation in Rhodopseudomonas palustris and Rhodobacter capsulatus respectively [ 2 , 58 ]. The reasons why the genes for ferrous iron oxidation were absent in acidophilic microbes were probably (i) due to the low abundance of microorganisms responsible for iron oxidation, resulting in the difficulty in the retrieval of relevant genes from metagenomeand (ii) due to limited knowledge of iron oxidation in microorganisms [ 2 , 11 ]. Genes for the redox proteins involved in Fe(II) oxidation in acidophiles (such as outer membrane cytochrome c or Cyc2, blue copper proteins rusticyanin and Cyc1, etc.) [ 2 , 11 , 58 , 59 ] were observed in this study (Fig. 4 ). The results indicated that rusticyanin genes ( rus ) existed in the metagenome, most of which were assigned to A. ferrooxidans -like species. Moreover, it also exhibited genes encoding iron oxidase in A. ferrivorans -like, though their number was not enriched. Thus, there were at least two different pathways for ferrous iron oxidation in the iron-oxidizing Acidithiobacillus spp.: (i) via rusticyanin (ii) or via high potential iron sulfur protein (HiPIP) encoded by the gene iro [ 11 ]. In acidophilic archaea, Fe(II) oxidation pathways were also proposed [ 11 , 52 , 60 ], indicating that electrons from Fe(II) oxidation by unknown mechanisms were transferred via sulfocyanin to a cbb 3 -type terminal oxidase (Fig. 4 ). In the tailings sample, only one sulfocyanin gene in the archaean genus Ferroplasma -like was identified (Additional file 5 ). Given the low relative abundance of Archaea in surface-layer of mine tailings, it was supposed that Archaea might play little role in the Fe(II) oxidation within the mine tailings. Fig. 4 Overview of the main known metabolic abilities (carbon fixation, ferrous iron oxidation and sulfur metabolism) of microbial community and environmental adaption in surface-layer mine tailings. This figure was adapted from the previous models [ 26 , 40 , 52 , 61 ]. All possible subsystems are depicted in the quarters of each image. In the CO 2 fixation, enzymes associated with rTCA cycle are indicated by numbers: 1, malate dehydrogenase; 2, fumarate hydratase; 3, fumarate reductase; 4, succinyl-CoA synthetase; 5, 2-oxoglutarate ferredoxin oxidoreductase; 6, isocitrate dehydrogenase; 7, aconitase hydratase 1; 8, citryl-CoA synthetase; 9, citryl-CoA lyase; 10, pyruvate ferredoxin oxidoreductase. The enzymes related to nitrogen metabolism, ferrous iron oxidation and sulfur metabolism as abbreviated forms are depicted. Abbreviations: R: Rusticyanin; SQR: sulfide quinone reductase; TQO: thiosulfate:quinone oxidoreductase; TetH: tetrathionate hydrolase; SOX: sulfur oxidizing protein; HDR: heterodisulfide reductase; SOR: sulfur oxygenase reductase; TST: thiosulfate sulfurtransferase; SO: sulfite oxidase; APR: adenylylsulfate reductase; SAT: sulfate adenylyltransferase; CysC: adenylylsulfate kinase; CysH: phosphoadenosine phosphosulfate reductase; CysJI: sulfite reductase (NADPH) flavoprotein; SIR: sulfite reductase (ferredoxin); DsrAB: sulfite reductase Sulfur metabolism Recently, there are a number of known enzymes that are related to the sulfur metabolism by microoorganisms [ 61 – 63 ]. Undoubtedly, the transformation of elemental sulfur with various oxidation states is more complicated than iron metabolism. In surface-layer of bioleaching heap, the genes encoding sulfur metabolic enzymes have not yet been elaborated. In this study, a total of 2479 CDSs were found to be linked to sulfur metabolism, mapping to 48 diverse genes (Additional file 5 ). Sulfur and reduced inorganic sulfur compounds (RISCs) might accumulate in areas where pyrite and other types of sulfide minerals were oxidized by Fe(III) ion [ 64 ]. Thus, the energetically favorable substrates could be utilized by acidophilic sulfur-oxidizing bacteria [ 62 , 63 ]. Microbial oxidation of sulfur or sulfur-compounds requires different enzymatic machineries, including sulfide quinone reductase (SQR), thiosulfate:quinone oxidoreductase (TQO) and tetrathionate hydrolase (TetH) (Fig. 4 ). As an intermediate forms during the RISC oxidation, sulfide was oxidized by SQR, the coding genes of which were enriched in the metagenome and were most closely identified as those of Acidithiobacillus -like populations, to generate elemental sulfur. In addition, the enriched genes encoding TQO, which catalyzed thiosulfate to produce tetrathionate, were also closely related to Acidithiobacillus -like genus. As the substrate of TetH, tetrathionate could be further hydrolyzed to thiosulfate, sulfate and elemental sulfur [ 61 ]. The genes encoding the truncated sulfur oxidation protein (Sox) system ( SoxAX , SoxB and SoxYZ ) were identified in Acidithiobacillus -like and Thiobacillus -like microorganisms. Although not enriched, SoxCD genes were also detected in the metagenome and identified in Rhodanobacter -like and Thiomonas -like populations. Sulfur oxygenase reductase (SOR) reported previously in certain bacteria such as Acidithiobacillus thiooxidans and A. caldus [ 61 , 65 ] and thermo-acidophilic archaea such as Acidianus tengchongensis and A. ambivalens [ 66 , 67 ] were found in the metagenome. In addition, the genes encoding heterodisulfide reductase ( hdrABC ) and thiosulfate sulfurtransferase ( tst ), which participated in the catalysis of thiosulfate to sulfite, were abundant in Acidithiobacillus -like populations. Sulfite would be oxidized directly by sulfite oxidase to sulfate or be catalyzed by adenylylsulfate (APS) reductase and sulfate adenylyltransferase (SAT) to generate sulfate via the reversed dissimilatory sulfuate reduction. Remarkably, the aprAB genes encoding APS reductase were enriched in the metagenome, most of which were assigned to those of Thiobacillus -like populations. No such gene, however, was found in Acidithiobacillus -like populations [ 61 , 65 , 68 ]. Indeed, previous reports showed that Thiobacillus species were widely found as the dominant member in some mining environments [ 1 ]. Reductive pathways of sulfate, which contained assimilatory sulfate reduction and dissimilatory sulfate reduction, ended with the formation of sulfide, and then entered into the biosynthesis of vital compounds. The coding genes for key enzymes of these processes were most closely related to Acidithiobacillus -like and Thiobacillus -like populations. Tolerant mechanisms to the extremely acidic environments We also focused on the research on extremely acidic environments, mainly because of the particularity of survival conditions that are harmful to most organisms and the adaptability of acidophiles that survive in these environments. Leach solution with chemical components which favor the formation of extremely acidic environments was sprayed onto the tailing dump, thus microbial species in the mine tailings probably possess the environmental adaptive capabilities. Heavy metal resistance Functional abundance profile analysis based on COG categories showed that COG categories [L] (Replication, recombination and repair; 9.44 %) were overrepresented in the metagenome dataset (Fig. 1 ). A reasonable explanation was that the concentration of toxic substance (e.g., heavy metals) in the extremely environment was higher than that of other ‘normal’ environments (e.g., agricultural soil), resulting in the accelerating rate of DNA injury [ 2 , 69 ]. The bioleaching system is rich in high concentration of toxic metal elements such as copper, mercury, zinc, arsenic, cadmium and cobalt [ 33 , 37 ] . Microbes in this environment might possess certain resistance systems to responses to the heavy concentrations of metal ions. Energy-dependent efflux pumps, to a large extent, play a crucial role in toxic metal ion resistance, including ATPases and other chemiosmotic ion/proton exchangers; on the other hand, fewer mechanisms associated with enzymatic transformation such as oxidation, reduction, methylation and demethylation, which converts metal ions from more toxic to less toxic forms, and metal-binding proteins (e.g., metallothionein, protein chaperone and periplasmic binding protein) are also important for microbes to adapt the environmental conditions [ 70 ]. Researches showed that the microbial community in copper mine tailings harbored a variety of heavy metal resistance systems, such as ars operon arsenate resistance/regulation ( arsC , COG1393; arsR, COG0640; arsB, COG1055; arsH , COG0431), mer operon mercuric resistance/regulation ( merR , COG0789; merA , COG1249; merP , NOG79562; merT , NOG11562; merC , NOG56224; merD , NOG40540), CzcD-like cobalt-zinc-cadmium efflux ( czcD , COG1230) and CzcABC cobalt-zinc-cadmium efflux ( czcA , COG3696; czcB , NOG01644; czcC , NOG19426) (Additional file 6 ). And the majority of sequences associated with heavy metal resistance were assigned to Acidithiobacillus -like and Thiomonas -like microorganisms. Our results probably supported the viewpoint that microbial community in bioleaching heap adopted aforementioned mechanisms to cope with heavy metal stress. Low pH adaptation Given that microorganisms inhabit the acidic environments, the resident microbes may possess several strategies to maintain the circumneutral intracellular pH. In order to acclimatize themselves to the extreme acidic environments, acidophiles share various structural and functional characteristics [ 17 ], mainly including (i) reversed membrane potential (ΔΨ) generated by a Donnan potential that creates a chemiosmotic barrier to inhabit the influx of protons; (ii) highly impermeable membranes that restrict proton influx into the cytoplasm; and (iii) active secondary transporters that drives transport by utilizing the transmembrane electrochemical gradient of protons or sodium ions. Functional abundance profile analysis dependent on COG catalogues indicated that the potassium-efflux system proteins including potassium-transporting ATPase subunit A (KdpA; COG2060), subunit B (KdpB; COG2216) and subunit C (KdpC; COG2156), voltage-gated potassium channels proteins (COG1226 and COG0667) and proton antiporters (CPA) (COG3263/COG0025, COG0475/COG1226 and COG3004) were found in the metagenome. Most microorganisms having those genes shared sequence identity with the Acidithiobacillus -related sequences (Additional file 6 ). A large amount of associated genes showed that one potential mechanism to generate a reversed ΔΨ was performed by potassium-transporting ATPases [ 17 ]. In addition, several genes encoding plasma-membrane proton-efflux P-type ATPase (COG0474), which were presumed to exclude intracellular redundant protons, were detected in the metagenome. Organic solvent tolerance Organic solvents contain a large number of compounds with different kinds of chemical structures, e.g., benzene rings as well as aliphatic alcohols. Many of these compounds are greatly harmful to all life forms including humans, animals, plants and microorganisms [ 71 ]. They accumulate in cell membranes and undermine membrane integrity, resulting in the functional loss of membrane as the permeability barrier and energy transducer, and further leading to the alteration of intracellular pH and membrane electrical potential, cellar metabolism disorder, growth inhibition and even, eventually cell death [ 72 – 74 ]. Both in Gram-negative and Gram-positive microorganisms, there are common resistant mechanisms including energy-dependent active efflux pumps, cis -to- trans isomerization of unsaturated fatty acids mediated by cis – trans isomerase and changes in phospholipid head groups, generation of membrane vesicles transferring toxic compounds, and the change rate of phospholipid biosynthesis to expedite modify process [ 75 ]. Given that extractant was used in metal extraction industry [ 76 ], microorganisms exposed to Lix984n (an organic extractant), which was a potential substrate for RND efflux pumps [ 77 ], might harbor the stress-response strategies to cope with organic solvents in the bioleaching system. As a member of RND family proteins, the organic solvent efflux pump composed of AcrB (transporter AcrB/AcrD/AcrF family protein; COG0841), AcrA (RND family efflux transporter MFP subunit; COG0845), and TolC (outer membrane efflux protein; COG1538) was presumed to transfer Lix984n as a potential substrate. Most of those CDSs were identified as Acidithiobacillus -like and Thiomonas -like sequences. Furthermore, a variety of COGs associated with toluene resistance, i.e., ABC transport system, including toluene tolerance protein (COG2854), Mce-related protein (substrate-binding protein) (NOG02063), toluene tolerance protein Ttg2B (permease protein) (COG0767) and toluene tolerance protein Ttg2A (ATP-binding protein) (COG1127), were identified in the metagenome. Moreover, functional abundance profile analysis based on COGs revealed that COG2067 (aromatic hydrocarbon degradation membrane protein) and COG1452 (organic solvent tolerance protein) were identified. Detoxification of hydroxyl radicals The reaction of ferrous iron (Fe(II)) with oxygen (Fenton reaction) could lead to the generation of hydroxyl radicals, which might cause the damage of biological macromolecules [ 11 , 12 ]. One strategy of avoiding the production of damaging free radicals is that electrons from Fe(II) substrates could be removed primarily using an outer membrane cytochrome c [ 11 ] . In addition, evidence showed that the enzymatic detoxification of hydroxyl radicals was identified in Leishmania chagasi and L. donovani [ 78 ]. Based on sequence identities (30 % identity cut-off; E-value ≤1e −5 ), sequence homologs for those peroxiredoxin coding genes ( LcPxn1 , LcPxn3 and LdPxn1 ) were mostly assigned to a Leptospirillum -like populations, indicating its key role in the detoxification of hydroxyl radicals. Moreover, other genes encoding antioxidant proteins such as peroxidase/catalase (COG0376) were also identified in the metagenome. Herein, the majority of sequences were assigned to the Euryarchaeota order Thermoplasmatales -like populations , especially Thermoplasmatales archaeon I-plasma-like and Thermoplasmatales archaeon Gpl-like groups, supporting the previous results that some extremely acidophilic genera belonging to order Thermoplasmatales were regularly found in bioleaching environments [ 79 , 80 ]."
} | 8,067 |
34427521 | PMC8409725 | pmc | 965 | {
"abstract": "ABSTRACT Microbial communities are constantly challenged with environmental stressors, such as antimicrobials, pollutants, and global warming. How do they respond to these changes? Answering this question is crucial given that microbial communities perform essential functions for life on Earth. Our research aims to understand and predict communities’ responses to change by addressing the following questions. (i) How do eco-evolutionary feedbacks influence microbial community dynamics? (ii) How do multiple interacting species in a microbial community alter evolutionary processes? (iii) To what extent do microbial communities respond to change by ecological versus evolutionary processes? To answer these questions, we use microbial communities of reduced complexity coupled with experimental evolution, genome sequencing, and mathematical modeling. The overall expectation from this integrative research approach is to generate general concepts that extend beyond specific bacterial species and provide fundamental insights into the consequences of evolution on the functioning of whole microbial communities."
} | 279 |
31804509 | PMC6895054 | pmc | 966 | {
"abstract": "Rare earth elements (REEs) are now considered emerging pollutants in the environment. Phytolacca americana , an REE hyperaccumulating plant, has been proposed for the remediation of REE-contaminated soils. However, there is no REE-related information for other Phytolacca species. Here, we examined five species ( P. americana , P. acinosa , P. clavigera , P. bogotensis , and P. icosandra ) for their response to REEs. REE accumulation and fractionation traits both occurred on the same order of magnitude among the five species. Heavy REEs were preferentially transferred to leaves relative to light REEs. Regardless of the species, lateral root length and chlorophyll content decreased under REE exposure, and lateral roots and foliar anthocyanins increased. However, plants did not experience or only slightly experienced oxidative stress. Finally, REE exposure strongly modulated the ionome of roots and, to a lesser extent, that of leaves, with a negative correlation between REE and Mn contents. In conclusion, our study provides new data on the response of several Phytolacca species to REEs. Moreover, we highlighted that the REE accumulation trait was conserved among Phytolacca species. Thus, we provide valuable information for the phytoremediation of REE-contaminated sites since the most appropriate Phytolacca species could be selected depending on the climatic/pedological area to be remediated.",
"conclusion": "Conclusions Phytolacca species are fast growing and high-biomass producing plants. P. americana is the first species of the genus that was demonstrated to be an REE accumulator. By testing five Phytolacca species under two different REE concentrations, this study therefore brings new information regarding the effect of REEs on the physiology and ionome of the REE-accumulating Phytolacca species. We used here hydroponic conditions to finely compare the potential of the different Phytolacca species to accumulate the whole set of REE species. Further research should be performed on the growth and REE accumulation potential of the different Phytolacca species with various REE-contaminated soils. These two parameters need to be evaluated in conditions of contamination to determine in situ the phytoextraction potential of the different Phytolacca species. It is interesting to note that the REE accumulating trait is conserved within the Phytolacca genus, as reported for the Carya genus 16 . Such information is very useful for the phytoremediation of REE-contaminated sites. Indeed, the different Phytolacca species tested here originate from different locations. For example, P. bogotensis is native to Colombia, P. icosandra to Mexico, and P. acinosa to Tibet 28 . Therefore, the conserved REE accumulation trait among these different species can be used to choose the most appropriate species according to the climatic and/or pedological context of the area to be remediated. Finally, to better understand the molecular and genetic mechanisms that underlie this particular trait, further work is needed.",
"introduction": "Introduction The Rare Earth Elements (REEs) are a group of 17 metallic elements including the 15 lanthanides plus yttrium and scandium. REEs can be further split into two groups known as the light REEs (LREEs, spanning from lanthanum (La) to europium (Eu), plus scandium (Sc)) and the heavy REEs (HREEs, spanning from gadolinium (Gd) to lutetium (Lu), plus yttrium (Y)). REEs are widely found in the Earth’s crust. Cerium (Ce), La, neodymium (Nd) and Y are the most abundant REEs and can be found at concentrations similar to those of Zn, Cu, Ni and Mo, among others 1 . A fast increase in the demand and subsequent production of REEs has been observed over the past decades because of their diverse use in new technologies, green energies and medical devices 2 . Moreover, the very low recyclability of these elements 3 , their use as food additives for livestock and poultry 4 , their spreading as fertilizers in agriculture 1 or as eutrophication regulators 5 , along with the high soil concentrations found in mining areas 6 , 7 lead REEs to be considered emerging pollutants. Although these elements are not essential to plants, REEs can be detected at moderate concentrations under non-contaminated conditions in plant tissues. The REE concentrations in leaves ranged from 0.0011 (Lu) to 0.33 mg/kg DW (Ce) 8 and even up to 5 mg REE/kg DW 9 . However, as reported for other metals, several plant species, mainly ferns, are able to accumulate REEs. Two fern species displayed the highest REE accumulation potential. Dicranopteris dichotoma (syn. D. linearis ) accumulated up to 0.7% DW of LREEs 6 , 10 and Pronephrium simplex accumulated 1.2 g REE/kg DW 11 , 12 . To a lesser extent, other ferns could also accumulate REEs, such as Dryopteris erythrosora 9 , Blechnum orientale 13 , 14 and Athyrium yokoscense 15 . Although ferns are highly represented, a few angiosperms are also known to accumulate REEs. Carya tomentosa (mockernut) accumulated REEs up to 859 mg/kg in a non-contaminated environment 14 , 16 . Phytolacca americana (pokeweed), first identified as a Mn hyperaccumulator 17 – 19 , with a Mn accumulation up to 2000 mg REE/kg DW on non-contaminated soils 17 , was further reported as an REE accumulator 15 , 20 . In comparison to other REE-accumulating species, P. americana is a fast growing and high-biomass producing plant that can reach 3 m in height. This ubiquitous weed of roadsides and disturbed areas in its native range of the southeastern United States is now distributed worldwide 17 . Notably, this species was found at an REE mining site in southern Jiangxi Province in China, with an average REE concentration of approximately 250 mg/kg in leaves, reaching up to 1,040 mg/kg 7 . Recently, several studies investigated the translocation and fractionation of REEs in P. americana . Yuan et al . (2017, 2018) observed a higher translocation of HREEs in the leaves compared to LREEs, while more LREEs accumulated in the roots and stems. Organic or amino acids have been implicated in complexing REEs and participating in the long-distance transport of these elements in P. americana 20 , 21 . Combined, these features would be of great interest for the phytoremediation of REE-contaminated soils. In addition, REEs extracted from plants could be further purified 22 or used for ecocatalysis 23 . Combining ecological and genetic analyses of P. americana , no genetic differentiation could be detected between populations from Mn-contaminated and uncontaminated sites 24 , suggesting that phenotypic plasticity is probably the major contributor to the successful colonization of marginal lands, such as metal-contaminated soils. The related species P. acinosa 25 , 26 is also a Mn hyperaccumulator, showing that the Mn hyperaccumulation trait is found in different species of Phytolacca . Similar findings were reported concerning the Ni hyperaccumulation trait in several species of Alyssum and Cochlearia 27 . Likewise, several Carya species accumulated similar REE concentrations in leaves 16 , therefore highlighting that the REE accumulation trait could be conserved throughout the Carya genus. However, the conservation of this REE accumulation potential has not yet been investigated for Phytolacca species other than P. americana . Consequently, we propose the hypothesis of a monophyletic REE accumulation trait within the Phytolacca genus. Therefore, we analysed the REE accumulation potential as well as the REE fractionation pattern of five Phytolacca species ( P. americana, P. acinosa, P. bogotensis, P. clavigera and P. icosandra ). Our analyses also aimed to reveal how REE accumulation in these species of Phytolacca modified the elemental composition of roots and shoots. In addition, analyses were performed to study the effects of REEs on a set of biomarkers related to modification of root architecture, leaf pigment composition and oxidative stress.",
"discussion": "Results and Discussion Molecular confirmation of the Phytolacca species Seeds from five species of Phytolacca were obtained from commercial or institutional origins as described in the materials and methods section and identified as P. americana , P. acinosa, P. bogotensis , P. clavigera , and P. icosandra . The identification of these species was based on morphological traits. However, due to the ambiguous taxonomy within the Phytolacca genus 28 , we performed a molecular investigation to confirm that the species tested were different. For this purpose, the ITS region was amplified, sequenced and compared with ITS sequences of Phytolacca species retrieved from GenBank. For most species, there was a single ITS sequence available in the database, and none were available for P. clavigera (Fig. 1 ). For P. bogotensis , P. icosandra and P. americana , our data matched the previously published ITS sequences for these species. However, it was less straightforward for P. acinosa and P. clavigera , where a mis-identification might have been made. Facing this problem, we analysed a second seed batch of P. acinosa obtained from a different supplier. Both plant morphology and ITS sequences were identical for the two batches, suggesting a correct identification of P. acinosa . Notably, different leaf morphologies were observed between P. acinosa and P. clavigera (Fig. 1 ), which shared the highest identity (97%). Leaves of P. acinosa were obovate, whereas the four other species had oval-shaped leaves. In conclusion, the molecular analyses confirmed that the five species tested were different. Sequence data were deposited in GenBank under the following accession numbers: MK602340 ( P. americana ), MK602343 ( P. acinosa ), MK602344 ( P. bogotensis ), MK602341 ( P. clavigera ), and MK602342 ( P. icosandra ). Figure 1 Maximum likelihood phylogram for the identification of Phytolacca species used based on ITS1-5.8S-ITS2 sequences. Aligned sequences using ClustalW were used to build a maximum likelihood tree based on the Kimura 2-parameter method. Bootstrap values over 50% (1000 replicates) are indicated below the branches. Analyses were conducted in MEGA7 54 . The ITS sequences of the Phytolacca species used in this study are indicated in bold and compared with other Phytolacca species available in GenBank. Monococcus echinophorus and Petiveria alliacea were used as outgroups. The leaf morphology of the tested species is shown. The REE accumulation trait was shared among Phytolacca species We first investigated the potential of the different Phytolacca species to accumulate REEs in their tissues. To this end, Phytolacca seedlings were grown hydroponically and exposed to a mixture of 16 REEs supplied at equimolar concentrations. Diatloff et al . (1996) reported that in uncontaminated sites, REE concentrations from the soil solution can reach the micromolar range 29 . Colim et al . (2019) analyzed the REE concentrations in surface water samples from the Lavras do Sul (Brazil) mining region. The total REE concentrations measured ranged from 27 to 279 µM 30 . Therefore, we decided to use two different REE concentrations, a low and a high concentration (10 and 100 µM). These concentrations were also pre-determined from preliminary experiments (data not shown), for which biomass was either not affected or reduced. Indeed, at the concentration of 10 µM REEs in the nutrient solution (REE10), no negative effect was observed for root, shoot and total plant biomass for any species (Fig. 2 ). A significant positive effect on the shoot biomass of P. icosandra and the root biomass of P. americana was observed even at this lowest concentration. Conversely, at 100 µM REEs (REE100), biomass tended to decrease by approximately 50% for most species, despite not always being significant (Fig. 2 ). These results suggest a similar REE tolerance level of the five Phytolacca species. Figure 2 Effect of REEs on the biomass of Phytolacca species. Biomass (DW) is expressed as a percentage of that of the control (no REE) in shoots ( a ), roots ( b ) and total plants ( c ) exposed to 10 µM or 100 µM REEs (REE10 and REE100, respectively). The data are the means (±SE) of n = 3 (control, REE10) or n = 4 (REE100) plants. Significant differences from the control condition are indicated by asterisks (P < 0.05, t-test). Then, REE accumulation was determined by ICP-MS, and the results are given in Fig. 3 . At REE10, the five species accumulated approximately 1500 mg/kg REEs in the roots, and no significant difference was observed between the species (Fig. 3a ). However, at REE100, an overall higher accumulation was found, and differences between species could be noted (Fig. 3a ). Phytolacca icosandra displayed the highest concentration of approximately 13,000 mg/kg REEs, while P. acinosa and P. bogotensis accumulated 2.4 and 2.3 times less REEs, respectively (Fig. 3a ). P. clavigera and P. americana had intermediate concentrations in their roots. Figure 3 REE accumulation by five Phytolacca species. Plants were exposed to a mixture of 10 µM or 100 µM REEs (REE10 and REE100, respectively). ( a ) REE concentrations in roots and leaves of Phytolacca species. ( b ) Translocation factor (TF) of REEs from the roots to the leaves. The data are the means (±SD) of n = 3 (REE10) or n = 4 (REE100). Within a given treatment, values with the same letter are not significantly different (P < 0.05, ANOVA, Tukey’s HSD). In leaves, the patterns of REE accumulation differed from those of roots. At REE10, P. bogotensis and P. acinosa were the two species accumulating the highest concentrations of REEs, with 279 and 261 mg/kg, respectively. Phytolacca americana and Phytolacca clavigera accumulated slightly less, with 174 and 157 mg/kg, respectively, while the lowest concentration was recorded for P. icosandra , with less than 100 mg/kg (Fig. 3a ). However, those variations were not observed at REE100. Despite a mean accumulation value ranging from 265 mg/kg ( P. clavigera ) to 837 mg/kg ( P. bogotensis ), these values were not significantly different (Fig. 3a ). Our results with P. americana are consistent with those from a previous study, carried out with the same species also grown hydroponically, where a similar REE accumulation capacity was reported 21 . As a comparison, P. americana accumulated in the leaves at approximately 300 and 500 mg/kg REEs when exposed to 10 and 100 µM REEs, respectively 21 . For each species, the translocation factor (TF) of REEs from roots to leaves was further calculated (Fig. 3b ). Under both REE concentrations, P. acinosa and P. bogotensis had the highest TFs, whereas P. clavigera and P. icosandra had the lowest TFs. However, the TF values were in the same range, and relatively low differences were found between the species (Fig. 3b ). Moreover, our data are in agreement with those described by Yuan et al . (2017) for P. americana , where a TF of approximately 0.1 was reported 21 . Overall, our results suggest that the accumulation of REEs in the leaves and their translocation from roots to shoots are of the same order of magnitude in the five Phytolacca species. HREEs were preferentially translocated to leaves in the five Phytolacca species We further investigated whether various REE fractionation patterns (LREEs vs HREEs) could be mediated by the different Phytolacca species. Therefore, we analysed the fractionation pattern of the whole set of REEs in both roots and shoots of the five species. To compare the fractionation process for various REE species mediated by different plant species, we used our hydroponic system to ensure equimolar concentrations of all REE species in the nutritive medium. The data of REE fractionation are given in Fig. 4a . REEs are ordered by their decreasing ionic radii at their trivalent oxidation state. The non-lanthanide Y, generally included in the HREEs, thus ranged between Dy and Ho. Interestingly, the five Phytolacca species displayed very similar REE accumulation patterns. In roots, the REE concentration decreased from La to Lu (Fig. 4a ). Conversely, in leaves, the REE concentration increased from La to Lu (Fig. 4a ). These conclusions were true regardless of the Phytolacca species and the REE concentration (REE10, REE100). The only exceptions were praseodynium (Pr) and ytterbium (Yb), for which anomalies could be demonstrated. However, these anomalies were similar among the five Phytolacca species. Figure 4 REE fractionation in roots and leaves of Phytolacca species. Plants were exposed to a mixture of 10 µM or 100 µM REEs (REE10 and REE100, respectively). ( a ) REE concentration patterns in roots and leaves of Phytolacca species (the linkage of data points does not indicate dependence). ( b ) Concentration ratios of LREEs (Sc, La to Eu) over HREEs (Gd to Lu, Y) among the species tested and exposed either to 10 µM or 100 µM as indicated. The data are the means (±SD) of n = 3 (REE10) or n = 4 (REE100). Within a given treatment, values with the same letter are not significantly different (P < 0.05, ANOVA, Tukey’s HSD). We further analysed the accumulation ratios between LREEs and HREEs (Fig. 4b ). At REE10, there was no particular enrichment of LREEs versus HREEs in roots. At REE100, and except for P. icosandra , a slight LREE enrichment in roots was found for the different Phytolacca species. Conversely, leaves were highly enriched with HREEs regardless of the Phytolacca species considered. At REE10, HREE concentration was indeed between 2 and 4 times higher than that of LREEs. This enrichment was more pronounced at REE100, where HREEs were from 5 ( P. bogotensis ) to 13 ( P. clavigera ) times more abundant than LREEs (Fig. 4b ). Altogether, the data indicated a preferential root-to-leaf translocation of HREEs vs LREEs, a trait that was conserved in the five Phytolacca species. The preferential translocation of HREEs to shoots had already been observed for P. americana grown either under hydroponic conditions or naturally in REE-mining areas 7 , 21 . In non-accumulating species, a higher transfer of HREEs vs LREEs has also been reported in several angiosperms, notably wheat 31 , 32 , soybean 33 and rice 34 . Conversely, LREE enrichment in shoots has been reported in fern species accumulating REEs. Likewise, non-accumulating ferns also had a higher content of LREEs than HREEs (Grosjean N., personal communication). Therefore, the different fractionation patterns described above are very unlikely to be related to the REE accumulation trait. More likely, angiosperms and pteridophytes, which are evolutionarily very distant, might not share the same mechanisms underlying REE translocation from roots to shoots. Such different fractionation processes could be explained by the production of specific compounds with distinct REE-chelating properties. This hypothesis was supported by previous works that reported that the REE-accumulating fern D. dichotoma produces a specific LREE-binding peptide 35 and that HREE enrichment in the shoots of P. americana was associated with the long-distance transport of HREE-organic acid complexes, such as HREE-citrate 21 . However, to test this hypothesis, further studies investigating the ligands associated with REEs in the xylem sap of a panel of angiosperms and pteridophytes are needed. REEs impacted root architecture It has already been reported that high concentrations of La can inhibit root elongation and induce lateral root development in Arabidopsis thaliana 36 . This mechanism was explained by the accumulation of reactive oxygen species (ROS) in the root tip, leading to the death of cells from the primary root tips and reorientation of auxin to lateral roots 36 , 37 . Thus, to verify whether root elongation and branching were also affected by a mixture of REEs, the root architecture of the five Phytolacca species was analysed after exposure to 10 and 100 µM REEs and compared to control plants (Fig. 5 ). The overall root architecture of the different species was relatively similar under the control condition (Fig. 5a ). The addition of REEs triggered modifications in the root architecture that were characterized by shorter lateral roots compared to the control condition (Fig. 5a ). At REE10, morphological differences were noticed for P. clavigera , P. icosandra , and P. acinosa when compared to non-exposed plants. These differences were much less obvious in P. bogotensis and P. americana . However, at REE100, the root architecture of all species was more impacted, even if P. americana was the less affected species (Fig. 5a ). The quantification of both the density and length of lateral roots confirmed the visual observations (Fig. 5b,c ). The number of lateral roots indeed increased under REE exposure. However, for both P. americana and P. bogotensis , the values were only significantly different when comparing the control and REE100 conditions (Fig. 5b ). Similarly, with increasing REE concentrations, the lateral root length decreased, with the exception of P. americana , which was not affected at REE10 when compared to the control (Fig. 5c ). In conclusion, analysis of the root architecture of the different species allowed us to show that, despite a similar REE accumulation rate for the species tested, P. americana was the least affected by REEs. Figure 5 Effect of REEs on the root architecture of Phytolacca species. ( a ) Architecture of representative roots (scale bars = 1 cm), ( b ) lateral root branching (expressed as the number of lateral roots per cm of primary root) and ( c ) lateral root length of the different Phytolacca species exposed to a mixture of 10 µM or 100 µM REEs (REE10 and REE100, respectively) or not exposed to REEs (control). Root systems of three (control, REE10) or four (REE100) plants per species and per treatment were analysed. Within a given species, significant differences between treatments are indicated by different letters (P < 0.05, Kruskal-Wallis). Phytolacca species did not experience or only slightly experienced oxidative stress under REE exposure We further selected two markers related to oxidative stress, namely, the malondialdehyde (MDA) concentration and the total antioxidant capacity (TAC), to investigate the effect of REE exposure in the five species (Table 1 ). Table 1 Total antioxidant capacity (TAC) and malondialdehyde (MDA) contents in leaves and roots of plants exposed to REEs. Treatment P. americana P. clavigera P. icosandra P. acinosa P. bogotensis Leaves TAC (nmol/mg FW) Control 5.06 ± 0.43 a 4.40 ± 0.67 b 4.66 ± 1.42 a 3.50 ± 1.98 a 2.92 ± 0.08 a REE - 10 6.82 ± 1.23 a 5.53 ± 0.09 a 6.94 ± 0.63 a 3.73 ± 0.72 a 2.65 ± 0.13 a REE - 100 6.05 ± 0.69 a 3.35 ± 0.30 c 6.53 ± 1.63 a 4.04 ± 1.10 a 2.62 ± 0.60 a MDA (nmol/mg FW) Control 1.04 ± 0.06 a 0.81 ± 0.05 a 0.66 ± 0.18 b 0.98 ± 0.13 a 1.16 ± 0.14 a REE - 10 1.09 ± 0.18 a 1.35 ± 0.71 a 1.08 ± 0.16 a 0.42 ± 0.05 b 0.47 ± 0.12 b REE - 100 0.86 ± 0.15 a 0.85 ± 0.07 a 0.94 ± 0.09 ab 0.84 ± 0.16 a 1.20 ± 0.12 a Treatment P. americana P. clavigera P. icosandra P. acinosa P. bogotensis Roots TAC (nmol/mg FW) Control 0.55 ± 0.07 a 0.60 ± 0.12 ab 0.47 ± 0.03 a 0.64 ± 0.06 b 0.48 ± 0.30 a REE - 10 0.63 ± 0.20 a 0.70 ± 0.15 a 0.46 ± 0.07 a 1.00 ± 0.05 a 0.41 ± 0.05 a REE - 100 0.99 ± 0.49 a 0.38 ± 0.05 b 0.33 ± 0.29 a 0.70 ± 0.11 b 0.62 ± 0.20 a MDA (nmol/mg FW) Control 0.16 ± 0.02 a 0.20 ± 0.09 a 0.09 ± 0.02 a 0.16 ± 0.03 a 0.17 ± 0.01 a REE - 10 0.30 ± 0.20 a 0.34 ± 0.13 a 0.11 ± 0.02 a 0.15 ± 0.05 a 0.09 ± 0.03 a REE - 100 0.18±0.04 a 0.19 ± 0.02 a 0.14 ± 0.02 a 0.23 ± 0.10 a 0.22 ± 0.05 a Phytolacca species were exposed to a mixture of 10 µM (REE10) (n = 3) or 100 µM REEs (REE100) (n = 4). Values are means ± SD. Within a given species, significant differences between treatments are indicated by different letters (P < 0.05, ANOVA Tukey’s HSD). TAC represents the non-enzymatic antioxidant capacity and is indicative of the ability to counteract oxidative stress-induced damage in cells. In both roots and leaves of the different species, only a few differences were found among the different treatments (Table 1 ). Indeed, REE exposure only impacted the TAC (1.3-fold compared to the control) in the leaves of P. clavigera . Similarly, in the roots of P. acinosa , the TAC content increased by 1.6-fold at REE10 when compared to the control (Table 1 ). The quantification of MDA, one of the end products of the chain reaction of lipid peroxidation, allows us to estimate potential oxidative damage. No significant change was observed in roots under REE exposure, regardless of the species considered. In the leaves of P. acinosa and P. bogotensis , the MDA concentration decreased by 2.5-fold at REE10 and to a lesser extent (1.3-fold) in P. acinosa at REE100. Conversely, it slightly increased (1.6-fold) in the leaves of P. icosandra at REE10 (Table 1 ). Combined, MDA and TAC analyses suggest that Phytolacca species were not affected or were only slightly affected by oxidative stress when exposed to REEs. Conversely, several studies have demonstrated the generation of REE-induced oxidative stress in plant species that are not REE accumulators (e.g., Nymphoides peltata and Hydrilla verticillata ), with, for example, a progressive increase in the concentration of MDA with increasing REE concentrations 38 – 40 . Therefore, our data suggest that the five Phytolacca species are most likely highly tolerant to REEs. However, additional studies will be needed to compare the REE tolerance levels of different Phytolacca species. REEs impacted the pigment contents and nitrogen balance index Given the high REE accumulation in leaves, we investigated its impact on the content of different pigments (chlorophyll, flavonoids, and anthocyanins) and on the nitrogen balance index (NBI). Differences between the treatments but also between the species were recorded (Fig. 6 ). Regardless of the species, the chlorophyll index moderately decreased at REE100 when compared to the control. Indeed, the chlorophyll content was reduced from 1.1 to 1.6 times in P. bogotensis and P. icosandra , respectively (Fig. 6a ). The lower chlorophyll content under REE exposure could either be due to a reduction of Mg supply 40 or to chlorophyll degradation caused by lipid peroxidation 39 . Since the MDA concentration did not increase at REE100 in the different Phytolacca species, this second hypothesis is very unlikely. However, it is noteworthy that the chlorophyll content increased for P. icosandra and P. americana at REE10 (Fig. 6a ). Such a result could suggest a stimulation of photosynthesis. Several studies reported growth stimulation in rice 41 , tobacco 42 and soybean 43 at low REE concentrations. Figure 6 Pigment contents and nitrogen balance index of Phytolacca species exposed to REEs. ( a ) Chlorophyll, ( b ) flavonoid, ( c ) anthocyanin, and ( d ) nitrogen balance (NBI) indexes were measured in the leaves of Phytolacca species exposed to a mixture of 10 µM or 100 µM REEs (REE10 and REE100, respectively). Twenty measurements of three (control, REE10) or four (REE100) plants per species and per treatment were performed. Within a given species, significant differences between treatments are indicated by different letters (P < 0.05, Kruskal-Wallis, Wilcoxon post hoc test). Compared to chlorophylls, the flavonoid content was poorly affected by REE exposure (Fig. 6b ). When compared to the control, it slightly increased at REE10 by 1.1 and 1.2 times in P. icosandra and P. acinosa , respectively. Conversely, it decreased in P. americana and P. bogotensis at either REE10 or REE100. The anthocyanin index, usually indicative of plant stress 44 , was also analysed and is reported in Fig. 6c . A similar pattern was found for the different species, where all species showed a significant increase in anthocyanins at REE100 when compared to the control (Fig. 6c ). These data suggest that this high dose of REEs, associated with a subsequent high REE accumulation in leaves (Fig. 3a ), could lead to a modification of anthocyanin biosynthesis. In addition, the increase in anthocyanin content could be caused by an increased limitation of phosphate in the plant due to the high reactivity of REEs with phosphates. Surprisingly, at REE10, P. americana was the only species for which there was a decrease of anthocyanins (Fig. 6c ), suggesting that at this low REE exposure, the leaf stress was less important to P. americana than to its close relatives. The different pigment indexes measured can provide indications of the health or stress status of plants under tested conditions. It has been demonstrated in Arabidopsis thaliana that an increase of anthocyanins combined with a decrease of the chlorophyll content was triggered in response to metallic stress 45 . Indeed, anthocyanins can play a major protective role against metal stresses by acting as antioxidants 45 . However, since no difference was obtained from the MDA and TCA measurements, strong oxidative stress is unlikely, and a higher anthocyanin index of all five species measured at the highest REE concentration might reflect phosphorus deficiency, as suggested above and as previously reported 46 . We also analysed the NBI in plants exposed to REEs. NBI is a marker that corresponds to the ratio of chlorophyll content to that of flavonoids and gives an indication of the nitrogen status of plants 47 . While no difference was noticed for P. bogotensis , the NBI slightly increased at REE10 for P. americana , P. clavigera , and P. icosandra (Fig. 6d ). Except for P. bogotensis , the NBI decreased at REE100 in the four other species. The relatively lower root branching, the increase of the chlorophyll content, along with the increased NBI for P. americana at REE10, could suggest that P. americana is slightly more tolerant to low REE exposure compared to the other species tested (Figs. 5 , 6 ). However, since the growth of other Phytolacca species was not more affected by REEs than that of P. americana for both REE10 and REE100 (Fig. 2 ), REE tolerance levels of the different species should be similar. Further dose-response studies would, however, be needed to shed more light on the tolerance levels of the different Phytolacca species. REEs modulate the ionome of Phytolacca species Finally, the ionome of the different species under the three different exposure conditions was established (Supplementary Tables S1 , S2 ). The two different compartments, leaves (Supplementary Tables S1a , S2a ) and roots (Supplementary Tables S1b , S2b ), were treated separately. Regarding the accumulation and fractionation of REEs, no differences were observed between the five species tested. Consequently, a global analysis was carried out using the ionomes of the different Phytolacca species to identify putative correlations between REE accumulation and essential element composition (Fig. 7 ). Two principal component analyses (PCA) (Fig. 7a,b ) and correlation matrices were generated for both leaves and roots (Fig. 7c,d ). Figure 7 Principal component analysis and Pearson correlation matrices of REE, micro- and macro-elemental composition of leaves ( a , c ), and roots ( b , d ) of Phytolacca species. Plants were exposed to a mixture of 10 µM or 100 µM REEs (REE10 and REE100, respectively) or left unexposed (control). REE, macro- and micro-element concentrations were used as quantitative variables. For the Pearson correlation matrices, only significant correlations are shown (P < 0.05, BH adjustment). In the PCA for the leaves, the two first principal components accounted for 78.4% of the observed variations, with the first principal component accounting for 54.6% (Fig. 7a ). The different conditions tested (control, REE10, and REE100) were well distinguished in the PCA. As expected, all the REEs were positively correlated together (Fig. 7a ). Interestingly, in the second principal component analysis, we observed that LREEs and HREEs were grouped separately (Fig. 7a ). This is in agreement with the HREE enrichment relative to the LREEs observed in the leaves (Fig. 4a ). The first axis of the PCA was mostly structured by the REEs, contributing more than 82% each but also by P with 57% and Mn for 40%. In the second axis, S and Mg were the most structuring variables, with 91% of explained inertia, followed by Cu (85%), Ca (84%), B (80%), K (70%), Fe and Mn (65%). The segregation of REEs (first axis) from the other elements (second axis) on the PCA (Fig. 7a ) is supported by the Pearson correlation matrix (Fig. 7c ). Indeed, relatively few correlations were found between REE accumulation and the concentration of essential elements. Whereas no significant correlation was found between any of the REEs and Fe or Na, negative correlations were found between the heaviest REEs (Tm to Lu) and Cu, Ca, Mg, and Zn (Fig. 7c ). A similar negative correlation was observed for Mn with HREEs (Gd excepted) (Fig. 7c ). Conversely, positive correlations were found between P and all REEs except La. Similar positive correlations were also found between some LREEs and B and S (Fig. 7c ). Interestingly, the Pearson’s r coefficient increased (P only) or decreased (the other elements), with the atomic number of REEs (Fig. 7c ). Again, Sc did not follow the global rule observed for the other REEs, supporting its difference compared to the lanthanides and Y. Correlations were more abundant and stronger between macro- or micronutrient elements and REEs in the root compartment (Fig. 7b–d ). In the PCA, the two first principal components accounted for 80.5% of the observed variations, with the first principal component accounting for 71.2% (Fig. 7b ). The first axis of the PCA was mostly structured by REEs, K, and P, contributing more than 84% each, and Mg contributed 68%. In the second axis, Ca was the most structuring variable, with 74% of explained inertia, followed by Na (65%), Mg (59%), B (56%), and S (47%). REEs covariated together and were also positively correlated with P in both roots and leaves. However, since (i) the nutritive solution lacked P during the exposure phase of REEs and (ii) growth at REE100 was reduced compared to the control, it was expected that P concentration in the biomass would be subjected to a dilution effect in the control treatment compared to the REE-exposed treatments. The observations reported in Supplementary Table S1 and Fig. 7c,d are in agreement with this hypothesis. While no correlation was found in the leaves between Fe or Na and REEs, positive correlations were observed in the roots (Fig. 7b–d ). The other elements, namely, Cu, Mn, S, Mg, Ca, Zn, K, and B, were all negatively correlated with REEs. These correlations were highly influenced by the highest exposure condition (REE100) (Fig. 7b–d , Supplementary Tables S1 , S2 ). REEs have a similar ionic radius to that of Ca. This characteristic has resulted in the use of REEs as Ca-channel blockers 48 , 49 , and later on, it was suggested that La and Eu could enter into plant cells through Ca-channels 50 . Therefore, the negative correlation found between Ca and REEs is not unexpected. Such antagonistic effects between Ca and La have been previously reported in plants 41 , 51 . Yuan et al . (2017) also demonstrated that increasing concentrations of Ca inhibited REE accumulation in P. americana 21 . Along with Ca, Mg also had a negative correlation with REE contents. Such deprivation had already been reported for P. americana exposed to REEs 21 but also in Brassica napus exposed to Ce 52 . Mg is necessary for chlorophyll synthesis and functioning; thus, the decrease in chlorophyll content could be related to the Mg reduction due to REE exposure. A negative correlation between REE and Mn was also found in leaves and roots (Fig. 7 ). Phytolacca americana 17 – 19 and P. acinosa 25 , 26 are two known Mn hyperaccumulators. It is interesting to note the reduced Mn content that occurred concomitantly with the accumulation of REEs. This might suggest a competitive uptake between these elements in Phytolacca species as in P. acinosa between Mn and Cd 53 ."
} | 9,111 |
31572336 | PMC6749037 | pmc | 967 | {
"abstract": "Despite the discovery of the first N -acyl homoserine lactone (AHL) based quorum sensing (QS) in the marine environment, relatively little is known about the abundance, nature and diversity of AHL QS systems in this diverse ecosystem. Establishing the prevalence and diversity of AHL QS systems and how they may influence population dynamics within the marine ecosystem, may give a greater insight into the evolution of AHLs as signaling molecules in this important and largely unexplored niche. Microbiome profiling of Stelletta normani and BD1268 sponge samples identified several potential QS active genera. Subsequent biosensor-based screening of a library of 650 marine sponge bacterial isolates identified 10 isolates that could activate at least one of three AHL biosensor strains. Each was further validated and profiled by Ultra-High Performance Liquid Chromatography Mass Spectrometry, with AHLs being detected in 8 out of 10 isolate extracts. Co-culture of QS active isolates with S. normani marine sponge samples led to the isolation of genera such as Pseudomonas and Paenibacillus , both of which were low abundance in the S. normani microbiome. Surprisingly however, addition of AHLs to isolates harvested following co-culture did not measurably affect either growth or biofilm of these strains. Addition of supernatants from QS active strains did however impact significantly on biofilm formation of the marine Bacillus sp. CH8a sporeforming strain suggesting a role for QS systems in moderating the microbe-microbe interaction in marine sponges. Genome sequencing and phylogenetic analysis of a QS positive Psychrobacter isolate identified several QS associated systems, although no classical QS synthase gene was identified. The stark contrast between the biodiverse sponge microbiome and the relatively limited diversity that was observed on standard culture media, even in the presence of QS active compounds, serves to underscore the extent of diversity that remains to be brought into culture.",
"introduction": "Introduction The marine ecosystem is considered to be an underexplored resource for the study of bacterial interactions within eukaryotic hosts. Despite a number of well-studied examples of bacterial interactions within marine hosts such as the density dependent production of luminescence by Aliivibrio fischeri within the light organ of Euprymna scolopes , relatively little is known about the interactions that occur within marine microbial communities ( Hmelo, 2017 ). This is particularly true in the case of the ancient invertebrate, the marine sponge. Marine sponges are sessile filter feeders that consume bacteria and other marine matter ( Taylor et al., 2007 ). Bacteria can inhabit the mesophyll matrix of these invertebrates with almost 60% of the biomass of a marine sponge being comprised of bacterial endosymbionts ( Wang, 2006 ). This symbiotic relationship is mutually beneficial whereby bacteria are provided with a sheltered nutrient rich environment and the marine sponges acquire limiting nutrients from the microflora ( Mohamed et al., 2008 ; Blunt et al., 2009 ; Mayer et al., 2010 , 2011 ). Within the dense polymicrobial environment of a marine sponge, bacteria can engage in a form of chemical communication termed quorum sensing (QS) ( Taylor et al., 2004a ; Diggle et al., 2007 ; Hmelo, 2017 ). Several classes of QS signaling system are known, with autoinducer peptides favored by gram positive bacteria while N -acyl homoserine lactones (AHLs) predominate within gram negative bacteria ( Whiteley et al., 2017 ). AHLs are capable of activating an autoinducing transcriptional regulator which controls the transcription of target genes involved in a wide variety of cellular processes including the production of virulence determinants ( Diggle et al., 2007 ). There are relatively few studies on the prevalence of functional AHL based QS systems within microorganisms inhabiting marine sponges ( Taylor et al., 2004b ; Mohamed et al., 2008 ; Cuadrado-Silva et al., 2013 ; Britstein et al., 2018 ). A number of studies have focused on the identification of homologs of genes associated with QS pathways ( Zan et al., 2011 ). However, sequence based approaches provide limited information on the functionality of these homologous systems, which for the most part remains to be determined. This homology-based approach is also limited by the lack of nucleotide sequence homology among AHL synthases and AHL responsive transcriptional regulators ( Steindler and Venturi, 2007 ). More recently, screening of marine sponges for AHL signals has revealed a rich diversity likely encoded by the microbial communities residing in those sponges ( Britstein et al., 2018 ). Given the difficulties faced in bringing marine sponge biodiversity into culture, it is intriguing to speculate that these signals may play a role in moderating the dynamics of the microbial communities within which they operate. To gain more insight into the relevance of AHL based QS systems within the microbiota inhabiting marine sponges, bacterial sponge isolates were screened for the production of AHLs using classical AHL reporter strains. A total of 10 QS producing isolates were identified and characterized for AHL production. Co-culture of QS positive isolates with marine sponge samples resulted in increased culturable plate diversity from these communities, although no new genera were identified. While addition of AHLs alone did not influence growth or biofilm in the marine sponge isolates, supernatants from several QS positive isolates suppressed biofilm formation in the marine sponge Bacillus sp. CH8a sporeforming strain. This suggests that the anti-biofilm activity of the QS active supernatants may be mediated downstream of intact QS signaling systems in the producing isolates. Genome sequencing of a QS positive Psychrobacter sp. isolate identified in this study revealed the presence of LuxR DNA binding domains. However, there was no evidence of a LuxR autoinducer domain or an AHL synthase domain in this or any other sequenced Psychrobacter genome. Further establishing the prevalence, structure and diversity of AHL based QS systems will give a better understanding of the role of AHL signaling in the marine ecosystem, potentially unlocking some of the natural biodiversity encoded therein.",
"discussion": "Discussion In this study, 650 marine sponge bacterial isolates were screened for the ability to produce AHLs. A total of 10 isolates were identified that were capable of activating AHL biosensor reporter strains. Mass spectrometry revealed that several of the isolates produced the same or similar AHLs (OC10–OC12 HSL). The capacity for AHL based signaling in the marine ecosystem has previously been reported. AHL signaling in marine snow was first described by Gram et al. (2002) , with species of Roseobacter shown to be QS active. More recently, Pantoea ananatis has been reported to produce a spectrum of AHL signals in marine snow, governing extracellular enzyme production in producing strains ( Jatt et al., 2015 ). Since the first description of AHL based QS in A. fischeri species ( Nealson and Hastings, 1979 ), where the LuxIR paradigm system was first identified, AHL signals have been found in a broad diversity of marine isolates ( Hmelo, 2017 ). Rasch et al. (2007) described AHL production in Aeromonas salmonicida isolates, while a number of studies profiled members of the Vibrionaceae for AHL production ( Yang et al., 2011 ; Purohit et al., 2013 ). The diversity of AHL signals that are encoded in the marine ecosystem has been highlighted by a recent study reporting AHLs with long (up to 19 carbons) and poly-hydroxylated acyl side chains ( Doberva et al., 2017 ). At the same time, studies reporting QS inhibition or quenching in the marine environment have also received considerable attention in recent years ( Romero et al., 2012 ; Gutierrez-Barranquero et al., 2017 ; Ma et al., 2018 ). Primarily produced by microbial species, host derived quorum quenching (QQ) has also been described ( Weiland-Brauer et al., 2019 ). Elucidating and profiling the extent of QS signaling within these environments is a key step in understanding the functional role played by QS in the host-microbe interaction. Interspecies communication within the microbial communities of the marine sponge may offer a competitive advantage through cross-genus coordinated behavior. If several species all produce the same or similar AHLs, then they potentially could adopt community like behaviors more rapidly than species that are not part of this interspecies signaling network. This could arise from the activity threshold being reached more rapidly if several different species produce the same signaling molecule. Of course, conservation within receptor systems would also be an integral factor in moderating these responses. The prevalence of orphan LuxR receptor systems in sequenced microbial genomes highlights the complexity of signaling interactions that remain to be identified and understood ( Patankar and Gonzalez, 2009 ). Adopting community-like behaviors through QS systems may offer a distinct competitive advantage as bacteria can attach to a form a biofilm like structure within the environment of the sponge. Community-based small molecular interactions may also be important with respect to intracellular sponge symbionts, such as the recently reported Candidatus Endohaliclona renieramycinifaciens intracellular interaction with Haliclona ( Tianero et al., 2019 ). The prevalence and diversity of AHLs being produced by the sponge bacterial isolates identified in this study suggests that mechanisms to inhibit these systems may also exist within the sponge microenvironment. The identification of novel compounds that are capable of inhibiting AHL based QS systems is one of the key areas of focus in the development of next generation antimicrobials. Previously, we have reported on the profiling of a subset of this collection of marine sponge isolates for quorum sensing inhibitory (QSI) or QQ activity. A total of 18/440 culturable isolates were found to encode QSI, being able to supress AHL signaling is an isolate dependent manner ( Gutierrez-Barranquero et al., 2017 ). It was interesting to note in that study that several species possessed dual QS and QSI activities. In this current study the finding that Psychrobacter sp. isolates from the same sponge collection were also capable of QS activity suggests that community level moderation of group behavior is a highly evolved trait in the marine ecosystem. Tan et al. (2015) previously showed how the dynamics of QS and QSI/QQ producing organisms can fluctuate in response to changes in environmental conditions. It is noticeable in this regard that two species cultured from QS treated sponge homogenate, Pseudomonas and Paenibacillus , are themselves known to possess AHL signaling systems ( Ma et al., 2016 ). Understanding the interplay between QS and QSI/QQ in the marine sponge ecosystem and the role of QQ in moderating community behavior will underpin advances in marine ecology and beyond. The dynamics of AHL production in marine microbial communities is seen as a mechanism to enhance culturability of rare genera, many of which encode valuable biosynthetic gene clusters for natural products such as antibiotics and anti-cancer drugs ( Reen et al., 2015 ). While co-culture with QS positive isolates did alter the profile of culturable bacteria isolated from marine sponge homogenates, they failed to introduce new genera into culture. This of course could be due to limitations in the culture conditions, including the general nature of the media used which is more conducive to the culture of fast-growing bacteria. Dilution based methods and modification of the growth conditions with regard to media, temperature, and time may provide the optimum conditions for culture of QS dependent organisms ( Rygaard et al., 2017 ). The absence of a LuxIR system in the QS positive Pychrobacter sp. 230 isolate would suggest that a hidden diversity to the molecular mechanisms underpinning QS signaling remains to be elucidated. This is consistent with previous reports of AinS and LuxM family autoinducer synthase enoding genes, quite distinct from their LuxI counterparts ( Venturi and Subramoni, 2009 ). Recently, a new LuxIR based system termed TswIR has been identified in an uncultured symbiont from the Red Sea Sponge Theonella swinhoei ( Britstein et al., 2016 ). The synthase protein TswI (COG3916) was annotated as both an autoinducer synthase and a GNAT acetyltransferase activity and while GNAT acetyltransferase proteins were identified in the Psychrobacter genomes, no members of the COG3916 family were found. Furthermore, the recent finding that LuxIR homologs can synthesize and respond to non-acyl HSL signals, serves to underscore the hidden complexity in these systems ( Ahlgren et al., 2011 ). Two orphan Photorhabdus LuxR proteins, PluR and PauR, sense alpha-pyrones and dialkylresorcinols, respectively ( Brameyer and Heermann, 2015 ). It is possible that other examples of non-AHL LuxR interactions may be uncovered in the future, something that would add greatly to the complexity of the signaling interactions as currently understood. The absence of homologs of these proteins in the Psychrobacter sp. 230 genome may necessitate a functional approach in order to elucidate the molecular mechanism through which AHL signaling is established in this and other marine genera."
} | 3,395 |
32145224 | null | s2 | 968 | {
"abstract": "High-throughput sequencing techniques such as metagenomic and metatranscriptomic technologies allow cataloguing of functional characteristics of microbial community members as well as their phylogenetic identity. Such studies have found that a community's makeup in terms of ecologically relevant functional traits or guilds can be conserved more strictly across varying settings than its composition is in terms of taxa. I use a standard ecological resource-consumer model to examine the dynamics of traits relevant to resource consumption, and analyze determinants of functional structure. This model demonstrates that interaction with essential resources can regulate the community-wide abundances of ecologically relevant traits, keeping them at consistent levels despite large changes in the abundances of the species housing those traits in response to changes in the environment, and across variation between communities in species composition. Functional structure is shown to be able to track differences in environmental conditions faithfully across differences in species composition. Mathematical conditions on consumers' vital rates and functional responses necessary and sufficient to produce conservation of functional community structure across differences in species composition in these models are presented. These conditions imply a nongeneric relation between biochemical rates, and avenues for further research are discussed."
} | 361 |
35575545 | PMC9239185 | pmc | 969 | {
"abstract": "ABSTRACT Ecotypic diversification and its associated cooperative behaviors are frequently observed in natural microbial populations whose access to resources is often sporadic. However, the extent to which fluctuations in resource availability influence the emergence of cooperative ecotypes is not fully understood. To determine how exposure to repeated resource limitation affects the establishment and long-term maintenance of ecotypes in a structured environment, we followed 32 populations of Escherichia coli evolving to either 1-day or 10-day feast/famine cycles for 900 days. Population-level analysis revealed that compared to populations evolving to 1-day cycles, 10-day populations evolved increased biofilm density, higher parallelism in mutational targets, and increased mutation rates. As previous investigations of evolution in structured environments have identified biofilm formation as the earliest observable phenotype associated with diversification of ecotypes, we revived cultures midway through the evolutionary process and conducted additional genomic, transcriptional, and phenotypic analyses of clones isolated from these evolving populations. We found not only that 10-day feast/famine cycles support multiple ecotypes but also that these ecotypes exhibit cooperative behavior. Consistent with the black queen hypothesis, or evolution of cooperation by gene loss, transcriptomic evidence suggests the evolution of bidirectional cross-feeding behaviors based on essential resources. These results provide insight into how analogous cooperative relationships may emerge in natural microbial communities.",
"conclusion": "Conclusions. Our results illustrate how long-term experimental evolution to cycles of feast and famine encourages rapid ecotypic diversification. Further analysis of a representative population reveals that evolved ecotypes may participate in cooperative behavior based on bidirectional cross-feeding, reminiscent of the black queen hypothesis. Here, ecotypes evolved mutations targeting costly processes such as fatty acid biosynthesis and iron sequestration, representing a substantial energy savings if these processes can be supplemented by environmental sources or via other microbial community members. Given that cooperative bidirectional cross-feeding is often observed in nature and that the investigation of how cooperative behaviors evolve has largely relied on engineered relationships ( 69 , 72 – 74 ), these populations evolved under 10-day feast/famine conditions represent a great resource for the study of how cooperative behaviors evolve de novo .",
"introduction": "INTRODUCTION Present in every colonizable habitat, microbes are the most resilient and ubiquitous organisms on the planet. This resilience is often attributed to the vast metabolic diversity present in microbial communities that allows these communities to withstand harsh conditions, such as prolonged resource limitation and rapid environmental disruptions. Although this metabolic diversity is often investigated on the species level by describing the roles of diverse taxa occupying a defined habitat, or microbiome, it has become clear that metabolic diversity can also evolve within a taxon and result in distinct subpopulations, or ecotypes. Microbiologists’ understanding of how ecotypic diversity evolves has been guided primarily by allopatric models, such as geographical or chemical barriers in a habitat ( 1 ) and spatially heterogeneous environments ( 2 ). However, more recently, studies describing sympatric diversification in relatively simple environments have drawn attention, as the metabolism of primary resources can result in the production of metabolic waste products, which make the environment much more complex ( 3 – 6 ). For example, the diversification and coexistence of ecotypes have been found in multiple cases of experimental evolution of Escherichia coli in glucose-limited media ( 7 – 10 ). In these cases, the evolution of incidental, or one-way, cross-feeding on metabolic waste products underlies ecotypic diversification, and the coexistence of the resulting ecotypes is thought to be maintained by negative frequency-dependent selection ( 11 – 14 ). Despite one-way cross-feeding repeatedly emerging in experimentally evolved cultures, the evolution of mutualistic, or bidirectional, cross-feeding is less commonly observed. In theory, the evolution of cooperative cross-feeding behaviors in a microbial population depends on the complementary changes of metabolic abilities between ecotypes and their combined fitness consequences, as the resulting combined fitness advantages can stabilize the cross-feeding interactions ( 15 , 16 ). These cross-feeding interactions can be further stabilized by gene loss, forcing the ecotypes to be codependent ( 17 , 18 ) in a manner described by the black queen hypothesis. Specifically, the black queen hypothesis proposes that reductive evolution involving the loss of biosynthetic abilities may provide benefits to the individual, as producing fewer metabolites and expressing fewer proteins potentially can save energy ( 19 ). However, a loss of metabolic ability could also be the product of neutral accumulation of degenerative mutations ( 20 ). Therefore, assessing the fitness of cross-feeders, in isolation and in the presence of their cross-feeding partners, will help us to better understand the evolutionary dynamics involved in the emergence of intraspecific metabolic diversity in microbial communities. Environmental and genetic context can also potentially affect the likelihood of ecotypic diversification and the emergence of cross-feeding behaviors. Microbes frequently encounter environments with fluctuating resource availability ( 21 , 22 ). For example, as an opportunistic pathogen, E. coli has a broad habitat ranging from soil and wastewater to the lower gut, often oscillating between feast and famine ( 23 – 25 ). Of the conditions encountered during feast/famine fluctuations, starvation has been observed to encourage diversification ( 26 , 27 ). However, it is less known how diversity is affected by environmental disruptions, such as the rapid replenishment of resources during feast/famine cycles. It has also been suggested that plasticity in the induction of stress responses can induce changes in molecular phenotypes such as mutation rates ( 28 – 30 ). If the evolution of cross-feeding depends on the spontaneous introduction of rare beneficial mutations, a higher mutation rate may facilitate efficiency in exploring the fitness landscape and rapid establishment of ecotypes due to the arrival of diverged lineages at isolated adaptive peaks ( 31 – 33 ). Thus, it is of particular interest to test how microbial populations respond to combinations of different feast/famine cycles and initial base genetic mutation rates. To study the effect of different feast/famine cycles on the evolution of diversity in microbial populations, we experimentally evolved E. coli in culture tubes containing LB broth under two different feast/famine cycle conditions: fresh LB broth supplied every 1 and 10 days. In addition, to understand how differences in genetic mutation rates further affect evolutionary dynamics, we utilized ancestral lines with two initial genetic backgrounds: a wild-type (WT) strain and a WT-derived strain with impaired methyl-directed mismatch repair (MMR − ; yielding an ~150× increase in the single nucleotide mutation rate by deleting mutL ) ( 34 ). Each combination of genetic background and feast/famine cycle condition was replicated in eight parallel replicates, which resulted in 32 experimental populations (8 × 2 × 2). Polymorphism data from periodic whole-genome sequencing allowed a general survey of subpopulation structure, parallelism within each treatment combination, and divergence between treatments. We also quantified traits that may have distinct patterns in evolutionary responses to feast/famine cycles, including biofilm formation and mutation rates. Using genomic and transcriptomic data of multiple single clones from the same population, we further proposed a detailed model on the emergence of cross-feeding and ecotypic diversification.",
"discussion": "DISCUSSION Through this study, we found that ecotypic diversification repeatedly evolves in Escherichia coli populations cultivated in 10-day feast/famine cycles. While long coexistence of diversified clades is a phenomenon that was also previously observed in populations cultivated in 1-day feast/famine cycles, 10-day-cycle populations additionally evolve to produce even thicker biofilms and harbor higher base genetic mutation rates than both the ancestral strains and the 1-day cycle populations. Together, the increases in evolved trait values and the genomic evidence reveal a stronger signal of mutational parallelism under 10-day feast/famine conditions. The 10-day feast/famine conditions provide a different and potentially more challenging environment for evolving E. coli populations than the 1-day feast/famine conditions, which might account for the increased mutational parallelism despite experiencing 1/10th fewer generations. It is important to note that the starvation conditions of 10-day feast/famine cycles also induce differences in population-size dynamics and mutation rate compared to 1-day conditions. Therefore, the effect of starvation, population size change, rate of evolution, or any combination of these might explain the unique evolutionary outcomes in 10-day cycles. Further investigation of clones isolated from a 10-day population illustrates that ecotypic diversification is rapid, results in individual fitness trade-offs with additive effects, and is based on differential investment in the biosynthesis or acquisition of key resources, such as fatty acids, iron, and phosphate. These individual fitness trade-offs are distinct to 10-day feast/famine cycles, as isolated clones from 1-day cycle populations all exhibited fitness increases in their evolved environment independent of being cocultured with their evolved community members ( 35 ). Thus, the stress invoked due to extended resource limitation experienced in 10-day cycles appears to encourage the evolution of bidirectional cost-sharing for metabolically expensive processes. Prior to this study, there were few investigations into how adaptation to resource limitation is shaped by feast/famine cycles ( 53 , 54 ). Instead, interpretations of how E. coli adapts to increasingly diminishing resources have been drawn primarily from studies focused exclusively on the famine component ( 55 , 56 ), such as investigations of the growth advantage in stationary phase (GASP) phenotype. However, in analysis of how the eventual feast component factors into adaptation, one can only observe mutations that allow E. coli to postpone death during starvation and famine conditions. As resources must eventually be replenished in nature or the population will ultimately go extinct, it is important to examine adaptation to starvation in the context of subsequent resource replenishment, which helps identify the starvation-associated traits and alleles that are beneficial when resources are scarce but are not overly detrimental when resources are abundant. A comparison of the genes that experienced parallel mutations during 10-day feast/famine cycles ( Fig. 2B and C ) to the genes reported as targets of parallel mutation in previous studies of long-term starvation ( 30 , 57 ) revealed 22 genes that overlap between two or more studies ( Fig. 5 ). Of these overlapping genes, four ( lrp , paaX , proQ , and putA ) were exclusively observed as targets of parallel mutation in studies of long-term starvation. Thus, mutations in these genes may be deleterious upon resource replenishment or may only confer benefits during deep starvation. Alternatively, 18 genes were targets of parallel mutation in both 10-day feast/famine cycle and under long-term starvation conditions, including global regulators and genes with broad effects on transcription and protein expression (i.e., crp , fusA , ompR , rpoS , and rpoB ). The continued presence and eventual fixation of these mutations following multiple 10-day resource replenishment cycles suggests that these mutations have a net benefit over the complete replenishment cycle. FIG 5 Overlapping targets of parallel mutation following evolution of E. coli populations to long-term starvation and 10-day feast/famine cycles. Venn diagram showing the number of overlapping genes experiencing parallel mutations as a result of evolution to 10-day feast/famine cycles (this study; blue) and long-term starvation (Ratib et al. [ 57 ] and Katz et al., 2021( 30 ); gray). Circles representing each gene set are not drawn to scale, and genes that overlap between multiple studies are listed in the offset labels. In addition to parallel mutations, we observed evolved changes affecting multiple traits that have been associated with adaptation to starvation in prior studies. One such trait is a trade-off between growth and survival under starvation conditions ( 30 , 58 ). When determining the time at which 10-day clones and the WT ancestor reach mid-log phase to ensure that our differential expression analysis would capture physiologically relevant comparisons, we observed that the midpoint of logarithmic growth occurred 105 min later for the evolved 10-day-cycle clones (see Fig. S2 in the supplemental material). Despite this more languid growth, both evolved 10-day-cycle clones exhibited increased starvation tolerance in coculture assays and outnumbered the WT ancestor by 4 days posttransfer ( Fig. 3C and D ). Recent study of E. coli growth and survival in batch culture compared to chemostat culture, where growth rates can be controlled, revealed that lower growth rates translated to lower death rates, as these cells required fewer resources per unit of time to maintain viability ( 50 ). Moreover, in addition to evolving lower growth rates, 10-day clones evolved changes in gene expression that were largely biased toward reduced expression and presumably result in lower cellular maintenance costs. A second trait that exhibited significant differences following evolution to feast/famine cycles was the base genetic mutation rate. Here, we observed that increased mutation rates occurred much more frequently in populations under 10-day cycles, a presumably more stressful environment, than 1-day cycles. Further, as most initially WT populations eventually evolved defects in MMR during evolution to 10-day feast/famine cycles, we observed few differences between initially WT and initially MMR − populations. Consistent with our results, the emergence of high mutation rates has been found in many other microbial evolution experiments in response to various conditions ( 29 , 59 , 60 ). Population-genetic models have also suggested that higher mutation rates can evolve in fluctuating environments by hitchhiking with other beneficial mutations ( 61 , 62 ). Mutators are also expected to be at their most prevalent when environmental fluctuations occur with intermediate rates of change, in contrast to rapid fluctuations where the benefit of a mutator is lost when environmental changes occur faster than the beneficial mutation rate ( 63 ). Under all conditions, however, as a population becomes better adapted to its environment, the pool of available beneficial mutations shrinks. At this point, the fitness costs associated with high mutation due to an increased fraction of deleterious mutations can outweigh the benefits of a rare beneficial mutation, ultimately leading to natural selection favoring compensatory evolution and reversion to lower mutation rates ( 64 ). Determining whether that will be also the case in the evolved 10-day feast/famine populations will require future follow-up as these populations continue to evolve under these fluctuating resource conditions. Other traits that are being increasingly associated with microbial adaptation to resource limitation are diversification and spatial organization. In response to 10-day feast/famine cycles, evolving populations exhibited rapid diversification and increased biofilm formation activity. This result echoes what has been observed in other studies, in which various degrees of resource limitation exhibited positive effects on spatial structure and diversity. Studies of patterning in bacterial colonies have revealed that spatial organization is a common outcome of resource limitation due to bottlenecks during colony expansion, and the speed at which bacteria spatially organize increases with resource limitation ( 65 ). Under heterogeneous culture conditions, unimodal relationships between maintenance of diversity and resource availability have been described, with peak diversity ultimately observed in intermediate nutrient concentrations ( 66 ). Further similarities between our E. coli system and these two studies include the utilization of an experimental ancestor strain that expresses poor motility ( 67 ), which was observed to increase spatial patterning in Pseudomonas aeruginosa ( 65 ), and biofilm formation at the surface-air interface, which underlies the diversity observed in the heterogeneous Pseudomonas fluorescens system ( 66 ). Further study is needed to parse the importance of reduced motility and increased biofilm formation to the generation and maintenance of diversity across resource-limited conditions. As a result of diversification, evidence of frequency-dependent cooperative behavior was observed between clones isolated from evolved populations. Here, isolated clones exhibited growth disadvantages when independently cocultured with the WT ancestor and when pooled at a 1:1 ratio. However, these growth disadvantages were minimized when a revived sample of the population was cocultured with the WT ancestor, with the ecotypes competing at a more favorable 1:2 ratio, and competition of the ecotypes against each other shows that the individual ecotypes exhibit increased selection rates when rare. As both ecotypes are viable in isolation, they have not evolved obligate dependence, and any evolved changes contributing to this cooperative behavior are likely to increase the metabolic efficiency of the population but at a cost to individual fitness. Mediating these costs to individual fitness has proven challenging to evolutionary biology. However, a relatively recent hypothesis, called the black queen hypothesis, has emerged as a leading theory describing the evolution of cooperation by genomic reduction or gene loss ( 19 ). Here, mutualism can arise based on overproduction of metabolic intermediates that leak from the cell, providing a public good. In response, other community members may lose the ability to biosynthesize (or reduce their investment in the biosynthesis of) the now freely available public good. If this exchange is bidirectional, then codependence may evolve. One major caveat to the black queen hypothesis is an innate vulnerability of mutualistic cross-feeding relationships to invasion by cheaters that benefit from the public goods without contributing ( 68 ). As such, mutualistic cross-feeding is expected to be (i) the most stable in structured environments ( 69 ), where access to the public goods can be limited; (ii) subject to Allee effects ( 70 ), where an optimal ratio of each cooperator type exists, maximizing metabolic efficiency and population size while also reducing an excess availability of public goods that may otherwise encourage the evolution of cheaters; and (iii) beneficial under resource-limited conditions ( 15 ), where cooperation can help maintain genetic variation. All of these conditions appear to be true for our experimental populations, as 10-day feast/famine cycles result in increased biofilm formation and an optimum 1:2 ratio exists between ecotypes for maximum population fitness. Lastly, the metabolic basis of cross-feeding observed between evolved ecotypes is likely parallel across our experimental populations, representing a common evolutionary outcome of culture under 10-day feast/famine cycle conditions. Differential expression analysis revealed that ecotype A evolved significantly reduced expression of essential fatty acid biosynthesis genes, fabD and fabH , likely due to the polar effect of an IS element insertion in the upstream gene, plsX . Similarly, fabH and plsX were common targets of mutations in our other 10-day populations, with fabH mutations arising in eight populations and plsX mutations arising in six populations ( Fig. 2 and Table S7 ). In a complementary fashion, ecotype B evolved significantly reduced expression of enterobactin exporter entS and other genes regulated by Fur-Fe 2+ during long-term stationary phase. Across the other 10-day populations, fur was also a common target of mutation (10 populations), with three populations containing nonsynonymous mutations resulting in substitutions of the R110 residue, which is proximal to Glu108 of Fur’s metal binding site 1 ( 71 ). The potential involvement of long-chain fatty acids and iron sequestration in cross-feeding mutualisms is significant, as both of these processes are energetically costly, specifically when considering that synthesizing palmitate (C 16:0 ), the precursor to other long-chain fatty acids, costs 8 acetyl-coenzyme A (CoA), 14 NADPH, and 7 ATP molecules ( 46 ). Thus, the evolution of cross-feeding to minimize the investment in palmitate production is consistent with the presumption of the black queen hypothesis that cross-feeding will evolve to conserve energy wasted by redundant synthesis of costly metabolites. Further study is needed to dissect the metabolic interactions between these evolved ecotypes as well as the stepwise process that leads to these cooperative interactions and their long-term evolutionary fate within populations. 10.1128/mbio.03467-21.10 TABLE S7 Mutations observed in plsX and fabH across evolved populations. Download Table S7, XLSX file, 0.01 MB . Copyright © 2022 Behringer et al. 2022 Behringer et al. https://creativecommons.org/licenses/by/4.0/ This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Conclusions. Our results illustrate how long-term experimental evolution to cycles of feast and famine encourages rapid ecotypic diversification. Further analysis of a representative population reveals that evolved ecotypes may participate in cooperative behavior based on bidirectional cross-feeding, reminiscent of the black queen hypothesis. Here, ecotypes evolved mutations targeting costly processes such as fatty acid biosynthesis and iron sequestration, representing a substantial energy savings if these processes can be supplemented by environmental sources or via other microbial community members. Given that cooperative bidirectional cross-feeding is often observed in nature and that the investigation of how cooperative behaviors evolve has largely relied on engineered relationships ( 69 , 72 – 74 ), these populations evolved under 10-day feast/famine conditions represent a great resource for the study of how cooperative behaviors evolve de novo ."
} | 5,827 |
32483306 | PMC7608354 | pmc | 971 | {
"abstract": "Biofilms are closely packed cells held and shielded by extracellular matrix composed of structural proteins and exopolysaccharides (EPS). As matrix components are costly to produce and shared within the population, EPS-deficient cells can act as cheaters by gaining benefits from the cooperative nature of EPS producers. Remarkably, genetically programmed EPS producers can also exhibit phenotypic heterogeneity at single-cell level. Previous studies have shown that spatial structure of biofilms limits the spread of cheaters, but the long-term influence of cheating on biofilm evolution is not well understood. Here, we examine the influence of EPS nonproducers on evolution of matrix production within the populations of EPS producers in a model biofilm-forming bacterium, Bacillus subtilis . We discovered that general adaptation to biofilm lifestyle leads to an increase in phenotypical heterogeneity of eps expression. However, prolonged exposure to EPS-deficient cheaters may result in different adaptive strategy, where eps expression increases uniformly within the population. We propose a molecular mechanism behind such adaptive strategy and demonstrate how it can benefit the EPS producers in the presence of cheaters. This study provides additional insights on how biofilms adapt and respond to stress caused by exploitation in long-term scenario.",
"introduction": "Introduction Cooperative interactions are prevalent for all life forms [ 1 ], even for simple microbes that often exist in communities of matrix bound surface-attached cells called biofilms [ 2 – 6 ]. However, when costly products such as siderophores [ 7 , 8 ], extracellular polymeric substances [ 9 , 10 ], digestive enzymes [ 11 ], and signaling molecules [ 12 , 13 ] are secreted and shared, cooperative behavior becomes susceptible to cheating [ 2 , 14 – 16 ], where mutants defective in cooperation can still benefit from cooperative community members [ 4 , 5 , 17 ]. It has been shown that spatially structured biofilms, where interactions with clone mates are common and diffusion of public goods is limited, may serve as natural defense against cheating [ 18 – 20 ]. However, long time scale studies have recently reported that biofilm defectors can spontaneously emerge and spread in biofilms by exploiting other matrix-proficient lineages [ 21 – 24 ]. In fact, a pioneering microbial evolution study on Pseudomonas fluorescens has already pointed toward dynamic evolutionary interplay between cooperation and exploitation in a biofilm mat [ 25 ], where emergence of cellulose overproducer (wrinkly) allowed mat formation, but also created an opportunity for exploitation by nonproducers (smooth), eventually leading to so-called “tragedy of the commons” [ 4 , 26 , 27 ]. Taken together, biofilms are a suitable model to understand social interactions in an evolutionary time scale [ 23 , 28 – 31 ]. Modeling and empirical data confirm that mutualism (beneficial to both actor and recipient) and altruism (beneficial to recipient but not to actor) play crucial roles in biofilm enhancement [ 32 ], but at the same time can lead to biofilm destabilization [ 25 ]. Can cooperators evolve tactics to evade exploitation and, in turn, can cheats utilize evolution to enhance their selfish actions? Recent studies showed that in well-mixed environment, cooperators adapt to cheats by reducing cooperation [ 14 , 15 , 33 ]. Such reduction could be achieved by various strategies, for instance decrease in motility [ 15 ], downregulation or minimal production in public goods [ 14 , 15 , 33 ], upregulation of other alternative public goods [ 14 ], or bi-stable expression in virulence gene [ 2 ]. Interestingly, populations of cooperators often exhibit phenotypic heterogeneity at the single-cell level [ 34 , 35 ]. Therefore, an alternative and simple mechanism to modulate levels of cooperation in a population would be through changes in phenotypic heterogeneity pattern. However, the long-term effects of cheats on costly goods’ expression at individual cell level have never been examined. Understanding how heterogeneity of gene expression within the population is affected in the presence of cheats would provide better insight on microbial adaptation and stress response mechanisms. Here, we address this question using pellicle biofilm model of Bacillus subtilis ( B. subtilis ) [ 36 , 37 ]. Pellicle formation in B. subtilis involves, among others, aerotaxis-driven motility and subsequent matrix production [ 38 ]. Aerotaxis is important for oxygen sensing to aid cells reach the surface, while matrix formation is significant to sustain cells to adhere to the surface and to each other. Exopolysaccharide (EPS) is a costly public good in B. subtilis biofilms [ 10 , 18 , 39 ] and is heterogeneously expressed during biofilm formation with ~40% of cells exhibiting the ON state [ 39 , 40 ]. We aimed to investigate the cheat-dependent alteration related to phenotypic heterogeneity in eps expression by the producer. We reveal that cheating mitigation by the EPS producers involves a shift in phenotypic heterogeneity toward stronger eps expression, which can be achieved by a loss-of-function mutation in a single regulatory gene. Our study uncovers an alternative anti-cheating mechanism based on changes in public goods’ expression pattern and highlights meandering trajectories prior cooperation collapse.",
"discussion": "Discussion Studies on evolution of cooperative behavior are important to understand how social behaviors are shaped in longer time scale. Moreover, exploring long-term consequences of exposure to cheating allows to better understand how cooperation prevails in nature where environmental stress and exploitation exist inherently. Here, we took a reductionist approach, focusing on evolution of a single cooperative trait—the expression of eps , which plays a crucial role in biofilm lifestyle of B. subtilis and other bacteria. As we focused on the single-cell level expression of eps in multiple single strain, isolated from the ancestral or evolved populations, we could obtain a multi-level insight into evolutionary changes in eps expression. Our study revealed previously observed population-level diversification of matrix genes expression, indicating the strain-independence and reproducibility of adaptation in biofilms [ 21 , 24 ]. Strikingly, next to population-level diversification, we also observe an increase in phenotypic heterogeneity of eps expression within single isolates. Based on coculture studies performed for WT and Δ eps (this work), as well as for WT and spontaneously evolved biofilm-deficient lineage [ 21 ], we believe that low- eps subpopulations may be acting as conditional cheater, supported by “hyper-cooperative” subpopulations of high eps . The observation that isolates with the most pronounced phenotypic diversification pattern tend to show lower average eps expression, clearly depicts the latter as a consequence of diversification, and specifically emergence of low eps . It remains to be determined whether increased/reduced levels of eps expression translate into higher/lower amount of released EPS, but based on previous studies it is likely to be the case [ 44 ]. Previous evolution studies on cheater-cooperator interactions in spatially structured environment showed cheater mitigation via minimization of the cooperative trait [ 2 , 14 , 15 ]. On the contrary, here we show that cooperators respond to cheating by intensifying the cooperative behaviors through uniform shift toward higher eps expression. Further molecular analysis of the high- eps isolates strongly suggests that this phenotype is triggered by loss-of-function mutation in rsiX gene. The product of rsiX represses the activity of ECF sigma factor, SigX that is involved in cell envelope stress response against cationic antimicrobial peptides [ 47 ]. Importantly, SigX has been previously shown to induce expression of epsA-O in B. subtilis via a complex regulatory pathway involving Abh and SlrR [ 48 ], explaining the observed enhanced in eps gene expression in rsiX mutant. Another example of matrix overproduction via ECF adaptation was also reported in gram-negative bacterium Pseudomonas aeruginosa ( P. aeruginosa ) where mutations in another ECF called AlgT led to alginate overproduction and increased resistance to antimicrobials [ 49 ]. Therefore, adaptive boosts in matrix production through modulation of ECF are not exclusive for B. subtilis , but seem to occur also in medically relevant gram-negative pathogens like P. aeruginosa . In contrast to previous studies that addressed long-term cheating on diffusible siderophores [ 50 – 53 ], we explored evolutionary interplay between biofilm producers and nonproducers in structured environment. Our results support previous observations on evolution of specific cheating-resisting mechanisms in cooperators, pointing toward ubiquity of this phenomenon. In addition, our work brings up three major findings: (1) matrix producers can adapt to matrix nonproducers by shifting phenotypic heterogeneity toward increased levels of matrix expression, (2) high- eps phenotype is associated with favorable positioning of the matrix producers in the biofilm in presence of cheats, thereby limiting their numbers, (3) high- eps anti-cheating strategy is a short-term solution followed by tragedy of the commons. As EPS-deficient strain took over in all but two mixed populations (including populations, without rsiX mutation and homogenous shift toward higher eps expression), we do not interpret the collapse as a direct consequence of mutation in rsiX gene. However, we argue that an emergence of several matrix overproducing lineages, may facilitate the spread of cheats [ 21 ], especially if a substantial number of cells within the high- eps lineage serves as facultative (phenotypic) cheaters. As recently demonstrated, EPS deficiency is not a dead-end strategy for B. subtilis population, because alternative EPS-independent biofilm formation strategies can emerge by single amino acid change is TasA [ 44 ]. It remains to be discovered whether shifts in phenotypic heterogeneity in response to long-term cheating is general phenomenon that applies to different types of public goods."
} | 2,581 |
32415218 | PMC7228921 | pmc | 972 | {
"abstract": "The ability to efficiently analyze the activities of biological neural networks can significantly promote our understanding of neural communications and functionalities. However, conventional neural signal analysis approaches need to transmit and store large amounts of raw recording data, followed by extensive processing offline, posing significant challenges to the hardware and preventing real-time analysis and feedback. Here, we demonstrate a memristor-based reservoir computing (RC) system that can potentially analyze neural signals in real-time. We show that the perovskite halide-based memristor can be directly driven by emulated neural spikes, where the memristor state reflects temporal features in the neural spike train. The RC system is successfully used to recognize neural firing patterns, monitor the transition of the firing patterns, and identify neural synchronization states among different neurons. Advanced neuroelectronic systems with such memristor networks can enable efficient neural signal analysis with high spatiotemporal precision, and possibly closed-loop feedback control.",
"introduction": "Introduction In nervous systems, the collective neuron activities and firing patterns control the function, consciousness, and memory formation 1 . Revealing the features encoded in neural spike trains from the neural network will significantly advance our understanding of the working mechanism of the nervous system 2 , 3 . Neural probe technologies, such as patch clamp 4 , nanowire-based field effect transistor 5 , microelectromechanical (MEMS) probes 6 , and complementary metal–oxide–semiconductor (CMOS) nanoelectrode arrays 7 , are often used to record intracellular and intercellular electrophysiological activities, i.e. neural spikes, from biological neurons. Data recorded over time and over different recording sites are then transmitted and stored, and later processed in an external data-processing system that may include conventional computers and more recently artificial neural networks (ANNs) 8 , 9 for analysis. Transmitting, digitizing, and storing the vast amounts of data pose severe power and throughput constraints on the neural probe design 10 – 12 , while the offline processing is a time-consuming process that do not allow real-time analysis. The ability to directly process neural signals at the recording sites, without having to go through pre-processing and storage, will significantly expand the capabilities of the neural probes and enhance our understanding of nervous systems, with the potential for real-time neural activity analysis and feedback in a closed loop. Reservoir computing (RC) is a concept originally developed from recurrent neural networks (RNNs) 13 and has recently been successfully used to implement a broad range of tasks, such as image pattern recognition, time series forecasting, and pattern generation 14 , 15 . Briefly, a reservoir is a dynamic system that can perform nonlinear transformations of the input signals, and project them to a high-dimensional space (represented as the reservoir states). The nonlinear transformation allows the original features, often in time domain, to be mapped as features in the reservoir states which can then be further processed by a small, trained linear neural network (termed the readout layer) 14 . A key feature of the reservoir is the fading memory 14 , i.e. short-term memory property, which states the reservoir state depends not only on the present inputs but also on inputs from the recent past (but not the far past). The fading memory effect is key for an RC system to extract and analyze temporal features in the input data. In particular, since information in neural spike trains is mainly encoded in the temporal domain, we expect neural signals to be well suited for RC hardware systems. More specifically, dynamic memristors with inherent short-term memory effects have recently been successfully utilized as reservoirs for temporal data processing 15 . Beyond benefits such as having a simple device structure that allows easy fabrication and integration, the characteristics of a memristor device can be tailored by carefully engineering the switching material and optimizing the device structure 16 – 19 , allowing one to develop memristors with desired operation voltages and dynamics for different applications. To this end, memristor-based RC systems offer intriguing opportunities to be integrated with the neural probe for on-site, real-time neural signal processing. In this work, we experimentally demonstrate the possibility of neural data analysis using a memristor-based RC system. A perovskite halide-based memristor with very low switching voltage (<100 mV) and current (1–100 nA) and an inherent short-term memory effect is developed and used as the reservoir. Using emulated neural spike signals, we show that an RC system based on such devices can potentially directly process neural spikes, and is capable of implementing important tasks, such as real-time recognition of neural firing patterns and neural synchronization states.",
"discussion": "Discussion In this work, we aim to show that memristor-based reservoirs with tailored operation conditions can be obtained by carefully choosing the switching material and designing the device structure. These systems and devices in turn create new opportunities for critical applications, such as neuroscience and engineering. Specifically, we demonstrated that by choosing proper ion species, ion migration energy barrier, and the intrinsic defect density, one can develop dynamic memristors, e.g. Ag/CsPbI 3 /Ag, with ultralow switching voltage (<100 mV) and switching current (~nA), where both the low programming voltage and low programming current are necessary if such devices need to be directly interfaced with biological neural signals. We further show that with these unique properties, these devices can be potentially directly driven by actual biological neural spikes and interfaced with neural probes for neural firing pattern recognition and neural synchronization analysis. Emulated neural spiking patterns were used for training and analysis in this study. Considering that actual neural spike patterns may contain diverse and complex temporal features at different timescales, dynamic memristors with different relaxation rates are desired and need to be carefully designed. Improvements in the readout layer, as well as training algorithms are excepted to further enhance the RC system performance. By integrating the low-voltage RC system with recently developed intracellular neural probes, for instance, large-scale nanoelectrode arrays that can simultaneously perform intracellular recordings from thousands of connected mammalian neurons 7 , 36 , real-time analysis of the interactions among many neurons in a large biological neural network may become feasible. Such large-scale implementations, however, would still require the development of new RC computing algorithms and the optimization of the memristor hardware. Beyond electrophysiological data recording and processing, the RC artificial network, which was originally inspired by neurobiology and offers functionalities resembling that of biological systems 37 , can possibly play a more active role though the interactions with the biological neural network 38 . For example, with the ability to be directly excited by neural spikes, these electronic networks may be considered as an extension of the biological neural network, and offer additional resources for tasks such as recognition and memory formation."
} | 1,892 |
39416694 | PMC11446362 | pmc | 973 | {
"abstract": "Addressing urgent environmental challenges, this commentary emphasizes the need for green, bio-based solutions in chemical production from renewable feedstocks. It highlights advanced metabolic engineering of microbial strains and the use of microbial consortia as innovative approaches for efficient resource recovery. These strategies aim to enhance the conversion of diverse renewable feedstocks, including agricultural residues, industrial by-products, and greenhouse gases, into value-added chemicals. This article discusses cutting-edge techniques in renewable feedstock upcycling, utilizing both engineered unicellular and multicellular systems. It advocates a paradigm shift in sustainable biomanufacturing, focusing on transforming renewable resources into value-added products. This approach is crucial for developing a circular bioeconomy, aligning with global efforts to mitigate environmental impacts.",
"conclusion": "4 Future perspective and conclusion The integration of microbial consortium-based approaches in renewable feedstock upcycling presents a groundbreaking strategy in sustainable bio-manufacturing, aligning with the goals of a circular bioeconomy. The single-strain approach, with its focus on metabolic pathway manipulation within individual strains, has demonstrated significant advancements due to the rapid evolution of synthetic biology and genetic engineering. Conversely, the microbial consortium-based approach utilizes the synergistic potential of multiple microbial species, offering a robust and versatile method for converting diverse renewable feedstock streams into valuable products. Moving forward, future research needs to focus on optimizing these approaches and maximising their synergy for higher efficiency and scalability. This includes improving the conversion rates of renewable by-product materials and designing systems that can be effectively scaled from the laboratory to industrial levels. In addition, expanding the range of feedstocks that can be repurposed is crucial which includes exploring the potential of these technologies to transform not only agricultural residues, discarded food and industrial by-products but also emerging expended material streams such as used textiles, 27 , 28 rubber 29 and plastic. 30 The field of synthetic biology is rapidly evolving, leveraging cutting-edge tools and techniques in genetic engineering, designing genetic biosensor, rewriting metabolic blueprints and incorporating machine learning. 31 , 32 , 33 These advances can lead to the development of more robust and efficient microbial strains capable of processing complex feedstocks. Furthermore, it is essential to assess the economic viability and environmental sustainability of these technologies which involves conducting life cycle analyses and cost-benefit assessments to ensure that these approaches are both environmentally sustainable and economically feasible. More importantly, developing a supportive regulatory framework and fostering public acceptance are key for the successful implementation of renewable feedstock upcycling technologies which can be achieved by addressing safety concerns, ethical considerations, and ensuring transparency in the development and deployment of these technologies. Finally, the complexity of renewable feedstock upcycling processes demands a collaborative and interdisciplinary approach, involving microbiologists, engineers, environmental scientists, and policymakers. Such collaborations can lead to more innovative solutions and effective implementation strategies. With continued research, innovation, and collaborations among countries where synthetic biology is flourishing, 34 , 35 , 36 , 37 , 38 , 39 , 40 technologies will undoubtedly be developed to contribute to resource conservation, environmental protection, and the transition towards a more sustainable and circular economy.",
"introduction": "1 Introduction In the pursuit of sustainable biomanufacturing, the upcycling of renewable feedstocks into value-added products is increasingly recognized as a critical component of a circular bioeconomy. 1 This approach mitigates the environmental impact of waste disposal and contributes to resource conservation and economic growth. 2 Among the various technologies for renewable by-product upcycling, microbial-based approaches are particularly noteworthy. These include both single-strain and consortia of multiple strains, each providing distinct and efficacious methodologies. The single-strain-based approach focuses on manipulating the metabolic pathways within a strain to enhance its ability to process both native and non-native feedstocks effectively, transforming them into value-added compounds. 3 This field has witnessed substantial advancements, driven by advances in metabolic engineering, and systems and synthetic biology. 4 , 5 On the other hand, the microbial consortium-based approach leverages the synergistic interactions of multiple microbial species to enhance the conversion of feedstocks into value-added products. 6 This method addresses some of the intrinsic limitations associated with single-strain-based engineering. It distributes complex metabolic pathways among different strains, optimizing the utilization of diverse renewable feedstock streams. 7 These two complementary approaches present compelling opportunities for the transformation of a broad spectrum of renewable materials — encompassing agricultural waste, industrial by-products, and atmospheric greenhouse gases such as CO and CO 2 into biofuels, pharmaceuticals, and value-added chemical products. In this commentary, we explore the latest advancements, applications, and prospects of metabolic pathway engineering and microbial consortium engineering in the realm of renewable feedstock upcycling, offering insights into their pivotal role in fostering future sustainable biomanufacturing strategies."
} | 1,474 |
19125816 | null | s2 | 975 | {
"abstract": "Biofilms transform independent cells into specialized cell communities. Here are presented some insights into biofilm formation ascertained with the best-characterized strain, Escherichia coli. Investigations of biofilm formation and inhibition with this strain using whole-transcriptome profiling coupled to phenotypic assays, in vivo DNA binding studies and isogenic mutants have led to discoveries related to the role of stress, to the role of intra- and interspecies cell signalling, to the impact of the environment on cell signalling, to biofilm inhibition by manipulating cell signalling, to the role of toxin/antitoxin genes in biofilm formation, and to the role of small RNAs on biofilm formation and dispersal. Hence, E. coli is an excellent resource for determining paradigms in biofilm formation and biofilm inhibition."
} | 207 |
34137988 | PMC7770917 | pmc | 976 | {
"abstract": "Highlights \n Owing to the great robustness, continuous conductivity, and geometric construction of a steel wire electrode, the FST–TENGs demonstrate high stability, stretchability, and even tailorability. By knitting several FST–TENGs to be a fabric or a bracelet worn on the human body, it enables to harvest human motion energy. The FST–TENGs can also be woven on dorsum of glove to monitor the movements of gesture. \n Electronic supplementary material The online version of this article (10.1007/s40820-019-0271-3) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusions In summary, a spiral steel wire-based TENG has been proposed for harvesting human motion energy and inspecting gesture. A single FST–TENG with the length of 6 cm and diameter of ~ 3 mm worked in the single-electrode mode at 2.5 Hz enables to generate V oc of ~ 59.7 V, Q sc of ~ 23.7 nC, maximum I sc of ~ 2.67 μA and average power of ~ 2.13 μW, respectively. Moreover, the FST–TENG can be used even after been twisted, bended and tailored into designed shape for enhancing user experience without impairing its performance. After knitting the FST–TENGs to an energy-harvesting FST–TENG fabric worn on the human body or a bracelet worn on wrist, the FST–TENG fabric with 1 × 1 knitting 12 fibers were demonstrated to light up at least 15 LEDs and charge a commercial capacitor to 2 V in ~ 68 s, and then drive an electronic watch. Meanwhile, the FST–TENG bracelet can light up 10 LEDs in series by hand tapping as well. With these excellent performances, the highly stretchable FST–TENG fabrics can be used as a wearable and convenient power source to harvest human motion energy for wearable electronics. Besides that, the FST–TENG can be woven on dorsum of glove to monitor the movements of gesture, which can recognize every single finger, different bending angle and numbers of bent finger by the voltage signals.",
"introduction": "Introduction The progress in human’s life has benefited from the rapid development of multipurpose wearable electronics, such as smart watches, smart glasses, electronic skins, etc. [ 1 , 2 ]. These electronics could mimic the characteristics of human skin and wireless sensing networks that monitor human health and motion track [ 3 – 5 ]. Several challenges have been raised for wearable devices that are desired to be flexible, lightweight, low price, and still stable [ 6 , 7 ]. Nevertheless, commercialized portable energy storage units including batteries and supercapacitors are relatively heavy, charged frequently, and face critical lifetime limitation [ 8 – 10 ]. To overcome these challenges, a series of advanced energy-harvesting technologies based on triboelectric, photovoltaic, and thermoelectric effects from ambient environment for sustainable and portable power source have been developed [ 11 – 15 ]. Compared with solar energy and thermal energy, mechanical energy harvesting is whenever and wherever possible that is independent of weather and environment. Triboelectric nanogenerator (TENG), based on coupling effects of triboelectrification and electrostatic induction, has been one of the most effective strategies to convert various types of mechanical energies into electricity, such as human motions, wind, water wave, rain drops, and vibrations [ 16 – 21 ]. It has achieved rapid progress as a sustainable power source with advantages of low-cost, high output, lightweight, and wide choice of materials for the usage in extensive devices [ 22 – 26 ]. Attempts have been made to fabricate flexible TENGs for potential applications in wearable power sources [ 27 – 31 ]. The ability to withstand complex mechanical deformation is limited owing to the Young’s modulus mismatch of interface compatibility of triboelectric layer and electrode layer. Recently, there have been three general strategies to make stretchable conductors to match stretchable triboelectric materials such as polydimethylsiloxane (PDMS) or silicone rubber [ 32 ]: including deterministic geometries of rigid materials to elongate, dispersing conductive particles in elastomer, and utilizing conductive materials that are intrinsically stretchable [ 28 , 33 – 35 ]. However, for the stretchable fiber-based power textiles, it is needed to be realized by designing reasonable geometrical shape. In this work, we present and tailorable (FST–TENG) to harvest human motion energy for wearable electronics. The spiral steel wire is selected as the electrodes, which is highly stretchable due to its special structure, while silicone rubber is employed to cover the spiral steel wire as the triboelectric layer. The geometric design enables every single FST–TENG to be stretchable and relatively stable within the range of ~ 50% stretch. In addition, fabrics and bracelet that worn on the body or wrists knitted with several FST–TENG fibers have been fabricated. To evaluate the performance of the FST–TENG fabric on power generation, the FST–TENG fabric is used to light up LEDs, charge commercial capacitors, and then drive an electronic watch for demonstration. Finally, the FST–TENG can also be applied for inspecting each single finger, different bending degree and identifying digital gestures by the smart glove with FST–TENGs attaching.",
"discussion": "Results and Discussion The schematic illustration and mechanical behaviors of the FST–TENG are schematically illustrated in Fig. 1 . Here, silicone rubber was chosen as the triboelectric and packaging materials of the FST–TENG due to excellent softness, toughness, stretchability, and strong tendency to gain electrons. A commercial steel wire that designed to be spiral shape was employed as the electrostatic electrode. For the fabrication of a FST–TENG, an acrylic tube was cut into two half-pipes as molds. Silicone rubber was coated evenly on the spiral steel wire and finally a fiber with controlled diameter can be obtained. After encapsulating the two molds and drying silicone rubber, the fiber demolded from the molds to get the final device (Fig. 1 a). The detail fabrication process can be found in the Experimental Section. From the cross-section optical microscope image of the FST–TENG, it can be observed that a spiral steel wire with the outer diameter ( D ) and wire diameter ( d ) of ~ 2 mm and ~ 200 μm, respectively, is located in the central part of the silicone rubber (Fig. 1 b). Moreover, a typical FST–TENG with the diameter of ~ 2.94 mm (Fig. 1 c) can be not only bended (Fig. 1 d) and knotted (Fig. 1 e) but also tailorable (Fig. 1 f), respectively. Since the silicone rubber and the spiral steel wire electrodes are terrific stable, for practical application, the FST–TENG can be tailored into arbitrary separate devices. All tailorable energy devices can still be used to collect the mechanical energy of the different parts of the human body [ 36 , 37 ]. Fig. 1 Schematic illustration and mechanical behaviors of the spiral steel wire electrode based fiber-shaped stretchable and tailorable triboelectric nanogenerator (FST–TENG). a Fabrication process of FST–TENG. b Cross-sectional optical microscope image (scale bar: 500 μm). c Photograph of the diameter measurement of FST–TENG (scale bar: 2 cm). Photograph of the FST–TENG at different states: d bended, e knotted, and f tailorable state (scale bars: 3 cm) The working mechanism and electrical output performances of the FST–TENG are shown in Fig. 2 . The spiral steel wire is connected to the ground and skin is acted as another triboelectric material for generating electricity. The single-electrode mode FST–TENG is based on the coupling effect of triboelectrification and electrostatic induction [ 38 ], as schematically illustrated in Fig. 2 a. In initial state, when skin contacts with the silicon rubber, the negative triboelectric charges would maintain on the silicone rubber surface and the generated positive charges on the skin (State I). When the skin detaches, the electrons will flow from the conductive steel wire to the ground under the short-circuit condition and the positive charges will be induced in the electrode (electrostatic induction effect) (State II). Until skin is quite far away from silicone rubber surface, the transferred charges from the spiral steel wire electrode to the ground will reach their maximum values (State III). Then the attaching skin causes the electrons to flow back in the reverse direction due to reverse electrostatic induction (State IV). When skin goes back to its original position, the charged surface comes into full contact again, while the triboelectric charge distribution of the FST–TENG returns to its original state. Through continuous circulation of the contact and separate process between the skin and FST–TENG, reciprocating motion of electrons between the spiral steel wire electrode and the ground can cause an alternating current and power output. To characterize the FST–TENG, the output performance was evaluated by a cyclic movement through a linear motor. Through mutual contacting and separation of different triboelectric materials with FST–TENG in single-electrode mode, corresponding transferred short-circuit charge ( Q sc ), open-circuit voltage ( V oc ), and short-circuit current ( I sc ) of these materials which reflect the triboelectric ability to lose electrons were recorded (Fig. S1). Compared with Cu, Al, and nylon with same area of 6 × 6 cm 2 , hog skin which simulates the human skin has the more electrostatic charges generated at the contacting interface and can also effectively improve the contact area due to its certain flexibility, and thus has the higher output Q sc [ 39 ]. Figure 2 b exhibits the results of V oc , I sc and Q sc for the FST–TENG under the motion frequencies ranging from 0.5 to 2.5 Hz, respectively. It can be seen that in Fig. 2 c, with the increase of the motion frequencies, the V oc (~ 59.7 V) and Q sc (~ 23.7 nC) keep almost constant, while the peak value of the I sc increases from 0.84 to 2.67 μA with the motion frequency. The average power of the FST–TENG measured at different external load resistances increases with the motion frequency. The maximum power value (~ 2.13 μW) can be achieved at the load resistance of 100 MΩ at 2.5 Hz. It is observed that the optimum resistance decreases with the increasing frequency, which can be explained by the decrease of the impedance of TENG with the increase of the motion frequency [ 20 ]. Also, for long-term application, the stability has been tested by contacting and separating the FST–TEMG for 5000 times, as illustrated in Fig. S2a. It is easy to see that the V oc during the first 50 cycles is close to the last 50 cycles, which verifies excellent stability and reliability. The durability of the FST–TENG was examined under circulating between 180° bended and 50% stretching strain motion for 5000 cycles by the linear motor, as shown in Fig. S2b. Normalized V oc , I sc , and Q sc values of FST–TENG were recorded after every 500 times of cycling, which shows no significant degradation in performance, confirming their excellent flexibility, stretchability, and stability. To investigate the tailorability of the FST–TENG, an 8 cm long fiber was cut into two parts of equal length, i.e., 4 cm. The V oc (~ 49.5 V), I sc (~ 1.01 μA), and Q sc (~ 17.3 nC) of both parts are almost half of the electric output of the original 8 cm long one, revealing that the tailoring process hardly affects the performance of the FST–TENG, as shown in Fig. S3. Fig. 2 Working mechanism and electrical output performance of FST-TENG at single-electrode mode. a Schematic illustration of working mechanism for generating electricity under short-circuit conditions. b Electrical outputs of the spiral steel wire based TENG with different motion frequencies ranging from 0.5 to 2.5 Hz, including V oc , I sc , and Q sc . c Dependence of the output average power under external load with different frequencies from 0.5 to 2.5 Hz The general performance of the stretchable FST–TENG under different strain level is schematically displayed in Fig. 3 . The Young’s modulus of steel wire (2 × 10 11 N m −2 ) is much higher than that of silicone rubber (1.2 × 10 6 N m −2 ) which means hard to stretch. However, by the geometric construction of the spiral steel wire, it can improve the stretchability of the electrode. Because the tensile properties of silicone rubber are much better than the spiral steel wire, the stretching degree of the entire device is mainly determined by the tensile properties of the spiral steel wire. From the geometrical diagram of the spiral steel wire (Fig. 3 a), the relationship between the geometric dimensions is: 1 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\alpha = \\tan^{ - 1} \\left( {\\frac{t}{{\\pi D_{2} }}} \\right)$$\\end{document} α = tan - 1 t π D 2 \n 2 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$D = D_{2} + d$$\\end{document} D = D 2 + d where D is outer diameter, D 2 is medium diameter, t is pitch, α is spiral angle and d is diameter of steel wire. Fig. 3 Output performance of the FST–TENG under different strain level. a Geometrical diagram of stretching spiral steel wire and b stress–strain curve of FST–TENG. The photograph of the spiral steel wire-based TENG c at original state and d stretched state (scale bar: 2 cm). e Dependence of the resistance and diameter of the spiral steel wire-based TENG under different strain levels (0–50%). Electrical output of the FST–TENG under various strain levels including f \n V oc , g \n I sc , and h \n Q sc \n In the linear elastic range of the spiral steel wire, the uniaxial tensile deformation is proportional to the external force, following Hooke’s law [ 40 ]: 3 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$H_{\\text{x}} - H_{0} = \\lambda = \\frac{{F_{\\text{x}} }}{k}$$\\end{document} H x - H 0 = λ = F x k where H 0 is the initial length, H x is the length of the spiral steel wire after being enlarged, λ is the deformation, F x is the tensile force and k is elastic modulus. When strain degree of the spiral steel wire reached 51.9%, it reaches the elastic limit, which means the recoverable stretch reaching the maximum (Fig. S4). Therefore, for a single device, it exhibits good stretchability, which can be enlarged as much as 53.4% (Fig. 3 b). Meanwhile, the elastic mold ( k ) of FTS-TENG (2.18 N cm −1 ) is close to that of spiral steel wire (1.94 N cm −1 ) in the stress–strain curves. Here, the degree of enlargement H is defined as: 4 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$H = \\frac{{H_{\\text{x}} - H_{0} }}{{H_{0} }} \\times 100\\% = \\frac{\\lambda }{{H_{0} }} \\times 100\\% .$$\\end{document} H = H x - H 0 H 0 × 100 % = λ H 0 × 100 % . \n As indicated by the photograph of the FST–TENG at original state (Fig. 3 c) and stretched state (Fig. 3 d), the diameter of the FST–TENG reduces from ~ 2.978 mm at original state to ~ 2.01 mm at 50% strain, while the resistance of the steel wire electrode was almost unchanged (Fig. 3 e). The output signals under the different stretched degree of FST–TENG were acquired, as shown in Fig. 3 f–h. With the elongation of the FST–TENG, V oc , I sc , and Q sc initially increase and recover at same value. The contact area between skin and the FST–TENG remains unchanged as shown in Fig. S5. Meanwhile, the positive effect of the thinner silicone rubber (small effective distance, d 0 ) during stretching affects the output, which can be explained by the Poisson’s effect [ 20 , 41 , 42 ]. Thus, the output increases fast from 0% to 30% strain and then the output increases slightly from 30 to 50%. To make the FST–TENG suitable for wearing on the human body, a FST–TENG based fabric was fabricated as self-charging wearable power source, as demonstrated in Fig. 4 . In order to connect several FST–TENGs in parallel that knitted into cloth, three knitting patterns of 1 × 1, 2 × 2, and 3 × 3 type of FST–TENG fabric were woven. The Q sc of the fabric increases with an increasing number of single devices due to the increase contact area (Fig. 4 a). Meanwhile, the I sc follows the same tendency (Fig. S6). When manually tapped 12 fibers with 1 × 1 knitting, the FST–TENG fabric can sufficiency light up 15 LEDs in series (Fig. 4 b and Movie S1). To systematically investigate the output performance, the as-fabricated FST–TENG fabric was evaluated at contact-separation motion in single-electrode mode. With the increase of the working frequency from 0.5 to 2.5 Hz, the I sc rises from 0.87 to 3.23 μA, while both the V oc and Q sc do not changes and stay at ~ 121.8 V and ~ 45.8 nC, respectively (Fig. 4 c). However, by increasing load resistance, the average output power of the FST–TENG fabric reaches the maximum value of 4.06 μW at an external load resistance of 100 MΩ (Fig. S7). Due to the pulsed and alternating output characteristic, the generated electricity of FST–TENG needs to be converted by the bridge rectifier from alternating to direct current and then stored into the storage devices. Figure S8 shows the working circuit of the self-charging powered system. After rectifying, the current output of FST–TENG remains almost same with the value before rectifying (inset of Fig. 4 d). The voltages of the commercial capacitors of 10 μF charged by FST–TENG fabric under different working frequencies were measured to evaluate the charging capacity. It takes ~ 68 s to charge the commercial capacitor to 2 V at 2.5 Hz. When tapping the FST–TENG fabric with hand, it takes ~ 40 s to charge commercial capacitor to drive the electronic watch (Fig. 4 e and Movie S2). Likewise, FST–TENG can also be knitted into bracelet worn on the wrist to harvest energy. By hand tapping the FST–TENG bracelet, which is knitted by three FST–TENGs, the V oc , I sc , and Q sc can reach ~ 73 V, ~ 1.12 μA, and ~ 26 nC, respectively (Figs. 4 f and S9). Moreover, the FST–TENG bracelet enables to light up 10 LEDs in series (inset of Fig. 4 f). Fig. 4 Demonstration of the FST–TENG-based fabric as wearable power source. a \n Q sc of the TENG fabric with knitting patterns of 1 × 1, 2 × 2, and 3 × 3 nets. b Photograph of the FST–TENG-based fabric, which enables to light up 15 green LEDs by hand tapping. c Electrical outputs of the TENG fabric under various motion frequencies ranging from 0.5 to 2.5 Hz, including V oc , I sc , and Q sc . d Charging curves of a capacitor (10 μF) charged by the FST–TENG fabric at various motion frequencies. Inset shows the current output of the FST–TENG fabric after rectification. e Photograph of the self-charging powered system driving an electronic watch. f \n V oc of the FST–TENG-based bracelet at the motion frequency of 2 Hz. Inset shows the photograph of 10 green LEDs lighted up by hand tapping Besides, the FST–TENG can be applied as active physiological motion and human–machine interface sensor due to its high sensitivity and rapid response/recovery time for low-frequency movements [ 43 – 46 ]. Here, it can be embodied in the gesture sensing when constructed into a smart glove, as demonstrated in Fig. 5 . Figure 5 a shows that the FST–TENGs are parallel connected and sew on the back of the glove corresponding to the position of five fingers. When any finger is bent and released, the contact area between the skin and FST–TENG increases and decreases at the same time. Thus, the real-time voltage could indicate the different gestures of testers, as shown in Fig. 5 b. It is remarkable that the fingers have very weak jitter when they are not bent, but it does not affect as the contrast (state I). Compared with slightly bending, the dramatically bending results in the larger contact area, which would induce larger change in the electric potential of the steel electrode. Therefore, the voltage output of the five fingers collected by bending in large angle was much higher than that in small angle (state II and III and Movie S1). To adapt to different frequencies of human motion, it was tested that the same voltage value can be maintained at different frequencies (Fig. S10), due to the constant contact area of the same bending degree. When connecting these 5 FTS-TENG devices to 5 signal acquisition terminals separately, the voltage signals of the five fingers are collected by bending them in turn from the thumb to the little finger and transmitted to the computer, as shown in Fig. 5 c. There is a phenomenon in the human finger that when a finger is bent, other fingers will also have slight resonant reactions. However, the V oc of the finger at the bending state are significantly higher than those at the immobile states, indicating that the smart glove can effectively identify real finger in bending. In addition, it is also obvious that the electrical outputs of this smart glove increased with the number of bent fingers increases from one (~ 3 V) to two (~ 9 V), three (~ 14 V), four (~ 20 V), and five (~ 32 V) (Fig. 5 d and Movie S2), further verifying its feasibility for use in recognition of different gestures. Fig. 5 Demonstration of the FST–TENGs woven in a smart glove as active gesture sensor. a Photograph of the smart glove with FST–TENGs woven on dorsum and demonstration of the operation mode. b Photograph and corresponding V oc of the smart glove under different degree of finger bending, including original state, slightly and dramatically bending. c \n V oc signals of the five fingers by buckling from the thumb to the little finger in turn. d Demonstrations of the smart glove for representing the numbers of “one”, “two”, “three”, “four” and “five” by different gestures"
} | 5,642 |
24671086 | PMC4139731 | pmc | 978 | {
"abstract": "Recently, a novel mode of sulphur oxidation was described in marine sediments, in which sulphide oxidation in deeper anoxic layers was electrically coupled to oxygen reduction at the sediment surface. Subsequent experimental evidence identified that long filamentous bacteria belonging to the family Desulfobulbaceae likely mediated the electron transport across the centimetre-scale distances. Such long-range electron transfer challenges some long-held views in microbial ecology and could have profound implications for sulphur cycling in marine sediments. But, so far, this process of electrogenic sulphur oxidation has been documented only in laboratory experiments and so its imprint on the seafloor remains unknown. Here we show that the geochemical signature of electrogenic sulphur oxidation occurs in a variety of coastal sediment environments, including a salt marsh, a seasonally hypoxic basin, and a subtidal coastal mud plain. In all cases, electrogenic sulphur oxidation was detected together with an abundance of Desulfobulbaceae filaments. Complementary laboratory experiments in intertidal sands demonstrated that mechanical disturbance by bioturbating fauna destroys the electrogenic sulphur oxidation signal. A survey of published geochemical data and 16S rRNA gene sequences identified that electrogenic sulphide oxidation is likely present in a variety of marine sediments with high sulphide generation and restricted bioturbation, such as mangrove swamps, aquaculture areas, seasonally hypoxic basins, cold sulphide seeps and possibly hydrothermal vent environments. This study shows for the first time that electrogenic sulphur oxidation occurs in a wide range of marine sediments and that bioturbation may exert a dominant control on its natural distribution.",
"introduction": "Introduction Sulphate reduction is the dominant mineralization pathway in coastal marine sediments, and leads to an accumulation of sulphide in deeper sediment horizons ( Jørgensen, 1982 ). Various respiratory pathways of microbial sulphur oxidation have evolved as a means of harvesting this stored chemical energy ( Jørgensen and Nelson, 2004 ). Recently, a novel sulphur oxidation process, herein called electrogenic sulphur oxidation, was discovered whereby the oxidation of free sulphide in deeper sediment horizons was observed to be directly coupled to reduction of oxygen near the sediment surface ( Nielsen et al., 2010 ). Such a spatial separation of redox half-reactions indicates there must be a direct pathway by which electrons originating from sulphide are conducted across centimetre-scale distances to oxygen ( Nielsen et al , 2010 ; Risgaard-Petersen et al., 2012 ). Experimental manipulation revealed that this electrical connection responded immediately (that is, within minutes) to oxygen depletion in the overlying water, and so the connection was too fast to be explained by diffusion of redox-active shuttles ( Nielsen et al., 2010 ). Metal-reducing bacteria are known to transport electrons externally via cell surface cytochromes, redox shuttles or conductive pili ( Reguera et al., 2005 ; Gorby et al., 2006 ; Lovley, 2008 ; Clarke et al., 2011 ), and such extracellular electron transport can bridge ∼100 micrometer distances in electrode biofilms ( Reguera et al., 2006 ; Logan and Rabaey, 2012 ). The electron transport associated with this electrogenic sulphur oxidation, by contrast, operates on scales of centimetres, thereby extending the known length scale of microbially mediated electron transport by two orders of magnitude. Electrogenic sulphur oxidation creates a unique geochemical signal in surface sediments, which is detectable by microsensor profiling. A key feature is that the cathodic half-reaction (that is, O 2 +4e − +4H + →2H 2 O) consumes protons, thereby creating a pH maximum in the oxic zone. Deeper in the sediment, protons are released by an anodic half-reaction (that is, electrogenic sulphide oxidation; such as ½H 2 S+2H 2 O→½SO 4 2− +4e − +5H + ) creating a broad pH minimum near the sulphide horizon, separated from the oxic zone by a suboxic zone ranging from several millimetres to approximately 2 cm wide. In laboratory experiments, the development of this characteristic geochemical fingerprint has been used as a sign of electrogenic sulphur oxidation activity ( Nielsen et al. , 2010 ; Pfeffer et al., 2012 ; Risgaard-Petersen et al., 2012 ). Compelling evidence indicates that electrogenic sulphur oxidation is likely carried out by long filamentous bacteria. First, in sediments with the geochemical signature of this process, detailed microscopic investigation revealed high densities of long multicellular filamentous bacteria, belonging to the family Desulfobulbaceae , which were spanning the entire length of the suboxic zone ( Pfeffer et al., 2012 ). Second, when a thin wire was passed horizontally through the suboxic zone, the electric circuitry was effectively ‘cut', as indicated by an immediate decrease in sediment oxygen consumption by a factor of 2, and a decrease in the signature pH maximum. This experiment identified that the electron conduction occurred along a continuous structure spanning the suboxic zone ( Pfeffer et al., 2012 ). Thirdly, when sediment was experimentally incubated with filters of various pore sizes which were inserted horizontally, the geochemical signature of electrogenic sulphur oxidation developed only in cores that had filters with pore sizes large enough for Desulfobulbaceae filaments to penetrate ( Pfeffer et al., 2012 ). Taken together, these experiments provide strong evidence that electrogenic sulphur oxidation is carried out by Desulfobulbaceae filaments. The capacity of Desulfobulbaceae filaments to oxidize sulphur by long-range electron transport represents an entirely novel microbial lifestyle ( Pfeffer et al., 2012 ). By conducting electrons across centimetre-scale distances, these filamentous bacteria appear capable of enabling a previously unrecognized strategy for resource competition in marine sediments ( Nielsen et al., 2010 ; Risgaard-Petersen et al. , 2012 ). Yet until now, electrogenic sulphur oxidation has only been induced in laboratory experiments that have used sieved and homogenized sediment from a deep sediment horizon obtained from one particular location in the Baltic Sea ( Nielsen et al., 2010 ; Pfeffer et al., 2012 ; Risgaard-Petersen et al., 2012 ). Our objective here is to identify whether electrogenic sulphur oxidation, together with Desulfobulbaceae filaments, are found in natural marine sediments. We further aim to identify major environmental constraints on the occurrence and distribution of electrogenic sulphur oxidation in marine sediments. Given the high growth rate and densities of these bacteria observed in laboratory experiments ( Pfeffer et al., 2012 ; Risgaard-Petersen et al., 2012 ), we hypothesize that electrogenic sulphur oxidation is a competitively successful strategy in natural marine sediments and therefore may be widespread in nature.",
"discussion": "Discussion Our results provide evidence that electrogenic sulphur oxidation, previously reported from laboratory incubations ( Nielsen et al. , 2010 ; Pfeffer et al., 2012 ; Risgaard-Petersen et al., 2012 ), occurs under natural conditions in the seafloor. We observed evidence of electrogenic sulphur oxidation, together with Desulfobulbaceae filaments, in a variety of coastal marine sediments (that is, a salt marsh drainage channel, a subtidal site recovering from seasonal hypoxia, a coastal area of high mud deposition). These sediments were characterized by high DOU and high organic matter content. In the narrow oxic zone (0.7–2.0 mm depth), a pH maximum was observed under dark conditions, indicative of cathodic oxygen consumption. Below this depth, there was a suboxic zone where oxygen and sulphide were not detectable, which was typically 5–15 mm deep (although sometimes extending deeper). Below the suboxic zone, these sediments exhibited a steep sulphide gradient, indicating high sulphide production. Repeated visits to these three study sites identified that electrogenic sulphur oxidation is a regularly occurring, although not necessarily permanent, feature at these field sites. Our observation of the geochemical signature of electrogenic sulphur oxidation always coincided with the presence of Desulfobulbaceae filaments that were closely related (that is, ⩾97% 16S rRNA gene sequence similarity) to those previously reported from the laboratory incubations of Pfeffer et al. (2012 ). Desulfobulbaceae filaments from the suboxic zone of our study sites revealed densities of 82.0 and 122.8 m filament per cubic centimetre sediment at RSM and BCZ, respectively, and no filaments were found at the reference site OSF. These densities are similar to those reported in Pfeffer et al. (2012) (127 m cm −3 ) and support the hypothesis that the long-range electron transport associated with electrogenic sulphur oxidation is mediated by the Desulfobulbaceae filaments ( Pfeffer et al., 2012 ). The filamentous bacteria retrieved from the three electrogenic sulphur oxidation study sites (RSM, BCZ and MLG) were long and unbranched, and retrieved fragments could exceed eight millimetres in length. The unusual morphological feature of raised ridges running in parallel along the length of the filaments, observed previously in laboratory incubations, were also prominent in the Desulfobulbaceae filaments isolated in this study. Pfeffer et al. (2012 ) reported filament diameters between 0.4 and 0.7 μm with fewer than 20 ridges per filament, whereas we herein report filament diameters between 0.9 and 3.0 μm with 16–58 ridges per filament. It remains to be seen if these morphological differences represent genotypic or phenotypic variation. A cosmopolitan distribution of electrogenic sulphur oxidation is supported by a number of published datasets in which the geochemical fingerprint of this process is evident, but has not been formally recognized ( Figure 1d ). Shallow subsurface pH maxima, characteristic of cathodic oxygen consumption, have been reported from in situ microsensor profiling in the Santa Barbara ( Cai et al., 2000 ) and Santa Monica Basins ( Reimers et al., 1996 ). These basins are permanently hypoxic, support high rates of sulphate reduction, and are devoid of bioturbating fauna. Similar pH profiles were also reported from a site situated within an oyster aquaculture park in the Thau lagoon (Mediterranean Sea, France; Dedieu, 2005 ) and from sediments near a fish farm in Pillan fjord (Chile; Mulsow et al., 2006 ), both of which experience high organic matter loading and having little or no burrowing infauna. Seasonally hypoxic sediments from Tokyo Bay have also demonstrated microprofile evidence of cathodic oxygen consumption ( Sayama, 2011 ), similar to those reported here from MLG. At a site characterized by high sulphide fluxes and devoid of large bioturbating infauna on the Mid-Atlantic ridge, high rates of sulphide oxidation were detected which could not be attributed to any other known sulphide removal process, such as sulphur oxidation by large nitrate-accumulating bacteria ( Schauer et al., 2011 ). One explanation is that electrogenic sulphur oxidations is responsible for the observed sulphide removal, however, direct evidence is needed to confirm this possibility. Gene sequence archives also support a cosmopolitan distribution of the conductive filamentous bacteria in locations with high rates of sulphide generation, via high organic matter loading or sulphide seepage, and a paucity of bioturbating animals. 16S rRNA gene sequences that were highly similar (⩾97%) to the Desulfobulbaceae at our field sites have been detected at cold seeps on the Hikurangi Margin ( Baco et al., 2010 ) and the Nile Deep Sea Fan ( Grünke et al., 2011 ) in areas devoid of fauna; in organic rich cohesive sediment of a subtropical mangrove swamp ( Liang et al., 2007 ); and at a sewage impacted site in the Bay of Cadiz ( Köchling et al., 2011 ). Previous laboratory experiments have shown that sulphur oxidation by long-range electron transport can play a dominant role in oxygen consumption and can exert a strong imprint on the overall sulphur cycling and pH dynamics in sediments ( Nielsen et al., 2010 ; Risgaard-Petersen et al., 2012 ). Our observations show that this process likely has a similarly profound effect on mineral cycling in some natural coastal sediments. We observed pH values as high as 8.8 in the oxic zone, and as low as 6.1 in the suboxic zone ( Figure 1 ). These pH extremes lead to dissolution of metal sulphides and calcium carbonate within the deep suboxic zone, followed by re-precipitation of iron (hydr)oxides and carbonate near the sediment–water interface ( Risgaard-Petersen et al., 2012 ). High alkalinity effluxes have been reported from muddy sites on the BCZ, including Station 130 ( Braeckman et al. , 2014 ), and such effluxes are consistent with cathodic oxygen consumption and elevated carbonate dissolution induced by electrogenic sulphur oxidation. Current densities estimated at the three field sites demonstrating electrogenic sulphur oxidation (15–23 mA m −2 ) are comparable with the values observed in previous laboratory experiments ( Nielsen et al., 2010 ; Risgaard-Petersen et al., 2012 ) and are similar to those reported from the anodes of sediment batteries colonized by natural microbial communities ( Tender et al. , 2002 ; Ryckelynck et al., 2006 ). We estimated that a minimum of 5.3–33.7% of the DOU in these sediments was due to electrogenic sulphur oxidation, substantiating that this process can be a dominant biogeochemical process in natural marine sediments. Our data further suggest that bioturbation may exert a major control on the natural distribution of electrogenic sulphur oxidation in coastal marine sediments. In heavily bioturbated sediment from OSF, the geochemical signature of electrogenic sulphide oxidation was not detected under field conditions, but could be induced in laboratory experiments, as long as bioturbation was excluded. In these sediments, a seed population of Desulfobulbaceae bacteria must have existed, from which these filaments were able to proliferate when burrowing animals are removed. We observed that sediment overturning by bioturbation caused a clear disruption of the electrochemical signal. This may arise from a direct breakage of the bacterial filaments, in a manner similar to the previously described cutting experiments of Pfeffer et al. (2012 ). In addition, sediment overturning may halt electrogenic sulphur oxidation by depositing an anoxic layer of sediment on top of the sediment surface, thereby depriving the resident filamentous bacteria from access to electron acceptors. In summary, we have shown that electrogenic sulphur oxidation is found in intact coastal marine sediments under natural conditions. We found this process, together with the Desulfobulbaceae filaments, in three depositional sediment areas (a salt marsh, a subtidal coastal mud plain and a hypoxic marine basin). These sediments were collectively characterized as being rich in organic matter, and consequently supporting high rates of sulphate reduction, and having an oxygenated overlying water column. These sediments were furthermore subject to little biomechanical mixing by macrofauna, which appears to be a key control on the distribution of electrogenic sulphur oxidation in natural settings. Until now, the occurrence of electrogenic sulphur oxidation has not been accounted for in studies of natural marine sediments. Given its widespread distribution, electrogenic sulphur oxidation may be a key process in the biogeochemistry and microbial ecology of the seafloor."
} | 3,972 |
26305687 | PMC5049663 | pmc | 980 | {
"abstract": "Summary \n Beneficial associations between plants and microbes play an important role in both natural and agricultural ecosystems. For example, associations between fungi of the genus Epichloë , and cool‐season grasses are known for their ability to increase resistance to insect pests, fungal pathogens and drought. However, little is known about the molecular changes induced by endophyte infection. To study the impact of endophyte infection, we compared the expression profiles, based on RNA sequencing, of perennial ryegrass infected with Epichloë festucae with noninfected plants. We show that infection causes dramatic changes in the expression of over one third of host genes. This is in stark contrast to mycorrhizal associations, where substantially fewer changes in host gene expression are observed, and is more similar to pathogenic interactions. We reveal that endophyte infection triggers reprogramming of host metabolism, favouring secondary metabolism at a cost to primary metabolism. Infection also induces changes in host development, particularly trichome formation and cell wall biogenesis. Importantly, this work sheds light on the mechanisms underlying enhanced resistance to drought and super‐infection by fungal pathogens provided by fungal endophyte infection. Finally, our study reveals that not all beneficial plant–microbe associations behave the same in terms of their effects on the host.",
"introduction": "Introduction Interactions between plants and beneficial microbes are crucial to the establishment and maintenance of stable ecosystems, particularly in the face of environmental stresses. An ideal model system for studying beneficial plant–fungal interactions is the association between fungi of the genus Epichloë and cool‐season grasses (Schardl et al ., 2013 ). Endophyte infection of grasses can increase host resistance to insect pests through protection from herbivory (Gallagher et al ., 1984 ; Rowan et al ., 1986 , 1990 ), enhance drought tolerance (Arachevaleta et al ., 1989 ; West et al ., 1993 ) and give protection from super‐infection by fungal pathogens (Tian et al ., 2008 ; Pańka et al ., 2013b ). Consequently, these associations have been widely exploited by the agricultural industry (Johnson et al ., 2013 ; Young et al ., 2013 ), and have been touted as a ‘perfect partnership’ (Christensen & Voisey, 2007 ). However, although much is known about the enhanced stress‐tolerance provided by fungal infection, little is known about what other effects endophyte infection may have on the host. A small number of microarray and RNAseq analyses are starting to shed light on the effects of plant–fungal symbiotic associations on the host transcriptome. For example, in both arbuscular mycorrhizal associations (Güimil et al ., 2005 ; Guether et al ., 2009 ; Zouari et al ., 2014 ), and ectomycorrhizal associations (Plett et al ., 2015 ), infection tends to change expression of a very small subset of host genes ( c . 1–3%). Similarly, in endophytic associations between Trichoderma and grapevine (Perazzolli et al ., 2012 ) or Arabidopsis (Morán‐Diez et al ., 2012 ), only c . 1% of host genes are altered. By contrast, the effects of pathogen infection tend to be more dramatic, with significantly more host genes differentially expressed ( c . 20%) (Doehlemann et al ., 2008 ; Kawahara et al ., 2012 ; De Cremer et al ., 2013 ). In order to fully understand the impact of Epichloë infection on perennial ryegrass gene expression, we performed deep mRNA sequencing (RNAseq) on endophyte‐infected and endophyte‐free plants. This study furthers an earlier SOLiD‐SAGE analysis of E. festucae in association with red fescue grass performed by Ambrose & Belanger ( 2012 ) that identified c . 200 host genes which are differentially expressed during a mutualistic association. We uncover major effects on host metabolism and development, and reveal that unlike other beneficial plant–fungal interactions, which affect expression of only a small number of host genes, Epichloë festucae infection, under controlled growth conditions, leads to differential expression of more than a third of host genes, with changes reflective of metabolic reprogramming.",
"discussion": "Discussion Plants infected with cool‐season grass endophytes, such as E. festucae , generally appear asymptomatic in the absence of any biotic or abiotic stress (Schardl, 2001 ) so it was surprising to find that over 38% of the host genes were differentially expressed between infected and uninfected samples grown under controlled environmental conditions. This is in contrast to a number of other recently analysed beneficial plant–fungal associations in which only c . 1–3% of the host gene set is differentially expressed in response to fungal colonisation. These interactions include the mycorrhizal associations between tomato and Funneliformis mosseae (Zouari et al ., 2014 ), Lotus japonicus and Gigaspora margarita (Guether et al ., 2009 ), Medicago truncatula and Glomus intraradices (Gomez et al ., 2009 ), rice and Glomus intraradices (Güimil et al ., 2005 ), and Trichoderma interactions with grapevine (Perazzolli et al ., 2012 ) and Arabidopsis (Morán‐Diez et al ., 2012 ). In the association between rice and G. intraradices , only c . 0.5% of the analysed host genes displayed altered expression upon infection (Güimil et al ., 2005 ). Interestingly, one of the genes upregulated in this association encodes a WRKY transcription factor, which is also upregulated when tomato is colonised by the phytopathogens Magnaporthe oryzae and Fusarium moniliforme , and is speculated to be a transcription factor involved in the basic plant defence response to many different fungal infections (Güimil et al ., 2005 ). This view is supported by our study, in which the homologous ryegrass WRKY gene (m.60140) is also upregulated in response to endophyte infection. In contrast to beneficial associations, plant–pathogenic associations tend to induce more significant changes in host gene expression, similar to that seen for Epichloë . For example, in the interactions between lettuce and Botrytis cinerea (De Cremer et al ., 2013 ), and maize and Ustilago maydis (Doehlemann et al ., 2008 ), c . 20% of the host genes are differentially expressed; and in the rice– Magnaporthe oryzae interaction over 10% of host genes are altered (Kawahara et al ., 2012 ). In contrast to the aforementioned mycorrhizal interactions, where almost all of the differentially expressed host genes are upregulated in response to fungal infection, the majority of the differentially expressed ryegrass genes were downregulated. Many of these genes encode proteins involved in primary metabolism, implying that endophyte infection represses primary metabolism. This is consistent with the results of Ambrose & Belanger ( 2012 ), which showed a general downregulation of genes involved in primary metabolism in response to infection of red fescue by E. festucae . In support of the hypothesis that endophyte infection represses primary metabolism, photosynthesis is reduced in infected plants. Surprisingly, this reduction in photosynthesis is contrary to the results of Ambrose & Belanger ( 2012 ), where genes involved in photosynthesis were found to be upregulated in endophyte‐infected red fescue. However, a reduction in photosynthesis is consistent with the results of Spiering et al . ( 2006 ), which showed that infection of L. perenne by E. festucae var. lolii leads to reduction in photosynthetic activity. This suggests that the effects of endophyte infection on host photosynthesis may vary depending on the endophyte and/or host species. This result suggests that the plant is under biotic stress (Bilgin et al ., 2010 ). In contrast to the reduction of photosynthesis, expression of secondary metabolism genes, particularly those encoding for enzymes producing bioprotective phenylpropanoids, is elevated. This observation suggests that endophyte‐infection leads to a reprogramming of host metabolism, or redistribution of resources, towards secondary metabolism, at a cost to primary metabolism. Recent work in the U. maydis –maize association has revealed that a fungal chorismate mutase enters host cells, leading to a reprogramming of host metabolism (Djamei et al ., 2011 ). However, E. festucae does not appear to possess a homologue of this secreted chorismate mutase, suggesting that it reprograms the host via an alternative method. Generally, the alternative branches of the phenylpropanoid biosynthesis pathway (lignin and anthocyanin) display different regulation states depending on the stress sensed by the plant. For example, wounding induces upregulation of the genes involved in the lignin biosynthesis, whereas low temperatures induce higher production of anthocyanins (Dixon & Paiva, 1995 ). Here, a general upregulation of both branches of the phenylpropanoid pathway is observed. This could be due to the fact that different tissue types were included in the samples used for RNAseq, including tissues both in contact with and distant from fungal hyphae. Contrary to the general downregulation of primary metabolism, genes associated with cell wall synthesis were upregulated in response to fungal infection. However, cell wall thickness was reduced in endophyte‐infected plants relative to uninfected plants. This may be due to enzymatic digestion of the host cell wall by the fungus, leading to reduced cell wall thickness despite the fact that expression of wall‐associated genes is upregulated. In support of this hypothesis, the genes encoding for a pectin methylesterase (PME) and endoxylanase are very highly expressed by E. festucae in planta (Eaton et al ., 2010 , 2015 ). The presence of a thicker cell wall adjacent to hyphae in infected plants suggests that some of the changes in wall‐associated gene expression may be required for physical attachment of hyphae to the host cell wall (Christensen et al ., 2002 ). The fungal PME may play a role in this, as demethylesterification of pectin by the PME can facilitate crosslinking of fungal and host cell walls (Pelloux et al ., 2007 ). Interestingly, in the synthetic obligate mutualistic association between the alga Chlamydomonas reinhardtii and the fungus Aspergillus nidulans , algal cell walls in contact with fungal hyphae are thinner, possibly due to the activity of fungal secreted cell wall remodelling enzymes (Hom & Murray, 2014 ). As endophyte infection seems to slow host growth (Hahn et al ., 2008 ), the cell wall thickness differences may also be due to different developmental stages in infected vs uninfected plants. In order to minimise the influence of this factor, comparable sections were harvested for each plant. Moreover, the measured plants were mature (11 wk post‐inoculation), and at this stage development differences tend to be minimal. Future studies will investigate whether this effect is also seen in seed infected (naturally infected) plants or other Epichloë –grass associations to determine whether this result is a feature of this particular association or a more general consequence of the symbiosis between Epichloë endophytes and their hosts. Despite major differences between endophyte and mycorrhizal effects on host gene expression, our study has identified a key similarity with respect to upregulation of genes encoding enzymes involved in biosynthesis of the phytohormone gibberellin in response to infection by both groups of fungi (Gomez et al ., 2009 ). Mycorrhizal colonisation has been shown to be enhanced in gibberellin‐deficient mutant hosts (Foo et al ., 2013 ), suggesting that host gibberellin restricts mycorrhizal growth in planta , the opposite of the role that plant strigolactones play in inducing branching of mycorrhizal fungi (Akiyama & Hayashi, 2006 ). Given that we have shown here that gibberellin biosynthetic genes are upregulated in response to endophyte infection, it is possible that gibberellin plays a role in controlling endophyte growth in planta . This putative upregulation of gibberellin signalling in endophyte‐infected plants is supported by the dramatic effect of infection on trichome development, as trichome development is controlled by gibberellin signalling in Arabidopsis (Perazza et al ., 1998 ). In addition to changes in hormone signalling, endophyte infection dramatically altered host pathways involved in the response to abiotic and biotic stresses. Drought‐related gene expression was lower in infected plants. However, levels of compatible solutes were increased, suggesting that endophyte infection may enhance drought tolerance by priming the host with compatible solutes to avoid imposition of drought stress. Additionally, larger trichomes and increased stomatal closure in endophyte‐infected plants will likely increase moisture retention. The increased production of compatible solutes is also linked to the altered temperature perception observed in infected plants, as genes encoding enzymes involved in production of raffinose family oligosaccharides, which act as compatible solutes and accumulate during cold acclimation in rye (Koster & Lynch, 1992 ), are upregulated in infected plants. Endophyte infection does not induce any obvious host defence response. Thus, it was hypothesised that either the fungal pathogen associated molecular patterns (PAMPs) are masked such that the fungus is not detected by the host, or that host defence is downregulated by the fungus or by the host itself (Schardl et al ., 2004 ). Our study revealed downregulation of many genes involved in the host biotic stress response in endophyte‐infected plants, supporting the latter two hypotheses. This is in contrast to E. festucae symbiotic mutants that have a pathogen‐like interaction with the host, and a dramatic induction of host defence‐related gene expression (Eaton et al ., 2010 , 2015 ). Small secreted proteins (SSPs) produced by the fungus are predicted to play a role in downregulating host defence and controlling mutualism, similar to the recently identified SSP of Laccaria bicolor , which is essential for establishment of a mutualistic interaction with poplar (Plett et al ., 2011 ). Interestingly, many of the most highly expressed E. festucae genes in planta encode putative SSPs (Eaton et al ., 2010 , 2015 ). In an exception to the downregulation of defence‐related genes, plant chitinase expression was elevated in endophyte‐infected plants. Moreover, one of the most highly expressed E. festucae genes in planta encodes a chitinase (Eaton et al ., 2010 , 2015 ). Chitin oligomers released via the action of chitinases are potent elicitors of host defence (Kaku et al ., 2006 ). However, E. festucae cell wall chitin appears to be masked when this fungus grows in planta , a result similar to the phytopathogen Cladosporium fulvum (van Esse et al ., 2007 ; de Jonge et al ., 2010 ). The highly expressed E. festucae SSPs may play a role in this masking, as they do in C. fulvum . Interestingly, chitinases are also upregulated in mycorrhizal associations (Gomez et al ., 2009 ). Given the apparent downregulation of host defence‐related gene expression, endophyte infection would be expected to increase susceptibility to microbial pathogens. However, this is not the case as infection by E. festucae var. lolii , has been shown to increase resistance to fungal pathogens (Tian et al ., 2008 ; Pańka et al ., 2013a ) due to production of phenolic compounds (Pańka et al ., 2013a ). This is consistent with the observed increased expression of several genes encoding enzymes of the phenylpropanoid pathway, which produces these bioprotective phenolics in infected plants. This suggests that whereas E. festucae infection leads to downregulation of genes directly involved in host defence, such as PR genes, there is likely an upregulation of the phenylpropanoid pathway and chitinases that would provide enhanced resistance to other microbes, possibly as a means of competitive exclusion. In the interaction between the biotrophic fungus U. maydis and maize, a fungal effector protein has been shown to enter host cells and activate genes encoding enzymes of the phenylpropanoid pathway, leading to an increase in anthocyanins and a decrease in lignin (Tanaka et al ., 2014 ). This is hypothesised to increase virulence by redistributing phenylpropanoid intermediates away from the production of host defence‐related compounds (phytoalexins) towards anthocyanin production. Despite being a mutualist, it is possible that E. festucae similarly upregulates host anthocyanin and flavonoid levels in order to reduce host production of defence compounds. This study uncovers the dramatic effects of endophyte infection on ryegrass gene expression. Endophyte infection has important effects not only on host responses to stress, but also leads to reprogramming of host metabolism, and substantially alters host development. Understanding the molecular basis for these endophyte‐induced plant changes will be a major focus of future research."
} | 4,305 |
39414777 | PMC11484764 | pmc | 982 | {
"abstract": "Microorganisms can be engineered to sustainably produce a variety of products including fuels, pharmaceuticals, materials, and food. However, highly engineered strains often result in low production yield, due to undesired effects such as metabolic burden and the toxicity of intermediates. Drawing inspiration from natural ecosystems, the construction of a synthetic community with division of labor can offer advantages for bioproduction. This approach involves dividing specific tasks among community members, thereby enhancing the functionality of each member. In this study, we identify six pairs out of fifteen composed of six auxotrophs of Yarrowia lipolytica that spontaneously form robust syntrophic and synergistic communities. We characterize the stability and growth dynamics of these communities. Furthermore, we validate the existence of syntrophic interactions between two yeast species, Y. lipolytica and Saccharomyces cerevisiae , and find a strain combination, Δtrp2 and Δtrp4 , forming a stable syntrophic community between two species. Subsequently, we introduce a 3-hydroxypropionic acid (3-HP) biosynthesis pathway into the syntrophic community by dividing the pathway among different strains. Our results demonstrate improved production of 3-HP in both intra- and interspecies communities compared to monocultures. Our results show the stable formation of synthetic syntrophic communities, and their potential in improving bioproduction processes.",
"introduction": "Introduction The advances in synthetic biology and metabolic engineering have led to improved biotechnology processes using microorganisms for the production of food, pharmaceuticals, biofuels, and biomaterials. Despite methodological advances in our capacities to improve microbial strains, some commonly found challenges remain, including metabolic burden due to the high level of pathway engineering, cofactor imbalance, or toxicity of intermediates and/or final products. To overcome the drawbacks of engineering single chassis strain, the establishment of synthetic microbial communities by engineering multiple strains that cooperate during the bioprocess has been proposed 1 – 3 . By dividing the labor among multiple strains, synthetic communities are able to improve the functionality of each member, reduce metabolic burden and engineering complexity, and accomplish high efficiency of production as found in natural consortia 1 , 2 , 4 . In natural communities, there are various cellular interactions that determine the dynamics of the consortia, such as competition, commensalism, mutualism, or neutralism 5 . When it comes to synthetic consortia, designing a proper interaction between members are crucial for constructing a stable and robust synthetic community 2 , 3 . A type of mutualistic interaction, cross-feeding or syntrophy, requires each population that relies on each other for survival, which can provide stable coexistence by tying together the members in the community 5 , 6 . One way to achieve cross-feeding is by using co-auxotrophic strains that exchange essential amino acids to allow each other to grow 7 , 8 . It has been generally regarded that yeast co-cultures were not as effective as bacterial ones in forming co-auxotrophic communities, except for strains engineered to produce higher amount of amino acids 8 . Recently, we performed high-throughput screening of syntrophic interactions in the model yeast Saccharomyces cerevisiae by using yeast knockout library 9 , 10 . From this study, 49 pairwise auxotroph combinations which is 3.6% of tested pairs were identified to spontaneously form syntrophic communities and some of them were tested for division of labor, leading to improved bioproduction 10 . This finding suggests that cross-feeding-based communities could be formed in other yeast species, including those with high industrial potential. Yarrowia lipolytica has been gaining interests as a host strain for bioproduction of chemicals, fuels, foods, and pharmaceuticals from both academia and industry 11 , 12 . Advantageous industrial features of this yeast include robustness, stress tolerance, being amenable by synthetic biology tools, and high cell density cultivation. Most research using Y. lipolytica have focused on engineering in a single strain. The studies on microbial communities using this yeast are so far limited. There are few studies of co-culture using Y. lipolytica with other species for bioremediation or feedstock utilization with the modulation of inoculation ratio or time among community members 13 – 20 . A study has explored division of labor for bioproduction of amorphadiene with Y. lipoltyica strains. A modular co-culture dividing the pathway for boosting precursor pools and amorphadiene synthesis resulted in the improved titers 21 . These works highlight the increasing interest in creating communities of Y. lipolytica . However, tools for controlling population dynamics, such as cross-feeding 9 , to maximize robustness and efficiency have not yet been developed in Y. lipolytica . In this study, we explored the creation of syntrophic communities of Y. lipolytica using auxotrophic strains and identified pairs exhibiting synergistic growths, which were further characterized. The Y. lipolytica auxotrophic strains were also evaluated for establishing the interspecies syntrophic growth with S. cerevisiae auxotrophs. We finally developed a division of labor strategy for the production of a bioplastic precursor, 3-hydroxypropionic acid, employing syntrophic intraspecies and interspecies communities, which resulted in increased bioproduction.",
"discussion": "Discussion In nature, many microorganisms are auxotrophs and therefore rely on external nutrients (including amino acids) for their growth 24 . This observation has inspired synthetic biologists to design synthetic communities using amino acid or nucleotide auxotrophic strains. The requirement on essential metabolites exchange promotes cooperative behaviors and beneficial interactions. Recent studies on synthetic communities often require a high level of engineering to maintain the stability of the coculture and control the population 25 , 26 , which limit the applicability and universality of these methods. Auxotrophic-based cross-feeding offers a simpler alternative to creating stable communities. However, identifying the adequate pairs of auxotrophs able to establish syntrophic interactions is challenging as metabolic costs and energy requirements for the synthesis of each amino acid or metabolite vary and their transport systems are not fully understood 27 – 29 . Here, we aimed to uncover spontaneous syntrophic communities in Y. lipolytica . Out of fifteen combinations involving six auxotroph strains, five exhibited robust syntrophic growth, and six demonstrated a slower but still discernible growth at an initial ratio of 1:1. Further investigation by modulating the initial inoculation ratio could potentially unveil additional auxotrophic pairs capable of establishing syntrophic communities. Generally, the success of syntrophic interaction is thought to be determined by the rates of import, export, and consumption of the involved metabolites, as the depletion of one of the metabolites before establishing the syntrophy can lead to the collapse of the community 30 . Therefore, pairs that failed to establish spontaneous syntrophic interactions might be attributed to low production or a limited transport system of specific metabolites that need to be provided to the other member of the community. Engineering strains to overproduce specific metabolites through the regulation of feed-back inhibition or the strengthening carbon flux towards their synthesis could be beneficial in promoting syntrophic interactions, as demonstrated independently in both E. coli and S. cerevisiae 8 , 9 , 31 . At a more fundamental level, it would be beneficial to study the transport mechanisms of metabolites, including amino acids in Y. lipolytica . Understanding the secretion or uptake of metabolites is pivotal in order to improve stable syntrophic interactions. Employing omics approaches, such as metagenomic sequencing 7 and exometabolomic analysis 32 , 33 could contribute to unravel some of these transport systems and better understand microbial cross-feeding within synthetic communities. In this work, we also demonstrated spontaneous syntrophic growth between two yeast species, Y. lipolytica and S. cerevisiae . A pair of Δtrp2 - Δtrp4 demonstrated successful syntrophic interaction between two strains regardless of the combination of the auxotroph pairs and the species. In S. cerevisiae communities, Δtrp2 - Δtrp4 has been described to exhibit an extremely unbalanced population distribution, with one strain dominating the coculture (over 95% of ∆ trp2 ) 10 . A similar trend was observed in the Y. lipolytica communities at inoculation ratios of 10:1 and 5:1, however, the population was more balanced (∆ trp2 :∆ trp4 of 1:0.9-1.5) at the ratio of 1:5 and 1:10 (Fig. 2f, i ). In the interspecies coculture of Δtrp2 - Δtrp4 , a balanced population was achieved in all tested strains, species, and ratios (Fig. 3e ), highlighting the potential of interspecies syntrophic communities to provide an additional level of control. In specific inoculation ratios (10:1 and 1:1) of the SC Δtrp2- YL Δtrp4 coculture, growth failed to occur, suggesting an insufficient exchange of metabolites in this experimental condition. Similarly, the pair of YL Δmet5 -SC Δtrp4 was unable to grow, while SC Δmet5 -YL Δtrp4 grew. This observation might also be explained by different exchange rates of metabolites in different species 7 , 10 . Interdependent cocultures for bioproduction have so far mostly been explored using model microorganisms 1 , 34 . Our results suggest a broader applicability of syntrophic interactions beyond model microorganisms, paving the way for designing synthetic communities of non-conventional yeasts for bioproduction. To study the effect of division of labor and cross-feeding in bioproduction by synthetic communities, we divided the biosynthetic pathway of 3-HP into two modules. The coculture of YLΔ trp2- B and YLΔ trp4 -P, with an initial ratio of 10:1, produced 19.3 times higher 3-HP (4.67 mM) than the WT monoculture harboring the complete 3-HP synthetic pathway in a single strain. Notably, this synthetic pathway converting β-alanine into 3-HP was investigated in Y. lipolytica for the first time in this study. The growth and metabolite analysis (Supplementary Fig. 16 ) suggests that the higher 3-HP production found in the co-cultures originated from a higher availability of pyruvate, a common precursor of 3-HP and citrate. This result underscores that the division of labor within a synthetic community can be used to validate undiscovered synthetic pathways, in addition to the traditional approach of embedding the entire pathway in a single strain. The production of 3-HP was further improved in the interspecies cross-feeding community, YL Δ trp2- B and SC Δ trp4 -P at a 10:1 inoculation ratio, reaching 3.96 mM of 3-HP. This is slightly higher than the reported 0.35 g/L (3.88 mM) of 3-HP production in Y. lipolytica harboring the alternative pathway from malonyl-CoA 22 . When it comes to MSA production in S. cerevisiae communities, the coculture of Δtrp2 -B: Δtrp4 -P = 10:1 produced the highest MSA among different inoculation ratios but also outperformed the monoculture, which is consistent with the MSA production in a previous study of S. cerevisiae communities (Supplementary Table 4 and Supplementary Figs. 14 and 18 ) 10 . However, the production of MSA in S. cerevisiae co-culture at Δtrp2 -B: Δtrp4 -P = 1:1 and 1:10 was negligible, although it was higher than the monoculture in the previous study. This might be due to the different promoters used for expressing BAPAT in each study, additional gene expression (YDFG) in this study, and different cultivation scales. In the case of 3-HP production, S. cerevisiae co-culture at specific inoculation ratio ( Δtrp2 -B: Δtrp4 -P = 1:1) performed better than the S. cerevisiae monoculture (Supplementary Fig. 14 , Supplementary Table 4 ). The level of total metabolites produced from the Δtrp2 -B strain (MSA and 3-HP) in the coculture of Δtrp2 -B: Δtrp4 -P = 10:1 is higher than the one from the monoculture (Supplementary Table 4 and Supplementary Fig. 14 ). In this study, we used the biosynthetic pathway of 3-HP as a proof of concept, but further modifications can lead to improve titers. Increased production is expected through additional engineering strategies such as promoter engineering, multi-copy integration, and precursor and/or cofactor supply. Overall, this work demonstrates that the combination of cross-feeding and inoculation ratio to control population dynamics in synthetic yeast communities with division of labor has the potential to improve the production of valuable molecules. It is worth noting that further research is required to understand the complex relationship between division of labor and bioproduction and fully correlate them both. The strategy described here could be expanded to multiple organisms (and their combination) and compounds of interest. In conclusion, we successfully demonstrated the establishment of a stable synthetic cross-feeding yeast community employing auxotrophs of Y. lipolytica , an yeast of high industrial interest. Synthetic communities of Y. lipolytica were characterized in terms of growth and population dynamics, considering different auxotrophic pairs and inoculation ratios. Our findings confirmed that specific auxotrophs can exchange metabolites with other members, facilitating spontaneous growth in both intraspecies ( Y. lipolytica ) and interspecies ( Y. lipolytica and S. cerevisiae ) communities. We further explored the division of labor and bioproduction of 3-HP within these syntrophic communities. Notably, we found a 3-HP production improvement by 19.3 and 18.6 times when labor was divided in intra- and interspecies communities compared to the Y. lipolytica monoculture, respectively. This study represents the first demonstration of a division of labor for biosynthetic heterologous pathway using syntrophic communities of Y. lipolytica . Our findings shed light on the potential of utilizing non-conventional microorganisms to form enhanced synthetic communities for bioproduction of various value-added molecules."
} | 3,632 |
34050025 | PMC8179235 | pmc | 983 | {
"abstract": "Significance Ocean warming has caused catastrophic losses of corals on reefs worldwide and is intensifying faster than the adaptive rate of most coral populations that remain. Human interventions, such as propagation of heat-resistant corals, may help maintain reef function and delay further devastation of these valuable ecosystems as society confronts the climate crisis. However, exposing adult corals to a complex suite of new environmental conditions could lead to tradeoffs that alter their heat stress responses, and empirical data are needed to test the utility of this approach. Here, we show that corals transplanted to novel reef conditions did not exhibit changes in their heat stress response or negative fitness tradeoffs, supporting the inclusion of this approach in our management arsenal.",
"discussion": "Discussion Coral Heat Stress Responses Unaffected by Transplantation. Transplantation of bleaching-resistant corals to a novel environment in situ did not alter their heat stress response, despite transplants exhibiting high levels of phenotypic plasticity for other traits. Because bleaching-resistant corals tend to have lower mortality ( 49 ) and higher reproductive success ( 7 , 52 , 53 ) than bleaching-sensitive conspecifics following a bleaching event, they have a clear selective advantage during and in the years following these events. Harnessing these natural advantages by propagating bleaching-resistant individuals is a promising approach to maintain reef function increasing the bleaching resistance of a population using native (i.e., endemic, local) coral stocks. Furthermore, relative bleaching resistance of M. capitata and P. compressa has persisted through multiple in situ bleaching events ( 54 , 55 ), indicating that bleaching resistance is retained and will likely continue during future heatwaves of similar magnitude. This, in combination with heat stress response being unaffected by both transplantation and acclimatization to a complex in situ environment, makes bleaching resistance a promising trait for selecting individuals to enhance resistance of coral populations to climate change. Finally, because M. capitata and P. compressa represent divergent lineages of two globally distributed coral genera, these patterns may be shared with species on reefs around the world. Fitness Consequences of Coral Acclimatization to Novel Environments. The identification of negative tradeoffs during acclimatization is important for informing trait-guided restoration. Indeed, corals acclimatizing to new thermal regimes can exhibit declines in growth and/or reproduction ( 38 , 40 ), reducing the potential benefits of their introduction. Here, despite corals exhibiting high levels of phenotypic plasticity across a range of traits including metabolism, feeding, growth, and reproduction following transplantation to reefs with distinct physicochemical conditions, negative tradeoffs were not observed for either species. In general, corals at the Outer Lagoon performed better overall, and improvements in any one trait did not come at the cost of another. These results are consistent with data from other reef systems that demonstrate an absence of tradeoffs between bleaching and reproduction ( 56 ) and between resistance traits against multiple stressors ( 57 ) and holds promise that these bleaching-resistant genets may also withstand additional stressors. Critically, bleaching-resistant cross-transplants maintained fitness equal to or higher than that of native corals, despite having acclimatized to substantially different environmental regimes. This indicates that these corals would have a neutral or positive effect on the fitness of recipient populations, even during a nonbleaching year, and would likely elevate the recipient population’s fitness during future heatwaves due to their greater bleaching resistance. The duration of elevated fitness in cross-transplants, which lasted at least 11 mo in this study, remains unknown and could be the result of a temporary carryover of the energetic benefits of having originated from a more favorable reef environment. This potential lag effect, as well as seasonal cycles, have been shown to affect corals on an annual cycle (e.g., refs. 58 and 59 ), and future work is needed to assess multiyear influences on coral fitness. However, even if this carryover were transient, the long-term fitness effects for recipient populations are likely net positive due to the transplants’ higher expected relative performance during increasingly common marine heatwaves. These results are a necessary first step to validate trait-guided approaches in reef restoration and adaptive management. Additional work is needed to determine the persistence of these traits in the population, which requires they be both heritable, as has been shown for several species ( 42 – 45 ), and introduced in sufficient abundance. Initial studies indicate that stress-resistant corals must be introduced in numbers equivalent to at least 2 to 5% of the population per year for several decades in order to achieve adaptive gains in heat tolerance that can keep pace with climate change ( 60 ). As such, work is needed to scale up these approaches if they are to have a meaningful impact on coral reef resistance to ocean warming. However, this approach cannot work in isolation, and it is imperative that investments in adaptive management are supported by strong local measures to maintain water quality and limit overfishing ( 61 , 62 ). In addition, these measures cannot replace global measures urgently needed to reduce carbon emissions and slow the intensification of climate change. Genotype–Environment Effects. Despite consistently higher mean coral performance at the Outer Lagoon reef, in many cases, these differences in performance between the two reefs for an individual metric were not significant due to a strong genotype–environment (GxE) effect. Growth in particular showed a strong GxE effect, aligning with recent work cautioning against using growth alone as a predictive trait for future coral performance as it can vary across time ( 63 , 64 ). Furthermore, heat tolerance in a stressful environment does not ensure rapid growth in a less stressful environment ( 65 ). Our results do support the need for a genetically diverse “planting stock” to account for the wide range of expressed phenotypes in different reef environments ( 66 ). In summary, this study indicates that heat stress response was not plastic in M. capitata and P. compressa , and past bleaching resistance is thus likely indicative of future coral performance. Biologically Guided Strategies for Coral Reef Restoration. There is mounting evidence that the current rate of ocean warming is outpacing the “natural” dispersal rate of heat-tolerant genets and the generation times required for adaptation to increase heat tolerance of coral populations ( 67 ). This reality underscores the need for scientifically informed human interventions in management and restoration. Here, we show that the heat stress response of bleaching-resistant corals was unaltered following transplantation into novel environments, and this was accomplished without incurring fitness costs. While more work is needed to determine how well bleaching resistance persists across generations, these results support the use of active restoration for promoting climate resilient reefs. Additional traits are also important when selecting individuals for restoration [e.g., ocean acidification tolerance, disease resistance, and genetic diversity ( 68 )], although the plasticity of many of these traits are not well described. Encouragingly, relative growth during acidification stress is consistent in several coral species ( 69 ) and thus, along with bleaching resistance, may be a useful selection marker for promoting climate change resilient reefs via active management. Site selection for nurseries and outplanting is also an important consideration to maximize restoration success, as water quality is critical for outplant survival ( 62 ) and can be managed at the local level. Here, we found that the reef with the greatest water flow, diel physicochemical variation, and distance from land resulted in higher coral growth and fitness. Sufficient water flow is generally beneficial for coral performance across reef systems ( 70 ), and both flow and temperature variability can mitigate bleaching responses ( 21 , 36 , 70 ), indicating that these may be generalizable environmental characteristics of reefs that promote coral fitness and bleaching resistance (although, for exception, see ref. 37 ). Our results highlight the importance of local management of water quality and ecosystem health (e.g., limiting fishing pressure) and the need to select sites for nurseries and outplanting that promote high coral fitness, as this could accelerate the successful establishment of corals. Furthermore, in situ nursery sites that promote faster growth would provide obvious logistical benefits, leading to shorter residence times for individuals and greater yields for outplanting. Assisted gene flow using climate change–resistant genets could complement traditional conservation measures such as marine protected areas, which could provide favorable habitat for stress-resistant outplants, and in coordination with less directed approaches [e.g., adaptation networks ( 71 )] to preserve genetic diversity. Restoration targets may include reefs damaged directly by human activity (e.g., ship groundings, dredging, etc.) or indirectly via bleaching-related mortality. While the former may not always be “high-fitness” sites, this study indicates that using corals from favorable sites or nurseries may still benefit the recipient population at a “low-fitness” reef because: 1) corals from a “high-fitness” reef had higher reproductive success than native corals, likely boosting the fitness of the recipient population, and 2) introduction of bleaching-resistant individuals would likely improve the fitness of that population during increasingly frequent marine heatwaves. Another strategy is to introduce bleaching-resistant genotypes into populations with lower bleaching thresholds (e.g., cooler or more stable mean temperatures), and evidence is accumulating that relative bleaching performance is maintained following acclimatization to both aquarium ( 40 ) and distinct in situ conditions [this study ( 46 )]. However, caution must be taken before moving heat-tolerant corals beyond the thermal regime to which they are adapted, as corals introduced to cooler climates can suffer significant cold stress in winter, exhibiting reductions in growth, reproduction, and survival ( 38 , 40 ). Promisingly, this study showed that bleaching-resistant corals exhibited increased fitness (growth, reproduction, and survival) following transplantation despite the energetic demands of acclimatizing to a complex suite of environmental conditions, a necessary prerequisite for assisted gene flow to successfully introduce heat-resistant alleles into recipient coral populations. Future work is needed to determine whether elevated fitness and bleaching resistance persist across generations and to increase the scalability of such efforts. Coral outplanting efforts are already occurring at a scale of tens of thousands of outplants each year in some regions ( 72 ), and initial analyses of these types of coral restoration efforts indicate positive returns on that investment for a diversity of coral species, indicating that this approach is both scalable and economically feasible ( 73 ). When accompanied by strong local measures to mitigate nonclimate related stressors, adaptive reef management could preserve species diversity and promote reef resilience to climate change, temporarily buying these invaluable ecosystems time as society struggles with reigning in the current climate catastrophe."
} | 2,989 |
36093244 | PMC9396637 | pmc | 984 | {
"abstract": "Electrospinning is a versatile and viable technique for generating ultrathin fibers. Remarkable progress has been made in techniques for creating electro-spun and non-electro-spun nanofibers. Nanofibers were the center of attention for industries and researchers due to their simplicity in manufacture and setup. The review discusses a thorough overview of both electrospinning and non-electrospinning processes, including their setup, fabrication process, components, and applications. The review starts with an overview of the field of nanotechnology, the background of electrospinning, the surge in demand for nanofiber production, the materials needed to make nanofibers, and the critical process variables that determine the characteristics of nanofibers. Additionally, the diverse applications of electrospun nanofibers, such as smart mats, catalytic supports, filtration membranes, energy storage/heritage components, electrical devices (batteries), and biomedical scaffolds, are then covered. Further, the review concentrates on the most recent and pertinent developments in nanofibers that are connected to the use of nanofibers, focusing on the most illustrative cases. Finally, challenges and their possible solutions, marketing, and the future prospects of nanofiber development are discussed.",
"conclusion": "5. Conclusion Over the past few decades electrospinning is grown rapidly and also nowadays it is the major attraction of many industrial and academic researchers for the therapeutic delivery system. The efforts of many researchers towards eliminating the drawbacks of the electrospinning process and inventing an advanced setup of electrospinning are appreciable. Major modifications have been made to neglect the problems in industrial fiber production and enhance fiber quality. Electrospinning is now capable to provide the controlled fiber diameter by using various polymer grades. Some major advancements are also made in the electrospinning setup to provide hollow nanofibers. The applications of nanofibers in various fields are increasing constantly due to their advantages over other nanotechnological therapies also the fact of consideration that the nanofibers are efficacious over microfibers. It is considered that besides their applications the potential of nanofibers is still to be explored and researchers continuously modify the electrospinning setup and the electrospinning parameters to assess the controlled unit fiber morphology of interest. The area of electrospun nanofiber is grabbed majorly by the polymer industries and textile engineering. In the future, electrospinning is supposed to be the bigger area for the fabrication of nanomaterials with the enumorus applications.",
"introduction": "1. Introduction Nanotechnology is one of the fastest growing fields of study and has entered the profession of medicine and other industries as well. Due to the wide variety of applications many researchers hailing from countries all over the world have shown their interest in the technology. With respect to the pharmaceutical industry, nanotechnology made a mark because of its efficacy and reduced adverse events. 1–5 Various fields of nanotechnology are discussed by Adlakha-Hutcheon et al. 6 and are diagrammatically shown in Fig. 1 . Fig. 1 Summarize the fields of nanotechnology and their classification. Nowadays a maximum number of drug targeting methods are based on nano-formulations because targeting therapy becomes easier when given in the form of a nano-formulation. 7–16 In the category of nano-formulations nanofibers have vast application in the field of medicines, 17–19,19–25 and amongst the nanofibers the magnetic nanofibers have different applications due to their characteristics and magnetic memory, as well as environmental treatment, catalysts, and biomedicine. 26,27 Nanofibers can be simply defined as the fibers of the polymers having the size in the range of nanometer where the temperature of calcination and holding time affects the shape, diameter, fiber morphology, and crystallite size of the nanofiber structure. There are plethora of polymers that are being used for the manufacturing of nanofibers and other than above mentioned factors the diameter of the nanofiber heavily depends on the type of polymer used for the fabrication. 28 When it comes the length of a unit nanofiber, it is almost around 60 nm, but the diameter does not vary much, it is found around 100 nm. Evaluations of the nanofibers are done using vibrating sample magnetometer, X-ray diffraction, Fourier transform infrared spectroscopy, scanning electron. 29 In the method of creation the electrospinning is very common and the basic principle of this method revolves around the static charges that are produced due to stretching of the polymer fiber repel each other which results in the formation of the fibers. 30 Application of the nanofibers heavily depends upon the drug loaded in them as is the case with all the formulation in the pharmaceutical industry. Some of the applications include wound healing, targeted delivery of drugs, pH-sensitive drug therapy, gene delivery, photodynamic therapy, veterinary delivery, stent coating, theragnostic delivery of drugs, ophthalmic delivery, etc. 31 Apart from this the general nanofibers have environmental applications also including, air filtration in which nanofiber beds/mats of mesh size 500 nm alone or in combination with any other filtration media used to filter the polluted air, dust particles. These filters work on the principle of absorption of dust particles over the individual nanofibers in the mesh which makes them highly efficacious. Also, these nanofiber filters are used widely in vehicle cabin filters and building filters. 32 Although they have some biomedical applications in terms of tissue engineering, wound dressing, drug delivery and fabrication of face masks. Now the fact is quiet known that the mortality rate is huge in the human population by the cancer, and thus nanofibers composed of anticancer drugs have caught the eye of crowd. As nanofibers provide a large surface area, high drug encapsulation efficiency due to their extreme porosity, good usability, low cost of fabrication with great benefits diverse research is carried out in cancer treatment via nanofiber delivery. 33 There are a lot of fabrication processes are available for the manufacturing of the nanofibers, but electrospinning is the widely used method for the processing of the nanofibers. Electrospinning was used to create nanofiber composite membranes in which electrospun nanofiber coated with carbon, metal oxide, or carbon-coated metal oxide, which were subsequently modified with a thermally induced chemical cross-linking procedure. They have wideband applications (W/B) fields like environmental-desalination, oil/water adsorption, and separation, membrane separation, and waste water treatment. Although, they have biochemical applications and applications in the field of defense and chemistry which are further explained in the application part. 34 Physically connected structures and chemically cross-linked networks between Polyvinyl Butyral (PVB) and Blocked Isocyanate Prepolymer (BIP) enhanced the mechanical and waterproof characteristics of the nanofibrous membranes. 35 I. Alghoraibi et al. explained different techniques to fabricate the nanofibers which are illustrated in below Fig. 2 . 36 Fig. 2 Illustrates the classification of different techniques used to fabricate the nanofibers. \n Fig. 3 depicts the survey of several publications per year from 2002 to 2022. This quantitative data in the literary survey is supported by the Scopus database with the keyword ‘Electrospun Nanofibers’. Fig. 3 Graphical representation of publications per year from 2002 to 2022 obtained from the Scopus database with the keyword ‘Electrospun Nanofiber’. The current review focuses on various fabrication techniques, applications, and challenges in the synthesis of electrospun nanofibers with compiled strategies to overcome challenges, additionally, it covers current available marketed products and future perspectives together which makes ease for readers to understand the concept."
} | 2,047 |
32960367 | PMC7509015 | pmc | 985 | {
"abstract": "Harvesting waste biomechanical energy has provided a promising approach to improve the power supplement of wearable devices for extending usage life. Surface morphology is a significant factor for enhancing output performance of triboelectric nanogenerator; however, there is a limitation for evaluating the morphology of the surface and its impact on power generation. To evaluate the relationship between the surface morphology and transfer charge, there is a mathematical theory that is the fractal geometry theory that has been proposed to analyze the characteristic of irregular surface morphology. This theory provided a good understanding of the contact area and roughness of the surface. We have designed three categories of knit structures with cord appearance by using a flat knitting machine and analyzed their surface characteristics. Meanwhile, the geometric structures can be demonstrated through the fractal dimension for evaluating the generated output performance during contacting and separation. The present research exhibits that, with the increasing number of knitted units, the triboelectric power-generation performance continued to reduce due to the available contact area decreasing. After calculating the fractal dimension of different knit structures, the m*n rib structures show the high transfer charge when the fractal dimension is close to number one, especially the fractal dimension of the 1*1 rib structure that can reach 0.99. The fractal theory can be further used as an approach to evaluate the influence on the output performance of irregular surface morphology, unrelated to the uniform convex unit distraction. The result of this research also demonstrated the feasibility of a knitted-based triboelectric nanogenerator in scavenging biomechanical energy for powering portable electronics integrated into garments.",
"conclusion": "Conclusion We have demonstrated that the knitted textile with high flexibility and excellent transfer charge can be applied in flexible TENGs for harvesting irregular and low-frequency biomechanical energy, which owns an outstanding output performance. To identify the relationship between surface morphology and output property, fractal theory has been used to quantify the surface geometry and used to evaluate its influence on the transfer charge ability of surface appearance. Different knit structures have been fabricated that can analyze their impact on energy harvesting. From the aspect of the knitted unit, the result shows that the maximum output of 1*1 rib structure can reach at 213 V with the minimum knitted unit. In addition, to further understand the working mechanism and the geometry of contact area, the various knit structures have been illustrated in a fractal dimension that is distinct from traditional dimension. Through calculation, different knitted structures with identical knit units can be used to obtain fractal dimension with the same knit units. The generated electrical output can be increased with the fractal dimension close to the value of one. Therefore, the difference between the fractal dimension and the value one can be used in the evaluation of transfer charge ability according to the irregular surface. In the near future, it is expected that an evaluation for generating output ability based on fractal theory in constructing a triboelectric nanogenerator, obtaining maximum output performance to optimize the flexible self-power system for harvesting wasted human motions in our daily life will be investigated.",
"introduction": "Introduction Advanced intelligence techniques have swept the global world and have brought out some novel flexible smart wearable devices, such as health tracking sensors [ 1 , 2 ], gesture-detecting devices [ 3 – 6 ], electronic skins (E-skins) [ 7 , 8 ], flexible circuits [ 9 , 10 ], and optical fiber wearables [ 11 , 12 ]. However, with disadvantages of mass weight, low conversion efficiency, serious environmental pollution, and short battery life, the power supplement is the enormous limitation for the development of electronics. Since the first triboelectric nanogenerator (TENG) has been developed successfully in 2012 [ 13 ], based on the characteristic of small scale, lightweight, various materials, safe, environmental virtues [ 14 ], and high efficiency, it has provided a promising and effective strategy to address above straits. Along with the rapid advent of TENGs working through a coupled effect of contact electrification and electrostatic induction [ 15 ], it has been conformed as one desirable approach to gain mechanical power [ 16 , 17 ] from our surrounding especially by harvesting low-frequency and irregular movements (including wind [ 18 , 19 ], waterdrop and human motion, biomechanical energy, etc. [ 20 – 22 ]), realizing data transmissions [ 23 – 25 ] and power supplement in the Internet of Things (IoT) [ 26 ]. For wearable devices, textiles are regarded as the best substrate, due to its structural retention and fatigue resistance, soft, integration, and high porosity. To date, an integration of a triboelectric nanogenerator and traditional textile [ 27 – 33 ] is one of the promising candidate for human-oriented wearable devices, such as self-powered flexible sensors [ 34 ], wearable energy harvesters, and textile-based energy storage systems. It is also endowed conventional textiles with functionality, intelligence, and high additional value . These electronic devices based on the textile that are satisfied with the requirement of lightweight, inexpensive, comfortable, breathable, portable, long-lasting, and washable for routine usage. In addition, it is facile to make textile with variable colors and abundant pattern designs which represent attractiveness for intelligent textiles. Especially, knit textiles with small strain and large deformation are sensitive to signal generation thus are ideal to be used for flexible sensors, overcoming movement resistance, and reducing energy loss [ 35 ]. Additionally, frictions and deformations of knit textile are common phenomena that are a thrilling opinion for constructing a triboelectric nanogenerator. As we all know, surface-morphology modification is a significant approach to enhance output performance of TENGs [ 36 – 39 ]. Most are purposed on increasing the available contact area and roughness of the surface. There are two primary methods which change the surface morphology, one being surface etching, the other being surface replication. However, the use of highly expensive, limited treatment area and multi-step manufacturing technique to generate surface appearance is difficult for industrial production. Herein, Li et al. [ 40 ] investigated a polydimethylsiloxane (PDMS) film with surface microstructures peeled off from the sandpaper, which was a one-process and low-cost method to prepare the difference roughness of the surface. The experimental results showed the generated maximum output of 46.52 V under the roughness class of 3000 detected by a 3D optical surface profile. Besides, too many microstructures can decrease severely the effective contact surface and result in the reduction of the ability of power performance. The size of TENGs was limited by the sandpaper area, which led to increase the fabrication cost. Nowadays, textile structures are receiving increasing attention due to formation of abundant surface appearances [ 38 ] without the complex fabrication procession and high cost. For fully understanding textile surface appearances, some factors need to be considered in terms of unique components and structure features, including the thread outlook, textile physical parameters, and knit structures. Then, Kwak et al. [ 41 ] investigated the contact area of three structures (including plain-, double-, and rib-fabric structures) during stretching and discussed the contribution for enhancing the potential. It was worth that rib-fabric can be strained up to 30%, enlarging contact area to 180 cm 2 . Depending on the middle region existing, rib fabric can be stretched largely, which can obtain a higher potential for increasing contact area. As the primary element of the textile structure, the characteristic of loops was analyzed that was also the significant factor for influencing surface appearance. Huang et al. [ 42 ] made a focus on the effect of basic parameters of textile (including loop legs, loop sinkers, and textile density) for confirming the difference on the output performance. The large stitch density fabric-based triboelectric nanogenerator could generate higher electric energy with a maximum peak power density of 203 mW m −2 at 80 MΩ, which makes a larger effective contact area. The result exhibited that the surface morphologies of various fabric structures had an influence on the electrical-output ability. In order to harvest much more energy for extending lifetime, 3D double-faced interlock stitch textiles [ 43 ] were knitted by double-needle bed flat, which exhibited the same output performance on front and back side. In addition, the TENGs based on the three-dimensional textile structure could generate a high-power density of 3.4 mW m −2 at the external resistance of 200 MΩ, demonstrating that the capacity of energy harvesting has been improved. However, the abovementioned surface appearances have little depiction on the geometry shape of the surface, and factors about generated transfer charge are still suffering from lack of specific explanations. There is no universal manner that can characterize surface appearance, which needs to find an evaluation of irregular morphology. Therefore, that is the limitation for fully understanding the transfer charge on the triboelectric nanogenerator currently. The purpose of surface analyzation is to characterize textiles’ geometric structures, which may be tested in two approaches of contact method and optical method [ 44 ]. The contact method can describe surface morphology well, but the time needed is much longer, and the needle leaves a trace on the surface. Compared to the contact method, with the benefits of short measurement time, low harness surface, and easy detection, the optical method has been used for detecting surface roughness. However, the false gaps and high level of noise may reduce the judgment of the real surface morphology. The mathematical tool is a theory analysis that can be used to quantify the extent of surface roughness. It is a novel approach to evaluate the irregular surface. With such an uneven surface, the conventional mathematical method of Euclidean geometry cannot be used because it is really hard to judge the quantitative geometry dimension and measurement accuracy, such as length of segment and weight of the object. However, fractal geometry, an approach named by Mandelbrot for describing irregular structures, has been provided to solve the issue and define the irregularity in nature [ 45 ], such as the physical properties of foams [ 46 ] and evaluation for fabric smoothness [ 47 ]. Almost all the rough surfaces can be divided into some self-similar parts which can be depicted by a non-integral dimension, named fractal dimension ( D f ). Based on the various geometric surfaces, the value of D f needs to be considered and analyzed that has an effect on the roughness and efficient contact area in the design of a triboelectric nanogenerator, optimizing the capacity of converting human motions into electrical. Herein, in this work, we present the various surface morphologies based on knit structures that are adopted as one of the dielectric layers. The knit-textile-based TENG was fabricated by using commercial threads and industrial knitting machine, which can realize the large-scale production and practical applications. To imitate the flapping hand movement, TENGs are designed in the contact-separate working mode (CS) which is the simplest working mechanism. The knit structures are formed in two kinds of approach, including structured- and shaped-based convex-concave surface morphology. Due to the diversity of knit structures, the resultant surface appearances can be systematically investigated and analyzed for confirming the relationship between surface morphology and knit structures. The D f of every fabric can be calculated through the appropriate fractal principle, evaluating the roughness of the fabric surface. The maximum transfer charge of surface appearance in 1*1 rib can reach up to 91.66 nC by flapping and releasing motion, which obtain the fractal dimension of 0.99. And an interesting phenomenon exhibits that with the value of D f closing to the number one, the transfer charge can be higher. Finally, using the fractal theory and knit structures can provide an effective method for quantity evaluating the transfer charge and are expected to be of help to design the knit-textile-based TENGs with more efficiency, industrial production, and inexpensive cost.",
"discussion": "Results and Discussion In order to confirm friction materials, the triboelectric order [ 49 ] is the significant reference, which quantified the triboelectric polarization of different common materials. The triboelectric order presents that one side shows gaining charge capacity and the other side owns a high ability to lose electrons, which have been defined as the fundamental material performance. To obtain the outstanding output performance, a couple of materials are selected that need to be attributed to the triboelectric series with a considerable distance, increasing the potential difference . Herein, one is the commercial, low-cost, excellent abrasion resistance and highly positively charged tendency (nylon) and the other one shows negatively charged tendency (PTFE). In this work, we selected the PTFE membrane without any treatments on the surface. Herein, the only factor is knit structures that can be analyzed by the performance of transferring charge. Another critical element is the electrode material that is copper foil with high flexibility, which can be pasted directly, that is a simple and one-step fabrication process. Compared to the precious metal of silver and gold, the price of copper foil is inexpensive and can be used to fabricate the economical products. So copper has been widely applied as flexible circuits and electrodes in the design of smart devices. At present, there are four universal working mode-operated TENGs corresponding to the different electrode structures and movements. With advantages of facile fabrication, the abundant material selection, reciprocating vertical direction movements, the CS TENGs are the first deeply investigated that have the potential ability for harvesting some biomechanical energy, such as flapping hands, walking, and running. Here, in order to investigate the influence principle of the surface structures, the triboelectric nanogenerators based on knitted textile (KNGs) have been designed, corresponding to the contact and separation between nylon fabric and PTFE film. The process of assembling the triboelectric nanogenerator is presented in Fig. 1 a, consisting of knitted fabrics, PTFE membrane, and copper foil. The versatility of flexible knitted fabric in terms of its capacity to crimping (Fig. 1 bi), bending (Fig. 1 bii), draping (Fig. 1 biii), and folding (Fig. 1 biv) in any direction is tailored in various scales depicted in Fig. 1 b. The KNGs can be designed based on the requirement of application position and the esthetics of clothing. The diversity of knit structures has been knitted with different surface appearances, and then these photographs of the textile surface have been shown in Fig. 1 c.\n Fig. 1 Schematic preparation, characteristic of KNG, and knit structure. a Fabrication process of KNG. b Images of KNG under various deformation. i, crimped; ii, bent; iii, draped; iv, folded. c All of the fabricated knit structures, from number 1 to 10 The operation mechanism of the KNGs is simply presented in Fig. 2 a. To measure the transfer charge, the maximum distance and the frequency of movements of the linear motor are set as 10 cm and at 0.3 Hz for simulating flapping hands motions, respectively. As for common monitorization, the open-circuit voltage (Voc), short-circuit current (Isc), and transfer charge (Qsc) are measured by a mechanical linear motor. In the original state (Fig. 2 ai), the nylon textile produced positive charges and PTFE film was charged with negative charges because of the electrostatic induction and conservation of charges. When the device was pressed (Fig. 2 aii), a shrinkage of the gap between the both contact surfaces will lead to the positive charge accumulating in the electrode pasted on the PTFE. The electrons flow from the external circuit for balancing the potential difference. It was worth noticing that the equivalent amount of electrons can be maintained on the surface of the contact area because both dielectric materials are insulators (Fig. 2 aiii). As the PTFE moves back (Fig. 2 aiv), the process reversed and the electric will be obtained balance once again between nylon textile and PTFE, reflecting the neutralization of charges. Consequently, the electrons will flow back for electrical potential differences. In this situation, the KNGs could generate Isc and Voc, which have a characteristic of periodical change, shown in Fig. 2 b and c. In Fig. 2 b and c, the inset is an enlarged graph which is described in one cycle.\n Fig. 2 Electrical power working mechanism and the output performance of KNG. a Operation mechanism of KNG using nylon fabric contact with PTFE member. b Voc of KNG and enlarged image for one cycle. c Isc of KNG and enlarged image of one cycle In order to fabricate convex structures on the textile surface, there are two kinds of methods used, including structure design and shape formation, as shown in Fig. 3 . The structure design is dependent on the different proportion of the face loop stitches and the reverse loop stitches. The total samples are designed in seven rib types, including the type of m*n ( m = n = 1, 2, 3, 4) in Fig. 3 a and 2*m ( m = 1, 2, 3, 4) shown in Fig. 3 b. The rib has a vertical cord appearance due to the face loop wales that tend to move over and in front of the reverse loop wales; then, the cord maximum height can arrive at 0.2 cm. The rib of m*n ( m = n = 1, 2, 3, 4) can be balanced by alternate wales of face loops on each side, so it lies flat without curl after tailoring. And both sides of the textile are the same appearance as shown in Fig. 3 e. However, the different proportions of face and reverse loops in 2*m rib structures, there is a distinction surface come out, as shown in Fig. 3 f. In addition, the stretching process of rib fabric is divided into two stages, including the reverse wales intermeshing on both sides until being stretched to reveal the reverse loop wales in between and then whole loops are continued to be stretched over twice as wide as an equivalent single fabric. Therefore, compared to plain fabrics, rib textiles have potential to increase the stretchable ability for harvesting flapping and stretching movements (transverse direction and longitudinal) during the contact-separation working mode. The other method for establishing raised structure is the shape deformation wherein the air layer is formed on the surface of the n ( n = 4, 5, 6) textile that is illustrated in Fig. 3 c. The thickness of cross-sectional area is in the range from 0.15 to 0.3 cm. The characteristic of the air layer is a prominent arch structure that can provide some space for accelerating electron separation when triggering motions. Above all mentioned, knitted textiles are designed through a computerized flat machine that can realize the knitting location accuracy, forming the whole garment and integrating smart materials into cloth perfectly. Such knitting-technique nomenclature has been marked on Fig. 3 d that depicts the features of structure correctly.\n Fig. 3 Schematic characteristics and components of knit structures. a Characteristics of m*n rib. b Characteristics of 2*m rib. c Characteristics of some needle horizontal cord. d Knitting-technique nomenclature. e Image of the face side and back side in the 1*1 rib structure. f Image of the face side and back side in 2* 1 rib structure Previous works [ 42 ] demonstrated the effective contact area of face side that was much more than the back of textile; result in transfer charge was twice as high as the output performance of back side. This is because the length of needle loop was longer than the sinker loop. Therefore, to enhance the output performance and to create only one influential factor, the contact-raised structures consist of face side loops. The outputs of the KNGs depending on the number of the convex units are plotted in Fig. 4 . A decreasing trend where the contact area of all experimented textiles was decreased with the number of the raised unit was formed. Also, the more significant electrical charges are in the sequential order of the 1*1 rib, 2*1 rib, and four needle shape-type structures (the first point of each line) with the values of 91.66 nC, 90.19 nC, and 69.64 nC, respectively.\n Fig. 4 The output performance changed through the number of the convex unit Then, the knit structure with the different surface morphology in aspects of diversity wale density, number of face side unit, and structures are investigated. All of the parameters of ten kinds of knitted textiles are tested and recorded in Table 2 . Notably, the course density is always constant because the cord appearance has grown along the vertical direction when analyzing sample Nos. 1–7. So, wale density as the main factor which needs to be discussed refers to the features of different knit structures. It is obvious that face loop and reverse loop have the same proportion in the Nos. 1-4, about 50%. These textiles show the same structures no matter what is the face or back side based on the double stitch knitting. The average thickness shows higher compared to the sample nos. 5–7 that consists of a different number of face stitches and reserve stitches. Texture no. 4 owns the largest repeat unit that its wale density is twice as large as no. 1. However, the number of face side units on the practical fabric is nearly a half decline than no. 1. This is because the more sinker loops are stretched with each other so that the column appearance can be formed. With the knitted unit increasing, the diameter of the column and the thickness of fabrics are enlarged, herein, decreasing the number of face side unit and the efficient contact area when triggering movements. In terms of rib structure with different proportions of face and reverse loop, the appearance exhibits the characteristic of single-faced structure obviously, with knitted repeat units increasing. Meanwhile, the wale density of no. 7 is as large as no. 1 and no. 5, but the number of face loop units has distinctive differences due to the number of the knitted unit is six loops that are much more than no. 1 (2 loops) and no. 5 (3 loops), so the output performance is lower than that of no. 1 and no. 5. As a result, the rib-knitted fabric no. 1 represents the most face loop units in nos. 1–10 during the contacting-separating movements.\n Table 2 Parameters of knitted textiles No. The proportion between face side and reverse side/% Wale density/in. Course density/in. Thickness/cm no. 1 50 26 18 2.93 no. 2 50 32 18 2.25 no. 3 50 36 19 2.72 no. 4 50 42 19 3.36 no. 5 66.7 25 19 2.83 no. 6 40 30 19 3.07 no. 7 33.3 26 19 2.20 no. 8 - 18 36 2.71 no. 9 - 18 39 2.83 no. 10 - 18 10 3.01 On the other side, the shape-type knitted textile has been designed through the different number of loops assembling into the whole fabric, forming arch structures. Due to the direction of cord length is horizontal, the wale density of the fabric shows approximate stability in the transverse direction. The arch structure provides an approach for separating charge on the surface, which has a hollow inner space. Thus, the efficiency of harvesting wasted mechanism energy has been improved. Generally, in order to enhance output performance, an arch type is made of flexible materials with perfect elastic and durability, such as silicon substrates, but it is tough to be knitted in industrial knitting machine for meeting the commercial requirements. When it comes to see the arch structure is based on the knitted textile in previous researches [ 24 , 41 , 50 ], the construction needs to be sewed or taped, which is a complex and time-consuming process. We presented a knitted-arch textile that is prepared through the whole forming technique without second manufacturing that endows the high efficiency of production. Among the horizontal cord structures, the 0.3-cm height shows the lowest charge output compared to four needle and five needle horizontal cord structure with the height of 0.15 cm and 0.2 cm, respectively, which can be influenced by the low stiffness of knitted textiles in a large distance between both ends fixed. The highest convex shape is hard to keep arch with force pressure and recover to pristine shape, which leads to some charges are neutralized. As a result, the decrease of the arch height can enhance the tolerance of convex structures. However, such shape-type cords reduce the effective contact area which is a line type that has little areas than real contact, decreasing the performance of electrical-output. Loops have irregular structures, so the evaluation of their geometry properties such as stitch size and surface shape is challenging. To identify the irregularity of the loops, traditional evaluation which is an integral dimension cannot be utilized. The fractal theory is suggested to analyze the category of irregularity in our surroundings and nature. The proposed concept of fractal dimension is an excellent tool for exhibiting complex morphology that presents the rules, the complexity, and the roughness of the textile surface. Because all fractals are not self-similar completely, the mathematical calculation is used to argue the geometry configuration. In order to understand the knitted structure’s surface, some images visualize the information carried in Fig. 5 d. As shown in Fig. 5 d, the characteristic of convex surface can be intuitively observed from different perspectives where the evidence for confirming the raised morphology is.\n Fig. 5 Fitting curve and some visual images for knitting textiles. a The m*n rib structure. b The 2*m rib structure. c Some needle horizontal cord structure. d The visual images from different aspects The uneven surface has been formed with the knit structure designed caused by the yarn morphology and structure design. The fractal geometry is an efficient calculation for evaluating the textile surface and understanding the characteristic of knitted structures and ability of triboelectric charge generation. In fact, with the increase of the raised unit, it can improve the uneven knitted textile owing to the surface shape modified. Although all of the knitted textile own convex structures in longitude and transverse direction, the degree of similarity is still not confirmed that is the significant reference value for whether using fractal dimension successfully or not. To estimate the feasibility of fractal dimension, all of the knitted fabrics are calculated through measuring the width of the convex unit, the size of loops in length, and width when textiles stay in stable size. Figure 5 a, b, and c show the fitting curve of fractal dimension of nos. 1–10 type fabrics, and slope of a line means the fractal dimension. The existence of the relationship is found in convex structures of the ten different types of knitted textiles, which confirms the fractal characteristic of ten knitted fabrics. Therefore, the fractal theory applied in the analysis of diversity knit structure that is practicable. Figure 6 a–f illustrates the generated Isc and Voc based on the practical applications of contact and separation working KNGs, based on the structure types and shape types. There is a trend that a decrease with the knit unit increases about the Isc and Voc as shown in Fig. 6 a–f. This is because the Isc is changed with the effective contact area which is affected by knit structures.\n Fig. 6 Schematic illustration of fractal dimension and generated Isc and Voc. a The Isc of m*n rib. b the Voc of the m*n rib. c The Isc of 2*m rib. d The Voc of the 2*m rib. e The Isc of n type. f The Voc of n type. g The D f -transfer charge curve. h The F value curve When calculating the D f of various knit structures, the investigated knit structure states that the different knit structures have an unequal value which is non-integral dimension due to the different components of convex as demonstrated in Fig. 6 g. As for Fig. 6 g, this is the image of the transfer charge versus fractal dimension curve of diversity structures. The rib structure presents desirable output performance and the fractal dimension near the value of one. The TENGs based on structure-type knitted-textiles have a higher transfer charge than shape type and the value of D f about the m*n rib type, 2*n rib type, and n type is in the range of 0–2, 0–1, and 1–2, respectively. Generally, the fractal dimension symbolizes the extent of surface roughness which is the roughness increasing with the large D f . However, the shape-type fabrics are designed in horizontal cord with small line-contact area, so the roughness has little influence on the transfer charge. In order to demonstrate the influence on D f of convex structure homogeneity in rib structures, the random side length is chosen and calculated. The result exhibits as follow:\n 1-4 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ \\varepsilon \\left(a\\ast b\\right)=M\\left(l\\ast b\\right) $$\\end{document} ε a ∗ b = M l ∗ b 1-5 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ N=\\frac{a}{l} $$\\end{document} N = a l where a is the length of the whole fabric, b is the width of the convex unit and is equal to the width of the whole fabric, l is the length of the convex unit, M is the number of the convex unit, N is the repeated multiple of self-similar units that is the length of convex units to the length of whole samples, and ε is the proportion of face loop and reverse loop, meaning the uniform of the convex distraction. Then, the calculation of M and N can be used in the formulation (1-3), the result shows that obtained D f is not the same with the D f that is calculated based on the length of actual measurement as shown in Table 3 . No matter how the raised structure is distributed, the value of D f is affected by the practical length and number of cords.\n Table 3 Compared to D f in random and actual length No. D f random D f actual no. 1 0.8488 0.9948 no. 2 0.8321 0.9793 no. 3 0.8190 0.974 no. 4 0.8168 1.0318 no. 5 1.1168 0.9884 no. 6 0.7784 0.9602 no. 7 0.7229 0.8979 It is noted that the fractal dimension of the 2*1 rib structure is close to the 1*1 rib reach at 0.99, and thus, the transfer charge is much the same as shown in Fig. 6 g. The generated electrical-output performance shows the highest when the D f is near the value of one. That has provided one guess if the fractal dimension can evaluate the surface morphology and character the output performance. To investigate the correlation of fractal and transfer charge, the difference between the fractal dimension and the value of one (named F value) has been illustrated in Fig. 6 h. The operating results show a trend that is decreased F value can boost the much higher Voc, taking evidence for potential application of fractal dimension. However, the F value is regarded as an evaluation of the roughness structures, which needs to consider the properties of the primary loop of the structure. Then, the influence on transfer charge is discussed comprehensively. The sample of no. 4 and no. 6 has a similar F value, but the massive difference exists on both of output performance. The surface morphology of no. 4 shows the planar structure due to the same number of face and reverse loops, so the transfer charge is low. But no. 6 has prominent appearance due to the reverse loops over the face stitches and the generated large transfer charge when contacting and separating. Therefore, the selection and design of the knitted structure of the textile based on the F value highly improved the generated total electrical charge, which is an indispensable requirement for construct a high-effective flexible self-power device based on the knitted textiles."
} | 8,302 |
31266255 | PMC6651624 | pmc | 986 | {
"abstract": "Hierarchical Temporal Memory (HTM) has been known as a software framework to model the brain’s neocortical operation. However, mimicking the brain’s neocortical operation by not software but hardware is more desirable, because the hardware can not only describe the neocortical operation, but can also employ the brain’s architectural advantages. To develop a hybrid circuit of memristor and Complementary Metal-Oxide-Semiconductor (CMOS) for realizing HTM’s spatial pooler (SP) by hardware, memristor defects such as stuck-at-faults and variations should be considered. For solving the defect problem, we first show that the boost-factor adjustment can make HTM’s SP defect-tolerant, because the false activation of defective columns are suppressed. Second, we propose a memristor-CMOS hybrid circuit with the boost-factor adjustment to realize this defect-tolerant SP by hardware. The proposed circuit does not rely on the conventional defect-aware mapping scheme, which cannot avoid the false activation of defective columns. For the Modified subset of National Institute of Standards and Technology (MNIST) vectors, the boost-factor adjusted crossbar with defects = 10% shows a rate loss of only ~0.6%, compared to the ideal crossbar with defects = 0%. On the contrary, the defect-aware mapping without the boost-factor adjustment demonstrates a significant rate loss of ~21.0%. The energy overhead of the boost-factor adjustment is only ~0.05% of the programming energy of memristor synapse crossbar.",
"conclusion": "5. Conclusions The SP of HTM has been known as the software framework to model human brain’s neocortical operation such as recognition, cognition, etc. However, mimicking the brain’s neocortical operation by hardware rather than software is more desirable, because the hardware not only describes the neocortical operation, but also employs the brain’s architectural advantages such as high energy efficiency, extreme parallel-computation, etc. To realize HTM’s SP by hardware, in this paper, we developed the memristor-CMOS hybrid circuit. One thing important for hardware implementation is that memristor defects such as stuck-at-faults, memristance variations, etc., should be considered in developing the memristor-CMOS hybrid circuit of SP. For considering memristor defects in hardware implementation, first, we showed that the boost-factor adjustment can make HTM’s SP defect-tolerant, because the false activation of defective columns can be suppressed. Second, we proposed the memristor-CMOS hybrid circuit with the boost-factor adjustment for realizing the defect-tolerant spatial-pooling in hardware. The proposed circuit does not rely on the conventional defect-aware mapping scheme, which cannot avoid the false activation of defective columns in spatial-pooling. For the MNIST data-set, the boost-factor adjusted crossbar with the defects = 10% was verified to have a rate loss as low as ~0.6%, compared to the ideal crossbar with the defects = 0%. On the contrary, the defect-aware mapping without the boost-factor adjustment demonstrated a significant rate loss, as much as ~21.0%. The energy overhead of the boost-factor adjustment was estimated to be as little as ~0.05% of the programming energy of the memristor synapse crossbar.",
"introduction": "1. Introduction The human brain’s neocortex covers the brain’s surficial area, which is known to carry out the most intelligence functions. The thickness of neocortex has been observed as thin as 2.5 mm, where six layers are stacked one-by-one [ 1 , 2 , 3 ]. The six neocortical layers seem to be columnar, in which the complicated vertical and horizontal synaptic connections are intertwined among neurons to form the 3-dimensional neuronal architecture [ 4 , 5 ]. The neocortical neurons collectively respond to human’s sensory information from retina, cochlea, and olfactory organ [ 6 ]. The collective activation of neocortical neurons are trained over and over with respect to time, by changing the synaptic connection’s strength according to the sensory stimuli. The neuronal activation and synaptic plasticity can be thought of as a fundamental aspect of human perception and cognition, which are computed in a different way from the conventional Von Neumann machines. As a software framework, Hierarchical Temporal Memory (HTM) has been developed to model the cognitive functions of neocortex [ 7 , 8 , 9 , 10 , 11 ]. By doing so, HTM can recognize and interpret various spatiotemporal patterns, mimicking how the human brain’s neocortex understands human’s sensory stimuli. The software framework of HTM is divided into two functional blocks: Spatial Pooler (SP) and Temporal Memory (TM). The role of SP is receiving and learning the sensory information. In SP, the sensory information is transformed into the collective activation of neocortical neurons. From the biological experiments, the neocortical neurons have been observed to be activated sparsely, not densely, in response to human sensory stimuli. The sparse activation of neocortical neurons is mathematically described as Sparse Distributed Representation (SDR) in HTM [ 1 ]. After SP learning the spatial features of the sensory stimuli, TM responds to the temporal sequences of SDR patterns generated from SP. By learning the temporal sequences of SDR patterns, TM can perform recognition and prediction for them. Figure 1 a shows a conceptual diagram of SP operation, where the input-space neurons are mapped to the SP neurons [ 8 ]. Here, the input-space and SP neurons refer to the neurons of sensory organ and neocortex, respectively. The sensory stimuli generated from the input-space neurons are connected with the neocortical neurons, as indicated in Figure 1 a. The lines between the input and the SP spaces represent the synaptic connections. Synaptic weights of the connections are trained according to Hebbian learning rule in HTM [ 8 ]. If an SP neuron becomes active, in response to an input-space stimulus, the synaptic weights belonging to this neuron are strengthened, and weakened otherwise [ 8 ]. The circle zone in the SP space represents a local inhibition area, within which only few neurons are allowed to be active. In HTM, the size of inhibition zone in the SP space can be decided to control the sparsity of neuronal activation. It has been known that the percentage of neuronal activation is as sparse as 2% on average in the brain’s neocortex. This low sparsity of neuronal activation may have something to do with high energy-efficiency of neocortical cognitive operation. In the previous publications, we developed hybrid CMOS-memristor circuits for implementing HTM, which was developed as the software framework originally, as mentioned earlier [ 12 , 13 ]. Memristors have been studied intensively for many years for their potential in neuromorphic hardware, since the first experimental demonstration [ 14 , 15 ]. This is because the memristive behaviors seem very similar with the experimental synaptic plasticity observed from biological neurons. From the biological experiments, the synaptic connections have been observed to be strengthened or weakened dynamically by electrical spiking signals applied to them [ 16 ]. Moreover, memristors can be fabricated to build 3-dimensional crossbar architecture using the CMOS-compatible Back-End-Of-Line (BEOL) process [ 17 , 18 ]. The 3-dimensional connectivity of memristor-synapses is very similar to the anatomical structure of the biological neocortex. In terms of cognitive functions, the memristor crossbar can perform vector-matrix multiplications in parallel, which can be considered very important in implementing energy-efficient computing like human brain’s cognition, unlike the state-of-the-art Von Neumann based computers [ 19 , 20 ]. One important thing to consider in the memristor crossbar is defects, as shown in Figure 1 b. In the real memristor crossbar, there are stuck-defects, such as stuck-at-0, stuck-at-1, etc. [ 21 ]. In addition, variation-related defects can also be considered, where each memristor can have different LRS and HRS values due to process variations [ 22 ]. Here, LRS and HRS mean Low Resistance State and High Resistance State, respectively. Figure 1 b compares the ideal crossbar (without defects) and the real one (with defects). The solid and open red circles with stars represent stuck-at-LRS and stuck-at-HRS defects, respectively. For the memristor defects such as stuck-at-faults and variations, these defects may be caused from the random nature of filamentary current path which can be formed or erased by the applied current and voltage to the memristor. The filamentary current path created or erased during the memristor programming can have statistical distributions like FLASH memory. Various statistical distributions by device-to-device, wafer-to-wafer, lot-to-lot, and process-to-process lead to the variations in memristance and stuck-at-faults [ 21 ]. To minimize a loss of recognition rate due to these memristor defects, we can consider the defect-tolerance scheme based on the conventional defect-aware mapping [ 21 ]. To explain the previous defect mapping scheme, the following logic function is assumed, f = X 1 X 2 + X 2 X 3 + X 3 X 1 + / X 1 / X 2 / X 3 is implemented in the crossbar [ 21 ]. In the logic function, /X 1 means the inversion of X 1 . Figure 2 a shows the real memristor crossbar (with defects). Here, I 1 , I 2 , etc. represent input columns. O 1 , O 2 , etc. are output rows. The gray circle indicates a good memristor cell, which can be programmed with HRS or LRS. The solid and open red circles represent stuck-at-1 and stuck-at-0 defects, respectively. Figure 2 b shows the direct mapping without considering the defect map. P 1 , P 2 , P 3 , and P 4 indicate the first, second, third, and fourth partial products in the target logic function. P 1 calculates X 1 X 2 . However, P 2 calculates X 1 X 2 X 3 , not X 2 X 3 defined in the logic function, because of the stuck-at-1 fault on the crossing point between X 1 and P 2 . P 4 also calculates the wrong partial product. The stuck-at-0 fault is found at the crossing point between /X 2 and P 4 . By doing so, P 4 calculates /X 1 /X 3 instead of the target product of /X 1 /X 2 /X 3 . Figure 2 c shows the defect-aware mapping, where the defects can be used in implementing the logic function according to the defect type and location. To do so, the crossbar’s rows in Figure 2 c are reordered to consider the defect type and location in calculating the partial products. For example, the first row in Figure 2 c is assigned to P 3 , not P 1 . P 1 is assigned to the second row to calculate X 1 X 2 . The stuck-at-1 fault on the second row can be used in calculating P 1 = X 1 X 2 . Similarly, the stuck-at-1 fault on P 4 can be employed to calculate P 4 = /X 1 /X 2 /X 3 . Moreover, the stuck-at-0 faults on P 2 and P 4 do not cause a wrong result for the calculation of partial products of P 2 and P 4 . As shown in Figure 2 c, the defects can be employed in implementing the target logic function according to the defect type and location. However, the defect-aware mapping scheme demands very complicated circuits, such as memory, processor, controller, etc., to be implemented in hardware. Figure 2 d shows the flowchart of crossbar training using the conventional defect-aware mapping. After fabricating the memristor crossbar, the defect map should be obtained by measuring the crossbar. As a post-fabrication configuration, the trained synaptic weighs can be transferred to the crossbar using the defect-aware mapping, as explained in Figure 2 c. To do so, however, the complicated digital circuits, such as memory, controller, processor, etc., are needed for implementing the defect-aware mapping in hardware, as mentioned earlier. Not using the defect-aware mapping, in this paper, we propose a simple memristor-CMOS hybrid circuit of defect-tolerant spatial-pooling, which does not need the complicated circuits of memory, controller, processor, etc., as shown in Figure 2 e, where, unlike in Figure 2 d, the crossbar’s defect map is not used. For developing the hybrid circuit of memristor-CMOS, we first show that the spatial-pooling based on Hebbian learning can be defect-tolerant, owing to the boost-factor adjustment, in Section 2 . Additionally, we propose a new memristor-CMOS hybrid circuit, where the winner-take-all circuit is implemented not using capacitors occupying large area. In Section 3 , the proposed hybrid circuit is verified to be able to recognize well Modified subset of National Institute of Standards and Technology (MNIST) hand-written digits, in spite of memristor defects such as stuck-at-faults, variations, etc. In Section 4 , we discuss and compare the following three cases: (1) Spatial-pooling without both the boost-factor adjustment and the defect-aware mapping, (2) spatial-pooling with the defect-aware mapping, and (3) spatial pooling with the boost-factor adjustment, in terms of hardware implementation, energy consumption, and recognition rate. Finally, in Section 5 , we summarize this paper.",
"discussion": "4. Discussion In this session, to understand the benefit of the proposed circuit exactly, we discuss and compare the following three SP schemes in Table 1 : (1) Spatial-pooling without both the boost-factor adjustment and the defect-aware mapping, (2) spatial-pooling with the defect-aware mapping, and (3) spatial-pooling with the boost-factor adjustment. First, we discuss the possibility of hardware implementation in Table 1 . As mentioned earlier, (1) and (3) can be implemented in hardware. However, the defect-aware mapping of (2), as indicated in Figure 2 d, demands very complicated circuits such as memory, processor, controller, etc. Second, the energy consumptions of the crossbar programming are compared among (1), (2), and (3) in Table 1 . The amount of programming energy is simulated during the training time of 10,000 MNIST vectors (1) and (2) consume 3.9 mJ for programming the crossbar with HRS and LRS, according to Hebbian learning rule, as explained in Figure 4 a. The energy overhead due to the boost-factor adjustment is less than ~0.05% of the crossbar programming energy. This is because each column has only one memristor for the boost-factor adjustment, compared to 400 cells per column for Hebbian learning. For the recognition rate, in Table 1 (1), without the boost-factor adjustment and defect-aware mapping, shows MNIST recognition rates of 77.3% and 55.6%, when the defects = 0% and 10%, respectively. Similarly, (2), with only the defect-aware mapping, shows the rates of 77.3% and 56.3%, when the defects = 0% and 10%, respectively. Without the boost-factor adjustment, the defective columns necessarily become activated frequently. The frequent activation of defective columns degrades the recognition rate significantly, as shown in (2) in Table 1 . On the contrary, (3) with the boost-factor adjustment shows the rates of 77.6% and 77%, when the defects = 0% and 10%, respectively. It has very little loss of the recognition rate, in spite of the defects = 10%. The gap between the defects = 0% and 10% is negligibly small for the crossbar with the boost-factor adjustment. We now discuss the relationship of this work to the previous works performed in HTM hardware realization. Actually, as a previous works of this paper, we developed the memristor crossbar circuits for performing the SP and TM operations of HTM, respectively [ 12 , 13 ]. However, in the previous works, we did not consider the memristor defects, which should be taken into account in the real memristor crossbar having defects of stuck-at-faults and variations. Thus, the SP hardware implemented with the real defective memristor crossbar can be an essential part of future HTM’s hardware system. Additionally, as a further work, we try to fabricate the crossbar having more than 100 memristors and combine the fabricated crossbar with the CMOS circuit to verify the SP operation by hardware, for testing the MNSIT vectors. Finally, we discuss possible applications of the memristor-CMOS hybrid circuit of HTM’s hardware. As Internet of Things (IoT) sensors become more popular in human life and environment, an amount of data generated from the sensors becomes enormous [ 28 , 29 , 30 ]. To handle this huge amount of data from the physical world, we can think of the integration of IoT sensors and memristor-CMOS hybrid circuit into one chip [ 31 , 32 ]. By doing so, the unstructured data from the sensors can be pre-processed and interpreted near the sensors by the integrated memristor-CMOS hybrid circuit of HTM hardware. If we deliver all the data generated from the IoT sensors to the cloud, without any pre-processing of the unstructured data near the IoT sensors, an amount of computing energy demanded at the cloud may be huge [ 33 ]. Thus, the memristor-CMOS hybrid circuit that can perform the pre-processing of the unstructured data from the IoT sensors can be very useful for energy-efficient computing in future."
} | 4,269 |
29601856 | null | s2 | 987 | {
"abstract": "Cyanobacteria are photosynthetic microorganisms whose metabolism can be modified through genetic engineering for production of a wide variety of molecules directly from CO"
} | 42 |
37571153 | PMC10422474 | pmc | 988 | {
"abstract": "Since the proposal of self-healing materials, numerous researchers have focused on exploring their potential applications in flexible sensors, bionic robots, satellites, etc. However, there have been few studies on the relationship between the morphology of the dynamic crosslink network and the comprehensive properties of self-healing polymer nanocomposites (PNCs). In this study, we designed a series of modified nanoparticles with different sphericity (η) to establish a supramolecular network, which provide the self-healing ability to PNCs. We analyzed the relationship between the morphology of the supramolecular network and the mechanical performance and self-healing behavior. We observed that as η increased, the distribution of the supramolecular network became more uniform in most cases. Examination of the segment dynamics of polymer chains showed that the completeness of the supramolecular network significantly hindered the mobility of polymer matrix chains. The mechanical performance and self-healing behavior of the PNCs showed that the supramolecular network mainly contributed to the mechanical performance, while the self-healing efficiency was dominated by the variation of η. We observed that appropriate grafting density is the proper way to effectively enhance the mechanical and self-healing performance of PNCs. This study provides a unique guideline for designing and fabricating self-healing PNCs with modified Nanoparticles (NPs).",
"conclusion": "4. Conclusions In this study, the effect of the structure of modified-NPs on the performance of PNCs is investigated through CGMDs. A series of PNCs with different sphericities (η) is established. First, the morphology of the supramolecular network and the dispersion of NPs are characterized. It is observed that as the η increases, the dispersion state of the NP is transferred from particle bridging to being sandwiched by polymer chains, and the increment of η enhances the completeness of the supramolecular network. Furthermore, the investigation of the segment’s dynamic shows that the mobility of the polymer chains was highly dependent on η, while the increment of η has a negative correlation with the mobility and relaxation behavior of polymer chains. The variation of the average number of entanglements per chain, < Z >, indicates that the completeness of the supramolecular network reinforces the entanglements of polymer chains significantly. The mechanical performance of the PNCs indicates that the contribution of the mechanical properties is mainly from the supramolecular network, and fully modified-NPs will damage the static mechanical performance of PNCs. The self-healing behavior of PNCs with different η is also investigated. The results show that the η has a positive correlation with the self-healing efficiency, and high temperature can improve efficiency as well. However, based on the study of the stress heatmap of PNCs, it is observed that high η will make the structure of PNCs unstable, which is against the stability of the supramolecular network. Furthermore, it is proven that the inference from the analysis of the mechanical responses of PNCs during triaxial deformation. This study offers valuable insights into the impact of modified NP structure on the self-healing behavior of PNCs. Furthermore, this study proposes a novel perspective on the relationship between the integrity of the dynamic crosslink network and the self-healing performance of PNCs.",
"introduction": "1. Introduction As one of the most significant soft materials, polymer nanocomposites (PNCs) exhibit remarkable comprehensive properties such as static elastic modulus, tensile strength, flexibility, abrasion performance, tear resistance, etc. By introducing nanoparticles and crosslinking networks [ 1 ], these unique performances make PNCs widely employed in numerous applications, especially for tire tread materials [ 2 , 3 , 4 ]. However, traditional PNCs are unfixable when damaged, and when they are no longer useful, it is challenging to recycle the materials into feedstock. Therefore, many scientists are exploring ways to endow PNCs with self-healing abilities to address these challenges [ 5 , 6 , 7 , 8 ]. There are typically two types of self-healing materials: extrinsic healing systems and intrinsic self-healing materials. Intrinsic self-healing materials are usually based on non-covalent chemistries or dynamic covalent chemistries [ 9 ]. For instance, Perera introduced a dynamic covalent adaptive network into hydrogel, designing a novel soft-material with self-healing, shape memory, and stimuli-induced stiffness changes [ 10 ]. Liu, et al., inserted sacrificial bond into covalent network to enhance the dynamics mechanical performance of PNCs. The highlight of this research is that the distribution of sacrificial bonds accelerates the rearrangement of network topology networks, significantly enhancing the overall mechanical properties of PNCs [ 11 ]. However, due to complex influencing factors, such as the fraction of nanoparticles (NPs), interfacial compatibility between matrix and fillers, the mechanism by which nanofillers affect the conformation of covalent networks is not fully understood, and the relationship between the distribution of the conformation of physical crosslinking networks is unclear. To discover the influence of nanofiller on the distribution of dynamic crosslinking network, numerous studies have been conducted [ 12 , 13 , 14 , 15 , 16 , 17 , 18 ]. An, et al., discovered that the moderate fraction of modified boron nitride would not affect the distribution of dynamic crosslinking network, which would significantly improve the tensile strength and the elongation of PNCs [ 19 ]. Wang, et al., found that when the weight fraction of modified silica NPs was approximately equal to 1%, the mechanical performance of hydrogel was maximized [ 20 ]. Park, et al., demonstrated that the introduction of graphene-oxide will stimulate the bond exchange reaction, significantly reinforcing the self-healing efficiency of PNCs [ 21 ]. Recently, the use of modified NPs as link-points to establish dynamic crosslinking networks has garnered attention from scientists [ 22 , 23 , 24 , 25 ]. Barrios, et al., combined sulfur (S) and zinc oxide (ZnO) as vulcanizing agents in a carboxylated nitrile rubber (XNBR) matrix to establish a dual network aimed at improving the material’s abrasion resistance. The mechanical performance of XNBR was significantly improved (up to 26%) [ 26 ]. Computer simulations offer several advantages for investigating the static and dynamic properties of polymer networks over traditional experimental approaches, including better control of network formation and precise knowledge of dynamic network features [ 27 ]. For instance, Wick et al., employed the molecular dynamics simulation (MD) method to create several dual-crosslinking systems and analyze which dynamic crosslinking density can endow the system with the greatest mechanical performance. The results indicated that at higher crosslinking densities, the material exhibited excellent self-healing performance and high tensile strain [ 28 ]. Fu et al., combined MD and experimental methods to investigate the thermal properties and microstructure morphology of A (DGEBA)/methyl tetrahydrophthalic anhydride (MTHPA) and DGEBA/nadic anhydride (NA). The experimental results confirmed the accuracy of the crosslinking model and indicated that the slight change in the curing agent structure significantly affected the synergy rotational energy barrier, cohesive energy density, and free volume fraction, thus affecting the glass transition temperature (T g ) and modulus of the system [ 29 ]. However, few studies have focused on the influence of NPs on the formation of dynamic crosslink networks and the performance of PNCs. Therefore, in this study, the coarse-grained molecular dynamics simulation (CGMD) method is employed to investigate the influence of nanoparticle (NP) structure on the formation of a dual network (chemical/supramolecular crosslink network). Various modified NPs with different structures are designed and introduced into a polymer matrix with a dynamic crosslink network. Taking the previous works as the reference [ 16 , 17 , 30 , 31 , 32 ], this supramolecular dynamic network is further constructed via the strong interaction between the end-groups of the grafted nanoparticles and grafted polymer chains. The influence of the completeness of the supramolecular network on the mechanical properties and self-healing efficiency of the corresponding composite materials is further studied, where the completeness is defined as the supramolecular cross-linked network structure is fully and uniformly represented throughout the system, indicating that there are no areas of the system that lack the cross-linked network or have disproportionately large clusters of the network. This term implies that the network is evenly distributed and integrated throughout the entire system. This research is expected to provide guidance for the design and fabrication of self-healing polymeric materials.",
"discussion": "3. Results and Discussion 3.1. Effect of η on the Structure and Dynamics of PNCs To characterize the morphology of a supramolecular network with different η values, various types of characterization are implemented. Based on the illustration in Figure 2 , we can see that as η increases, numerous modified groups are introduced into the PNCs, causing the supramolecular network to become denser. To quantitatively analyze the formation of the supramolecular network, the number of clusters (N c ) and the average number of single clusters (N sc ) are calculated, as shown in Figure 2 e. Based on the results, it is observed that as η increases, the size of the clusters decreases. However, when η = 1.0, the size of the cluster increases suddenly. The reason is that when η = 1.0, the excessive concentration of modified groups within the material causes some beads to agglomerate. According to the analysis of the distribution of modified groups ( Figure S1 ), some large clusters (N sc > 90) are formed, while other systems just obtain small clusters (N sc < 10). Therefore, in order to establish a uniform supramolecular network, a moderate number of modified groups are necessary. To characterize the dispersion of core-NPs, the radial distribution function (RDF) between core-NPs with varying values of η is calculated, and shown in Figure 3 b. For η = 0 and η = 0.3, a pronounced peak appears at r = 5.14σ, indicating that the NPs are dispersed via segmental-level tight particle bridging. For η = 0.5 and η = 0.8, peaks appear at r = 5.75σ and 7.87σ, indicating that there are two states of NP dispersion: segmental-level tight particle bridging and adsorbed layers coexisting with longer range bridging. For η = 1.0, only one peak appears at r = 7.92σ, indicating that the core-NPs are sandwiched by grafted chains and polymer matrix. Meanwhile, based on the results in Figure 3 b, when 0 < η < 1, the maximum value of g ( r ) is reduced as η increases, signaling an improvement in the dispersion of core-NPs. In general, the characterization of RDF proves that the dispersion of NPs is governed by the configuration of the supramolecular network. The effect of η on T g is investigated and the results are shown in Figure 4 a, which were obtained where we identified the temperature at which the specific volume varied. The curves of specific volume-temperature are shown in Figure S3 . It is found that as η increases, the value of T g increases synchronously and slowly, indicating that the completeness of the supramolecular network affects the mobility of polymer chains. This suggests that the dynamic crosslink network creates a hindrance effect, which hinders the mobility of polymer chains. To quantitatively analyse the relationship between η and the mobility of polymer chains, the mean-square displacement (MSD) is calculated, which is shown in Figure 4 b,c. As η increases, the value of MSD decreases obviously, which proves the inference summarized in Figure 4 a. Meanwhile, the bond autocorrelation function, C b ( t ), as a function of η, is calculated to further measure the dynamics of polymer chains. The equation of C b ( t ) is exhibited as follows: (7) C b ( t ) = ( μ ( t ) × μ ( t 0 ) ) \nwhere μ(t) denotes a unit vector characterizing the orientation at a time. Figure 5 shown after a prolonged relaxation process, the decay of C b ( t ) decreases in tandem with an increase in η, signifying the significant impact of the supramolecular network on the relaxation behavior of polymer chains. Furthermore, it is observed that when η > 0.5, the minimum value of C b ( t ) is higher than 1/e, which further proves the negative correlation between the completeness of the supramolecular network and the relaxation of polymer chains. The entanglement of the polymer chains was characterized using the Z1 code to determine the confinement effect of η on segment dynamics [ 43 , 44 , 45 , 46 ]. The entanglement network analysis for different PNCs is summarized in Table 2 . Based on the results, the increase of η corresponds positively to the mean-squared end-to-end distance R 2 , indicating that the modified NPs absorb the polymer chains on the surface. The value of the average number of entanglements per chain, denoted by < Z >, is also enhanced with an increase in η, signifying that the formation of the supramolecular network significantly affects the entanglement network. Meanwhile, it is observed that although the introduction of modified groups accelerates the entanglement behavior of polymer chains, when η > 0, the value of the mean contour length of the polymer chains ( L p ) decreases and exhibits an irregular tendency with an increase in η. This suggests that the supramolecular network affects the flexibility of the polymer chains, and the effect of η on the flexibility of polymer chains is unpredictable. This unpredictability may affect the process ability of PNCs. 3.2. The Static Mechanical Performance and Self-Healing Behavior of PNCs To investigate the influence of η on the mechanical performance of PNCs, uniaxial deformation tests of PNCs were carried out. As η increases, the tensile stress of PNCs increases synchronously, indicating that the supramolecular network reinforces the mechanical performance of PNCs as shown in Figure 6 a. It is noteworthy that when η = 1.0, the curve breaks when the tensile strain is equal to 5, indicating that excessive dynamic crosslink nodes can affect the mechanical performance of PNCs. Except for this special case, when η = 0.8, the PNC exhibits the best static mechanical property. The chain orientation behavior during uniaxial deformation process is also examined by employing the second-order Legendre polynomials, which is exhibited as follows [ 33 , 47 ]: (8) < P 2 ( cos θ ) > = 1 2 [ 3 ( cos 2 θ ) − 1 ] \nwhere θ denotes the angle between a given element (two adjoining monomers in the chains) and the reference stretching direction. The possible values of < P 2 ( cos θ ) > range from −0.5 to 1, and the values of −0.5, 1 and 0 indicate a perfection orientation perpendicular to the reference direction, a perfection orientation parallel to the reference direction or randomly oriented, respectively. Based on the results in Figure 6 b, the value of < P 2 ( cos θ ) > is negatively correlated with η. Especially when the tensile strain is relatively large (>5), the difference in < P 2 ( cos θ ) > is apparent. This result illustrates that the supramolecular network highly affects the orientation behavior of polymer chains, and the major contribution to the mechanical performance of PNCs belongs to the supramolecular network, not the reinforcement of NPs or polymer matrix chains. To understand the self-healing behavior of PNCs, the stress-strain curve of PNCs with various values of η was studied as a function of self-healing temperature under triaxial deformation. Firstly, the 1st triaxial deformation process was carried out, which is shown in Figure 7 a. Based on the results, it was observed that when the tensile strain is relatively small (<1), the tensile stress is positively correlated with the increment of η. This indicates that the completeness of the supramolecular network enhances the mechanical performance of PNCs, which is consistent with the conclusion obtained from Figure 6 . However, as the triaxial deformation is conducted, the system with a high η exhibits low tensile stress. This means that when the crosslink network is broken, the number of dynamic crosslink nodes is not necessary for the mechanical performance of PNCs. Then the influence of temperature on the self-healing efficiency, ψ sh , of PNCs is studied. ψ sh is defined as the σ max,z (the maximum tensile stress along z direction with different self-healing temperature in 2nd triaxial deformation) over σ max,o (the maximum tensile stress at 1st triaxial deformation). The PNCs were deformed by the triaxial deformation and compressed afterward. Then, they were kept in a range of fixed temperatures for a period of time. Based on the results in Figure 4 a, all systems obtain T g lower than 1.0, so the self-healing temperatures are set as 1.0, 1.5, 2.0, 2.5 to study the self-healing behavior of PNC, which ensure the PNCs are in rubbery state. Based on the results in Figure 8 , it is observed that the value of ψsh is positively correlated with the temperature, signaling that the increase in temperature enhances the self-healing process of PNCs. It is observed that when η = 0.3, the value of ψsh is decreased. This phenomenon indicates that low content of modified-end groups cannot enhance the re-establishment of supramolecular crosslink network, but hinder the re-build process of crosslink network. Therefore, it is necessary to ensure the moderate modified-end groups to accelerate the self-healing behavior of PNCs. Meanwhile, it can be seen that for η ≥ 0.5, when T = 2.0 and 2.5, the difference in ψ sh is not distinct, indicating that the self-healing efficiency has a limitation value dependent on the self-healing temperature when the completeness of supramolecular crosslink network is relatively high. Meanwhile, as the increment of η, the efficiency is generally enhanced, indicating that the improvement of the supramolecular network benefits the self-healing efficiency. Nonetheless, when η > 0.8, the value of ψ sh stops increasing, and even decreases slightly, indicating that the excessive modified groups are a shortcoming of the self-healing behavior. This inference proves that the supramolecular network with moderate completeness is the most suitable selection for designing self-healing PNCs. To investigate the morphology of PNCs during triaxial deformation, we examine the contribution of each component to the total stress, σ T, during two triaxial deformations. We evaluate the value of σ T using color, where redder denotes higher σ T and bluer denotes lower σ T . To make the argument more representative, we consider systems with η values of 0, 0.5, and 1.0, which represent low, medium, and high completeness of the supramolecular network, respectively. As shown in Figure 9 , we observe that as η increases, the stress distribution becomes narrower during either the 1st or 2nd triaxial deformation, which confirms the analysis based on Figure 7 a. Furthermore, when the completeness of the supramolecular network is low or medium, the structure of the PNC does not fully collapse during the 2nd triaxial deformation. Meanwhile, it is observed that when η = 1.0, the stress distribution becomes narrower than η = 0.5, and still contain the crosslink structure stable under high self-healing temperature, while the structure is totally collapse under low self-healing temperature, signaling that for the relatively high grafted density, the high temperature is a necessary condition to achieve remarkable self-healing efficiency."
} | 5,014 |
31646156 | PMC6804398 | pmc | 990 | {
"abstract": "This paper presents the raw data of biogas production and composition (relative pressures and concentrations of each of the biogas constituents) for batch experiments to evaluate the anaerobic digestion of xylose. Also, metagenomic sequencing data and analysis were reported. All data is available at Mendeley Data. 16S DNA sequencing data and metadata is available at MG-RAST (metagenomics.anl.gov/linkin.cgi?project = 9961). For further discussion, please refer to the scientific article entitled \"Effect of acidic and thermal pretreatments on a microbial inoculum for hydrogen and volatile fatty acids production through xylose anaerobic acidogenic metabolism\" (Mockaitis et al., 2020)."
} | 172 |
27858972 | null | s2 | 993 | {
"abstract": "Anaerobic microorganisms play a central role in several environmental processes and regulate global biogeochemical cycling of nutrients and minerals. Many anaerobic microorganisms are important for the production of bioenergy and biofuels. However, the major hurdle in studying anaerobic microorganisms in the laboratory is the requirement for sophisticated and expensive gassing stations and glove boxes to create and maintain the anaerobic environment. This appendix presents a simple design for a gassing station that can be used readily by an inexperienced investigator for cultivation of anaerobic microorganisms. In addition, this appendix also details the low-cost assembly of bioelectrochemical systems and outlines a simplified procedure for cultivating and analyzing bacterial cell cultures and biofilms that produce electric current, using Geobacter sulfurreducens as a model organism. © 2016 by John Wiley & Sons, Inc."
} | 232 |
23840768 | PMC3694090 | pmc | 994 | {
"abstract": "Mass mortality events of benthic invertebrates in the Mediterranean Sea are becoming an increasing concern with catastrophic effects on the coastal marine environment. Sea surface temperature anomalies leading to physiological stress, starvation and microbial infections were identified as major factors triggering animal mortality. However the highest occurrence of mortality episodes in particular geographic areas and occasionally in low temperature deep environments suggest that other factors play a role as well. We conducted a comparative analysis of bacterial communities associated with the purple gorgonian Paramuricea clavata , one of the most affected species, collected at different geographic locations and depth, showing contrasting levels of anthropogenic disturbance and health status. Using massive parallel 16SrDNA gene pyrosequencing we showed that the bacterial community associated with healthy P. clavata in pristine locations was dominated by a single genus Endozoicomonas within the order Oceanospirillales which represented ∼90% of the overall bacterial community. P. clavata samples collected in human impacted areas and during disease events had higher bacterial diversity and abundance of disease-related bacteria, such as vibrios, than samples collected in pristine locations whilst showed a reduced dominance of Endozoicomonas spp. In contrast, bacterial symbionts exhibited remarkable stability in P. clavata collected both at euphotic and mesophotic depths in pristine locations suggesting that fluctuations in environmental parameters such as temperature have limited effect in structuring the bacterial holobiont. Interestingly the coral pathogen Vibrio coralliilyticus was not found on diseased corals collected during a deep mortality episode suggesting that neither temperature anomalies nor recognized microbial pathogens are solely sufficient to explain for the events. Overall our data suggest that anthropogenic influence may play a significant role in determining the coral health status by affecting the composition of the associated microbial community. Environmental stressful events and microbial infections may thus be superimposed to compromise immunity and trigger mortality outbreaks.",
"conclusion": "Conclusions Overall our data suggest that influence by direct human activities, such as the mechanical injuries considered in the present study, may play a significant role in determining the coral health status by leading to a more unstable microbial community (e.g. altering the composition and diversity of the associated bacterial community). Changes in the structure of coral associated microbial communities such as the ones observed in healthy coral populations affected by direct anthropogenic influence (Po30IH) and similarly, in diseased corals (Pa63ID, Ta30ID) under comparable human pressure, point to a link among alteration in bacterial holobiont members, direct human influence and susceptibility of corals to microbial pathogens and associated diseases. According to this scenario, a decrease in the contribution of dominant Endozoicomonas spp. in P. clavata from populations subjected to human disturbance might possibly explain for the colonization of disease-related bacteria such as vibrios, and/or other not yet recognized pathogens, by the opening of new available niches. Environmental and climate linked stressful events such as the occurrence of temperature anomalies and starvation periods as well as the presence of highly virulent microbial strains may thus be superimposed to compromise immunity and trigger mortality outbreaks.",
"introduction": "Introduction Mass mortality events of benthic invertebrates from different phyla (sponges, cnidarians, molluscs, ascidians, bryozoans) occurred in recent years in the temperate Mediterranean Sea with catastrophic effects on the coastal marine ecosystem [1] , [2] . Scientific evidence gathered so far indicated that sea surface temperature anomalies linked to global warming are among the primary triggering factors explaining such events [2] , [3] . In particular, long and hot summer periods recorded during the recent years were observed to place a prolonged energetic constraints on benthic suspension feeders, which as a result, go into dormancy or decreased activity, leading to mortality in late summer and early fall [4] , [5] . In addition nonresident thermodependant bacterial pathogens may take advantage both from high temperature and compromised host conditions contributing significantly to coral disease and mortality [6] , [7] . For instance, we recently reported that Vibrio -TAV24 strain, belonging to the species Vibrio coralliilyticus, is involved in mass mortality events of the purple gorgonian P. clavata in the NW Mediterranean Sea [7] . In the above scenario, data coming from scleractinian corals suggest that alterations in bacterial holobiont members due to environmental stressors can disrupt coral’s immune system and enhance coral susceptibility to bacterial infections and diseases [8] . In fact, symbiotic bacteria are considered an important component of the “coral’s immune system” by preventing the colonization of nonresident, possibly pathogenic, bacteria [9] . Studies conducted on tropical corals have shown that, under stressful conditions, resident microbes critical to the healthy functioning of the coral organism are outcompeted by pathogenic microbes and usually in the context of environmental disruptions such as ‘heat waves’ during the warm summer months [10] , [11] . However, despite acquired knowledge, we are still far from having a complete understanding of the causes triggering mass mortality events of benthic invertebrates in the Mediterranean Sea, especially concerning environmental stressors, other than temperature, which could be possibly involved. For example, occurrence of the mortality episodes is more frequent in some geographic areas of the Mediterranean Sea (e.g. Ligurian Sea and Adriatic Sea) compared to others, regardless of the seasonal temperature regime [12] , [13] . In addition, mortality events were occasionally observed in the deep well below the thermocline layer [14] . The mesophotic zone comprised between 60 and 100–120 m depth is less subjected to climate driven changes than the upper ocean, and seasonal temperature follows lower fluctuations [15] . Clearly, as it appears from these considerations, additional factors other than seasurface temperature anomalies are likely to be involved in the occurrence of mortality events and still need to be addressed. Impact by human activities, such as fisheries and recreational activities (e.g. diving), ship traffic and water pollution in coastal areas can lead local benthic populations to a general state of physiological stress. In particular, recent studies have shown that direct anthropogenic impact such as diving damage on P. clavata corals dramatically increase the natural mortality rate [16] as well as the time to quasi-extinction for populations subjected to realistic frequencies of mass mortality events [17] . Nevertheless, to date, little information is available on this subject, also possibly due to the fact that experimentally assessing the role of human factors in contributing to disease outbreaks is not trivial to pursue. For example, the use of laboratory model system (e.g. manipulative experiments based on coral growth in aquaria), largely employed in reductionist ecological studies, have a limited application to this specific purpose because unable to reproduce the complexity of the natural ecosystem and the global synergic influence of natural and anthropogenic factors in affecting coral health status in the field. To overcome these concerns and explore additional factors possibly involved in the occurrence of mass mortality episodes in the Mediterranean Sea, we conducted a infield comparative study of bacterial communities associated with the purple gorgonian P. clavata , one of the most affected species, collected at different geographic locations and depth, showing contrasting levels of anthropogenic disturbance and health status. Our rationale considers the fact that alteration of the coral holobiont may be investigated as a proxy of coral health status and sensitivity to nonresident pathogens. The objective of this study was to assess whether anthropogenic influence affect the composition of the bacterial holobiont and if this could be linked to disease outbreaks.",
"discussion": "Results and Discussion Overall Bacterial Diversity and Community Structure Associated with the Purple Gorgonian P. clavata in the Mediterranean Sea We sequenced 133.853 PCR amplicons spanning the V6 hypervariable region of 16SrDNA gene from P. clavata genomic DNA extracted from different coral populations living in different geographic areas, at different depth and showing different health status condition ( \n Table 1 \n ). The number of reads per sample ranged from 10.642 to 15.240 sequences. To minimise the bias associated to random sequencing errors, a stringent trimming procedure was conducted, by eliminating reads that contained one or more ambiguous bases, had errors in the barcode or primer sequence, were atypically short (<70 bp), and had an average quality score <30 [20] ; on average this step reduced the size of the data set by 15%. Species richness of the bacterial communities associated with P. clavata was highest in coral samples collected from populations subjected to anthropogenic disturbance and disease events such as those of Po30IH, Pa63ID, and Ta30ID samples, whilst declined to less than half in corals collected from pristine populations (samples Pa30PH, Ta30PH, Po50PH, Pa90PH) ( \n Fig. 3 \n ). The bacterial diversity associated with P. clavata estimated by the Shannon diversity index from the OTU data was also high for human impacted samples varying from 2.2 to 5.0 with an average of 3.2 (SD = 1.6). In contrast, diversity was lower in coral samples collected in pristine areas that showed comparable values of the Shannon index ranging from 0.8 to 1.3 with an average of 1.1 (SD = 0.2) ( \n \n Fig. 3 \n \n ) . In these latter rarefaction curves, analysis revealed that the diversity had almost reached a stable value ( \n \n Fig. 3 \n \n ) . In contrast, in coral samples from human impacted populations, rarefaction curves showed a diverse pattern, which failed to reach a plateau. This was evident in Po30IH, Pa63ID, and Ta30ID samples suggesting that a large number of unseen OTUs still existed in the original samples and more sequencing effort should be required to detect additional phylotypes. According to this, the number of OTUs in these samples estimated by ACE, Chao1 and jackknife richness estimator was considerably higher than the number of observed OTUs (sobs) representing less than 50% of the estimated richness ( \n Fig. 3 \n ). This means, for example, that for OTU numbers generated by the Chao1 estimator there were between 546 and 1550 additional phylotypes with a final richness of between 1069 and 2501 in each sample. Considering the current sequencing effort, discordance between estimated and observed richness in these samples may be due to the effects of rare species as it is well know that microbial communities naturally contain a large number of rare species and a small number of abundant ones [20] . 10.1371/journal.pone.0067745.g003 Figure 3 Bacterial diversity associated with P. \n clavata . Alpha diversity metrics derived from 16S rDNA pyrosequencing of bacteria associated with P. clavata samples: (a) Rarefaction curves; (b) number of OTUs predicted (Chao1, Ace, jackknife) and observed (sobs); (c) Shannon-Weiner diversity of bacteria from each coral sample. OTU’s were grouped at >97% similarity based on mothur classified results. To investigate most abundant members of the coral holobiont phylogenetic identity of generated sequences from coral samples were evaluated using the Metagenome Analysis Software (MEGAN). Each trimmed read sequence was BLASTed against a reference database of nearly 400.000 rRNA genes for the bacterial domain as described in the method session. The results of BLASTN were then used to estimate the taxonomic content of the data set, using NCBI taxonomy with MEGAN [26] . Only the most abundant reads (those occurring at least 20 times in the trimmed data set) were assigned to bacterial taxa and included in the results. Using this threshold and quality filters described in the method session the lowest common ancestor algorithm assigned 85 231 reads to the domain Bacteria, while 26 remained unassigned, either because the bit-score of their matches or the minimum number of reads for taxon assignment fell below the threshold. Bacterial community were dominated by the class of Gammaproteobacteria that accounted on average for >95% of the overall community structure in healthy corals from pristine populations. Among this class the most dominant bacterial genus in the microbiome of P. clavata was Endozoicomona s within the order Oceanospirillales which represented ∼90% of the overall bacterial community. This genus is frequently found in association with marine invertebrates and the type strain Endozoicomonas elysicola was firstly isolated from the sea slug Elysia ornata off the coast of Izu-Miyake Island, Japan [27] . The species E. montiporae was later reported from the encrusting pore coral Montipora aequituberculata in Taiwan [28] . Clade of bacteria in the Oceanospirillaceae is also widely distributed in Porites spp. and other hermatypic corals [29] . Interestingly, it has been recently shown that Endozoicomonas represents the dominant bacterial genus associated to the Mediterranean gorgonian coral Eunicella cavolinii , a species closely related to P. clavata \n [30] . Other bacterial classes consistently found in the microbiome of P. clavata were alphaproteobacteria (∼5%), betaproteobacteria (∼0.2%), and actinobacteria (∼0.3%). The genus Vibrio which includes species pathogenic for corals accounted on average for less than 1% of the healthy coral bacterial community. Comparative Analysis of the Bacterial Communities Associated with P. clavata Corals Collected from Different Geographic Areas and Showing Contrasting Levels of Anthropogenic Disturbance and Health Status We used a multiple-comparative analysis in MEGAN to compare the bacterial communities associated with P. clavata corals collected as previously described ( \n \n Table 1 \n \n ) . To minimize potential bias arising from differences in absolute read counts taxon data were normalised over all reads, such that each data set had 100 000 reads [31] . As detected by OTU diversity analyses we observed that shifts in bacterial assemblages associated with P. clavata occurred mainly in relation to the level of human impact and health status of the coral samples. Accordingly Po30IH, Pa63ID and Ta30ID samples were clearly separated from other samples by CLUSTER and MDS analyses ( \n Fig. 4 \n ). Interestingly, a relative decrease in the dominant bacterial symbionts Endozoicomonas spp. was observed in these samples. In particular, contribution by Endozoicomonas genus to the overall bacterial community decreased from more than 90% dominance in P. clavata samples from pristine populations (Pa30PH, Ta30PH, Po50PH, Pa90PH) to less than 70% dominance in Po30IH and Pa63ID to 5% in Ta30ID. In contrast, bacterial symbionts exhibited remarkable stability in P. clavata collected at euphotic and mesophotic depth in pristine locations (e.g. Pa30PH and Pa90PH samples, dominance >90%) suggesting that fluctuations in environmental parameters such as irradiance and temperature (typically exhibiting strong seasonal variations in the surface euphotic layer and almost constant low values in the deep mesophotic layer) have limited effect in structuring the bacterial holobiont. 10.1371/journal.pone.0067745.g004 Figure 4 Comparative analysis of the bacterial communities associated with P. \n clavata . 16S rDNA pyrosequencing-based comparative analysis of dominant bacterial groups associated with P. clavata samples collected in different geographic areas, at different depth and showing different health status condition. (a) a comparative map is shown, where the numbers of normalised reads taken by each taxon (the tree is collapsed to the ‘genus’ level) in each year are represented as colour bar. The cumulative number of normalised reads across the different coral samples is also shown for each taxon [29] . Genus shared across all samples are in bold (b) agglomerative hierarchical clustering (CLUSTER analysis) and (c) non-Metric multi-Dimensional Scaling (nMDS) of the different sample datasets. The decrease in dominance by Endozoicomonas members of the bacterial holobiont in diseased P. clavata samples (Pa63ID and Ta30ID) and in samples collected in shallow environments at Portofino Promontory (Po30IH), one of the most affected area in the Mediterranean Sea by recurring mortality events [32] suggests that this dominant genus may contribute to the health status of the coral. Accordingly, coral symbionts are thought to provide benefit to their hosts by supplying food compounds such as organic carbon, nitrogen or secondary metabolites [33] . In particular, it has been recently shown that member of the order of Oceanospirillales can utilize constituents of coral tissues and mucus such as organic acids, amino acids, ammonium, and dimethylsulfoniopropionate (DMSP) [34] . High concentrations of DMSP and DMS have been found within animals that harbor symbiotic algae, such as scleractinian corals, providing a potential link between DMSP-degrading Oceanospirillales and corals [35] , [36] . Nevertheless, the functional and ecological role of the dominant Endozoicomonas genus in the azooxanthellate purple gorgonian P. clavata is still largely unknown. Microbial symbionts may also protect corals from disease, for instance by producing antimicrobial chemical defenses targeted at pathogens or other potentially deleterious microorganisms and by preventing unwanted microbial colonization through the occupation of otherwise available niches [37] , [38] , [39] . Accordingly, the study of the Vibrio community in our samples clearly showed that VAI index was significantly higher in diseased than in healthy corals concomitant with the decrease of Endozoicomonas spp. (ANOVA, p<0.05, Table S1 ). In addition, bacterial taxa such as Bacteroidetes , Bdellovibrionales , Campylobacterales as well as unclassified environmental bacteria were only found in diseased animals although their possible role in coral pathogenesis is unknown ( \n \n Fig. 4 \n \n ) . Vibrios are a large group of indigenous marine bacteria with over 80 species described, also including several species capable of causing infection in humans and animals [40] . In a previous study we recently conducted to assess the potential role of microbial pathogens in contributing to mass mortality events of P. clavata in the NW Mediterranean Sea, we observed that Vibrio species are generally pathogenic toward P. clavata corals being capable to produce tissue damage during experimental infection conducted in aquaria [7] . In particular, we described a TAV24 strain later identified as V. coralliilyticus which satisfied Koch’s postulates and was linked thus to coral disease observed during mortality episodes [7] . The TAV 24 strain is a non resident pathogen as in a larger survey conducted from 2008 to 2011 in three different areas of the NW Mediterranean Sea affected by mortality episodes (Portofino Promontory, Tavolara Island and Capo Mortola Marine Reserve) it was consistently found in association to disease corals but never found in healthy corals (data not shown). Interestingly although V. coralliilyticus was detected on diseased corals collected at Tavolara Island in 2008 it was not found on diseased corals collected during a deep mortality episodes occurred at Pantelleria Island in 2011. To our knowledge, this is the first study investigating a deep mortality episode of P. clavata occurred in the mesophotic layer of the Mediterranean Sea ( \n \n Fig. 2 \n \n ) . Temperature recorded at this depth is almost constant all year round and rarely exceed 18°C which can be taken as a threshold above which physiological stress and microbial proliferation could occur in P. clavata \n [7] \n [41] . This suggests that neither temperature anomalies nor recognized microbial pathogens such as V. coralliilyticus TAV24 strain which are currently considered among the main casual factors triggering mortality outbreak in P. clavata populations are solely sufficient to fully explain the occurrence of these events. As a speculation the increased dominance of Vibrio spp. (non- V. coralliilyticus species ) observed in Pa63ID samples compared to healthy coral samples may have contributed to disease as well as the presence of yet unrecognized pathogens. It is, nevertheless, worth mentioning that infection by TAV24 strain was associated to a more severe disease event at Tavolara (almost 100% of affected colonies) than at Pantelleria (less than 10% of affected colonies) where the strain was not found. Accordingly the alteration of the bacterial holobiont members (e.g. increased specie richness and diversity, \n Fig. 3 \n ) and dominance of Vibrio spp. ( \n Fig. 5 \n ) was greater in Ta30ID than Pa63ID. Vibrio species may thus take advantage of the shift in bacterial holobiont members and, when conditions are favorable (e.g. high sea surface temperature) possibly trigger the occurrence of diseases. The same conditions may favors coral infection by highly virulent strains, such as TAV24 strain, that may be responsible for larger disease outbreaks such as the one occurred at Tavolara island in October 2008. 10.1371/journal.pone.0067745.g005 Figure 5 Analysis of Vibrio populations. \n Vibrio relative abundance index (VAI) calculated on P. clavata samples collected in different geographic areas, at different depth and showing different health status condition. Z values are obtained by subtracting the population mean and dividing the difference by the s.d. * ANOVA p<0.05;+presence of the species V. coralliilyticus . Conclusions Overall our data suggest that influence by direct human activities, such as the mechanical injuries considered in the present study, may play a significant role in determining the coral health status by leading to a more unstable microbial community (e.g. altering the composition and diversity of the associated bacterial community). Changes in the structure of coral associated microbial communities such as the ones observed in healthy coral populations affected by direct anthropogenic influence (Po30IH) and similarly, in diseased corals (Pa63ID, Ta30ID) under comparable human pressure, point to a link among alteration in bacterial holobiont members, direct human influence and susceptibility of corals to microbial pathogens and associated diseases. According to this scenario, a decrease in the contribution of dominant Endozoicomonas spp. in P. clavata from populations subjected to human disturbance might possibly explain for the colonization of disease-related bacteria such as vibrios, and/or other not yet recognized pathogens, by the opening of new available niches. Environmental and climate linked stressful events such as the occurrence of temperature anomalies and starvation periods as well as the presence of highly virulent microbial strains may thus be superimposed to compromise immunity and trigger mortality outbreaks."
} | 5,927 |
33102450 | PMC7545826 | pmc | 996 | {
"abstract": "Upgrading of furanic aldehydes to their corresponding furancarboxylic acids has received considerable interest recently. Herein we reported selective oxidation of furfural (FAL) to furoic acid (FA) with quantitative yield using whole-cells of Pseudomonas putida KT2440. The biocatalytic capacity could be substantially promoted through adding 5-hydroxymethylfurfural into media at the middle exponential growth phase. The reaction pH and cell dosage had notable impacts on both FA titer and selectivity. Based on the validation of key factors for FAL conversion, the capacity of P. putida KT2440 to produce FAL was substantially improved. In batch bioconversion, 170 mM FA was produced with selectivity nearly 100% in 2 h, whereas 204 mM FA was produced with selectivity above 97% in 3 h in fed-batch bioconversion. Particularly, the role of molybdate transporter in oxidation of FAL and 5-hydroxymethylfurfural was demonstrated for the first time. The furancarboxylic acids synthesis was repressed markedly by destroying molybdate transporter, which implied Mo-dependent enzyme/molybdoenzyme played pivotal role in such oxidation reactions. This research further highlights the potential of P. putida KT2440 as next generation industrial workhorse and provides a novel understanding of molybdoenzyme in oxidation of furanic aldehydes.",
"conclusion": "Conclusions In summary, we have established an improved approach for selective synthesis of FA from FAL by P. putida KT2440. High FA titer was achieved in both batch conversion and fed-batch conversion, which not only addressed the challenge of FAL inhibition and toxicity, but also found an economically competitive process for FA production in large-scale. It is of interest to uncover the pivotal role of molybdate transporter in furanic aldehyde oxidation. The molybdoenzyme involved in the selective oxidation of formyl on furan ring can be expected to be developed as promising biocatalyst for the production of a series of furan-based carboxylic acids.",
"introduction": "Introduction Lignocellulosic biomass is the most abundant and sustainable resource from which many kinds of platform chemicals can be derived (Den et al., 2018 ). 5-hydroxymethylfurfural (HMF) is accessible from glucose and cellulose while furfural (FAL) derives from xylose and C 5 -rich hemicellulose in the presence of acid catalysts (Mika et al., 2018 ). Both FAL and HMF are versatile platform molecules that can be converted into a variety of important chemicals, due to the presence of active groups such as primary hydroxyl and formyl (Mariscal et al., 2016 ; Hu et al., 2018 ). Selective oxidation of FAL and HMF to a variety of highly functionalized carboxylic acids and hydroxyl acids was an important aspect in their upgrading (Mariscal et al., 2016 ; Hu et al., 2018 ). FAL can be oxidized to furoic acid (FA) (Mariscal et al., 2016 ; Shi et al., 2019 ; Wang et al., 2020 ), while HMF to 5-hydroxymethylfuroic acid (HMFCA), 2,5-diformylfuran, 5-formylfuroic acid, and 2,5-furandicarboxylic acid (FDCA) (Qin et al., 2015 ; Hu et al., 2018 ). Up to now, chemocatalytic approaches for selective oxidation of FAL and HMF have been extensively studied, which often involve the use of toxic and hazardous reagents and solvents, the harsh conditions, as well as the low selectivity (Mariscal et al., 2016 ; Sajid et al., 2018 ). In the case of FA, it is currently produced from FAL industrially via a Cannizzaro reaction with NaOH, accompanied by the byproduct furfuryl alcohol (FOL) formation, resulting in a very low selectivity (Mariscal et al., 2016 ). An attractive alternative exists in the biocatalytic transformation that performs reactions with much higher selectivity under ambient conditions. Efforts with regards to biocatalysis have been made at achieving selective oxidation of FAL to FA. Based on catalytic promiscuity of enzymes, recombinant Escherichia coli strains harboring various heterogeneous enzymes were reported for FA production. These enzymes included 3-succinoylsemialdehyde-pyridine dehydrogenase (SAPDH) (Shi et al., 2019 ), horse liver alcohol dehydrogenase (Peng et al., 2019 ), and HMF oxidase (HMFO) variant (Wang et al., 2020 ). Likewise, microbes resistant to FAL were also isolated and implemented for FA production. Nocardia corallina B-276 oxidized FAL of 9 g/L to FA with yield of 92% using resting cells (Pérez et al., 2009 ). Gluconobacter oxydans ATCC 621H converted FAL into FA with concentration exceeding 40 g/L and yield close to 100% using a compressed oxygen supply-sealed and stirred tank reactor (Zhou et al., 2017 ). In view of their good performances, it could be envisioned that these bacteria harbored relevant enzymes with high activities toward FAL and other furanic aldehydes. However, researches focusing on identification of such enzymes were limited except for few reports (Koopman et al., 2010 ; Wang X. et al., 2015 ). The trace element molybdenum is essential in the prokaryotic and eukaryotic organisms, and it participates in the biosynthesis of molybdenum cofactor that forms the catalytic center of a vast variety of molybdoenzymes (Schwarz et al., 2009 ). In nature, molybdenum is bioavailable in the form of molybdate ( MoO 4 2 - ), which is transported by an ABC-type transporter comprising three proteins, ModA (molybdate binding protein), ModB (membrane protein) and ModC (the ATPase) (Self et al., 2001 ). Once entering into the cell, molybdate is subsequently incorporated into the complex pathway of molybdenum cofactor biosynthesis and finally to activate molybdoenzymes (Hille et al., 2014 ; Mendel and Leimkühler, 2015 ; Leimkühler and Iobbi-Nivol, 2016 ). Researches had showed that ModABC played a critical role in the detoxification of aromatic compounds, such as nicotine and quinoline (Blaschke et al., 1991 ; Ganas et al., 2008 ; Xia et al., 2018 ). Moreover, molybdate uptake was also involved in aerobic degradation of FA, as a 2-furoyl-CoA dehydrogenase which converted 2-furoyl-CoA into 5-hydroxy-2-furoyl-CoA was confirmed as molybdoenzyme (Koenig and Andreesen, 1989 , 1990 ; Koopman et al., 2010 ). Despite the importance of molybdate transporter in metabolism of aromatic compounds, there has been no work focusing on its functionality in biotransformation of furanic aldehydes to furancarboxylic acids. As well-known, bacteria from the genus Pseudomonas exhibit an intrinsic resistance to a variety of toxic compounds. Among them, the Pseudomonas putida KT2440 has a profound impact on biotechnology through its use in the degradation of aromatic compounds and production of value-added products (Jiménez et al., 2002 ; Belda et al., 2016 ). In this paper, the application potential of P. putida KT2440 was developed as outstanding FA producer. The catalytic performance of this strain was substantially enhanced upon HMF adaption and process optimization. Moreover, it was for the first time to uncover that a molybdate transporter (PP_3828–PP_3830) played an indispensable role in oxidation of formyl groups of FAL and other furanic aldehydes, which would guide us to discover unknown molybdoenzyme as novel biocatalyst for oxidative upgrading of bio-based furanic aldehydes.",
"discussion": "Results and Discussion Addition of FAL and FAL Analogs to Boost the Catalytic Performance of Whole-Cell Biocatalyst Efficient whole-cell biocatalysts are critical for economically feasible biocatalysis process. Addition of substrate or substrate analogs at very low concentration to the medium had been proved as an effective approach to enhance the bioconversion performance of biocatalyst (Wen et al., 2020 ). The additives might induce the expression of certain enzyme(s) responsible for substrate transformation. Therefore, to boost the activity of FAL oxidation in P. putida KT2440, FOL, FAL, and HMF were added during cultivation and the bioconversion abilities of resultant whole-cells were determined. The results showed all three kinds of furan additives exerted positive influence on bioconversion, since all tested parameters obtained from adapted cells were superior to that from the control ( Figure 1A ). Therein, HMF exhibited the best positive effect on conversion, yield, and V 0 , followed by FAL and then FOL. The conversion was significantly improved from 27.4 to 49.2%, the yield from 13.9 to 38.7%, and the V 0 from 39.9 to 112.6 mM/h. It should be pointed out that all aforementioned data were calculated on the basis of samples withdrawn at 10 min, and given enough time, FAL would be completely oxidized to FA in all cases. Besides, it was worth noting that the cell density reduced by ~10% during cultivation with HMF and FOL addition, while decreased by more than 20% in the presence of FAL, which might be due to the most serious toxicity of FAL to cells. Zhang et al. ( 2019 ) reported that the FAL-adapted yeast cells showed no improvements in the synthesis of FOL compared to cells free of substrate adaptation. All these results indicated that FAL and FOL might not be an appropriate option as additive into media. Figure 1 Effect of induction on catalytic performances of P. putida KT2440. Reaction conditions: phosphate buffer (200 mM, pH 7.0), 50 mM FAL, 7 g/L cells (dry weight), CaCO 3 at half molar concentration of FAL, 30°C, 200 rpm, 10 min. (A) Various inducers were added with 3 mM at the early exponential growth phase; (B) 3 mM HMF was added at different growth phases as inducer; (C) HMF was added with designated concentrations at the middle exponential growth phase. The influence of HMF addition time was shown in Figure 1B , and it was found that addition at early exponential growth phase and middle exponential growth phase was vastly superior to addition at the end exponential growth phase. However, when HMF was added at early exponential growth phase, the final cell density was very low. Therefore, addition at the middle exponential growth phase was favorable to the biotransformation process. Given these facts, HMF was added at middle exponential growth phase in the range of 1.5–8 mM to promote whole-cell catalytic synthesis of FA. Figure 1C showed that 4.5 mM was the most suitable concentration accompanied by the best catalytic performance of cells, while HMF with both 1.5–3 mM and 6–8 mM gave second stimulation ( Figure 1B ). Low concentration of HMF with 1.5–3 mM might do not guarantee abundant enzyme activities, whereas high concentration of HMF resulted in inevitable cell damage because of the formation of reactive oxygen species (ROS), causing ROS-associated damage to proteins, nucleic acids, and cell organelles (Wierckx et al., 2011 ; Zhu et al., 2019 ). For P. putida KT2440, such damage might occur at 6–8 mM HMF, which, in turn, reduced catalytic activity of cells. Nevertheless, compared with control, the cells adapted to HMF with any tested concentration were all conducive to FAL oxidation, since all results in Figure 1C were better than that of control in Figure 1A . Optimization of Biotransformation Conditions for FA Production A major advantage of whole-cells is that cells provide a natural environment for the enzymes, preventing loss of activity in complex reaction systems. However, contrarily to pure enzymes, side-reactions are inevitable for whole-cell biotransformation, which always is the most important issue to be considered (de Carvalho, 2011 ). In terms of the whole-cell bioconversion of FAL, oxidizing to FA and reducing to FOL usually occurred together in cells, and the concomitant FOL would greatly reduce the selectivity of FA (Wang et al., 2020 ). Figure 2A showed the effects of reaction pH on FA and FOL production when pH values varied from 5.5 to 7.5. The trend of FA concentration was inverted U-shape curve, whereas the trend of FOL kept rising, likely due to the different optimal pHs of enzymes responsible for oxidation and reduction of FAL. With the increased pH, the maximum titer of FA was raised from 32.6 mM (pH 5.5) to 37.0 mM (pH 6.0) and then back to 25.3 mM (pH 7.5). In terms of selectivity, it dropped from 96.5 to 84.5% with increased pHs. We also found that all FOL would be further oxidized into FA with enough time, assuming the redox reaction between FOL and FAL was reversible, which was consistent with previous reports on Cupriavidus basilensis HMF14 and Amorphotheca resinae ZN1 (Koopman et al., 2010 ; Wang X. et al., 2015 ). However, the accumulated FOL throughout the bioconversion process was much lower for P. putida KT2440 than that for C. basilensis HMF14 and A. resinae ZN1. For the latter two strains, the intermediate FOL could reach half of the starting FAL (Koopman et al., 2010 ; Wang X. et al., 2015 ). Based on these, it can be predicted that P. putida KT2440 harbors weak enzymatic activity for FAL reduction, which make it an outstanding workhorse for FAL oxidation. Figure 2 Effect of pH (A) and cell concentration (B) on biocatalytic oxidation of FAL. Reaction conditions: phosphate buffer (200 mM) with designated pH for (A) and pH 6.0 for (B) , 70 mM FAL, 7 g/L cells (dry weight) for (A) and designated cell concentrations (dry weight) for (B) . Other parameters were same as Figure 1 . Effects of cell dosage on FAL bioconversion were investigated at pH 6.0. Figure 2B showed that, similar to Figure 2A , the concentration of FA rose first and then fell with the increasing of cell dosage. The highest concentration of 42.2 mM was obtained at cell loading of 10.5 g/L. In contrast, the concentration of FOL kept rising. It was plausible that oxygen became a limiting factor as a result of the enhanced cell biomass and the oxidation reaction toward FAL was, to a large extent, necessary for oxygen although the enzyme(s) involved in this process was uncovered. Unlike the oxidation of FAL, its reduction was often efficiently occurred in facultative anaerobic strains, such as Bacillus coagulans and Saccharomyces cerevisiae (Yan et al., 2018 , 2019 ), indicating the non-critical role of oxygen. On the whole, the impact of cell dosage on the selectivity was slightly lower than that of pH. The following experiments for FAL bioconversion were conducted with cell dosage of 10.5 g/L. Bioconversion of FAL With High Loading Under Optimal Conditions Many studies have focused on the inhibition of furans toward microorganisms (Zaldivar et al., 1999 ; Franden et al., 2013 ). In these works, inhibitory activity of furans is to a large extent ascribed to their hydrophobicities. Among them, FAL is generally recognized as the most harmful substance, which might be explained by its strong hydrophobicity (log P FAL = 0.41, log P HMF = −0.37) (Franden et al., 2013 ). Therefore, the tolerance of strains to FAL is extremely essential to the high production of FAL-derived chemicals. To evaluate the tolerance of P. putida KT2440 toward FAL and its bioconversion ability to produce FA, cells were prepared and exposed to FAL in a wide range of 20–175 mM. Unexpectedly, the final FA concentration was remarkably improved linearly upon the addition amount of FAL, steadily rising to 170 mM with yield of 97.1%. Trace of FOL was formed in the initial reaction period, which was re-oxidized to FA as no FOL could be detected in the end. And as a consequence, the selectivity in all investigations was close to 100% ( Figure 3 ). Given these facts, it could be concluded that the tolerant level of P. putida KT2440 was >175 mM under the cultivation and reaction conditions of this study. Of the recently reported strains, each one displayed a distinct tolerant profile. The oxidation of FAL by E. coli harboring SAPDH became very sluggish when FAL concentration was improved to 125 mM (Shi et al., 2019 ; Cheng et al., 2020 ), whereas the yield of FA sharply decreased to 32.1% using 75 mM FAL as initial substrate in the case of whole-cells of E. coli harboring HMFO variant (Wang et al., 2020 ). The substrate-adapted Comamonas testosteroni SC1588 cells could convert 50 mM FAL to the desired product with yield of 96% in 24 h, but only yield of 64% from 100 mM FAL even with the prolonged reaction time of 120 h (Wen et al., 2020 ). The performance of G. oxydans ATCC 621H was comparatively better, which gave satisfactory yield from 104 mM FAL, and the yield decreased by 60% when FAL concentration was increased to 158 mM (Zhou et al., 2017 ). Figure 3 Effect of FAL concentration on FA production. Reaction conditions: phosphate buffer (200 mM, pH 6.0), designated FAL concentrations, 10.5 g/L cells (dry weight), CaCO 3 at half molar concentration of FAL, 30°C, 200 rpm. The FA concentrations in conversion of 20, 50, 70, 100, and 125 mM FAL were determined on the basis of samples withdrawn at 1 h. The FA concentrations in conversion of 140 and 175 mM FAL were determined on the basis of samples withdrawn at 2 h. Although FAL concentration had no significant effect on the maximal yields (above 95%) and selectivities (nearly 100%), its influence on the reaction rates was substantial. As shown in Figure 3 , the highest initial reaction rate of 164.1 mM/h was observed when FAL concentration was 70 mM. Further increase in FAL concentration resulted in a moderately substrate inhibition, since the initial reaction rate dropped to 131.7–89.4 mM/h at the FAL concentrations of 100–175 mM. Anyhow, the initial reaction rate remained satisfactory even at high FAL concentration of 175 mM. As previously described, FAL has more inhibitive effect on microorganisms than HMF, but the V 0 in this study was considerably high than those for HMF in our previous results (Xu et al., 2020 ). A reasonable explanation was that the strategy developed in this study to enhance the catalytic performances of cells was valid and reliable, which was also evidenced by the difference between the crude activities of inducer-free cells and HMF-induced cells ( Supplementary Figure 1 ). Compared to the microorganisms reported previously (Zhou et al., 2017 ; Shi et al., 2019 ; Cheng et al., 2020 ; Wang et al., 2020 ; Wen et al., 2020 ), biocatalysts of this study appeared to be much more prevailing in synthesis of FA from FAL, because of its high tolerance toward the substrate as well as good catalytic performances. High product titers and productivities are eminently desirable in the industrial processes. In view of the inhibition of FAL against enzymes and microorganisms, high loadings of FAL would encounter low biocatalytic efficiency, thus, fed-batch was an effective approach to cope with such inhibition. With the periodical addition of FAL at ~60–70 mM, FA concentration increased rapidly and steadily, demonstrating the strong viability of cells and feasibility of fed-batch strategy. Up to 204 mM FA was produced within 3 h after three-batch feeding of FAL, with total conversion of 100% and yield of 97.5%. The productivities were changed from 100.5 to 48.5 mM/h ( Figure 4 ). The FA produced was not further degraded in contrast to reports on certain strains of genus Pseudomonas , which was proved to utilize FA as sole carbon (Wierckx et al., 2011 ). In addition, only trace of FOL was formed as the sole byproduct and, hence the selectivity of the desired product reached >97% ( Figure 4 ). Guarnieri et al. ( 2017 ) cultivated P. putida KT2440 in M9 medium supplemented with 1 g/L FAL as sole carbon, and they observed 50% reduction in FAL accompanied by notable accumulation of FOL but no FA after 16 h. This is obvious differently from our results. Generally, FOL accumulation was more commonly occurred in Saccharomyces cerevisiae, Corynebacterium glutamicum , etc. under anaerobic conditions (Ishii et al., 2013 ; Tsuge et al., 2014 ). Figure 4 Profiles of selective oxidation of FAL to FA in fed-batch conversion. We also investigated the potential of cells to biosynthesize of FA from simply prepared and partially purified substrate. In this case, FAL was self-made via dilute acid-catalyzed dehydration of corncob followed by dichloromethane extraction. It was found that FAL within 80 mM could be completely depleted with good yields of above 90%. FOL was accumulated in the initial phase of the reaction and re-oxidized finally, leading to FA with selectivity of nearly 100%. We eventually increased the substrate concentration to 100 mM, but under such conditions, both the yield and productivity were profoundly decreased ( Figure 5 ). It might be attributed to the toxic effects of the residual solvent and other undetectable detrimental compounds present in the partially purified FAL on biocatalytic activities of cells, since the crude FAL was directly obtained from corncob through one-step treatment. Currently, except for limited examples, the FAL concentrations used in whole-cell biocatalysis system are usually low or it takes a long time to achieve high yields. Compared with recent reports, the results of this study afforded a promising biocatalytic approach to FA production and further demonstrate the feasibility of biocatalysis in synthesis of biobased furan-derived chemicals ( Supplementary Table 2 ). Figure 5 Profiles of selective oxidation of partially purified FAL (Cyan, 65 mM; Dark yellow, 80 mM; Purple, 100 mM) to FA under optimized biotransformation conditions. The Role of Molybdate Transporter in Furanic Aldehyde Oxidation After demonstrating that P. putida KT2440 was a good biocatalyst for conversion of FAL and HMF to their corresponding carboxylic acids, we expected to better understand the enzyme involved in such oxidation reactions. Previously, the HMF/furfural oxidoreductase (HmfH) from C. basilensis HMF14 and HMFO from Methylovorus sp. strain MP688 were proved with activities toward both FAL and HMF, yielding FA and FDCA as products (Koopman et al., 2010 ; Dijkman and Fraaije, 2014 ). And the synthetic potential of these biocatalysts in furancarboxylic acids has been fully exploited and developed in recent years (Yuan et al., 2018 ; Wang et al., 2020 ). Nevertheless, the target enzyme in our research was likely different from them, as only the formyl group was oxidized by P. putida KT2440, evidenced by the conversion of HMF to HMFCA (Xu et al., 2020 ). G. oxydans DSM 50049 and C. testosteroni SC1588 enabled similar selective oxidations as P. putida KT2440. Certain membrane bound enzyme was presumably suggested responsible for this reaction in G. oxydans DSM 50049 (Sayed et al., 2019 ), while several aldehyde dehydrogenase from C. testosteroni SC1588 were identified for the oxidation of furanic aldehydes into corresponding furancarboxylic acids in E. coli (Zhang X. Y. et al., 2020 ). In our research, based on genome annotation and sequence alignment, several candidate genes of P. putida KT2440 were selected and their functions in FAL oxidation were assessed through gene disruption. Regrettably, no substantial progress has been made up to now. During the inspection of the genome, a molybdate transporter (ModABC, PP_3828–PP_3830, Figure 6A ) attracted our attention. As mentioned above, molybdate transporter had been evidenced critically in detoxification of many aromatic compounds (Blaschke et al., 1991 ; Ganas et al., 2008 ; Xia et al., 2018 ). In view of the facts that FAL and HMF were typical inhibitors to strains (Zaldivar et al., 1999 ; Franden et al., 2013 ), we thus wonder whether molybdate transporter was related to the biotransformation/detoxification of furanic aldehydes. We constructed a modA -disrupted mutant by homologous recombination ( Figure 6B ), which was routinely cultivated and prepared for FAL and HMF conversion. Figure 7 showed that the modA -disrupted mutant appeared to have lost the ability to oxidize substrates in a large extent. Although trace amounts of FA and HMFCA were detected, it was plausible that they were produced by the catalysis of certain unspecific aldehyde dehydrogenases. In addition, FOL and BHMF were evidently accumulated and couldn't be re-oxidized, which might be ascribed to the existence of excess furanic aldehydes. The key role of molybdate transporter suggested that one or more molybdoenzyme was involved in furanic aldehyde oxidation, which is essential for production of furancarboxylic acid. It is interesting to note that inactivation of molybdate transporter of P. putida KT2440 would substantially abolish its activity for vanillin oxidation (Graf and Altenbuchner, 2014 ), but it is still unclear whether these oxidation reactions were catalyzed by the same molybdoenzyme. Although molybdoenzyme had been proofed associated with aerobic degradation of FA in P. putida Fu1 and C. basilensis HMF14 (Koenig and Andreesen, 1989 , 1990 ; Koopman et al., 2010 ), our study demonstrated its pivotal role in oxidation of FAL and HMF for the first time. Figure 6 Organization of molybdate transporter operon in P. putida KT2440 (A) and construction of modA -disrupted mutant strain (B) . Figure 7 Time course of FAL conversion (A) and HMF conversion (B) by P. putida KT2440 wide-type (WT) and molybdate transporter mutant. Reaction conditions: phosphate buffer (200 mM, pH 6.0), 50 mM FAL (A) or HMF (B) , 10.5 g/L cells (dry weight), 25 mM CaCO 3 , 30°C, 200 rpm. To effectively synthesize furan-based carboxylic acids via biocatalysis approach, an important challenge is to seek robust biocatalyst and characterize the functional enzyme(s). Presently, strains with strong ability to oxidize furanic aldehydes had been verified in several different species, covering C. testosteroni, G. oxydans, N. coralline, Brevibacterium lutescens etc. (Pérez et al., 2009 ; Zhang et al., 2017 ; Zhou et al., 2017 ; Sayed et al., 2019 ; Wen et al., 2020 ; Zhang R. Q. et al., 2020 ). On the contrary, there is a limited amount of researches aiming at identification and characterization of the relevant enzymes. The 3-succinoylsemialdehyde-pyridine dehydrogenase and vanillin dehydrogenase from C. testosteroni SC1588 enabled recombinant E. coli to oxidize FAL and HMF into FA and HMFCA, respectively (Shi et al., 2019 ; Zhang X. Y. et al., 2020 ), which was likely due to the wide substrate scope of 3-succinoylsemialdehyde-pyridine dehydrogenase and vanillin dehydrogenase instead of their main physiological functions in vivo . The enzyme responsible for furanic aldehydes oxidation seems intriguing to unravel both in vivo and in vitro . Identifying the molybdoenzyme involved in furanic aldehyde conversion in P. putida KT2440 was still ongoing in our laboratory, which would provide valuable enzyme resources for furan-based carboxylic acids production, as well as contribute to construction of recombinant strains with high furanic aldehydes tolerance."
} | 6,693 |
30382153 | PMC6208414 | pmc | 997 | {
"abstract": "Comparative analysis of the expanding genomic resources for scleractinian corals may provide insights into the evolution of these organisms, with implications for their continued persistence under global climate change. Here, we sequenced and annotated the genome of Pocillopora damicornis , one of the most abundant and widespread corals in the world. We compared this genome, based on protein-coding gene orthology, with other publicly available coral genomes (Cnidaria, Anthozoa, Scleractinia), as well as genomes from other anthozoan groups (Actiniaria, Corallimorpharia), and two basal metazoan outgroup phlya (Porifera, Ctenophora). We found that 46.6% of P. damicornis genes had orthologs in all other scleractinians, defining a coral ‘core’ genome enriched in basic housekeeping functions. Of these core genes, 3.7% were unique to scleractinians and were enriched in immune functionality, suggesting an important role of the immune system in coral evolution. Genes occurring only in P. damicornis were enriched in cellular signaling and stress response pathways, and we found similar immune-related gene family expansions in each coral species, indicating that immune system diversification may be a prominent feature of scleractinian coral evolution at multiple taxonomic levels. Diversification of the immune gene repertoire may underlie scleractinian adaptations to symbiosis, pathogen interactions, and environmental stress.",
"conclusion": "Conclusions This comparative analysis revealed significant expansion of immune-related pathways within the Scleractinia, and further lineage-specific diversification within each scleractinian species. Different immune genes were diversified in each species (e.g., Nod-like and tachylectin-like receptors in A. digitifera and S. pistillata , and caspase-like and JNK signaling genes in O. faveolata and P. damicornis ), suggesting diverse adaptive roles for innate immune pathways. Indeed, immune pathways govern the interactions between corals and their algal endosymbionts 15 , 79 , the susceptibility of corals to disease 80 , and their responses to environmental stress 72 . Therefore, prominent diversification of immune-related functionality across the Scleractinia is not surprising, and may underlie responses to selection involving symbiosis, self-defense, and stress-susceptibility. The function and diversity of both the Scleractinia-specific and the species-specific immune repertoires deserve further study as they could prove to be critical for coral survival in the face of climate change. Indeed, factors placing high selection pressure on corals, such as bleaching and disease, both involve challenges to the immune system. Lineage-specific adaptations indicate corals continue to evolve novel immune-related functionality in response to niche-specific selection pressures. These results suggest that evolution of the innate immune system has been a defining feature of the success of scleractinian corals, and likewise may mediate their continued success under climate change scenarios.",
"introduction": "Introduction Scleractinian corals serve the critical ecological role of building reefs that provide billions of dollars annually in goods and services 1 and sustain high levels of biodiversity 2 . However, corals are declining rapidly as ocean acidification impairs coral calcification and interferes with metabolism 3 , ocean warming disrupts their symbiosis with photosynthetic dinoflagellates (family Symbiodiniaceae) 4 , and outbreaks of coral disease lead to mortality 5 . As basal metazoans, corals provide a model for studying the evolution of biomineralization 6 , symbiosis 7 , and immunity 8 , 9 - key traits which mediate ecological responses to these stressors. Understanding the genomic architecture of these traits is therefore critical to understanding corals’ success over evolutionary time 10 and under future environmental scenarios. In particular, there is great interest in whether corals possess the genes and genetic variation required to acclimatize and/or adapt to rapid climate change 11 – 13 . Addressing these questions relies on the growing genomic resources available for corals, and establishes a fundamental role for comparative genomic analysis in these organisms. Genomic resources for corals have expanded rapidly in recent years, with genomic or transcriptomic information now available for at least 20 coral species 10 . Comparative genomics in corals has identified genes important in biomineralization, symbiosis, and environmental stress response 10 , and highlighted the evolution of specific immune gene repertoires in corals 14 , 15 However, complete genome sequences have only been analyzed and compared for two coral species, Acropora digitifera 16 and Stylophora pistillata 17 , revealing extensive differences in genomic architecture and content. Therefore, additional complete coral genomes and more comprehensive comparative analysis may be transformative in our understanding of the genomic content and evolutionary history of reef-building corals, as well as the importance of specific gene repertoires and diversification within coral lineages. Here, we present the genome of Pocillopora damicornis , one of the most abundant and widespread reef-building corals in the world 18 . This ecologically important coral is a model species and is commonly used in experimental biology and physiology. It is also the subject of a large body of research on speciation 19 – 21 , population genetics 22 – 25 , symbiosis ecology 26 – 28 , and reproduction 29 – 31 . Consequently, the P. damicornis genome sequence advances a number of fields in biology, ecology, and evolution, and provides a direct foundation for future studies in transcriptomics, population genomics, and functional genomics of corals. Using the P. damicornis genome and other publicly available genomes of cnidarians and basal metazoans, we performed a comparative genomic analysis within the Scleractinia. Using this analysis, we address the following critical questions: (1) which genes are specific to or diversified within the scleractinian lineage, (2) which genes are specific to or diversified within individual scleractinian coral species, and (3) which features distinguish the P. damicornis genome from those of other corals. We address these questions based on orthology of protein-coding genes, which generalizes the approaches taken by Bhattacharya et al . 10 and Voolstra et al . 17 to a larger set of complete genomes to describe both shared and unique adaptations in the Scleractinia. In comparing these genomes, we reveal prominent diversification and expansion of immune-related genes, demonstrating that immune pathways are the subject of diverse evolutionary adaptations in corals.",
"discussion": "Results and Discussion P. damicornis genome assembly and annotation The estimated genome size of P. damicornis is 349 Mb, smaller than other scleractinian genomes analyzed to date (Table 1 ). The size of the final assembly produced here was 234 Mb, and likely lacks high-identity repetitive content (estimated ~25% of the genome based on 31-mers) that could not be assembled. Total non-repetitive 31mer content was estimated at 262 Mb and the sum of contigs was 226 Mb, indicating that up to 14% of non-repetitive content may also be missing from the assembly, likely due to high heterozygosity (Dovetail Genomics, personal communication). However, the assembly comprises 96.3% contiguous sequence, and has the highest contig N50 (28.5 kb) of any cnidarian genome assembly (Table 1 ). We identified 26,077 gene models, which is consistent with the gene content of other scleractinian and cnidarian genomes (Table 1 ). Among these genes, 59.7% had identifiable homologs (E-value \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$\\le $$\\end{document} ≤ 10 −5 ) in the SwissProt database, 73% contained identifiable homologs in at least one of the other 10 genomes, and 83.7% contained protein domains annotated by InterProScan. Genome completeness was evaluated using BUSCO, which found that 88.4% of metazoan single-copy orthologs were present and complete (0.5% were duplicated), 2.9% were present but fragmented, and 8.7% were missing. Together, these statistics indicate the P. damicornis genome assembly is of high quality and mostly complete (Table 1 ). Table 1 Assembly and annotation statistics for the P. damicornis genome (Pdam) and others used for comparative analysis. (Spis = Stylophora pistillata 17 ; Adig = Acropora digitifera 16 ; Ofav = Orbicella faveolata 100 ; Disc = Discosoma spp . 101 ; Afen = Amplexidiscus fenestrafer 101 ; Aipt = Aiptasia 85 ; Nema = Nematostella vectensis 102 ; Hydr = Hydra vulgaris 103 ; Mlei = Mnemiopsis leidyi 104 ; Aque = Amphimedon queenslandica 105 ). *Re-annotated using present pipeline (github.com/jrcunning/ofav-genome). Pdam Spis Adig Ofav* Disc Afen Aipt Nema Hydr Mlei Aque Genome size (Mb) 349† 434 420 428 350 260 329 1300 190 Assembly size (Mb) 234 400 419 486 445 370 258 356 852 156 167 Total contig size (Mb) 226 358 365 356 364 306 213 297 785 150 145 Contig/Assembly (%) 96.3 89.5 87 73.3 81.9 82.6 82.5 83.4 92.2 96.5 86.8 Contig N50 (kb) 25.9 14.9 10.9 7.4 18.7 20.1 14.9 19.8 9.7 11.9 11.2 Scaffold N50 (kb) 326 457 191 1162 770 510 440 472 92.5 187 120 No. gene models 26077 25769 ‡ 23668 37660 23199 21372 29269 27273 31452 16554 29867 No. complete gene models 21389 25563 ‡ 16434 29679 16082‡ 15552‡ 26658 13343 BUSCO completeness (%) 88.4 72.2 34.3 71 Mean exon length (bp) 245 262 ‡ 230 240 226 218 354 208 314 Mean intron length (bp) 667 917 ‡ 952 1146 1119 1047 638 800 898 80 Protein length (mean aa) 455 615 ‡ 424 413 450‡ 475 ‡ 517 331 154 280 † Computed as total occurrences for non-error 31-mers divided by homozygous-peak depth (Dovetail Genomics). ‡ Calculated in this study using GAG based on publicly available data. Genomic feature frequency phylogeny Feature frequency profiling shows the phylogenetic relationships among the genomes analyzed here (Fig. 1 ). This genome-scale analysis resolves the Complexa ( A. digitifera ) and Robusta branches ( P. damicornis, S. pistillata, O. faveolata ) of the scleractinians as a monophyletic sister clade to the corallimorpharians, lending further support to the conclusions of Lin et al . 32 that corallimorpharians are not ‘naked corals’. Figure 1 Genome phylogeny. Feature frequency profiling of protein-coding gene models produces a genome scale phylogeny that supports a monophyletic scleractinian clade (blue) with corallimorpharians as a sister clade (red). Bootstrap support from 100 pseudoreplicates was 100% at every node. Scleractinian gene content and core function Gene families were identified by ortholog clustering across the 11 genomes in Table 1 (Supplementary Data S1 ). Across all four scleractinian genomes, we identified 43,580 ortholog groups ranging in size from 1 to 566 genes, with 14,653 of these gene families present in more than one coral species. That only a third of ortholog groups occurred in multiple genomes suggests high divergence among scleractinians, consistent with the findings of Voolstra et al . 17 . The highest number of shared ortholog groups occurred between P. damicornis and S. pistillata , the two most closely related species, and a dendrogram based on shared gene content 33 reproduces the evolutionary relationships among the four corals 34 (Fig. 2 ). Although gene content in O. faveolata is more similar overall to the other robust corals than the complex A. digitifera, O. faveolata also has the highest number of species-specific gene families, which may reflect its large genome size and/or adaptations to the Atlantic Ocean. Figure 2 Species-specific and shared gene families across four scleractinian genomes. Numbers indicate gene families, including both single-copy genes and multi-copy gene families. Dendrogram is based on shared gene content, following 33 . A total of 7,536 ortholog groups were found in all four scleractinian genomes, constituting putative coral ‘core’ genes. Members of these core gene families comprised 46.6% of all P. damicornis genes, and functional profiling of the core genome revealed significant enrichment of 44 GO terms associated with basic cellular and metabolic functions, including nucleic acid synthesis and processing, cellular signaling and transport, and lipid, carbohydrate, and protein metabolism (Supplementary Data S2 ). This basic functionality explains why >30% of these gene families were also found in all other cnidarians, and 96.3% had orthologs in at least one non-coral. This is consistent with the identification of basic housekeeping functions in the core (shared) protein sets in other comparative studies 10 , 17 . Genes specific to and diversified in scleractinians are enriched for immune functionality A subset of the coral core gene families (n = 278; 3.7%) had no orthologs outside the Scleractinia, suggesting they may reflect important evolutionary innovations within this group 35 . (We refer to these genes as ‘coral-specific’ since they did not have orthologs in the non-scleractinian genomes analyzed here, yet these may still show sequence or domain similarity with genes in other organisms and in reference databases.) Other gene families (n = 21) were significantly larger in scleractinians than other anthozoans (Fisher’s exact test, p < 0.01; Fig. 3 , Supplementary Table S1 ), indicating gene family expansion that may underlie adaptation to the scleractinian condition 36 , 37 . Genes belonging to both coral-specific and coral-diversified families in P. damicornis were enriched for GO terms involved in cellular signaling and immunity (Table 2 ), and showed significant similarity to proteins with known immune function (Fig. 3 , Supplementary Table S1 ). Figure 3 Heatmap showing gene ortholog groups that were larger in scleractinians compared to other cnidarians (Pd = P. damicornis , Sp = S. pistillata , Of = O. faveolata , Ad = A. digitifera , Nv = N. vectensis , Ap = A. pallida , Ds = Discosoma sp ., Af = A. fenestrafer ). For each ortholog group, the longest protein sequence from P. damicornis was compared to the UniProt-SwissProt database using blastp, and the top hit was selected based on the lowest E-value (if < 1e-10). Uniprot accession numbers are shown in brackets. Sequences with no annotation had no hits to the SwissProt database with E < 1e-10. Gene family sizes and E-values for SwissProt hits can be found in Supplementary Table S1 . Table 2 Enrichment of GO terms in the coral-specific and coral-diversified genes in P. damicornis . Of the coral-specific genes (n = 349), 184 (53%) had GO annotations. Of the coral-diversified genes (n = 339), 229 (68%) had GO annotations. Gene set GO Accession GO Term Name Observed Expected p Coral-specific GO:0035434 copper ion transmembrane transport 2 0.10 0.004000 GO:0007165 signal transduction 32 23.78 0.004200 GO:0046654 tetrahydrofolate biosynthetic process 1 0.01 0.012300 GO:0051092 NF-kappaB activation 1 0.02 0.024500 GO:0070836 caveola assembly 1 0.04 0.036500 GO:0051607 defense response to virus 2 0.32 0.040300 GO:0006355 regulation of gene-specific transcription 12 6.06 0.045600 GO:0000183 chromatin silencing at rDNA 1 0.05 0.048400 GO:0045132 meiotic chromosome segregation 1 0.05 0.048400 Coral-diversified GO:0006857 oligopeptide transport 3 0.05 0.000015 GO:0007264 small GTPase mediated signal transduction 6 1.05 0.000600 GO:0001822 kidney development 2 0.05 0.000850 GO:0006468 protein phosphorylation 11 3.82 0.001960 GO:0007219 Notch signaling pathway 2 0.11 0.004980 GO:0033151 V(D)J recombination 1 0.01 0.007680 GO:0045737 positive regulation of CDK activity 1 0.01 0.007680 GO:0008152 metabolic process 31 30.87 0.012710 GO:0016255 attachment of GPI anchor to protein 1 0.04 0.037840 Immune-related GO terms that were significantly enriched in the coral-specific gene set included viral defense, signal transduction, and NF-κB pathway regulation (Table 2 ). NF-κB signaling plays a central role in innate immunity 38 , 39 , and was recently demonstrated to be conserved and responsive to immune challenge in the coral O. faveolata 40 . Signal transduction was associated with 32 coral-specific genes that showed significant similarity to dopamine receptors, neuropeptide receptors, G-protein coupled receptors, and tumor necrosis factor (TNF) receptor-associated factors (TRAFs) (Supplementary Data S3 ), potentially representing other coral-specific immune pathways. Indeed, the TNF receptor superfamily is more diverse in corals than any organism described thus far other than choanoflagellates 9 , 41 , with 40 proteins in A. digitifera . The P. damicornis genome contained 39 proteins with TNFR cysteine-rich domains, suggesting that diversification of this repertoire may be a common feature of corals. Another enriched GO term in the coral-specific gene set was caveola assembly–the formation of structures in cell membranes that anchor transmembrane proteins–which may also play a role in signal transduction and immunity 42 . The coral-diversified gene families (Fig. 3 , Supplementary Table S1 ) showed high similarity to receptors for pathogen recognition, such as a C-type lectin, G-protein-coupled receptors (GPCRs), and both Notch and Wnt-signaling receptors (lipoprotein receptor-related protein). Notch and Wnt signaling are critical developmental gene pathways with potentially diverse roles in coral biology, but which also have a role in coral innate immunity 43 , particularly in wound-healing processes 39 , 44 . Other coral-diversified genes were similar to Ras-related proteins with leucine-rich repeats, and a tetratricopeptide repeat-containing protein, which may play roles in signal transduction 45 . Many of these tetratricopeptide repeat proteins also contained a CHAT domain characteristic of caspases 46 , indicating a potential role in apoptotic signaling and/or coral bleaching. Other coral-diversified genes were similar to Poly (ADP-ribose) polymerase, which may act as an anti-apoptotic signal transducer 47 , and lactadherin, which may be involved in phagocytosis and clearance of apoptotic cells 48 . Genes previously found to be differentially expressed in corals under stress or immune challenge were also found in the coral-diversified gene set, including the HSP70 co-chaperone sacsin 49 , the oligopeptide transporter solute carrier family 15 50 , and NFX1-type zinc finger protein 51 . Together, these results suggest that corals as a group have evolved a diverse set of immune signaling genes for interacting with and responding to pathogens and the environment. Importantly, the immune repertoire in corals also contains many other important gene families that are not discussed here (e.g., Toll-like receptors) 39 , 40 , since we focus only on those that are specific to or diversified in corals. In addition to defending against pathogens, many of the immune pathways highlighted here may mediate the establishment and maintenance of symbiosis, including with beneficial bacteria and the Symbiodiniaceae 15 . Indeed, lectins and other pattern recognition receptors have previously been shown to regulate symbiont uptake and specificity 7 , 52 , 53 , while caspases have previously been shown to mediate bleaching and symbiont removal through apoptosis 54 . We also found that copper ion transmembrane transport was highly enriched in the coral-specific gene set (Table 2 ), which may reflect an important role for delivery of copper to endosymbionts, where it is a critical component of photosynthetic proteins (plastocyanin) and antioxidants (superoxide dismutase) 55 . For example, in mycorrhizal symbioses, fungi are known to deliver copper to their photosynthetic plant partners 56 , and shortage of trace metals such as copper has recently been linked to coral bleaching 57 . In addition to immunity and symbiosis, the genes specific to and diversified within corals may underlie other unique scleractinian traits. Calcium carbonate skeleton formation, for example, may be linked to the diversification of calcium ion channels (e.g., polycystins) and cell adhesion proteins (e.g., coadhesin, Fig. 3 ), which have previously been identified as components of the skeletal organic matrix 6 , 58 . Corals may also have diversified mechanisms for controlling gene expression, evidenced by the enrichment of transcriptional regulation and chromatin silencing functions in the coral-specific gene set (Table 2 ), and the diversification of a histone demethylation protein family (Fig. 3 ). Finally, we note that some enriched GO annotations do not translate directly to corals (e.g., kidney development), and/or are only represented by a single gene (Table 2 ), and should therefore be interpreted with caution. Within-species gene diversification also highlights immune function in scleractinians The expansion of gene families within individual lineages may represent an important mechanism of molecular evolution driving adaptation and speciation 36 . Consistent with patterns of gene family size in other organisms 59 , the number of coral gene families decreased exponentially as gene family size increased (Fig. 4 ). P. damicornis had smaller gene families overall, and the fewest large gene families (n = 3 with size > 32, max size = 75), while A. digitifera had the most large gene families (n = 25 with size > 32, max size = 255), consistent with pervasive gene duplication in this species suggested by Voolstra et al . 17 . However, statistical comparison of shared gene family sizes across the four coral species, accounting for differences in total gene content, indicated that S. pistillata had the most significantly expanded gene families (n = 16), followed by A. digitifera (n = 11). Even though O. faveolata had the highest number of non-shared gene families (Fig. 2 ), only one shared gene family was significantly expanded, suggesting that its large genome size is the result of species-specific genes and/or even expansion of shared gene content. Finally, P. damicornis had no significantly expanded gene families relative to the other scleractinians, confirming that uneven gene family size 17 and lineage-specific gene family expansion is common in the Scleractinia. Figure 4 Gene family size distribution in four coral genomes. Pdam = P. damicornis , Spis = S. pistillata , Ofav = O. faveolata , Adig = A. digitifera . Bars represent the total number of gene families in a given size class using exponential binning, with each interval open on the left (i.e., the first interval contains gene families of size 1, the second interval contains gene families of size 2 and 3, etc.). Among the gene families showing lineage-specific expansions in corals, several were similar to reverse transcriptases and transposable elements (Supplementary Table S2 ); these may represent ‘genetic parasites’ that propagate across the genome 60 , but they may also play crucial roles in genome evolution and the regulation of gene expression 61 . Annotations of other expanded gene families (Supplementary Table S2 ) suggest important roles in interacting with the environment, cellular signaling, and immunity. While uneven gene family size could also reflect variation in assembly completeness and quality of genome assemblies 62 , these annotations are consistent with categories of genes known to undergo lineage-specific expansion across eukaryotes 45 . One expanded gene family in A. digitifera was similar to NOD-like receptors (NLRs), which are cytoplasmic pattern recognition receptors that play a key role in pathogen detection and immune activation 63 . Characterized by the presence of NACHT domains, NLR genes are highly diversified and variable in number in the genomes of cnidarians 64 and other species 60 . The expansion of this gene family in A. digitifera is consistent with these observations, and may represent adaptation to a new pathogen environment 65 , or to species-specific symbiotic interactions with microbial eukaryotes and prokaryotes 39 . Another expanded gene family in A. digitifera was similar to ephrin-like receptors, which may mediate signaling cascades and cell to cell communication 66 . In S. pistillata , one expanded gene family was similar to tachylectin-2, a pattern recognition receptor that has been identified in many cnidarians 39 . Previously, a tachylectin-2 homolog was found to be under selection in the coral Oculina 67 , providing more evidence that such genes are involved in adaptive evolution in corals. The one significantly expanded gene family in O. faveolata did not have a strong hit in the SwissProt database, but did contain a caspase-like domain suggesting a role in apoptosis, which was recently linked to disease susceptibility in this species 68 . Overall, differential expansion of genes related to the immune system is consistent with the findings of Voolstra et al . 17 , and suggests that this phenomenon is a general attribute of corals. Lineage-specific immune diversification in corals and other taxa may reflect interactions with specific consortia of eukaryotic and prokaryotic microbial symbionts 69 . In addition to putative immune-related function, genes that have undergone lineage-specific expansions in corals may also play roles in biomineralization, which could contribute to variation in growth and morphology among coral species. For example, one significantly expanded gene family in A. digitifera was similar to a CUB and peptidase domain-containing protein that was found to be secreted in the skeletal organic matrix 58 , and another in S. pistillata was similar to fibrillar collagen with roles in biomineralization 70 . Although the P. damicornis genome did not contain any gene families that were significantly expanded relative to the other corals, it did contain many genes (n = 6,966, 26.7%) with no orthologs in other genomes. While most of these P. damicornis -specific genes were unannotatable, protein domain homology revealed significant enrichment for 11 GO terms, including G-protein coupled receptor (GPCR) signaling pathway, bioluminescence, activation of NF-κB-inducing kinase, and positive regulation of JNK cascade (Table 3 ). The mitogen-activated protein kinase JNK plays a role in responses to stress stimuli, inflammation, and apoptosis 71 . JNK prevents the accumulation of reactive oxygen species (ROS) in corals in response to thermal and UV stress, and inhibition of JNK leads to coral bleaching and cell death 72 . The NF-κB transcription factor may also link oxidative stress and apoptosis involved in coral bleaching 73 , in addition to its central role in innate immunity 39 . The occurrence of lineage-specific genes that may function in these pathways indicates that P. damicornis may have evolved unique immune strategies for coping with environmental stress. Table 3 Enrichment of GO terms in the P. damicornis -specific gene set. This gene set included 6,966 genes, of which 1,498 (22%) had GO annotations. GO Accession GO Term Name Observed Expected p GO:0007186 G-protein coupled receptor signaling pathway 407 149.36 <1e-30 GO:0008218 bioluminescence 15 3.24 6.00E-08 GO:0007250 activation of NF-kappaB-inducing kinase activity 10 3.48 0.0014 GO:0046330 positive regulation of JNK cascade 10 3.48 0.0014 GO:0007165 signal transduction 478 231.92 0.0018 GO:0035278 miRNA mediated inhibition of translation 5 1.68 0.0196 GO:0007411 axon guidance 3 0.72 0.0261 GO:0035385 Roundabout signaling pathway 3 0.72 0.0261 GO:0006814 sodium ion transport 11 6.25 0.0419 GO:0050482 arachidonic acid secretion 5 2.04 0.0446 GO:0015074 DNA integration 11 6.37 0.0474 An expanded role of immunity in P. damicornis may explain how Pocillopora has achieved such a widespread distribution 18 , 19 . Indeed, Pocillopora corals function as fast-growing and weedy pioneer species in Hawaii 74 , on the Great Barrier Reef 75 , and in the eastern tropical Pacific (ETP) 76 . In fact, in the ETP, where the coral used in this study was collected from, Pocillopora thrives in marginal habitats, often dealing with elevated turbidity and reduced salinity after heavy rainfall events, subaerial exposure during extreme low tides, and both warm- and cold-water stress due to ENSO events and periodic upwelling 77 . A diversified immune system may also allow for flexibility in symbiosis, which may further contribute to the success of Pocillopora 26 , 78 . While the wide distribution of P. damicornis suggests there may be considerable variation in its genome that is not captured by our sample from the ETP, this work provides a foundation for future genomic analysis in this important coral species."
} | 7,261 |
40082732 | PMC11963765 | pmc | 1,000 | {
"abstract": "Abstract Yeasts have emerged as well-suited microbial cell factory for the sustainable production of biofuels, organic acids, terpenoids, and specialty chemicals. This ability is bolstered by advances in genetic engineering tools, including CRISPR–Cas systems and modular cloning in both conventional ( Saccharomyces cerevisiae ) and non-conventional ( Yarrowia lipolytica, Rhodotorula toruloides, Candida krusei ) yeasts. Additionally, genome-scale metabolic models and machine learning approaches have accelerated efforts to create a broad range of compounds that help reduce dependency on fossil fuels, mitigate climate change, and offer sustainable alternatives to petrochemical-derived counterparts. In this review, we highlight the cutting-edge genetic tools driving yeast metabolic engineering and then explore the diverse applications of yeast-based platforms for producing value-added products. Collectively, this review underscores the pivotal role of yeast biotechnology in efforts to build a sustainable bioeconomy.",
"conclusion": "Concluding remarks Yeast has moved far beyond basic genetic tools and ethanol production. Advances in genetic engineering tools, such as CRISPR–Cas systems, MoClo, and RNA-based regulation, have led to newfound control of metabolic pathways. These innovations, when combined with computational approaches, have expanded the metabolic landscape of traditional yeast models like S. cerevisiae and emerging non-conventional yeasts such as Y. lipolytica, R. toruloides , and C. krusei for industrial applications (Patra et al. 2021 , Koh et al. 2024 ). Despite these breakthroughs, challenges remain in addressing metabolic burden, product toxicity, and the limitations of genetic tools for non-conventional yeasts (Mao et al. 2024 ). Computational modeling and machine learning offer promising solutions, but their full potential has yet to be realized in these hosts, especially when datasets can be limited (Lu et al. 2022 ). Additionally, achieving economic scalability remains critical, emphasizing the need for cost-effective feedstocks, efficient downstream processes, and strong industry-academia collaboration (Makepa and Chihobo 2024 ). Looking ahead, yeast-based biomanufacturing is poised to play a pivotal role in driving sustainable solutions for biofuels, organic acids, terpenoids, and specialty chemicals. The future looks bright for yeast-based hosts to serve as a major contender for bio-based production of sustainable chemicals.",
"introduction": "Introduction Yeasts have played a pivotal role in our history even before understanding their existence as they enabled the production of essential metabolites like ethanol, acetic acid, and lactic acid through natural fermentation processes (Maicas 2020 , Voidarou et al. 2021 ). These native metabolites have been a mainstay in the food and beverage industries for centuries, contributing to products like bread, beer, and yogurt. However, with the rise of genetic engineering and sequencing technologies, yeast has evolved beyond this traditional food fermentation role (Jin et al. 2020 ). Specifically, modern tools now enable precise genome manipulation and the introduction of heterologous pathways thereby transforming yeast into a versatile microbial cell factory capable of producing high-value chemicals tailored to industrial needs (Lee et al. 2021 , Koh et al. 2024 ). At the recent forefront of this transformation are tools such as CRISPR–Cas9, which provides programmable genome editing, and auxiliary systems like CRISPR interference and CRISPR activation for tunable gene regulation (Deaner and Alper 2019 , Bowman et al. 2020 , Liu et al. 2022 ). Additionally, synthetic biology has advanced efforts to introduce modular cloning (MoClo) systems, synthetic promoters and terminators, create dynamic regulatory circuits, and facilitate the construction of efficient metabolic pathways. These advances, combined with systems biology approaches, have made yeasts a rather potent and tractable set of host organisms (Perrot et al. 2024 , Yuan et al. 2024 ). While Saccharomyces cerevisiae remains a cornerstone of yeast biotechnology due to its genetic tractability and industrial relevance/track record, there are many limitations that remain including substrate utilization profile and stress tolerance that have thus driven interest in non-conventional yeasts such as Yarrowia lipolytica, Rhodotorula toruloides , and Candida krusei ( Issatchenkia orientalis ), among others. These emerging yeast hosts exhibit unique metabolic traits, robust growth under stressful conditions, and versatile carbon source utilization (Blazeck et al. 2014 , Wagner and Alper 2016 , Koh et al. 2023 ). Despite these advances, many key challenges in the engineering of these hosts. With these contexts in mind, this review explores the transformative impact of advanced genetic engineering tools on yeast biotechnology. First, we highlight how CRISPR systems, MoClo frameworks, and RNA-based regulatory strategies have expanded the capabilities of both traditional and non-conventional yeasts. Next, we examine trends in the production of biofuels, organic acids, terpenoids, and specialty chemicals, offering an overview of how cutting-edge innovations are shaping the future of yeast-based biomanufacturing. Finally, we look conclude with visions for the future of yeast host organisms to drive sustainable solutions for global industrial needs."
} | 1,359 |
26987552 | PMC4985588 | pmc | 1,001 | {
"abstract": "Abstract Methanogenic inhibitors are often used to study methanogenesis in complex microbial communities or inhibit methanogens in the gastrointestinal tract of livestock. However, the resulting structural and functional changes in archaeal and bacterial communities are poorly understood. We characterized microbial community structure and activity in mesocosms seeded with cow dung and municipal wastewater treatment plant anaerobic digester sludge after exposure to two methanogenic inhibitors, 2‐bromoethanesulfonate ( BES ) and propynoic acid ( PA ). Methane production was reduced by 89% (0.5 mmol/L BES ), 100% (10 mmol/ LBES ), 24% (0.1 mmol/ LPA ), and 95% (10 mmol/ LPA ). Using modified primers targeting the methyl‐coenzyme M reductase ( mcrA ) gene, changes in mcrA gene expression were found to correspond with changes in methane production and the relative activity of methanogens. Methanogenic activity was determined by the relative abundance of methanogen 16S rRNA cDNA as a percentage of the total community 16S rRNA cDNA . Overall, methanogenic activity was lower when mesocosms were exposed to higher concentrations of both inhibitors, and aceticlastic methanogens were inhibited to a greater extent than hydrogenotrophic methanogens. Syntrophic bacterial activity, measured by 16S rRNA cDNA , was also reduced following exposure to both inhibitors, but the overall structure of the active bacterial community was not significantly affected.",
"introduction": "Introduction Methane can be viewed as a potent greenhouse gas, an energy source, a dangerous, and explosive byproduct of anaerobic biodegradation, a waste product diverting energy from animal feed, or a driver of microbial carbon cycling (Hallam et al. 2003 ; Dupont and Accorsi 2006 ; Knittel and Boetius 2009 ; Appels et al. 2011 ; Chowdhury and Dick 2013 ; IPCC 2013 ; Patra and Yu 2013 ). Due to the importance of methane in fields ranging from climate science to animal husbandry, much research has focused on understanding the activity of methanogenic archaea under anaerobic conditions (Reeve et al. 1997 ; Conrad 2007 ). Aerobic methane generation has also been identified and may be an important source of methane from oceans (Karl et al. 2008 ); however, this study focuses on methane production under anaerobic conditions. All known methanogenic archaea contain genes that encode for the methyl‐coenzyme M reductase (MCR), which catalyzes the final step of methanogenesis. There are two isoenzymes, MCRI and MCRII, and the mcrA and mrtA genes encode for the α ‐subunit of each of these isoenzymes, respectively (Reeve et al. 1997 ). The mcrA / mrtA genes have been a common target for measuring methanogen abundance, activity, and diversity. Distinctions between mcrA and mrtA genes often are not made in the literature and hereafter we use mcrA to refer to the combination of both genes, unless specified otherwise. The agreement between phylogenetic trees based on 16S rRNA genes and mcrA genes has helped to support the use of the mcrA gene as a methanogen‐specific phylogenetic target (Luton et al. 2002 ). Compounds that inhibit methanogenesis have been important in research to study pure cultures of methanogens (Ungerfeld et al. 2004 ; Watkins et al. 2012 ), carbon cycling in soils (Sugimoto and Wada 1993 ; Wu et al. 2001 ), ruminal methanogens (Ungerfeld et al. 2006 ; Zhou et al. 2011b ), dechlorination (Perkins et al. 1994 ; Chiu and Lee 2001 ), mercury methylation (Han et al. 2010 ; Avramescu et al. 2011 ), production of volatile fatty acids (Zhang et al. 2013 ; Jung et al. 2015 ), anaerobic digestion (Zinder et al. 1984 ; Navarro et al. 2014 ), and the degradation of nitrosamines (Tezel et al. 2011 ) and methanethiol (Sun et al. 2015 ). Further, inhibitors have been useful in elucidating the activity of methanogens related to metal and metalloid methylation (Meyer et al. 2008 ; Thomas et al. 2011 ). A variety of chemicals have been applied to inhibit methanogenesis in livestock to either reduce methane emissions or to direct more of the feed energy to animals for increased agricultural output (i.e., milk and meat) (Machmüller and Kreuzer 1999 ; Boadi et al. 2004 ; Beauchemin et al. 2009 ). Regardless of the intended use, when methanogenic inhibitors are used in mixed communities, detailed characterization of inhibitor‐induced changes to both archaeal and bacterial populations is needed to ensure that the observed effects can be accurately ascribed to the inhibition of methanogenic activity and to elucidate any indirect effects. This is especially important given that a wide diversity of methanogenic inhibitors with varying properties and mechanisms of action are available. Methanogenic inhibitors can be divided into several categories (as reviewed by (Liu et al. 2011 )), including analogs of coenzyme M (Gunsalus et al. 1978 ; Zinder et al. 1984 ), inhibitors of methanopterin biosynthesis (Dumitru et al. 2003 ), medium‐ and long‐chain fatty acids (Prins et al. 1972 ; Soliva et al. 2003 ), nitrocompounds (Zhou et al. 2011b ), halogenated hydrocarbons (Denman et al. 2007 ), ethylene (Oremland and Taylor 1975 ), acetylene (Oremland and Taylor 1975 ; Sprott et al. 1982 ), and unsaturated analogs of propionate and butyrate (Ungerfeld et al. 2003 , 2004 , 2006 ; Zhou et al. 2011b ). While many inhibitors are considered methanogen‐specific, various studies have found that other microorganisms can be affected. The most commonly used methanogenesis inhibitor, 2‐bromoethanesulfonate (BES), a coenzyme M analog, has been found to also inhibit dechlorinating bacteria (Loffler et al. 1997 ; Chiu and Lee 2001 ) and to affect bacterial growth on aliphatic alkenes (Boyd et al. 2006 ). Propynoic acid (PA), an unsaturated propionate analog with one triple carbon bond, is also an effective inhibitor of methanogenesis (Ungerfeld et al. 2004 ; Zhou et al. 2011b ). However, limited studies have been performed on the effects of PA on the structure of microbial communities (Patra and Yu 2013 ). To date, studies of the impacts of methanogenic inhibitors on bacterial and archaeal communities have relied on clone libraries, denaturing gradient gel electrophoresis (DGGE), or terminal restriction fragment length polymorphism (TRFLP) targeting the 16S rRNA gene (Chiu and Lee 2001 ; Xu et al. 2010a , b ; Patra and Yu 2013 ; Lins et al. 2015 ) and the mcrA gene (Denman et al. 2007 ). Results from DGGE‐based evaluations of the impact of inhibitors have shown changes in the overall community structure, but did not yield insights into how specific populations were impacted (Chiu and Lee 2001 ; Patra and Yu 2013 ). Studies using TRFLP and clone libraries of the 16S rRNA gene have reported decreases in the relative abundance of aceticlastic methanogens and syntrophic bacteria and increases in the relative abundance of homoacetogens after exposure of mesophilic anaerobic digester sludge to BES and chloroform (Xu et al. 2010a , b ). In a study of cow rumen communities, mcrA gene clone libraries and quantitative PCR revealed a decrease in the most abundant methanogenic genus, Methanobrevibacter , under BES inhibited conditions (Denman et al. 2007 ). Since these studies relied on DNA‐based techniques (Chiu and Lee 2001 ; Denman et al. 2007 ; Xu et al. 2010a , b ; Patra and Yu 2013 ; Lins et al. 2015 ), they may not have revealed short‐term changes in microbial activity in batch mesocosms or in systems with low yield, because of low growth rates and the retention of dead or inactive biomass and extracellular DNA (Chiao et al. 2014 ; Smith et al. 2015a ). In this study, we evaluated a modification to commonly used PCR primer sets targeting the mcrA gene to expand their coverage. We then applied this primer set to track the expression of mcrA genes by using reverse transcriptase quantitative PCR (RT‐qPCR) in mixed communities seeded with anaerobic digester sludge and cow dung at different levels of inhibition by either BES or PA. The effects of BES and PA on methanogenic and bacterial populations were characterized through a combination of DNA‐ and RNA‐based Illumina sequencing targeting the V4 region of the 16S rRNA gene and 16S rRNA cDNA, and the mcrA gene and mcrA transcript cDNA.",
"discussion": "Results and Discussion \n mcrA primer design and mock community characterization To target the mcrA gene in methanogens, the mlas forward primer described by Steinberg and Regan ( 2009 ) was modified with additional degeneracies and used with the previously reported mcrA‐rev reverse primer (Steinberg and Regan 2008 ). These modifications improved the predicted amplification for 10 of the 32 methanogens with complete genomes available (Table S1). Amplification was confirmed using 10 DNA extracts from pure cultures of methanogens (Tables S2, S3). These DNA extracts were pooled to create two mock communities A and B, to represent either a relatively even community (A) or an uneven community (B) with relative methanogen DNA abundances similar to those found in an anaerobic digester (Smith et al. 2013 ). For mock communities A and B, both the 16S rRNA genes and mcrA genes were sequenced. A third mock community, mock community A‐PCR was created by pooling the PCR products from individually amplified mcrA genes for each methanogen. Calculated relative abundances were determined based on pooled concentrations and the experimental sequencing results are compared in Figure 1 . When comparing the results obtained for mock communities A and B, the trends were similar for both genes although some differences in the percent relative abundances were observed (Fig. 1 ). A previous comparison of methanogen mock communities with TRFLP noted greater differences between expected and observed communities based on the mcrA gene as compared to the 16S rRNA gene, which were attributed to the higher number of degeneracies in the primers used for the mcrA gene (Lueders and Friedrich 2003 ). Comparing our calculated and experimentally measured communities using the θ \n yc community dissimilarity metric, we observed a lower community dissimilarity based on the mcrA gene ( θ \n yc of 0.48, 0.33, and 0.40 for mock communities A‐PCR, A, and B, respectively) compared to the dissimilarity based on the 16S rRNA gene ( θ \n yc of 0.58 and 0.72 for mock communities A and B, respectively). These differences may result, in part, from challenges in quantification using amplicon sequencing due to gene target‐specific biases, PCR conditions, quantification method, and primers used (Suzuki and Giovannoni 1996 ; Zhou et al. 2011a ; Pinto and Raskin 2012 ). The relative abundance of Methanobacterium was much greater, while the relative abundance of Methanosaeta was much lower than predicted for both the 16S rRNA and mcrA genes (Fig. 1 ). However, both genera were more abundant in mock community B compared to mock community A for both genes, as expected. For Methanobrevibacter, Methanococcus, and Methanosphaera , the relative abundance as measured by the mcrA gene was much lower in mock communities A and B as compared to the predicted values and those measured by the 16S rRNA gene. Obvious PCR biases were not responsible for this underrepresentation as the primers have no mismatches with their target sequences for these organisms (Table S3) and mock community A‐PCR, which was generated by pooling individually amplified PCR products of the mcrA gene from each strain, exhibited similar results (Fig. 1 ). Other factors that can affect sequencing errors include template concentration (Kennedy et al. 2014 ) and library preparation method (Schirmer et al. 2015 ). Errors during Illumina sequencing can be related to certain motifs, which can vary based on library preparation method (Schirmer et al. 2015 ). The differences between the predicted and the experimental sequencing results observed for the mock communities can be useful in guiding the analyses of mesocosm samples, as described below. Previous studies that compared the methanogen community structures using sequencing of the 16S rRNA gene, mcrA gene, and other functional genes related to methanogenesis have found some quantitative differences depending on the gene sequenced (Dziewit et al. 2015 ; Wilkins et al. 2015 ), but did not include mock communities for comparison. Given the observations made with the mock communities, we note that our interpretation of sequencing results from unknown mesocosm samples focuses on the comparison of relative abundances between samples. Inhibition reduced methane production, mcrA expression, and 16S rRNA of methanogens To characterize short‐term changes in mixed communities induced by methanogenic inhibitors, biomass samples were collected from cow dung and anaerobic digester sludge mesocosms operated for 9 days at varying levels of methanogenic activity controlled through the addition of BES and PA. Methanogenic activity was monitored through the measurement of methane production and mcrA gene expression. The microbial communities and their activities were characterized using sequencing of the 16S rRNA gene, 16S rRNA cDNA, mcrA genes, and mcrA transcript cDNA. As expected, with increasing concentrations of the methanogen inhibitors BES and PA, the rate of methane production and cumulative methane produced decreased (Figs. 2 and S1). Expression of the mcrA gene corresponded to the rate of methane production (Fig 2 ). This finding is important, as relationships between the expression of genes and the resulting function are often assumed but rarely confirmed (Rocca et al. 2015 ). Similarly, higher total methane production was associated with a higher proportion of active methanogens as measured by 16S rRNA cDNA sequences (referred to here as “relative activity”) of methanogens over the total community (including Bacteria and Archaea ) (Fig. 2 ). This finding is consistent with other observations linking these measurements in an anaerobic membrane bioreactor (Smith et al. 2015b ) and anaerobic digesters (Wilkins et al. 2015 ). There are well‐recognized biases associated with quantifying 16S rRNA cDNA to measure activity, including differences in rrn operon copy numbers and lifestyle strategies among different populations. These biases highlight the importance of comparing rRNA levels with other measures of metabolic activity (Blazewicz et al. 2013 ). Here, the observed correlation between methanogen 16S rRNA cDNA concentrations and expression levels of a functional gene specific to methanogens (Pearson matrix correlation r = 0.93) (Fig. 2 ) indicates that 16S rRNA activity can be a reliable metric for methanogen activity, at least for the current conditions. Figure 2 Cumulative methane production and molecular characterization of methanogens in cow dung and anaerobic digester sludge mesocosms after 9 days of incubation. Relative methanogen activity based on methanogen 16S rRNA cDNA as a % of the total community (including Bacteria and Archaea ) (bars), mcrA expression normalized by 16S rRNA cDNA (diamonds) determined with RT ‐ qPCR , and cumulative methane production (circles). Error bars for methane production volume represent the propagated uncertainty in methane concentration measurements. mcrA expression is displayed as the averages and standard deviations of triplicate RT ‐ qPCR reactions. Duplicates shown represent duplicate biomass samples from the same reactors. No inhibitor was added in control conditions. Differences in the mesocosms for different inhibition conditions were evaluated by sequencing the 16S rRNA gene, 16S rRNA cDNA, mcrA gene, and mcrA transcript cDNA. As expected, given the short duration of the experiment, differences in the archaeal DNA‐based sequencing results for the five conditions were modest (Fig. 3 A and C). In contrast, the RNA‐based sequencing results (Figure 3 B and D), revealed substantial differences in the five mesocosms. These results highlight changes to the methanogenic community structure, but do not reflect changes in absolute abundance or activity. Based on the 16S rRNA cDNA quantification (Fig. 2 ), the methanogenic community was shown to become less active with increasing inhibitor concentration. As with the mock communities, the broad trends in relative abundance and activity across inhibition conditions within a given methanogenic genus were similar for the two different genes sequenced (Fig. 3 A and B compared to Fig. 3 C and D). However, the actual values for percent relative abundance and activity for the two genes were quite different. Similar to the results from the mock communities, Methanosaeta spp. appeared to be more abundant and active when mcrA ‐based sequencing was used, while Methanospirillum spp. were more abundant and active according to 16S rRNA‐based sequencing. Figure 3 Relative abundance ( DNA ) and activity ( RNA ) of methanogens in anaerobic mesocosms after 9 days of incubation based on 16S rRNA genes (A), 16S rRNA cDNA (B), mcrA genes (C), and mcrA transcript cDNA (D), sequencing. Sequences from duplicate samples for each condition are combined (duplicates are shown in Figure S4). \n Methanosaeta spp. were the most abundant and active methanogens in the control samples, representing 38% of the archaeal 16S rRNA gene and 71% of the archaeal 16S rRNA cDNA sequences (Fig. 3 ). Results from mcrA gene and transcript cDNA sequencing of the control samples also show Methanosaeta spp. were the most abundant and active methanogens, representing 86% and 93% of the methanogen community and active methanogen community, respectively. Further, the activity of Methanosaeta spp. was reduced in both BES and PA 10 mmol/L inhibition conditions, shown by both 16S rRNA cDNA and mcrA transcript cDNA results (Fig. 3 B and D). Little difference was observed between Methanosaeta spp. activity in PA 0.1 mmol/L compared to the control condition. This is consistent with the methane generation results since, among the four inhibited conditions, the most methane was generated in the PA 0.1 mmol/L treatment (Fig. 2 ). Results from both the 16S rRNA gene and 16S rRNA cDNA sequencing indicated that Methanosphaera spp. and Methanobrevibacter spp. represented a greater fraction of the archaeal community and active archaeal community under all inhibited conditions compared to the control (Fig. 3 A and B). These genera made up a smaller fraction of the mcrA ‐based communities, though Methanobrevibacter spp. was found to be more active for the most inhibited conditions as compared to the control based on mcrA transcript cDNA (Fig. 3 C). Methanoregula spp. constituted 15–33% of the archaeal community according to 16S rRNA gene sequencing, but its activity represented a much smaller fraction, between 2 and 6%, based on 16S rRNA cDNA sequencing for all conditions. Using mcrA ‐based sequencing, Methanoregula spp. represented less than 2% of the abundance and activity of methanogens under all conditions. Differences between Methanoregula 16S rRNA genes and cDNA sequencing have been previously reported (Smith et al. 2015a , b ), but little is known about how these levels translate to activity. These results could indicate that Methanoregula was present in the inoculum, but not active in the mesocosms or could result from differences in the relationship of activity to rRNA levels within the cells of this genus. Interestingly, Methanoregula has only one copy of the 16S rRNA gene, while most other methanogens have two or more. This is further support of the possible different lifestyle strategy of Methanoregula compared to other methanogens."
} | 4,948 |
33328652 | PMC8114936 | pmc | 1,003 | {
"abstract": "In principle, iron oxidation can fuel significant primary productivity and nutrient cycling in dark environments such as the deep sea. However, we have an extremely limited understanding of the ecology of iron-based ecosystems, and thus the linkages between iron oxidation, carbon cycling, and nitrate reduction. Here we investigate iron microbial mats from hydrothermal vents at Lōʻihi Seamount, Hawaiʻi, using genome-resolved metagenomics and metatranscriptomics to reconstruct potential microbial roles and interactions. Our results show that the aerobic iron-oxidizing Zetaproteobacteria are the primary producers, concentrated at the oxic mat surface. Their fixed carbon supports heterotrophs deeper in the mat, notably the second most abundant organism, Candidatus Ferristratum sp. (uncultivated gen. nov.) from the uncharacterized DTB120 phylum. Candidatus Ferristratum sp., described using nine high-quality metagenome-assembled genomes with similar distributions of genes, expressed nitrate reduction genes narGH and the iron oxidation gene cyc2 in situ and in response to Fe(II) in a shipboard incubation, suggesting it is an anaerobic nitrate-reducing iron oxidizer. Candidatus Ferristratum sp. lacks a full denitrification pathway, relying on Zetaproteobacteria to remove intermediates like nitrite. Thus, at Lōʻihi, anaerobic iron oxidizers coexist with and are dependent on aerobic iron oxidizers. In total, our work shows how key community members work together to connect iron oxidation with carbon and nitrogen cycling, thus driving the biogeochemistry of exported fluids.",
"conclusion": "Conclusions and implications At Lōʻihi Seamount, energy from iron oxidation fuels the growth and ecological interactions of a diverse microbial community. The well-known Zetaproteobacteria colonize iron-rich vents [ 18 , 97 ], oxidize Fe(II) aerobically, and produce Fe(III) oxyhydroxide stalks to create the physical framework of the mat [ 21 ]. The Zetaproteobacteria use energy and electrons from Fe(II) to fix carbon, some of which binds to mat biominerals. Organic carbon is also made available through the viral lysis of Zetaproteobacteria, as they are hosts to Mu-like lysogenic phages and in contact with a diverse and active viral assemblage. Zetaproteobacteria oxygen consumption creates anaerobic zones and thus, in these various ways, Zetaproteobacteria create the physical and chemical niche for the nitrate-reducing heterotrophic iron oxidizer Candidatus Ferristratum . Because this metabolism is inferred from genomes and transcriptomes, the logical next step would be to attempt isolation. However, it is not clear that complete isolation would be successful, as Candidatus Ferristratum sp. appears to require others to remove byproducts to prevent chemodenitrification and toxicity. Instead, we can use our results as a starting point for more specific probing of ecological interactions and metabolite exchange within the mat, which may be unique in iron mats because of the affinity of organics for iron. In addition to affecting local biogeochemistry, organic-bound iron and other metabolites are carried by diffusely venting fluid moving through the mat, exported from Lōʻihi in buoyant plumes for 100 s to 1000+ km [ 98 ]. These exported fluids fertilize iron-depleted waters, connecting microbial iron and nutrient cycles across ocean basins. Description of Candidatus Ferristratum gen. nov Ferristratum (fer.ri.stra’tum. L. neut. n. ferrum iron; L. neut. n. stratum mat/cover or a layer; N.L. neut. n. Ferristratum from an iron mat layer). Genus defined from nine metagenome-assembled genomes with >70% completeness and <20% redundancy with an average 77% pairwise AAI. Genome sources from three unique samples. Type material: MAG S6_Bacteria1 (partial 16S rRNA gene present). Five of the highest-quality (>90% complete, <5% redundant) Candidatus Ferristratum sp. genomes have been submitted to IMG under the analysis project IDs Ga0454285-Ga0454287, Ga0454293, and Ga0454316 (type material). Physiological inferences from genome annotation: Heterotrophic. Facultative anaerobic. Capable of respiring nitrate coupled to organic carbon or Fe(II) oxidation for energy. Capable of nitric oxide reduction and nitrate assimilation. Found within Fe(II)-rich hydrothermal vent bulk/deep mat and sediment environments.",
"introduction": "Introduction Chemolithotrophy fuels primary production and nutrient cycling in dark environments (e.g., [ 1 – 3 ]). This has been well demonstrated for deep sea sulfur-oxidizing ecosystems (e.g., [ 4 , 5 ]), yet at the bottom of the ocean, the most abundant source of energy for chemolithotrophy is iron that originates from basaltic ocean crust [ 6 ]. Iron-oxidizing microbial communities can be found associated with widespread hydrothermal vents at the ocean floor (i.e., [ 7 – 17 ]). The best-studied example is Lōʻihi Seamount (also in publication as Loihi Seamount), a submarine volcano near Hawaiʻi with extensive iron microbial mats associated with low- to mid-temperature vents [ 11 , 14 , 18 – 20 ]. These distinctive biomineral mats at Lōʻihi are produced by the Zetaproteobacteria [ 21 , 22 ], a class of aerobic autotrophic iron oxidizers that are the only known iron oxidizers in the mat. While the Zetaproteobacteria are relatively well-studied [ 7 , 23 , 24 ], the ecology of their microbial mats is poorly explored. Metabolic predictions are largely based on isolate physiology studies [ 25 , 26 ] and genomic potential [ 19 , 20 , 24 ] of the Zetaproteobacteria, while the functions of other, flanking members of the microbial community have largely been inferred from 16S rRNA gene taxonomy, which assumes metabolism is tied with taxonomic affiliation [ 8 , 11 , 18 ]. However, this approach overlooks the roles of uncharacterized taxa such as the DTB120 phylum, found at Lōʻihi [ 11 , 20 ], as well as viral communities that may moderate mat ecology and mediate nutrient fluxes [ 27 ]. Thus, major questions remain about the metabolisms and biogeochemical roles of these iron oxidation-driven ecosystems. One key question is how iron oxidation drives carbon cycling throughout the iron mat. Lōʻihi mats are somewhat enriched in 13 C [ 28 ], consistent with primary productivity. A study using quantitative PCR showed that the Calvin-Benson-Bassham (CBB) pathway gene cbbM/rbcL is much more abundant than aclB (reductive tricarboxylic acid pathway), suggesting that the CBB pathway is the dominant carbon fixation pathway in the mats [ 29 ]. However, not all carbon fixation pathways were investigated and the responsible organisms were not identified. Zetaproteobacteria are often abundant in the mats (ranging from 1 to 96%) [ 19 , 24 ], and all have CBB pathway genes, based on isolate and environmental genomes [ 19 , 20 , 24 – 26 , 30 – 33 ]. Thus, due to abundance and genetic potential, Zetaproteobacteria are the presumed primary producers in the Lōʻihi iron mats, yet this has not been definitively shown. Particularly since Zetaproteobacteria are aerobic and oxygen is typically depleted within the first few mm’s to cm’s of the mat surface [ 21 , 34 ], there are large anoxic portions of the iron mat where we do not know the source of fixed carbon or the trophic structure of the community. To understand how carbon flow structures the ecosystem, we need to determine how both aerobic and anaerobic organisms contribute to carbon cycling. Compared to carbon, we know even less about nitrogen cycling in iron-rich mats and vents. The main nitrogen sources at Lōʻihi are vent fluids containing ammonia (0.28–7.5 µM) and the surrounding ocean water containing nitrate (36–43 µM) [ 35 – 37 ]. Sylvan et al. [ 35 ] showed that Lōʻihi vent fluids have elevated nitrate N and O isotope ratios, with patterns that suggest denitrification combined with nitrification and/or ammonia oxidation. Some Zetaproteobacteria have genes for nitrate assimilation ( nasA ) and denitrification ( napA / nirK/nirS ) [ 7 , 19 , 29 ], but it is unclear if Zetaproteobacteria are the primary drivers of denitrification in the mats. Denitrification within the anaerobic portions of the mat is expected, though it is unknown whether this denitrification is coupled to the oxidation of organic carbon, Fe(II), or both. The presence of both nitrate and Fe(II) at Lōʻihi suggests there is a niche for nitrate-reducing iron oxidizers. However, there has been much debate about whether denitrifying organisms can enzymatically conserve energy from iron oxidation as opposed to chemodenitrification, in which organotrophic denitrification produces nitrite that oxidizes Fe(II) (see review by Bryce et al. [ 38 ]). While nitrate reducers that clearly enzymatically oxidize iron largely elude isolation, a number of studies have demonstrated the relevance of coupled iron oxidation and denitrification in coastal marine environments [ 39 , 40 ], suggesting that it may also be important in other marine environments. These issues are important to resolve if we are to understand how iron, carbon, and nitrogen cycling are linked in deep sea iron systems. To reconstruct microbial interactions that connect iron oxidation with nutrient cycling, we conducted a genome-resolved metagenomics and metatranscriptomics study at Lōʻihi Seamount. We aimed to better understand the balance of metabolic processes and contributions of specific organisms throughout the mat, including organisms like viruses that have not otherwise been surveyed in iron mats. We collected a surficial mat sample to represent aerobic processes, as well as two bulk samples, which include deeper, anaerobic portions of the mats. The surface sample and one bulk sample were preserved in situ for metatranscriptome studies. We used the other bulk sample in a shipboard incubation experiment in which we added Fe(II) and oxygen to stimulate aerobic iron oxidation and monitored the transcriptomic response of the community. Our results reveal a fuller picture of the microbial ecology and geochemical cycling in iron microbial mats, including viral influences on dominant community members, nitrogen cycling by Zetaproteobacteria, and a role for a potential nitrate-reducing iron-oxidizing Candidatus Ferristratum sp., from the uncharacterized DTB120 phylum.",
"discussion": "Discussion Iron-oxidizing microorganisms can strongly influence the biogeochemical cycles of a wide range of elements. Previous studies have focused primarily on the biomineral byproducts, iron oxyhydroxides, which adsorb and coprecipitate various elements (e.g., [ 33 , 83 – 86 ]). However, iron oxidizer physiology and ecological interactions are just as likely to affect biogeochemical cycling. Here, we have explored these interactions within three iron mat samples from Lōʻihi Seamount, comparing community composition and expression in aerobic surface mat (S1) with bulk mat that includes anaerobic niches (S6 and S19). In the S6 bulk mat microcosm, we were able to observe which organisms and processes are most primed to respond to Fe(II). With these samples, we explored the capabilities of individual iron mat members and reconstructed ecological interactions and community impacts on iron, carbon, and nitrogen cycles. The Lōʻihi Seamount iron mat communities are dominated by a few key players, primarily Bacteria, but also include Archaea and viruses (Figs. 1 and 7 ). The microbial community is structured by a gradient of oxygen, which becomes undetectable within the first few mm’s to cm’s of the mat surface [ 21 , 34 ] Depending on oxygen levels, we find microorganisms involved in either aerobic iron oxidation (Zetaproteobacteria) or anaerobic iron oxidation ( Candidatus Ferristratum sp.). These iron oxidizers (Fig. 7 , process 1) form metabolic products (Fe(III), C org , NO 2 − ) that support the metabolisms of flanking community microorganisms, thus either directly or indirectly driving most biogeochemical processes in the mat, including iron reduction, carbon fixation, fermentation, nitrate assimilation, and denitrification (Figs. 2 and 7 ). Fig. 7 Cartoon showing the contribution of different members of the microbial community to iron, carbon, and nitrogen cycling within the aerobic/anaerobic gradient of an iron mat. The Zetaproteobacteria influence nearly every metabolic process in the mat, and support/are supported by a diverse flanking community, including the anaerobic Candidatus Ferristratum sp. The mat community is also supported by other metabolisms, such as oxidation of methane and H 2 . Zetaproteobacteria are known to be autotrophs based on culture work and inference from MAGs [ 19 , 20 , 25 , 77 ], but a definitive link to primary production in iron mats has not been demonstrated. Our work shows that the Zetaproteobacteria are the primary producers in the surface mat, where nearly all gene expression from known carbon fixation pathways can be attributed to RuBisCO Form II gene expression in the Zetaproteobacteria (Fig. 2 ). This suggests that carbon fixation is concentrated near the aerobic surface of actively growing mats. This is consistent with microscopy evidence that mats accrete as iron oxidizers form biomineral stalks, positioning the cells at the surface of the mat [ 21 ]. The iron oxyhydroxide stalk structures of the Zetaproteobacteria contain polysaccharides [ 87 ] and adsorbed organic exudates [ 88 ], thus the biominerals act as reservoirs of organic carbon for use by heterotrophs and fermenters. Viral induced lysis is another key mechanism for organic nutrient recycling, such as the viral shunt that maintains a pool of dissolved organic matter driving oceanic carbon cycling in the water column [ 89 ]. Given the diversity of phage, high abundance of viral transcripts in the metatranscriptomes, and evidence of repeated interaction between viruses and the dominant microbial populations in the mat, it is quite likely that viruses may be playing a similar role mediating carbon bioavailability in the iron mats (Supplementary Tables 3 and 4 ). Zetaproteobacteria are aerobes, as no Zetaproteobacteria culture can grow anaerobically [ 7 , 20 ] and all sequenced Zetaproteobacteria genomes include aerobic terminal oxidase genes [ 24 ]. However, some have the genetic potential for using nitrate reduction to live within an aerobic/anaerobic transition zone. A few Zetaproteobacteria genomes from the S6 bulk microcosm have the dissimilatory nitrate reductase genes napAB (Figs. 4 and 6 ). These organisms may conduct aerobic nitrate reduction, using nitrate as a backup electron acceptor under oxygen limiting conditions [ 90 ]. This is consistent with the concurrent expression of genes for terminal oxidase ( ccoNO ) and nitrate reduction ( napAB ) (Fig. 6 ) within the well-mixed S6 microcosm. Furthermore, different Zetaproteobacteria MAGs encode various other parts of the dissimilatory denitrification pathway (Fig. 6 ). Only the dissimilatory nitrate reductases are known to conserve energy in this pathway, though eNOR may also conserve energy via a proposed proton pump [ 82 ]. Thus, Zetaproteobacteria engaging in only part of the pathway may do so as a means of detoxifying intermediates, such as nitric oxide [ 91 ], or removing nitrite to avoid abiotic iron oxidation and encrustation in an Fe(II)-rich environment [ 38 , 92 ]. Together with evidence of nitrate assimilation (including in [ 19 , 20 ]), these results show that within the Zetaproteobacteria themselves, iron oxidation and nitrate reduction are coupled in multiple ways. Since Zetaproteobacteria are primarily aerobes, this leaves anaerobic niches open for other organisms to thrive. Candidatus Ferristratum was overall the second most abundant taxa within the iron mats (Fig. 1 ) and their high expression of narGH (Figs. 2 and 4 ) and Cluster 1 Cyc2 homologs (Fig. 3 ) suggests they are active anaerobic denitrifiers with the ability to oxidize Fe(II). Given the rapid and parallel responses of cyc2 and narG genes after Fe(II) stimulus (Fig. 5 ), Candidatus Ferristratum sp. may represent a novel nitrate-reducing iron-oxidizing taxon. Although oxygen was available in the microcosm, the injection of Fe(II) may have promoted transient anaerobic niches within the mat material. In the environment, Candidatus Ferristratum sp. are much more abundant in bulk mat samples (Fig. 1 ), including those from previous studies at Lōʻihi Seamount [ 11 , 20 ]. Our findings suggest there may be two partitioned niches of iron oxidation in iron mats: The Zetaproteobacteria oxidizing iron in the shallow, aerobic zone, and Candidatus Ferristratum sp. conducting nitrate-reducing iron oxidation in the primarily anoxic zone. The Lōʻihi iron mats should be an ideal place to find nitrate-reducing iron oxidizers, due to gradients of Fe(II), O 2, and NO 3 - in the mat [ 21 , 34 , 35 ]. Nitrate reduction coupled to iron oxidation is theoretically possible for a single organism, but no isolate aside from the hyperthermophilic Archaea Ferroglobus placidus [ 93 ] has been shown to do so unequivocally [ 38 , 94 ]. Indeed, the model system for nitrate-reducing iron oxidizing bacteria is the KS enrichment culture containing an autotrophic, nitrate-reducing iron-oxidizing Gallionellaceae partnered with heterotrophs that complement the Gallionellaceae’s denitrification pathway [ 94 – 96 ]. At Lōʻihi, Zetaproteobacteria produce organic carbon and consume oxygen, creating a niche for nitrate-reducing iron oxidizers. However, it is clear that Zetaproteobacteria themselves do not fill this niche, though they can assimilate nitrate using electrons from Fe(II) (Fig. 8 , Zetaprotobacteria cell A), and some may conduct aerobic denitrification for redox balance (Zetaproteobacteria cell C). The anaerobic iron oxidation niche is open for another organism, such as the dominant denitrifier Candidatus Ferristratum sp., which may couple nitrate reduction to either iron (Fig. 8 , Candidatus Ferristratum cell A) and/or organic carbon oxidation ( Candidatus Ferristratum cell B). Either way, the high Candidatus Ferristratum sp. narG expression (Figs. 2 and 4 ) suggests that significant quantities of nitrite are produced in the mats. In theory, this could rapidly oxidize Fe(II) through chemodenitrification, which would result in the encrustation of nitrite-producing cells [ 38 , 92 ]. However, Zetaproteobacteria expressed nirK highly, and increased nirK expression in the Fe(II)-amended bulk microcosm after peak narG expression (Fig. 5 ), suggesting that the Zetaproteobacteria actively remove nitrite produced by Candidatus Ferristratum sp. (Fig. 8 , Zetaproteobacteria cell A). Zetaproteobacteria (cell B; Fig. 8 ) and Candidatus Ferristratum sp. (cell B) also express genes to reduce NO via eNOR and cNOR, respectively (Fig. 6 ), suggesting the need to detoxify NO. Finally, some Zetaproteobacteria can complete denitrification from N 2 O to N 2 (cell B; Fig. 8 ). In this partner approach, a nitrate-reducing iron-oxidizing organism is relieved of the burden of having to synthesize all four separate enzyme complexes to denitrify to N 2 because other taxa help with detoxifying byproducts. This cooperative approach to denitrification may explain why it has been challenging to isolate nitrate-reducing iron oxidizers. Fig. 8 Cartoon model showing potential C and N cycling between the aerobic Zetaproteobacteria and anaerobic Candidatus Ferristratum sp. within the gradient of the iron mat. Letters denote either different Zetaproteobacteria taxa or different potential metabolic strategies in a single Candidatus Ferristratum sp. cell. Zetaproteobacteria A (ZA): Capable of Fe(II) oxidation using oxygen, assimilating nitrate, reducing nitrite, and fixing carbon. ZB: Capable of reducing nitrogen intermediates. All have NOR, though ones with NIR don’t have NOS. (see Fig. 6 ). ZC: A few Zetaproteobacteria with NapA may be able to couple nitrate reduction with iron oxidation (these Zetaproteobacteria also capable of the metabolism in ZA). Candidatus Ferristratum A: Capable of nitrate reduction coupled to iron oxidation. Candidatus Ferristratum B: Capable of nitrate reduction coupled to organic C oxidation."
} | 5,065 |
28358919 | PMC5391122 | pmc | 1,005 | {
"abstract": "Metabolic cross-feeding interactions between microbial strains are common in nature, and emerge during evolution experiments in the laboratory, even in homogeneous environments providing a single carbon source. In sympatry, when the environment is well-mixed, the reasons why emerging cross-feeding interactions may sometimes become stable and lead to monophyletic genotypic clusters occupying specific niches, named ecotypes, remain unclear. As an alternative to evolution experiments in the laboratory, we developed Evo 2 Sim, a multi-scale model of in silico experimental evolution, equipped with the whole tool case of experimental setups, competition assays, phylogenetic analysis, and, most importantly, allowing for evolvable ecological interactions. Digital organisms with an evolvable genome structure encoding an evolvable metabolic network evolved for tens of thousands of generations in environments mimicking the dynamics of real controlled environments, including chemostat or batch culture providing a single limiting resource. We show here that the evolution of stable cross-feeding interactions requires seasonal batch conditions. In this case, adaptive diversification events result in two stably co-existing ecotypes, with one feeding on the primary resource and the other on by-products. We show that the regularity of serial transfers is essential for the maintenance of the polymorphism, as it allows for at least two stable seasons and thus two temporal niches. A first season is externally generated by the transfer into fresh medium, while a second one is internally generated by niche construction as the provided nutrient is replaced by secreted by-products derived from bacterial growth. In chemostat conditions, even if cross-feeding interactions emerge, they are not stable on the long-term because fitter mutants eventually invade the whole population. We also show that the long-term evolution of the two stable ecotypes leads to character displacement, at the level of the metabolic network but also of the genome structure. This difference of genome structure between both ecotypes impacts the stability of the cross-feeding interaction, when the population is propagated in chemostat conditions. This study shows the crucial role played by seasonality in temporal niche partitioning and in promoting cross-feeding subgroups into stable ecotypes, a premise to sympatric speciation.",
"conclusion": "Conclusion Using a multi-scale computational model of ISEE, we studied the evolution and stability of cross-feeding interactions in well-mixed environments, providing a single limiting resource periodically or continuously, as in batch cultures or chemostat devices. Our results led us to consider a stable cross-feeding polymorphism as the stable coexistence of different ecotypes, defined as different monophyletic clusters undergoing independent periodic selection events in their own ecological niche [ 14 ]. We observed that, even if cross-feeding polymorphism systematically appears in all the simulations, the evolution of stable ecotypes coexisting via cross-feeding is favored in the periodic environment, similarly to the S/L polymorphism observed in the LTEE [ 11 ]. In the continuous environment, competitive exclusion precludes the stabilization of cross-feeding interactions, in apparent contradiction with wet experiments. Indeed, while ecotypes interacting via cross-feeding can temporarily coexist, a mutant always eventually outcompetes them. Then, we suggest to study the evolution of cross-feeding polymorphism by fully integrating the notion of ecotype, and distinguishing between ecological stability and evolutionary stability, the latter including long-term evolutionary dynamics such as periodic selection. Our results contributed to understand temporal niche partitioning, by modeling various mechanisms such as cross-feeding, niche construction and seasonality. At a more general scale, our results may contribute to the study of the evolution of bacterial communities, by deciphering the conditions of sympatric speciation in asexual populations.",
"introduction": "Introduction Stable metabolic cross-feeding interactions between microbial strains are commonly observed in nature [ 1 – 4 ]. For example, nitrification, an important step of the nitrogen cycle, is carried out in consecutive steps by several bacterial species maintaining cross-feeding interactions [ 3 ]. In laboratory experiments, microbial populations also demonstrated their ability to quickly establish metabolic cross-feeding interactions between morphotypes [ 5 – 13 ]. An important question, at the crossroads between ecology and evolution, is the evolutionary stability of such cross-feeding polymorphisms, because they are often considered to be the first steps toward speciation. According to Cohan [ 14 ], the species concept in bacteria should not rely on the named species of systematics but on the notion of ecotype , which itself relies on the ecological and evolutionary dynamics of the subpopulations. Two bacterial subpopulations may be considered as different ecotypes if they form monophyletic clusters, occupy different ecological niches and if periodic selection purges diversity in one subpopulation independently from the other [ 14 ]. A cross-feeding polymorphism therefore leads to adaptive diversification and ultimately to speciation when it is stable enough to resist the invasion of a mutant that would otherwise take over the whole population. If the environment is spatially structured, the stabilization of new ecotypes that emerged after an adaptive diversification event is facilitated by the locality of environmental conditions and frequency-dependent interactions. This mechanism of allopatric (or micro-allopatric) divergence is well-known, since ecotypes can escape competitive exclusion in their local niches [ 14 ]. For example, Pseudomonas fluorescens populations have been shown to produce adaptive diversification events in spatially heterogeneous environments, but not in homogenized conditions [ 5 , 6 ]. Microbial populations can also exhibit adaptive diversification in sympatry, when the environment is homogeneous with a single carbon source. In this case, the stability of ecotypes is maintained by frequency-dependent interactions, often due to cross-feeding interactions, as observed in the Long-Term Evolution Experiment with Escherichia coli (LTEE [ 15 ]). In this ongoing experiment, 12 populations are being independently propagated in a constant glucose-limited environment in batch culture since 1988. The experiment reached 66,000 generations at the time of this writing. Every day, 1% of the population is transferred into fresh medium such that each population experiences a daily cycle of feast and famine phases. In one of the 12 populations, a long-term polymorphism has been observed [ 11 ]. Two ecotypes, named S and L (for Small and Large, related to their respective colony sizes on plate), evolved from a common ancestor before generation 6,500. The L ecotype grows efficiently on glucose, while the S ecotype mainly grows on acetate, a by-product secreted by L [ 16 ]. Experiments showed that the interaction between S and L ecotypes relies on negative frequency-dependent selection, each ecotype having a selective advantage when rare. This balanced polymorphism is now stable for more than 55,000 generations [ 11 ]. It was also shown that S and L ecotypes specialized in their own niches, the L ecotype increasing its ability to grow on glucose but not on acetate, and conversely for the S ecotype [ 16 ]. The evolutionary stability of this polymorphism may be explained by the temporal niche partitioning that arises from the periodic transfers into fresh medium [ 17 ]. A first season starts immediately after a transfer, when the environment contains mostly glucose. The L ecotype grows during this season, consumes glucose and secretes acetate, thereby generating a second season where the environment contains mostly acetate and supports the growth of the S ecotype. Yet several experiments have shown that microbial populations can also evolve cross-feeding interactions in a chemostat in a few tens of generations [ 7 , 8 , 10 ]. Those interactions appear to be stable over a few hundreds of generations [ 7 , 8 , 10 ]. In chemostat, there is no obvious spatial or temporal niche partitioning and it is thus intriguing that the dynamics predicted by the competitive exclusion principle has not been observed so far. Indeed, one would expect a mutant to eventually appear, which would either completely degrade glucose or feed on both glucose and acetate, thereby outcompeting the specialized ecotypes. It has been proposed that energy constraints and flux optimization principles prevent competitive exclusion, thereby stabilizing the polymorphism [ 18 , 19 ]. However, experimental evolution in chemostat has generally been performed for only a few hundreds of generations (up to 1,900 generations in [ 7 ]), precluding the possibility to confirm this statement on a longer term. Thus, as a step to better understand how cross-feeding, niche construction and seasonality contribute to microbial diversification, we addressed here the following question: What makes emerging cross-feeding interactions stable in the long-term, in single carbon source batch culture or chemostat experiments? While experimental evolution provides a very precise picture of evolution, it remains a long and costly process. An alternative approach consists in simulating evolution in a computer. In Silico Experimental Evolution (ISEE), where digital organisms are evolved for tens of thousands of generations, reproduces the environmental conditions of experimental evolution [ 20 ]. Like in the wet approach, it is possible to simulate several independent populations to understand the respective importance of general laws and historical contingencies. In addition, ISEE provides an exhaustive fossil record and, more importantly, allows for “impossible experiments” [ 21 ], like saving the fitness at full resolution for tens of thousands of generations, or changing any parameter (mutation rates, environment fluxes) at will. We developed Evo 2 Sim, a multi-scale computational model of in silico experimental evolution. Evo 2 Sim allows us to address many questions raised by experimental evolution [ 20 ]. Typically, we can use it to investigate how evolution shapes the different organization levels of an organism ( e.g. , genome size, complexity of the regulation network and metabolic network) and of an ecosystem (polymorphism, speciation) depending on global parameters such as environmental conditions or mutation rates. Here, we tested which environmental conditions can lead to stable adaptive diversification events, by reproducing the resource dynamics of experimental evolution setups like chemostat and batch culture. Previous mathematical works have already studied the conditions of interspecific coexistence via resource partitioning [ 22 ], and of cross-feeding interactions [ 18 , 23 , 24 ], during one or more competition episodes. Stewart and Levin [ 22 ] studied the conditions of coexistence of several ecotypes in batch culture and chemostat. However, they focused on a single episode of competition between preexisting strains without modeling a random mutational process. Moreover, the strains were not allowed to cross-feed on by-products of other strains. Rozen et al. [ 13 ] and Ribeck and Lenski [ 25 ] modeled analytically the cross-feeding interaction between S and L ecotypes in the LTEE, showing the existence of negative frequency-dependence in batch conditions. These models also did not include a mutational process. Gudelj and colleagues [ 24 ] studied the short-term dynamics of two competitors in various environmental conditions including batch and chemostat, and showed that stable cross-feeding was possible, depending on initial competitors frequency and resource abundance. Again, this model did not include the mutational process. Other mathematical studies introduced a simplified evolutionary dynamics, by computing successions of competition episodes and introduction of fit mutants. For example, Pfeiffer and Bonhoeffer [ 18 ] studied the conditions of emergence of stable cross-feeding in chemostat conditions, when a trade-off on ATP production is introduced on abstract metabolic pathways. Doebeli [ 23 ] compared the conditions of emergence of cross-feeding polymorphism in chemostat and batch culture. The authors concluded that the evolution of cross-feeding is more likely in chemostat than in batch culture. However, this model forced a trade-off between consumption rates of glucose and acetate, forbidding the emergence of a generalist mutant. Two rates are evolvable but only the glucose consumption rate is mutable, as the acetate rate is deduced from the glucose rate. The rate at which acetate is secreted is constant ( i.e. , it does not depend on glucose consumption, which could affect the generality of the conclusions). Thus, none of the previous models take into account a realistic random mutational process, and none of them explicitly models the genomic level. Indeed, it is difficult to include a competition process as well as realistic mutational dynamics in a single mathematical model. Another approach consists in simulating evolution with individual-based models. Computational models of in silico experimental evolution have already been used to explore the evolution of cross-feeding interactions. Johnson and Wilke [ 26 ] used the Avida software [ 27 ] to study the evolution of resource competition between two digital species coexisting via mutualistic cross-feeding in a closed environment, with only two possible metabolites. However, they did not test the influence of the environmental dynamics. Williams and Lenton [ 28 ] used an individual-based evolutionary model to explore the stability of connected ecosystems undergoing cross-feeding and “evolutionary regime shifts”. Yet, the genotype-to-phenotype mapping of their organisms was rather simple (fixed size arrays defining the affinity of the organism for each resource), thus not allowing to study the effects of ecological dynamics on genome and metabolic network structures. Crombach and Hogeweg [ 29 ] and Boyle et al. [ 30 ] studied the evolution of resource cycling and its stability. In the first model [ 29 ], the resource cycling was imposed by the system. In the second model [ 30 ], the environment was strongly structured (patches of individuals with random migration events), such that it was not possible to study sympatric diversification. Chow and colleagues [ 31 ] used Avida [ 27 ] to explore the relation between productivity and diversity in a digital ecosystem under mixed influx of nine pre-defined resources, while Gerlee and Lundh [ 19 ] explained the maintenance of cross-feeding interactions in a microbial population by energy and efficiency constraints on metabolic fluxes. To do so, they developed an individual-based model evolving simple binary strings, thereby precluding evolvable interactions between the different organization levels of an organism, and their possible effects on the ecological dynamics. Gerlee and Lundh [ 32 ] also related ecosystem productivity to energy-uptake efficiency, with the same type of individual-based model as in [ 19 ]. Recently, Liu and Sumpter [ 33 ] used an individual-based model evolving artificial ecosystems relying on a “number soup”: In this model, each species perform one modular addition transforming specific numbers into others, immediately available for other species. With their model, authors showed that artificial ecosystems always self-organize to consume all the available resources. While stable cross-feeding, and reciprocal cross-feeding, are common evolutionary outcomes in their model, authors also show that whole population extinctions sometimes occur, even without external perturbations. Yet, the absence of complex and evolvable genotype-to-phenotype map in their model precludes the possibility to get insights into the influence of ecosystem evolution on the structure of the organisms. Finally, Großkopf et al. [ 16 ] predicted the adaptive diversification event leading to S and L ecotypes in the LTEE, by mixing flux balance analysis (FBA) and in silico evolution in a single model. By modeling the evolution of reaction rates in the metabolic network of Escherichia coli , they demonstrated that the emergence of a stable cross-feeding similar to S and L interaction is highly probable in the LTEE conditions. However, in their model, digital organisms are highly constrained (there is no innovation, e.g. new by-products cannot appear in the evolutionary process). To the best of our knowledge, none of these individual-based models compared the evolution of stable cross-feeding in different experimental setups, such as batch culture or chemostat. To sum up, we were not able to find in the literature models that combine: (i) an explicit mutational process along with the modeling of natural selection and drift, (ii) evolvability at all organization levels (genome structure, metabolic network, number of reactions, number of metabolites, reaction rates, …), and (iii) a comparison between batch culture and chemostat. Our results show that stable cross-feeding interactions are favored in batch culture, owing to the seasonality of the environment. In continuous culture, the absence of seasonality precludes niche construction and leads to competitive exclusion, even if the population is initially composed of two ecotypes maintaining frequency-dependent interactions. We also demonstrate that the long-term evolution of a stable cross-feeding interaction in batch culture leads to character displacement [ 16 , 34 ], at the level of the metabolic network but also of the genome structure. This difference of genome structure between the two ecotypes has an impact on the further stability of the cross-feeding interaction when the population is propagated into continuous culture. Model Evo 2 Sim is a multi-scale and individual-based computational model. Digital bacterial-like organisms own a coarse-grained genome that contains genomic units encoding a simplified metabolic network. The organisms evolve on a two-dimensional toroidal grid (the environment), uptaking, transforming and releasing metabolites, and dividing in the presence of empty spots or dying. Extracellular metabolites diffuse across the grid spots. In this model, metabolites are implicit molecules identified by a tag ∈ N * . The model is described in more details below, and summarized in Fig 1 . The source code is written in C++. All the material necessary to replay experiments (software, parameter files, strain backups, …) is freely available at http://www.evoevo.eu/adaptive-diversification-simulations/ . The latest version of Evo 2 Sim is available at http://www.evoevo.eu/evo2sim-software/ . 10.1371/journal.pcbi.1005459.g001 Fig 1 Presentation of the model. The genotype-to-phenotype mapping, as well as the population and environment, are schematized here. (A) Description of the genotype-to-phenotype mapping. Organisms own a coarse-grained genome that contains genomic units. (A.1) Non-coding units (NC, grey circles) are not functional. The arrangement of the genomic units on the circular single strand defines functional regions, where a promoter (P, blue cross, A.2 ) controls the expression of all contiguous enzyme units (E, red circles), thereby allowing for operons. (A.3) When enzyme units are expressed, they contribute to the metabolic network. (A.4) Enzymes perform metabolic reactions in the cytoplasm, or pump metabolites in or out (see the description of the metabolic network below). The score of an organism is computed from its “essential metabolites” (see the description of the score function below). Lethal toxicity thresholds are applied to each metabolic concentration and preclude organisms to accumulate resources. (B) Description of the population and environment levels. Organisms are placed on a 2D toroidal grid, and compete for resources and space. (B.1) When an organism dies, it leaves its grid spot empty and organisms in the Moore neighborhood (if any) compete to divide in the available spot. The competition is based on scores, a minimal threshold being applied on scores to preclude worst organisms to divide. At division, daughters share cytoplasm content (enzymes and metabolites). At death, metabolites from the cytoplasm are released in the local environment and diffuse on the grid. (B.2) At the largest scale, the population evolves in the environment by uptaking, transforming and releasing metabolites. Metabolites then diffuse and are optionally degraded. This interaction between the population and its environment allows for the evolution of complex ecological situations. Genome structure The genome is a circular single-stranded sequence of genomic units, inspired from [ 35 , 36 ]. Genomic units belong to three different types: non-coding units (NC), promoter units (P), and enzyme coding units (E). The order of the units in the genome determines the existence of functional regions, meaning that not all sequences of units are functional. The functional regions of a genome are those that have the following pattern: a promoter (P) followed by one or more enzyme coding units (E). A promoter can thus control several coding units, as bacterial operons. The first genomic unit that is not enzyme coding interrupts transcription and marks the end of the functional region. Non-coding units (NC) have no particular function. They constitute the non-coding part of the genome. Promoter units (P) contain a floating-point number β ∈ [0.0, 1.0] representing the production rate of the protein(s) depending on the promoter. All the parameters and their units are listed in S1 Table . Enzyme coding units (E) contain two integers s and p ∈ N * , indicating the tag of the substrate and product respectively, two floating-point numbers k cat ∈ ±[10 −3 , 10 −1 ], and the ratio k cat / K M ∈ [10 −5 , 10 −3 ] describing the enzymatic kinetics (see the description of the metabolic network below). In the special case where s = p , the enzyme is considered as a pump, actively pumping in (or out) the metabolite s if k cat is positive (or negative, respectively). Initial genomes of 50 genomic units are generated. These genomes contain ten P and ten E, all with random positions and attribute values. Upon cell division, the parental genome is replicated with mutations in the two daughter cells. Each genomic unit can undergo point mutations, meaning here changes in the numbers it contains, like the values of s , p , k cat and k cat / K M for an E. Each unit attribute mutates at a rate of 10 −3 per attribute per replication. For the substrate/product tags, a mutation consists in randomly incrementing/decrementing s or p respectively. For k cat or k cat / K M , a random number drawn from N ( 0 , 0 . 1 ) is added to the decimal logarithm of the parameter. β mutates by adding a random number drawn from N ( 0 , 0 . 1 ) . A genomic unit can also undergo a type transition from any unit type to any other at a predefined rate, set here to 10 −3 per genomic unit per replication. All types of genomic units are actually implemented as a tuple containing all possible attributes, like (unit_type, β , s , p , k cat , k cat / K M ). The unit type tells us which parameters are functionally relevant and the others are free to mutate neutrally. The genome can also undergo rearrangements affecting segments of any number of genomic units. There are four types of rearrangements: duplications, deletions, translocations and inversions. All rearrangement rates are set to 10 −3 per genomic unit per replication, hence the number of rearrangements is related to the genome size thereby limiting genome expansion [ 37 ]. The breakpoints for each rearrangement are randomly drawn in the whole genome. In real genomes, spontaneous rearrangement breakpoints have no reason to lie exactly between two of our genomic units and could thus break our genomic units. To model that with our coarse-grained genome representation, we alter the content of the two genomic units that are adjacent to a rearrangement breakpoint. Suppose for example that a deletion joins two genomic units, one containing the attributes (unit_type 1 , β 1 , s 1 , p 1 , k cat 1 , ( k cat / K M ) 1 ) and the other the attributes (unit_type 2 , β 2 , s 2 , p 2 , k cat 2 , ( k cat / K M ) 2 ). Then for each attribute, there is a probability of 10 −3 for the value in unit 1 to be exchanged with the value in unit 2. Both units could for example exchange their values of s , thereby leading to (unit_type 1 , β 1 , s 2 , p 1 , k cat 1 , ( k cat / K M ) 1 ) and (unit_type 2 , β 2 , s 1 , p 2 , k cat 2 , ( k cat / K M ) 2 ). Metabolic network Gene products can either be pumps, pumping metabolites from or to the growth medium, or enzymes performing catalytic transformations in the metabolic space. Let us consider an enzyme in the cytoplasm, that catalyzes one specific reaction s → p , with s ∈ N * and p ∈ N * being the substrate and the product of a Michaelis-Menten-like reaction, respectively. The variation in concentrations [ E ], [ s ] and [ p ] over time are then driven by Eq 1 :\n d [ E ] d t = β - ϕ [ E ] d [ s ] d t = - k c a t [ E ] [ s ] K M + [ s ] d [ p ] d t = k c a t [ E ] [ s ] K M + [ s ] (1) \nwhere β is the basal production rate specified in the promoter unit, ϕ is the enzyme degradation rate (set to 0.1 per centi-time-step for all enzymes here, with 1 centi-time-step = 0.01 time-steps), K M and k cat are the kinetic attributes of the enzyme ( K M being deduced from k cat and k cat / K M attributes). Pumps are treated here as special enzymes for which [ s ] and [ p ] describe the internal and external concentrations of the same metabolite. If k cat is positive (resp. negative), [ s ] is the external (resp. internal) concentration of the metabolite and [ p ] the internal (resp. external) concentration. The dynamics of metabolic concentrations [ s ] and [ p ] are thus also driven by Eq 1 when the gene product is a pump. Each organism has an ODE (Ordinary Differential Equation) system that keeps track of: (i) the concentrations of all metabolites inside the organism, i.e. , internal concentrations, (ii) the concentrations of all metabolites at the organism’s location on the grid, i.e. , external concentrations, and (iii) the concentrations of all proteins (pumps and enzymes) in the cytoplasm. For a very simple organism whose genome merely encodes one pump importing metabolite #10 into the cell, and one enzyme converting #10 to #7, the ODE system would read:\n d [ Pump ] d t = β Pump - ϕ [ Pump ] d [ Enzyme ] d t = β Enzyme - ϕ [ Enzyme ] d [ # 10 external ] d t = - k c a t Pump [ Pump ] [ # 10 external ] K M Pump + [ # 10 external ] d [ # 10 internal ] d t = k c a t Pump [ Pump ] [ # 10 external ] K M Pump + [ # 10 external ] - k c a t Enzyme [ Enzyme ] [ # 10 internal ] K M Enzyme + [ # 10 internal ] d [ # 7 internal ] d t = k c a t Enzyme [ Enzyme ] [ # 10 internal ] K M Enzyme + [ # 10 internal ] (2) The number of equations in the ODE system generally differs across individuals within a population because it depends on the number of functional genes, and chromosomal rearrangements like duplications and deletions can alter gene number. In practice, the size of the ODE system goes from tens to thousands of equations depending on the individual. Similarly, the parameter values of the ODE system also vary across individuals, as they are encoded in the organism’s genes and thus result from the mutation process. Initially, in the individuals used to seed a run at time-step 0, each protein starts at its equilibrium concentration β / ϕ , and each metabolite starts with an internal concentration of 0.0 ACU (Arbitrary Concentration Unit). At time-step 0, for all grid spots, external concentrations are initialized to 0.0 ACU for all nutrients except for metabolite #10 (the exogenous carbon source). Between time-steps 0 and 1, the ODE system computes the dynamics of the metabolite and protein concentrations, using the adaptive Runge-Kutta-Cash-Karp method (RKCK), during 100 centi-time-steps. In organisms that possess a pump for metabolite #10, this metabolite will enter the cell. If the genome of this organism also encodes an enzyme to transform #10 into #7 for example, then the internal concentrations will show an accumulation of #7. At time-step 1, each organism will either die, divide or just survive (see paragraph “Population and environment” below for details). If the organism merely survives without dividing, its current internal concentrations are used as initial conditions for the computation of the next 100 centi-time-steps ( i.e. , for the transition from time-step 1 to 2). If the organism divides, each of the two daughter cells inherits half of each metabolite and each protein amounts. These will constitute the initial conditions for each cell’s ODE system for the next 100 centi-time-steps. If the organism dies, its internal content is released into the environment, thereby increasing the local external concentrations. As the metabolites can diffuse across the grid, the metabolites produced by the dead cell, like metabolite #7, will become available to the neighboring cells, which will thus be able to feed on both #10 and #7, if they own the corresponding pumps. This process is repeated for each transition from time-step t to time-step t + 1. Thus, when e.g. , a 32 × 32 grid is full of organisms (see the description of the experimental protocol below), a time-step involves the computation of about a thousand different ODE systems, each of them containing from tens to thousands of equations depending on gene number. Score function Some metabolites are essential for an organism’s replication. Here, we arbitrarily define as essential the metabolites whose tag is a prime number. The score of an organism is then simply defined as the sum of its internal concentrations of essential metabolites. However, to prevent organisms from producing a single specific prime number in huge quantities, we also define lethal toxicity thresholds for both essential and non essential metabolites. Here these toxicity thresholds are set to 1.0 ACU for all metabolites. Population and environment Organisms evolve on a two-dimensional toroidal grid, each spot containing at most one organism. The physical environment is described at the grid level: each grid spot contains external metabolites, each with its concentration. These external metabolites diffuse with a diffusion parameter D = 0.1 gridstep 2 .time-step -1 , meaning that a fraction D of each metabolite present at one location will diffuse to each of the eight neighboring grid spots at each time-step. The discrete diffusion equation we are using is inspired from [ 38 ]. External metabolites are also degraded with a degradation rate D g , meaning that a fraction D g of each metabolite at each location will disappear at each time-step. We make the simplifying assumption that there are no enzymatic reaction in the environment, and thus that metabolite transformation only occurs inside the organisms. Organisms compete for the external metabolites to produce offspring in empty spots. They interact with their local environment by pumping metabolites in and out and releasing their metabolic content at death. At each time-step, organisms are evaluated and either killed, updated or replicated depending on their current state: If the organism does not die and cannot divide ( e.g. , because there is no free space in its neighborhood), its metabolic network is updated, and its score is computed. If lethal toxicity thresholds are reached, the organism dies (see point 2); Organisms can also die randomly with a probability following a Poisson law of parameter p death = 0.02 per organism per time-step. At death, the metabolic content is released into the local environment; For each empty grid spot, all living organisms in the Moore neighborhood whose score is higher than a minimum score of 10 −3 ACU compete. The organism having the best score in the neighborhood is allowed to divide if it did not replicate previously at the same time-step (such that any dividing cell generates at most two daughters per time-step). Experimental protocol In all our simulations, the environment provided one primary resource with tag m exo = #10. To initialize an evolutionary run, the entire grid was populated with individuals having random genomes (different for each individual). This initial population was allowed to evolve for 500 time-steps, at which point its viability is assessed. We repeated this procedure until a viable population was found, i.e. , with at least 500 viable individuals after the 500 time-steps. In this case, some organisms possess at least one pump to internalize m exo , and (because m exo is not a prime number, see the description of the score function above) one enzyme to transform m exo into a prime number, thereby producing an “essential metabolite”. Up to a few hundred trials were usually needed to find a viable population, which was then used to seed the evolutionary run. Each evolutionary run was seeded with a different viable population. These organisms grow on the primary resource and start to release by-products (mostly at death), hence modifying their environment. Populations evolved in two different environments: The periodic environment , in which the resource dynamics of the LTEE [ 15 ] was mimicked. The environment was periodically refreshed by removing all the external metabolites and introducing m exo at concentration f in = 10.0 ACU per grid spot. Internal metabolites were not affected by the refresh event. The refresh period was Δ t = 333 time-steps. We call a “cycle” this time interval between two environmental resets. The value of Δ t was calibrated to let the organisms live for approximately 7 generations per cycle, as in the LTEE. Within each cycle, the metabolites in the environment were conserved ( D g = 0 per time-step). Note that we mimicked the resource dynamics of the LTEE but not the 1% population subsampling occurring during serial transfers, because it would have implied transferring populations of 10 individuals or fewer. Such a low population size would have implied dramatic genetic drift and impeded adaptive evolution (in the LTEE, where the population size before sampling is very large, the 1% subsampling still leaves the population large enough to keep genetic drift reasonably low). To simulate subsampling, a significantly larger grid would have been needed, making the whole campaign impossible to compute in a reasonable time. The continuous environment , in which the resource dynamics of a chemostat environment was mimicked. The medium was constantly provided with a small influx of the primary resource. All the external metabolites were slowly degraded. Specifically, at each time-step, a concentration Δ f in = 0.03 ACU of m exo was added in every grid spot, and external metabolites were degraded at rate D g = 0.003 per time-step. For each environment, 12 independent populations were propagated for 500,000 time-steps (approximately 50,000 generations). On the long-term, the quantity of resources available in the system was equivalent in both environments. The grid size is 32 × 32. Complementary experiments were also run in a randomized batch environment similar to the periodic environment except that the environment reset intervals followed a Poisson law of parameter Δ t = 333 time-steps instead of the exact regular period of 333 time-steps. The simulation parameters common to all the simulations are described in S1 Table . Cross-feeding interactions In order to detect the potential cross-feeding interactions in the population, the metabolic activity of each individual was evaluated at each time-step. For each organism, a “trophic profile” was computed from its metabolic network activity. The trophic profile is a binary sequence summarizing the uptake, production and release activity of an organism. The length of the binary string was defined by the largest metabolite tag present in the system at time t . For example, if an organism uptakes metabolite #4, produces #3 from #4 and releases #3, knowing that the largest metabolite tag in the whole grid is #5, then its profile is |00010|00100|00100|. We classified organisms in two trophic groups depending on their trophic profiles: “Group A” pumps in m exo , and possibly other metabolites, “Group B” pumps in group A by-products, and possibly other metabolites, but not the primary resource m exo . A trophic group is considered an ecotype if the organisms of the group form a monophyletic cluster (see below). Phylogenetic relationships Phylogenetic relationships were exhaustively recorded during each simulation. Since organisms can only divide once per time-step, phylogenetic trees are binary trees. It was possible to recover the line of descent of any organism, and to compare the phylogenetic tree structure with the distribution of the trophic groups in the population. In particular, we can determine if groups A and B are monophyletic, and thus can be considered as ecotypes. To this aim, we computed a phylogenetic structure score ( PS score) to identify the degree of monophyly of both groups. This phylogenetic structure score was defined as PS = | f 1 − f 2 |, where f 1 and f 2 are the relative frequencies of group B in both subtrees rooted to the last common ancestor of the whole final population. A high PS value indicates a strong clustering of groups A and B in the phylogenetic tree, i.e. , that groups A and B are two different ecotypes. A low PS value indicates a random distribution or the absence of polymorphism. Sensitivity analysis We tested variations of our parameters set (see S1 Table ), by changing the death probability p death , the external metabolites diffusion rate, the mutation rates, the toxicity thresholds, the “migration rate” (a parameter controlling the fraction of exchanged pairs among all possible pairs of individuals), and the grid size. Details and results are described in S1 Appendix .",
"discussion": "Discussion Using in silico experimental evolution, we have shown that the long-term maintenance of cross-feeding interactions is favored in a seasonal environment, where the environment is reset and primary resource is supplied at regular intervals. In this environment, 5 simulations over 12 evolved a stable cross-feeding interaction at the end of the simulations, with two monophyletic ecotypes coexisting via a negative frequency-dependent interaction. At each cycle, ecotype A grows during the first season, feeding on the primary resource and releasing by-products, while ecotype B exclusively feeds on by-products during the second season. The stable coexistence of ecotypes A and B is then based on niche construction, followed by a negative frequency-dependent interaction, as the S and L ecotypes in the LTEE. According to our model, batch culture experiments seem to especially favor the evolution of stable cross-feeding polymorphisms owing to the cyclic nature of the environment that generates the conditions for the existence of at least two stable seasons: a first season is externally generated by the cyclic mechanism (thus being intrinsically stable) while the second one is generated by the replacement of the exogeneously-provided nutrient by the secreted by-products through a mechanism of niche construction. In the continuous environment, where the primary resource is constantly provided (like in a chemostat), cross-feeding interactions emerged, but were not stable because of competitive exclusion. In this case, organisms enriched their environment via their metabolic activity, such that mutants were temporarily able to feed on by-products. But the absence of seasonality precludes any possibility for the stabilization of cross-feeding interactions. Our multi-scale model allowed us to investigate the impact of resource dynamics on the organization of genome ( e.g. , gene amplification) and of the metabolic network. It also allowed us to dissect the precise mechanism behind the evolved robustness of the cross-feeding interaction. We demonstrated that those results are robust to model parameter variation. Indeed, stable cross-feeding interactions emerged in the periodic environment for a wide range of parameter values, including well-mixed populations and infinite diffusion rate, while they never appeared in the continuous environment, thus reinforcing our conclusions ( S1 Appendix ). Previous wet experiments in chemostat demonstrated the emergence of cross-feeding interactions [ 7 , 8 , 10 ]. In those experiments, E. coli populations have been propagated in a chemostat with glucose as a single limiting carbon source for at most 1,900 generations. When isolated and evolved together in competition experiments, the different mutants identified to contribute to the cross-feeding interactions reached a stable equilibrium owing to frequency-dependent interactions [ 8 ]. Several reasons were invoked to explain why cross-feeding interactions could be stable in chemostat, despite the competitive exclusion principle. According to Pfeiffer and Bonhoeffer [ 18 ], cross-feeding may evolve in microbial populations as a consequence of the maximization of ATP production, and the minimization of enzyme concentrations and intermediate products. Those constraints may hinder the emergence of mutants completely degrading glucose (or uptaking glucose and acetate), and outcompeting other cells by competitive exclusion. In our model, organisms do not need to explicitly produce energy carriers. However, competition for resources, toxicity thresholds and division impose metabolic flux optimization. Based on the same conclusions, Doebeli [ 23 ] also suggested that this trade-off between uptake efficiency on the primary and the secondary resources should favor the emergence of cross-feeding polymorphism in chemostat but not in batch culture, because in a chemostat, by-products are more abundant and constantly provided. However, the limit of this model is that the rate of by-product production did not rely on the rate of primary resource consumption. Besides, a more recent theoretical work concluded that, in a continuous and well-mixed environment, the diversity of cross-feeding polymorphism was negatively correlated with primary resource abundance [ 32 ]. Our results shed a new light on this question. First, in our model, cross-feeding polymorphisms emerged both in the periodic and continuous environments. However the stabilization of the cross-feeding interactions was favored in the periodic environment, leading to the evolution of specialized ecotypes. Cohan [ 14 ] defined an ecotype as an independent monophyletic cluster occupying a specific ecological niche. Ecotypes are at the heart of the bacterial species concept: what makes the genetic cohesion of an asexual bacterial species is periodic selection that regularly purges the genetic diversity in the same ecological niche [ 14 ]. As a consequence, ecotypes occupying different niches independently experience selective sweeps, the mutants from one niche not invading the ones from the other niche. Thus, the stability of a cross-feeding polymorphism should only be analyzed in the light of the robustness of each ecotype against selective sweeps by other ecotypes [ 14 ]. This mechanism is observed in the LTEE, as well as in our model. In the periodic environment, ecotypes A and B independently experience periodic selection events. In the continuous environment, competitive exclusion implied that only one ecotype evolved in this environment. Secondly, when ecotypes A and B evolved in the periodic environment were transferred in the continuous environment, they retained their negative frequency-dependent interaction for hundreds of generations, until a selective sweep purged the whole population diversity, and destroyed the cross-feeding interaction. Moreover, ecotypes A and B that evolved for a long time in the periodic environment had a more robust interaction in continuous conditions, because of niche specialization and character displacement on the long-term. In the light of those results, we suggest to distinguish between ecological stability and evolutionary stability. Even if different monophyletic clusters, related by cross-feeding interactions, have frequency-dependent interactions, they are not necessarily robust to competitive exclusion on the long-term. In this sense, ecotypes A and B are no longer ecotypes in the continuous environment. By contrast, in the periodic environment, A and B ecotypes can be considered as proto-species. Those remarks lead us to hypothesize that the S and L interaction observed in the LTEE, which is still at an early stage, should not be stable in a chemostat on the long-term, even if it could become more and more stable. We also hypothesize that the S/L polymorphism is an ongoing speciation event. On the long run, the S ecotype could even loose the ability to consume glucose. In a more general view, what we observed is strongly related to known results about temporal niche partitioning in ecology [ 17 ]. Bacterial communities commonly undergo adaptive diversification or niche specialization in sympatry, when the environment is seasonal. For example, this mechanism has been observed in marine microbial communities [ 42 ], and in lake phytoplankton [ 43 ]. In the LTEE [ 13 ] and in our model, seasonality of glucose originates from the serial transfer, but the seasonality of acetate is due to cross-feeding and niche construction. Moreover, we demonstrated in our model than negative frequency-dependent cross-feeding is not enough to stabilize the interaction between multiple ecotypes. External factors are necessary, such a regular serial transfer. While the environment is intentionally simplified in those experiments, we can expect much more complex environmental conditions in nature. Such complex interactions between external factors, emergent cross-feeding interactions and niche construction are therefore of primary importance to understand the evolution of microbial communities in well-mixed environments. Using a computational model of ISEE to decipher those interactions seems to be a rich complementary approach to wet experiments and mathematical modeling."
} | 11,719 |
33627505 | PMC8544882 | pmc | 1,006 | {
"abstract": "ABSTRACT There is a growing interest in the endolithic microbial biofilms inhabiting skeletons of living corals because of their contribution to coral reef bioerosion and the reputed benefits they provide to live coral hosts. Here, we sought to identify possible correlations between coral interspecific patterns in skeletal morphology and variability in the biomass of, and chlorophyll concentrations within, the endolithic biofilm. We measured five morphological characteristics of five coral species and the biomasses/chlorophyll concentrations of their endolithic microbiome, and we compare interspecific patterns in these variables. We propose that the specific density of a coral’s skeleton and its capacity for capturing and scattering incident light are the main correlates of endolithic microbial biomass. Our data suggest that the correlation between light capture and endolithic biomass is likely influenced by how the green microalgae (obligatory microborers) respond to skeletal variability. These results demonstrate that coral species differ significantly in their endolithic microbial biomass and that their skeletal structure could be used to predict these interspecific differences. Further exploring how and why the endolithic microbiome varies between coral species is vital in defining the role of these microbes on coral reefs, both now and in the future. IMPORTANCE Microbial communities living inside the skeletons of living corals play a variety of important roles within the coral meta-organism, both symbiotic and parasitic. Properly contextualizing the contribution of these enigmatic microbes to the life history of coral reefs requires knowledge of how these endolithic biofilms vary between coral species. To this effect, we measured differences in the morphology of five coral species and correlate these with variability in the biomass of the skeletal biofilms. We found that the density of the skeleton and its capacity to trap incoming light, as opposed to scattering it back into the surrounding water, both significantly correlated with skeletal microbial biomass. These patterns are likely driven by how dominant green microalgae in the endolithic niche, such as Ostreobium spp., are responding to the skeletal morphology. This study highlights that the structure of a coral’s skeleton could be used to predict the biomass of its resident endolithic biofilm.",
"conclusion": "Conclusions. Here, we identify correlations between interspecific variability in coral endolithic microbial biomass and skeletal morphology. Our results suggest that differences in skeletal density and a coral species’ capacity for capturing and scattering light on the outer surface of its skeleton influence the variability in skeletal microbial biomass, predominantly of dominant euendolithic green algae (e.g., Ostreobium spp.). Further studies exploring the responses of individual endolithic taxa to different aspects of coral skeletons are essential for defining the variable ecological role of this microbiome on coral reefs and identifying coral types that may be differentially affected (both positively and negatively) by the endolithic community.",
"introduction": "INTRODUCTION Marine endolithic (i.e., living within rock) microbial biofilms are cosmopolitan in their distribution and inhabit a wide range of substrates, including the shells and skeletons of calcifying organisms such as hard corals, bivalves, and foraminifera ( 1 – 4 ). The microbes in these communities can be subdivided based on their niche endolithic lifestyle ( 1 ): euendoliths actively and chemically bore into rock, cryptoendoliths grow within structural cavities within porous rocks, and chasmoendoliths inhabit fissures and cracks. Although they are not active borers, cryptoendoliths and chasmoendoliths nonetheless degrade substrates through the alteration of pore water chemistry (i.e., biocorrosion) ( 5 – 7 ). When these organisms colonize the shell or skeleton of a marine calcifier, they can be parasitic and reduce the survival and growth of the host by degrading the its protective shell ( 3 , 8 ) or creating lesions ( 9 , 10 ). But they can also be beneficial, for example, by offsetting high temperature or light stress to the host ( 11 – 13 ) or by providing nutrients during times of environmental stress ( 14 – 16 ). This dual capacity for symbiosis and parasitism ( 12 ) underlies the growing interest in the potential role of endolithic biofilms in the future sustainability of marine calcifiers under climate change ( 3 , 17 – 19 ). The influence of endolithic biofilms on host responses to environmental change is of keen interest for research focusing on hard corals (Scleractinia, Cnidaria) which act as ecosystem engineers for tropical coral reefs ( 17 , 18 , 20 , 21 ). The endolithic microbiome of hard corals is diverse and composed of both eukaryotic (microalga and fungi) and prokaryotic (cyanobacteria, bacteria, and archaea) microbes, as well as marine viruses ( 18 , 22 ). Evidence to date indicates that the biomass of these microbial assemblages is predominantly composed of euendolithic siphonous unicellular algae in the genus Ostreobium ( 17 , 21 ). Microfungi ( 4 , 23 ) and cyanobacteria are also important bioeroders and often the pioneer colonizers of newly denuded substrates ( 24 ). Other important functional groups within the coral endolithic microbiome include diazotrophic and other bacterial taxa ( 25 ) which participate in nitrogen and phosphorous regeneration of the skeletal pore water ( 26 , 27 ). Given their capacity to affect the health of corals as keystone species, coral endolithic microbes can affect whole ecosystem function. Endolithic biofilms in dead coral substrates have previously been identified as significantly contributing to whole-reef primary productivity ( 28 ). This is in addition to being major contributors to reef bioerosion, total carbonate budgets and bathymetric structural complexity ( 29 – 31 ). Endolithic microfungi have also been observed parasitizing their cnidarian hosts ( 32 ), which could reduce coral fitness. As such, the coral endolithic microbiome can be parasitic and symbiotic to its host, providing the coral with nutrients but simultaneously damaging their supportive skeleton, and through these behaviors they can affect whole ecosystem characteristics. Variation in the macro- and microstructure of the coral skeleton is often highlighted as a potentially important factor affecting endolithic microbial biomass and the contribution of microbial endoliths to coral health and degradation ( 33 – 36 ). Vogel et al. ( 36 ) previously showed that the rate of microbial bioerosion varies between types of calcifier shells as well as between different mineral phases of calcium carbonate (e.g., calcite versus micrite). For hard corals, there are many possible host skeletal characteristics to which endolithic biofilms may respond, but interspecific variability in endolithic community composition and biomass is not well known. For example, massive or mounding corals, whose microbial endoliths are most commonly studied, have a large volume of substrate to colonize which could lead to greater colony-specific biomass for these corals. However, the quality of the substrate, as well as quantity, needs to be considered. Ralph et al. ( 37 ) note that endolithic microalgae within corals from reef lagoons with high irradiance show contrasting behavior to those from deeper reef slopes ( 38 ). Namely, microborers burrow downwards (i.e., positive geotropism) in high irradiance habitats while they grow “upward” toward the host tissue in deeper habitats. This suggests that the compensation depth for photosynthetically active radiation inside the coral skeleton varies across habitats. Therefore, a massive coral on a deep reef may have a large absolute volume of substrate, but only a small portion of this can support the phototrophic growth of dominant microalgae in the skeleton. In contrast, tabular corals have evolved to maximize light capture for the endosymbiotic algae within their tissues (i.e., zooxanthellae [ 39 ]); this may also increase the internal skeletal light field in the endolithic microenvironment. An expected result may be a higher endolithic biomass compared to massive corals but in high irradiance habitats, light enhancement may lead to photodamage/photoinhibition in Ostreobium spp. adapted to extreme shade but lacking typical photoprotective mechanisms designed to prevent damage from excess light ( 37 , 40 , 41 ). Therefore, tabular coral skeleton may in fact be a poor-quality habitat on reef flats but a good-quality habitat on reef slopes. Variation in coral biomechanical and morphological features has been previously shown to correlate with the diversity of coral-associated bacterial communities ( 42 , 43 ). For the endolithic component of the coral microbiome, interspecific variability in coral skeletal porosity and density are expected to affect the biomass of the resident endolithic biofilm. Crypto- and chasmoendolithic taxa may benefit from high porosity since it represents greater available space for colonization. Euendolithic taxa would, in theory, be more sensitive to interspecific variability in skeletal microdensity since it is expected to effect the energetic cost of boring into the coral skeleton. Variation in coral skeletal density arises from (i) the organic matrix intercalating the skeleton, (ii) aragonite grain size and orientation, and (iii) small amounts of nonaragonitic calcium carbonates such as calcite and micrite. Vogel et al. ( 36 ) demonstrated that bioerosion rates differ between aragonite, calcite, and micrite, while Iha et al. ( 41 ) recently presented data that suggests euendolithic Ostreobium feeds on coral organic matrices. As such, evidence suggests that coral microdensity is a key skeletal characteristic affecting euendolithic taxa distribution and therefore overall microbial biomass inside the skeleton. Our ability to model the impact of endolithic microbiomes on whole coral reefs, however, is limited by an incomplete understanding of the nature and drivers of interspecific variability in the coral endolithic microbiome. Here, we aimed to test whether coral skeletal characteristics correlate with variability in the biomass and chlorophyll concentrations of the endolithic microbiome of five species of tropical hard coral ( Fig. 1 ): Goniastrea retiformis , Isopora palifera , Montipora digitata , Porites cylindrica , and Porites cf. mayeri . We measured two skeletal porosity and microdensity, as well as two microstructural elements which can affect light capture/scattering on the surface of the coral colony: the size of calices relative to intercorallite space and corallite complexity ( Fig. 1 ) ( 44 ). Based on previous studies of coral light capture ( 44 – 46 ), larger and more complex corallites are considered more effective “photon traps,” which increases light availability for tissue-associated zooxanthellae. Finally, we measured tissue thickness ( Fig. 1 ) according to the hypothesis of Shashar et al. ( 47 ) that coral tissue thickness affects irradiance within the endolithic microhabitat. FIG 1 Coral species selected for this study: Goniastrea retiformis (A), Montipora digitata (B), Porites cf. mayeri (C), Isopora palifera (D), and Porites cylindrica (E; white arrow). (F) For each coral species, five morphological parameters were measured for their hypothesized influence upon endolithic microbial biomass. Measurements 1 to 3 are used to calculate the ratio of calice width to coenosteum width; measurements 4 and 5 are used to calculate corallite complexity. Tissue thickness is measured from a skeletal cross-section (inset). For details on measurements of microdensity and porosity, see Fig. S2 and Text S1 . Photographs A and C to F were provided by Francesco Ricci and are used with permission. 10.1128/mSphere.00060-21.1 TEXT S1 Output of diagnostic tests applied to statistical models used in this study. Download Text S1, PDF file, 0.1 MB . Copyright © 2021 Fordyce et al. 2021 Fordyce et al. https://creativecommons.org/licenses/by/4.0/ This content is distributed under the terms of the Creative Commons Attribution 4.0 International license . Simultaneously, we measured the biomass of the endolithic biofilm in each coral species (by ash-free dry weight [AFDW]) and the concentrations of chlorophylls a , b , c , and d (by spectrophotometry). Chlorophyll a is the primary pigment used in oxygenic photosynthesis. Chlorophyll b is the major accessory pigment used by green algae (Chlorophyta) such as Ostreobium spp. Chlorophyll c (here encompassing both c 1 and c 2) is used by a wide range of microbial phototrophs, including dinoflagellates, diatoms, and brown algae, all of which have been previously identified from the endolithic microbiome ( 22 ). Chlorophyll d is the primary pigment in Acaryochloris marina and Acaryochloris -like cyanobacteria ( 48 – 50 ), which have been identified from endolithic habitats at this location ( 48 ). We then used principal-component regression to combine the five morphological parameters and test their relationships with the biomass of and chlorophyll concentrations within the endolithic biofilm.",
"discussion": "DISCUSSION The endolithic biofilms of coral skeletons play a key, but as-yet-undercharacterized role in the coral metaorganism and may be important in defining how corals respond to climate change. As such, we set out to explore whether interspecific variation in coral skeletal morphology could be used to predict the biomass of and chlorophyll concentrations within the endolithic microbiome. This information could form the basis of a framework for identifying coral species, based on their morphometric traits, which are likely to be significantly impacted by endolithic community responses. Five aspects of coral morphology were measured ( Fig. 1 and 4 ): tissue thickness, skeletal microdensity, porosity, corallite complexity, and the ratio of calice width to coenosteum width ( 44 , 51 ). Several significant positive relationships were identified between biomass/chlorophyll concentrations and variables produced by a PCA of the morphological data (PC1 and PC2). These provide evidence for coral skeletal morphology structuring the biomass and phototrophic composition of endolithic biofilms. The biomass of coral endolithic microbiomes is commonly dominated by euendolithic (i.e., “true-boring”) phototrophs such as Ostreobium spp.; therefore, coral skeletal microdensity was hypothesized to influence the rate of algal boring and therefore microbial biomass inside the coral skeleton. Of the five coral species we compared, we found that G. retiformis had the lowest skeletal microdensity ( Fig. 4 ), the second highest endolithic biomass ( Fig. 2 ), and the highest concentrations of chlorophylls a and b in their microbiomes ( Fig. 3 , Chl a and Chl b ). Further, the interspecific patterns in the chlorophylls a and b ( Fig. 3 , Chl a and Chl b ) mirror those in skeletal microdensity ( Fig. 4 ) with less-dense skeletons harboring biofilms with higher concentrations of chlorophylls a and b . Given that only green algae (Chlorophyta) such as Ostreobium utilize chlorophyll b as an accessory pigment, this suggests that euendolithic green algae benefit from low skeletal density, possibly due to a decreased energetic cost of boring. It has also been recently proposed that Ostreobium spp. feed on coral skeletal organic matrices to supplement autotrophy ( 41 ). A higher relative mass of organic matrix would decrease microdensity and act as a food source for dominant microborers; therefore, this relationship may be due to more abundant food sources for Ostreobium in low-density skeletons. It is possible that anoxygenic green sulfur bacteria, which were found to be the dominant endolithic phototrophs in I. palifera from the South China Sea ( 25 ), are influencing the relationship between chlorophyll concentrations and skeletal structure for our I. palifera samples. However, recent work has shown that the I. palifera endolithic microbiome at this location is primarily composed of oxygenic phototrophs, including Ostreobium ( 52 ), and it is not yet known whether these two groups of phototrophs can coexist in the coral skeleton. Nonetheless, the presence of bacteriochlorophylls used by green sulfur bacteria has been found to lead to minor overestimations of chlorophyll a concentrations in freshwater lake sediments ( 53 ). G. retiformis also had the largest (relative to coenosteum width) and most complex corallites, indicating that they are more effective “photon traps” than the other species studied here ( 44 , 46 , 54 ). In contrast, I. palifera and M. digitata have relatively high microdensities (∼2.8 g cm −3 ), low corallite complexity, and low calice/coenosteum width ratios. They also have significantly lower microbial biomasses and concentrations of chlorophylls a and b ( Fig. 2 and 3 , Chl a and Chl b ). These observations suggest that variability in coral skeletal morphology is affecting the dominant phototrophs in the endolithic microbiome and is therefore an important driver of interspecific variability in endolithic microbial biomass. Specifically, corals with low-density skeletons that are more effective at capturing light support great abundances of green microalgae. This is reflected in the results of the principal component regression. PC1, which was loaded by the above factors ( Fig. 5 ), is significantly positively correlated with endolithic biomass and chlorophylls a and b ( Fig. 6 and 7 ). FIG 7 Scatterplots showing the statistically significant relationships, across all species ( n = 50), between principal components 1 (PC1; light blue lines) and 2 (PC2; purple line), and accessory chlorophylls b (left) and d (right). These were modeled using gamma regression. In all cases, the regression lines represent the predicted effect of, for example, PC1 upon the value of AFDW when PC2 is held at its mean and vice versa. Insets show the the log-linked gamma means (μ) and dispersion parameters (ϕ) for each model. To calculate the gamma distribution shape (α) and rate (β) parameters, α = 1/ϕ and β = α/μ. P. cylindrica had the highest recorded endolithic biomass in our species, and yet its skeleton was more dense than its congeneric species P. mayeri . However, it did have a larger calice/coenosteum width ratio and more complex corallites. It also has a submassive/branching macromorphology, which makes P. cylindrica more effective at scattering light within the whole colony than P. mayeri ( 39 ). This suggests that light-capture parameters are relatively more important than microdensity in driving the observed relationship between PC1 and biomass/chlorophyll. Effective light capture at the external surface of the coral skeleton could either increase the light intensity in the endolithic environment or decrease it by scattering light more effectively in the photosymbiont-rich coral tissue. Higher light intensity inside the endolithic habitat might be expected to increase phototrophic growth. However, endolithic algae like Ostreobium spp. can be photochemically saturated at very low light intensities (<7 μmol photons m −2 s −1 [ 37 ]), and a recent genomic analysis revealed an absence of photoprotective genes in this shade-adapted algae ( 41 ). Therefore, they may be particularly susceptible to photodamage at high light; this could also explain why Ralph et al. ( 37 ) identified endolithic algae as boring away from the coral surface (i.e., negative phototaxis) in a shallow reef flat environment. In contrast to microdensity and corallite morphology, porosity and tissue thickness showed no significant correlations with chlorophyll concentrations in the endolithic biofilm, and the relationship between these two variables and microbial biomass is difficult to interpret. Porosity was hypothesized as being a key driver of endolithic microbial biomass because higher porosity would represent more space for cryptoendoliths that inhabit pores and chasms in the coral skeleton. However, skeletons with high porosity also have relatively less substrate for euendoliths to colonize. Consequently, we observed pairs of species with similar porosities (e.g., G. retiformis / I. palifera or P. cylindrica / M. digitata ) having very different microbial biomasses inside their skeletons ( Fig. 2 and 4 ). Similarly, I. palifera and P. cylindrica had approximately the same tissue thicknesses but the lowest and highest biomasses, respectively ( Fig. 2 and 4 ). Thicker tissues were hypothesized by Shashar et al. ( 47 ) as being the cause of lower light intensity measured in massive Favia spp. coral skeletons compared to massive Porites spp. skeletons. In this study, species with thin tissue (i.e., G. retiformis and M. digitata ) were observed harboring biofilms with both high and low biomasses ( Fig. 2 and 4 ). Conclusions. Here, we identify correlations between interspecific variability in coral endolithic microbial biomass and skeletal morphology. Our results suggest that differences in skeletal density and a coral species’ capacity for capturing and scattering light on the outer surface of its skeleton influence the variability in skeletal microbial biomass, predominantly of dominant euendolithic green algae (e.g., Ostreobium spp.). Further studies exploring the responses of individual endolithic taxa to different aspects of coral skeletons are essential for defining the variable ecological role of this microbiome on coral reefs and identifying coral types that may be differentially affected (both positively and negatively) by the endolithic community."
} | 5,469 |
25470793 | PMC4254613 | pmc | 1,007 | {
"abstract": "In microbial ecosystems, bacteria are dependent on dynamic interspecific interactions related to carbon and energy flow. Substrates and end-metabolites are rapidly converted to other compounds, which protects the community from high concentrations of inhibitory molecules. In biotechnological applications, pure cultures are preferred because of the more straight-forward metabolic engineering and bioprocess control. However, the accumulation of unwanted side products can limit the cell growth and process efficiency. In this study, a rationally engineered coculture with a carbon channeling system was constructed using two well-characterized model strains Escherichia coli K12 and Acinetobacter baylyi ADP1. The directed carbon flow resulted in efficient acetate removal, and the coculture showed symbiotic nature in terms of substrate utilization and growth. Recombinant protein production was used as a proof-of-principle example to demonstrate the coculture utility and the effects on product formation. As a result, the biomass and recombinant protein titers of E. coli were enhanced in both minimal and rich medium simple batch cocultures. Finally, harnessing both the strains to the production resulted in enhanced recombinant protein titers. The study demonstrates the potential of rationally engineered cocultures for synthetic biology applications.",
"conclusion": "Conclusions Synthetic biology broadens the possibilities to tune sophisticated production platforms, and coculturing is seen as a promising new frontier for taking the bioproduction to the next level [35] , [36] . This study demonstrates that rationally engineered synthetic cocultures can improve biomass production, culture viability, and product formation in simple unoptimized batch conditions. The study extends applications of novel type of bioprocess optimization, and provides clues for the development of functional, readily engineered, and dynamic cocultures for synthetic biology applications.",
"introduction": "Introduction In a microbial ecosystem, bacterial species cooperate by producing metabolites that serve as substrates for other species of a community. Dynamic and symbiotic interactions between different species enable an efficient carbon and energy flow. Furthermore, the interactions balance rapid environmental changes and high concentrations of substrates, intermediates, or other potentially harmful compounds. Common end-metabolites, such as organic acids, are rapidly converted to other compounds. Thus, tolerance to high concentrations of inhibitory compounds is not essential for individual members of a community. In biotechnological applications, however, engineered pure cultures are typically utilized as production platforms because of the easier bioprocess control and more straight-forward genetic engineering. Nevertheless, establishing a sustainable and economical bioprocess is a challenge, for which metabolic engineering and synthetic biology are seeking solutions [1] – [4] . For example, accumulation of unwanted side products and inefficient carbon and energy fluxes are common issues. Since Escherichia coli is a widely used host in bioprocesses, considerable effort is dedicated to controlling the overflow metabolism. In this phenomenon, carbon is excessively transported into the cells, and the capacity of tricarboxylic acid (TCA) cycle is exceeded. As a result, the metabolism is shifted to produce acetate from acetyl coenzyme A (CoA) [5] . The accumulation of acetate results in the induction of the stress response, growth inhibition, significant carbon loss, and eventually reduced productivity [6] , [7] . Previous strategies to neglect the negative effects of acetate include both genetic engineering of E. coli to circumvent the acetate production and comprehensive optimization of fed-batch bioprocesses [8] – [10] . These approaches, however, potentially decrease the growth rate or require a substantial amount of optimization and engineering work. “Pure cocultures” typically consisting of two known strains are a more recent approach to improve the production platforms [11] , [12] , and a trend towards more controllable and tunable cocultures involving systems biology and synthetic biology solutions is growing stronger [13] , [14] . For example, Eiteman et al. [15] have described the co-fermentation of hexoses and pentoses by a community of engineered E. coli. Tsai et al. [16] established a system for ethanol production from cellulose by a synthetic yeast community. Although the obtained ethanol yields are low, the study is an elegant example of the utilization of an engineered community for bioenergy production. The culture systems described above are based on exploiting strains of single species, but they represent the first steps towards consolidated processing of sustainable substrate to products by an engineered consortium. Employing different strains of a single species broadens the engineering possibilities; however, the strains often possess very similar carbon utilization patterns, thus leaving unresolved the issues related to by-product accumulation and the inefficient carbon flow. On the other hand, genetic engineering is very challenging for mixed populations which efficiently degrade complex substrates and mixed sugars [17] . In the present study, the advantages of both mixed populations and readily engineered model strains are exploited; the well-characterized model organisms E. coli K12 and Acinetobacter baylyi ADP1 were employed for the construction of a synthetic and symbiotic microecosystem with directed carbon flow for improved culture performance. In the first stage of the study, the strains were cultured in a minimal medium batch cultivation to monitor the coculture compatibility, performance of individual strains, and carbon flow. Secondly, E. coli strain was engineered to produce recombinant protein, and batch cocultures with different substrate concentrations were conducted to demonstrate the effects of accelerated growth to recombinant protein production. Finally, both strains were engineered to produce the same recombinant protein, and cocultivations in a bioreactor were carried out.",
"discussion": "Results and Discussion Establishing a synthetic coculture In natural ecosystems, different species interact through diverse forms of cooperation, such as mutualism, parasitism, or symbiosis, creating a complex network of integrated metabolic pathways. From a synthetic biology point of view, cocultures can provide several advantages over monocultures by performing multi-step tasks and being more catabolically versatile [25] – [27] . In this study, a synthetic coculture was constructed using two strains of different species, E. coli K12 and A. baylyi ADP1, and the effects of connecting the carbon metabolism of these strains were investigated. For the construction of a functional, synthetic coculture, the bacterial species for the community were chosen based on characteristics supporting one another. To promote straight-forward system design and engineering, the availability of readily applicable engineering tools and well-characterized metabolic networks was considered essential here [28] , [29] . In microbiological terms, the strains also share similar preferences for growth conditions. \n A. baylyi ADP1 has been recently introduced as a competitive model organism for genetic studies [18] and metabolic engineering purposes [22] , [23] , and genetic tools are widely available for the strain [21] , [30] , [31] . A. baylyi ADP1 exhibits a relatively wide substrate range and is known to efficiently utilize organic acids such as acetate as a sole carbon and energy source [32] , although the detailed mechanism for the uptake and utilization of acetate has not yet been characterized. Moreover, our recent studies suggest that ADP1 neither produces harmful overflow metabolites nor exhibits substrate inhibition (S. Santala and V. Santala, unpublished data). In order to establish a functional symbiotic coculture system with directed carbon flow, and to produce unambiguous data for analyzing the coculture performance, a mutant strain of A. baylyi ADP1 made deficient of utilizing glucose was employed; according to the metabolic model of ADP1, the disruption of a high-affinity gluconate permease ( gntT , ACIAD0544) blocks the glucose pathway. Due to the exceptional glucose utilization pathway of ADP1, a glucose molecule is oxidized to gluconate on the outer surface of the inner membrane by an electron carrier associated to glucose dehydrogenase, pyrroloquinoline quinine (PQQ). Subsequently, the high-affinity gluconate permease GntT transports gluconate into cells [32] . Here, a knock-out mutant strain A. baylyi ADP1Δ gntT :: Kan r /tdk \n [18] was employed. When glucose is sufficiently available in the coculture, E. coli cells uptake glucose into the cells in excess and return a significant amount of the carbon into the medium in the form of acetate. Thus, in a coculture, the glucose negative mutant strain of ADP1 is dependent on the end-metabolites of E. coli, enabling an experimental validation of the carbon flow and fate. A schematic illustration of the proposed carbon flow is presented in Figure 1 . 10.1371/journal.pone.0113786.g001 Figure 1 The proposed carbon flow in the wild type Escherichia coli culture and in the coculture of engineered Acinetobacter baylyi ADP1 and E. coli . A) E. coli culture supplied with excess glucose readily shifts to an overflow metabolism (*), producing large amounts of acetate into the culture medium. Acetate inhibits growth and reduces product formation, and carbon flow is directed off the product route. B) In a coculture involving directed carbon flow, the strain A. baylyi ADP1 is made deficient in glucose utilization by a knock-out of gluconate permease gntT and is solely dependent on the end-metabolites (acetate) of E. coli . Carbon is further metabolized and can be directed to biomass and the product of interest. Metabolic pathways in the figure are simplified and only the main products are shown. For a functional community, the compatibility of the two strains is evidently essential. In order to monitor the performance and growth of ADP1Δ gntT::Kan r /tdk in the coculture, pVKK81-T- lux containing a bacterial luciferase genes luxCDABE (from P. luminescens \n [33] ) were transformed into the strain by natural transformation. A luminescent colony was selected from a plate containing 10 µg/ml tetracycline. The resulting strain ADP1Δ gntT :: Kan r /tdk[lux_tet r ] was designated as ABlux. In order to monitor the growth dynamics of E. coli and ABlux, the strains were cultivated both separately and in cocultures for 12 h in batch cultures. A minimal salts medium MA/9 supplemented with 50 mM glucose was used in the experiment. Biomass and luminescence were determined from the cultivations hourly. Selective plate count was also applied in order to determine the individual cell numbers of E. coli and ABlux. The total biomass of the coculture and the proportion of ABlux of the total biomass are presented in Figure 2 . The use of luminescent construct enabled specific real-time monitoring of ABlux performance during the cocultivation ( Figure 2 , the inlet), and allowed the observation of potential fluctuations and problem situations in the cultures, not readily detectable by other means. According to the luminescence signal data, ABlux starts to grow in the coculture after a short lag phase, which was subsequently confirmed by the 5-hour timepoint plate count. After 11 h of cultivation, the strain reached an approximately 6% proportion of the total biomass. 10.1371/journal.pone.0113786.g002 Figure 2 The total biomass of the coculture of E. coli and ADP1Δ gntT :: Kan r /tdk [ lux_tet r ] (ABlux), and the proportion of ABlux in the total biomass. The strain ABlux is deficient in utilizing glucose. During the cultivation, the growth of ABlux was monitored in real-time via a luminescence reporter luxCDABE (the inlet). The cells were cultivated in a minimal salts medium supplied with glucose for 12 h. The mean and standard deviation of the two independent cultures are shown. Note the logarithmic scale in biomass and luminescence y-axes. Total biomass – line with squares; proportion of ABlux cells in total biomass – columns. The biomass of E. coli in the monoculture and in the coculture was calculated according to optical densities and plate counts: the data is presented in Figure 3A . In the E. coli monoculture, the growth shifted to an exponential phase after 5–6 hours of lag phase and reached a total cell number of ∼6.15·10 8 ml −1 after 12 h of cultivation. In the coculture, the lag phase for E. coli was shorter, 4–5 h, and the culture reached a slightly higher cell number of 6.63·10 8 ml −1 at 12 h timepoint. 10.1371/journal.pone.0113786.g003 Figure 3 \n E. coli biomasses and substrate/end-metabolite concentrations. The cells were cultivated in a minimal salts medium supplied with glucose for 12 h. A) E. coli cell number in monoculture (empty squares) and in coculture (filled squares). B) Glucose (squares) and acetate (circles) concentrations in monoculture (empty symbols) and in coculture (filled symbols). The mean and standard deviation of two independent culture samples are shown. Note the logarithmic scale in biomass y-axis. The consumption of glucose and accumulation of end-metabolites in E. coli monoculture and coculture were analyzed by HPLC ( Figure 3B ). The consumption of glucose in E. coli monoculture was nearly linear after 7 h of cultivation, with a total decrease of 6 mM in the glucose concentration. Acetate could be reliably detected after an 8 h timepoint, the concentration reaching 4 mM at the end of the cultivation. End-metabolites other than acetate were not detected. In the coculture, the glucose was utilized more rapidly and efficiently, with the total decrease in glucose concentration being 9 mM. Detectable amounts of acetate could not be observed in the coculture until the 11 h timepoint, indicating efficient acetate utilization by ABlux. The acetate concentration in the end of the cultivation (2 mM) was lower compared to that of the monoculture of E. coli . The strain ABlux did not exhibit growth in the monoculture due to the lack of suitable carbon source, and neither consumption of glucose nor accumulation of end-metabolites was detected (data not shown). We discovered that the coculture was beneficial for E. coli , as the cells grew faster in the cocultivation compared to the monoculture in the studied conditions. As expected, the proportion of ADP1 cells in the coculture increased as a result of E. coli acetate production. According to the E. coli growth curve, the consumption of acetate by ABlux positively affected the E. coli growth, which could be observed as a shorter lag phase and more sufficient growth. Even though the concentration of acetate did not reach the detection limit until the 8–10 h timepoint the cultures already showed differential performance after the 5 h timepoint, indicating that acetate already hinders growth at very low concentrations. Eventually, the coculture resulted in more efficient glucose utilization by E. coli and a lower overall acetate concentration in the culture. Thus, it could be concluded that the constructed carbon channeling system diminished the negative effects of acetate and favorably affected the E. coli growth, and a relatively low proportion of the supporting strain ABlux in the community was able to benefit the growth. Production of recombinant protein by E. coli in the synthetic coculture It was demonstrated that the coculture is beneficial for E. coli growth and biomass production. The next step was to study the effects of the coculture on production of the heterologous product that competes with the cell building blocks, demonstrated here by recombinant protein production. Green fluorescent protein (GFP) was chosen for the proof-of-principle production platform in order to enable real-time monitoring of production dynamics and comparability between the mono and cocultures. In addition, GFP has previously been successfully expressed both in E. coli and A. baylyi ADP1 [21] . A plasmid containing a superfolder variant of GFP (BBa_I746909) under inducible lactose promoter was constructed. The resulting plasmid sfGFP/pAK400c was transformed to E. coli, the clones were selected from chloramphenicol plates, and the strain was designated as ECsf. To enable A. baylyi ADP1 growth in the coculture, plasmid pBAV1C- ara (described elsewhere [22] ) without an insert was transformed to the mutant strain ADP1Δ gntT::Kan r /tdk and selected from the chloramphenicol plates, resulting in a strain designated as ABc. The potential support of ABc for ECsf growth and protein production in a coculture was studied in minimal medium batch cultures supplied with different glucose concentrations: 50 mM, 100 mM, and 250 mM. The amount of GFP produced was estimated by fluorescence determination. The end-point biomasses (OD 600 ) and fluorescence signals for monocultures and cocultures are presented in Figure 4 . The cocultures grew to higher optical densities and produced more fluorescence with all studied glucose concentrations compared to the respective monocultures, and the highest optical density and fluorescence signal were obtained in the culture containing 50 mM glucose. Interestingly, the more glucose is present in the medium, the more significant is the difference between the mono- and cocultures. This indicates that high substrate concentrations may hinder E. coli growth by direct substrate inhibition and by accelerated overflow metabolism. In the cultures containing 250 mM glucose, a fluorescence signal of approximately 2-fold and an optical density of 3-fold higher were obtained for the coculture compared to the monoculture. Thus, the batch experiments further emphasized the coculture relevance for improved culture performance and productivity in variable substrate concentrations. 10.1371/journal.pone.0113786.g004 Figure 4 Biomasses and fluorescence of E. coli expressing sfGFP/pAK400c (ECsf) and A. baylyi ADP1Δ gntT::Kan r /tdk expressing empty plasmid pBAV1C- ara (ABc) in mono and cocultures. Fluorescence is produced solely by E. coli expressing GFP. The strains were cultivated in a minimal salts medium supplemented with 50; 100; or 250; mM glucose for 24 h. A) Biomasses (optical density, OD 600 ) of ECsf monocultures and cocultures (ECsfABc). B) Fluorescence signals measured for monocultures and cocultures demonstrating the production of recombinant protein in cultures. The mean and standard deviation of two independent cultures are shown. Production of recombinant protein by the synthetic coculture in a bioreactor The previous experiments demonstrated that ABc efficiently supports ECsf growth in cocultures, being especially beneficial in cultures containing high substrate concentrations by neglecting the negative effects of acetate. In an ideal case, the secreted acetate could be directed to the product by the supportive strain. For such an approach to be practical, a single expression construct that is functional in both strains should be used. Indeed, an expression vector exploitable both in E. coli and A. baylyi ADP1 have been described [21] . Derived from the original vector, a plasmid pBAV1C-T5-GFP was constructed and transformed to E. coli and A. baylyi ADP1Δ gntT :: Kan r / tdk. The vector contains the gene gfp under a constitutive strong promoter (T5), and a chloramphenicol resistance gene ( cat ). The resulting strains were designated as ECg and ABg, respectively. To provide sufficient supply of nutrients and building blocks for the strains under strong over-expression, and to study the coculture functionality in simple unoptimized batch conditions without pH buffering or control, cultivations were performed in a rich medium supplied with glucose. The batch cultivations for the ECg and ABg coculture and ECg monoculture were carried out in a rich medium supplied with 100 mM glucose and aeration. The cultivations were conducted in a bioreactor in order to reveal more information about the culture characteristics and dynamics; oxygen partial pressure, pH, biomass production (OD), proportions of the coculture strains (%, CFU), substrate and end-metabolite concentrations, and fluorescence signals were monitored throughout the 10-hour cultivations ( Figure 5 ). The coculture started to grow clearly faster after four hours of cultivation, reaching a final optical density of 13, which was ∼3-fold higher compared to the biomass of ECg monoculture (OD 4.3). The initial proportions of the two strains were equal but the proportion of ABg seemed to increase fastest between the 6–8 h timepoints, assumingly in parallel with the exponential growth phase of ECg. It was noted that when cultured in a rich medium, the proportion of ABg cells (20–30%) was much higher compared to the minimal medium coculture where the proportion of ABlux was approximately 6% by the end of the cultivation. The relatively high biomass of ABg did not, however, negatively affect ECg growth, which further supports the hypothesis of ADP1 not producing compounds inhibiting E. coli growth. 10.1371/journal.pone.0113786.g005 Figure 5 Monitored bioreactor cultivations of the monoculture ECg and the coculture ECgABg. Both strains produce GFP by expressing the same genetic construct pBAV1C-T5-GFP. The cultivations were performed in a bioreactor in a rich medium supplied with 100 mM glucose and aeration. A) Culture parameters; pH (large circles) and oxygen partial pressure (pO2 %, small circles) for ECg monoculture (empty circles) and ECgABg coculture (filled circles) B) Total biomasses (OD 600 ) of the cultures and proportion of ABg cells (%, CFU) in the coculture. C) Fluorescence signals (cts) for the monoculture ECg and coculture ECgABg demonstrating the production of recombinant protein in the cultures. The mean and standard deviation of 2–4 independent culture samples are shown. ECg - E. coli expressing pBAV1C-T5-GFP, ABg - A. baylyi ADP1ΔgntT::Kan r /tdk expressing pBAV1C-T5-GFP. The fluorescence signals differentiated after 4 hours of cultivation, the final fluorescence signals being approximately 4-fold higher in the coculture compared to the ECg monoculture. The higher biomass and consequently the higher fluorescence signal of the coculture can be explained by the more propitious growth conditions; in both cultures, pH started to drop after 3–4 hours of cultivation, but the decline of pH in ECg monoculture was faster compared to the coculture. In the monoculture, the final pH of 5.4 was reached at 8–9 h timepoint. In the coculture, pH decreased more moderately, but increased rapidly after the 8 h timepoint, reaching a final pH of 6.5, which is very close to the initial pH of the culture. This implies an efficient internal buffering system in the coculture, which retains favorable growth conditions. Acetate concentrations started to differentiate after the 6 h timepoint. At the end of the cultivation, the acetate concentration in the ECg monoculture was 13 mM, which probably affected the growth and cell performance negatively. In the coculture, the acetate concentration was only 3 mM, indicating successful recycle of acetate from the medium to the biomass and the product. The final glucose concentrations were 65 mM in the coculture and 90 mM in the monoculture. For the biomass samples collected at the end of the cultivations, the obtained CDWs were 2.1 g/l and 5.1 g/l for ECg and ECgABg cultures, respectively. Control cultivations without glucose supplementation were carried out in batch bottles. It was found that the coculture does not grow sufficiently in such conditions (data not shown), suggesting that the relevance of the coculture lies for the most part in the interactive glucose-acetate metabolism. The previous minimal medium cultivations constituted of ECsf and ABc strains demonstrated that both biomass and product titers could be improved, even though ABc was not contributing to the protein production. In a minimal medium, ABc is completely dependent on the carbon provided by ECsf growth, and the proportion of ABc cells is low, whereas rich medium enables more rapid and independent growth of ABc (as for ABg), resulting in significantly higher cell proportions consuming valuable building blocks from protein production. Thus, we were interested to see how the increased number of ABc cells affects the protein production in a rich medium when only ECg contributes to protein production ( Figure S1 ). In short, it was observed that the bioprocess trends, acetate production, glucose consumption, and biomass production were very similar compared to the ECgABg bioreactor coculture. According to the fluorescence signals, the recombinant protein titer was 50% higher in the ECgABc coculture compared to the ECg monoculture, which is expectedly less than improvements gained in ECgABg coculture. It can be concluded that as protein overexpression is highly dependent on the available biocomponents consumed by both the strains, it is not energetically affordable to maintain the system in a rich medium with high proportions of the supportive strain, without it participating in the production. Thus, the availability of genetic tools and the possibility to rationally engineer and commit both the strains to the production are crucial to take full advantage of the carbon redirection, as demonstrated here. Furthermore, the product example (recombinant protein) is constituted of amino acids instead of carbon based molecules, and thus the benefits of carbon rerouting could be potentially better exploited in production of hydrocarbons, such as fatty acid derived compounds for bioenergy [22] , [23] . Our findings suggest it is possible to tune the coculture balance, nature, and dynamics by altering the medium, substrate concentration, and genetic constructs. The accumulation of acetate is a widely recognized problem in bioprocessing of E. coli . This study demonstrates that by exploiting a coculture with optimally chosen strains, the growth, carbon utilization, and product formation can be improved through symbiotic interactions, and the negative effects of acetate can be diminished. Employing the coculture enabled the use of a high substrate concentration in a simple batch culture, and the over-flow metabolism of E. coli was successfully exploited in producing the protein of interest. Fluctuations in sugar concentrations in industrial feedstocks can impose restrictions for single strain cultures (6), whereas high substrate concentration is not an issue for the coculture system presented here. Thus, with regards to sustainable and economical carbon sources, exploiting a well-designed coculture may broaden the possibilities to exploit challenging liquors with high concentrations of carbohydrates. Nevertheless, cocultures are complex systems involving numerous interactions. Despite the possibility to reduce these interactions, it is very difficult to identify all the mechanisms that are potentially beneficial for the culture. E. coli based bioprocesses have demanded decades of optimization work to build up a cost-effective bioproduction platform [34] . Thus, improvement of the utility, profitability, and viability of cocultures requires significant research efforts. It is impossible to exceed the production rates and yields of highly tuned and optimized systems at a single cell level, but the power of cocultures most probably lie in processes that cannot be readily optimized (e.g. processes exploiting diverse industrial streams) or in processes involving variable conditions."
} | 6,983 |
24688663 | PMC3962192 | pmc | 1,008 | {
"abstract": "Lactic acid bacteria (LAB) are receiving increased attention for use as cell factories for the production of metabolites with wide use by the food and pharmaceutical industries. The availability of efficient tools for genetic modification of LAB during the past decade permitted the application of metabolic engineering strategies at the levels of both the primary and the more complex secondary metabolism. The recent developments in the area with a focus on the production of industrially important metabolites will be discussed in this review.",
"conclusion": "Conclusions The cases presented shortly in this review reveal a variety of approaches for metabolic engineering of LAB including mutagenesis, classical gene inactivation and overexpression, redox engineering or engineering of primary carbon metabolism, as well as predictive approaches for improving cellular phenotypes. They also exemplify the power of metabolic engineering in LAB for the improved and often efficient production of a number of industrially important metabolites with wide applications in the food and pharmaceutical industries. It is expected however that progress in metabolic modelling and Systems Biology approaches will provide the means for engineering complex (biosynthetic) pathways for the efficient production of metabolites such as vitamins, antioxidants and other nutraceuticals by LAB.",
"introduction": "Introduction Lactic acid bacteria are used worldwide in the industrial manufacture of fermented foods. Their most important application in this respect is in the dairy industry with an enormous variety of fermented dairy products, while next to that is the fermented meat and vegetable products industry. Besides food production, LAB are used in a variety of other industrial applications such as the production of lactic acid, high-value metabolites involved in flavor and texture development or health applications, probiotic products, and antimicrobial peptides. Characteristics such as a rather simple energy and carbon metabolism and a small genome size (∼2-3 Mb), make LAB important candidates for metabolic engineering strategies. Such strategies have mainly focused on rerouting of pyruvate metabolism to produce important fermentation end-products e.g. sweeteners, flavors, aroma compounds [ 1 , 2 ], and on more complex biosynthetic pathways leading to the production of exopolysaccharides and vitamins [ 3 , 4 ], while attempts to manipulate the central carbon metabolism (CCM) are rather limited in number [ 5 ]. Being one of the model organisms in microbial metabolism, Lactococcus lactis has been the main target of metabolic engineering among LAB. The knowledge of its complete genome sequence [ 6 ], the availability of numerous genetic tools for this microorganism [ 7 ], and its industrial relevance, facilitated its use in the development of efficient cell factories [ 8 ]. The present work aims to give an overview of the recent advances in engineering the metabolism of LAB for the production of industrially important compounds."
} | 753 |
35878024 | PMC9351376 | pmc | 1,009 | {
"abstract": "Significance Nacre has inspired scientists because of its ability to reach high stiffness and fracture toughness using relatively weak mineral building blocks. Suppressing vibration is another remarkable feature of nacre that has been less explored and is yet to be translated into synthetic composites. We unveil the damping principles of tough biological materials by investigating structure-property relationships of model nacre-like composites. Our bio-inspired structures display a loss modulus 2.4-fold higher than natural nacre and 1.4-fold higher than lightweight flax-based composites used for damping applications, despite the less organized brick-and-mortar architecture compared to nacre. Since such properties arise primarily from the multiscale structure of the material, this work opens an enticing pathway toward high-performance composites using more sustainable, environmentally friendly building blocks."
} | 230 |
33739376 | PMC8175051 | pmc | 1,010 | {
"abstract": "Abstract The last eukaryote common ancestor (LECA) possessed mitochondria and all key traits that make eukaryotic cells more complex than their prokaryotic ancestors, yet the timing of mitochondrial acquisition and the role of mitochondria in the origin of eukaryote complexity remain debated. Here, we report evidence from gene duplications in LECA indicating an early origin of mitochondria. Among 163,545 duplications in 24,571 gene trees spanning 150 sequenced eukaryotic genomes, we identify 713 gene duplication events that occurred in LECA. LECA’s bacterial-derived genes include numerous mitochondrial functions and were duplicated significantly more often than archaeal-derived and eukaryote-specific genes. The surplus of bacterial-derived duplications in LECA most likely reflects the serial copying of genes from the mitochondrial endosymbiont to the archaeal host’s chromosomes. Clustering, phylogenies and likelihood ratio tests for 22.4 million genes from 5,655 prokaryotic and 150 eukaryotic genomes reveal no evidence for lineage-specific gene acquisitions in eukaryotes, except from the plastid in the plant lineage. That finding, and the functions of bacterial genes duplicated in LECA, suggests that the bacterial genes in eukaryotes are acquisitions from the mitochondrion, followed by vertical gene evolution and differential loss across eukaryotic lineages, flanked by concomitant lateral gene transfer among prokaryotes. Overall, the data indicate that recurrent gene transfer via the copying of genes from a resident mitochondrial endosymbiont to archaeal host chromosomes preceded the onset of eukaryotic cellular complexity, favoring mitochondria-early over mitochondria-late hypotheses for eukaryote origin.",
"conclusion": "Conclusion Serial transfers of mitochondrial DNA to the chromosomes of the host are not only a mechanism of gene duplication, they are a form of endosymbiont genome duplication in which an original copy is retained in the organelle and remains functional. Gene duplications in LECA support an early origin of mitochondria and record the onset of the eukaryotic gene duplication process, a hallmark of genome evolution in mitosing cells ( Ohno 1970 ; Scannell et al. 2006 ; Hittinger and Carroll 2007 ; Van De Peer 2009 ; Treangen and Rocha 2011 ).",
"introduction": "Introduction The last eukaryote common ancestor (LECA) lived about 1.6 Ba ( Betts et al. 2018 ; Javaux and Lepot 2018 ). It possessed bacterial lipids, nuclei, sex, an endomembrane system, mitochondria, and all other key traits that make eukaryotic cells more complex than their prokaryotic ancestors ( Speijer et al. 2015 ; Gould et al. 2016 ; Zachar and Szathmáry 2017 ; Barlow et al. 2018 ; Betts et al. 2018 ). The closest known relatives of the host lineage that acquired the mitochondrion are, however, small obligately symbiotic archaea from enrichment cultures that lack any semblance of eukaryotic cell complexity ( Imachi et al. 2020 ). This steep evolutionary grade separating prokaryotes from eukaryotes increasingly implicates mitochondrial symbiosis at eukaryote origin ( Gould et al. 2016 ; Imachi et al. 2020 ). Yet despite the availability of thousands of genome sequences, and five decades to ponder Margulis ( Margulis et al. 2006 ) resurrection of endosymbiotic theory (Mereschkowsky 1910; Wallin 1925 ), the timing, and evolutionary significance of mitochondrial origin remains a polarized debate. Gradualist theories contend that eukaryotes arose from archaea by slow accumulation of eukaryotic traits ( Cavalier-Smith 2002 ; Booth and Doolittle 2015 ; Hampl et al. 2019 ) with mitochondria arriving late ( Pittis and Gabaldón 2016 ), whereas symbiotic theories have it that mitochondria initiated the onset of eukaryote complexity in a nonnucleated archaeal host ( Imachi et al. 2020 ) by gene transfers from the organelle ( Martin and Müller 1998 ; Lane and Martin 2010 ; Gould et al. 2016 ; Martin et al. 2017 ). Information from gene duplications can help to resolve this debate. Gene and genome duplications are a genomic proxy for biological complexity and are the hallmark of eukaryotic genome evolution ( Ohno 1970 ). Gene families that were duplicated during the transition from the first eukaryote common ancestor (FECA) to LECA could potentially shed light on the relative timing of mitochondrial acquisition and eukaryote complexity if they could be inferred in a quantitative rather than piecemeal manner. Duplications of individual gene families ( Hittinger and Carroll 2007 ) and whole genomes ( Scannell et al. 2006 ; Van De Peer et al. 2009 ) have occurred throughout eukaryote evolution. This is in stark contrast to the situation in prokaryotes, where gene duplications are rare at best ( Treangen and Rocha 2011 ) and whole-genome duplications of the kind found in eukaryotes are altogether unknown. In an earlier study, Makarova et al. (2005) used a liberal criterion and attributed any gene present in two major eukaryotic lineages as present in LECA. Their approach overlooks eukaryotic lineage phylogeny, leading to the inference of 4,137 families that might have been duplicated in LECA. More recently, Vosseberg et al. (2021) examined nodes in trees derived from protein domains that could be scored as duplications among the 7,447–21,840 genes that they estimated to have been present in LECA and used branch lengths to estimate the timing of duplication events. However, they did not report integer numbers for duplications because of their approach based on the analyses of very large protein-domain trees instead of discrete protein-coding gene trees. Here, we addressed the problem of which, what kind of, and how many genes were duplicated in LECA and discuss the implications of our findings for the mitochondria-early versus mitochondria-late debate.",
"discussion": "Results and Discussion To ascertain when the process of gene duplication in eukaryote genome evolution commenced and whether mitochondria might have been involved in that process, we inferred all gene duplications among the 1,848,936 protein-coding genes present in 150 sequenced eukaryotic genomes. For this, we first clustered all eukaryotic proteins using a low stringency clustering threshold of 25% global amino acid identity (see Materials and Methods) in order to recover the full spectrum of eukaryotic gene duplications in both highly conserved and poorly conserved gene families. We emphasize that we employed a clustering threshold of 25% amino acid identity because our procedure was designed to allow for the construction of alignments and phylogenetic trees for each cluster. The 25% threshold keeps the alignments and trees out of the “twilight zone” of sequence identity ( Jeffroy et al. 2006 ), where alignment and phylogeny artifacts based on comparisons of nonhomologous amino acid positions arise. We then identified all genes that were duplicated across 150 sequenced eukaryotic genomes. In principle, genes present only in one copy in any genome could have also undergone duplication, with losses leading to single-copy status. Quantifying duplications in such cases are extremely topology-dependent. We therefore focused our attention on genes for which topology-independent evidence for duplications existed, that is, genes that were present in more than one copy in at least one genome. Eukaryotic gene duplications were found in all six supergroups: Archaeplastida, Opisthokonta, Mycetozoa, Hacrobia, SAR, and Excavata ( Adl et al. 2012 ), whereby 941,268 of all eukaryotic protein-coding genes, or nearly half the total, exist as multiple copies in at least one genome. These are distributed across 239,012 gene families, which we designate as multicopy gene families. However, 89.7% of these gene families harbor only recent gene duplications, restricted to a single eukaryotic genome (inparalogs). The remaining 24,571 families (10.3%) harbor multiple copies in at least two eukaryotic genomes, with variable distribution across the supergroups ( fig. 1 ). Opisthokonts (animals and fungi) together harbor a total of 22,410 multicopy gene families present in at least two genomes. The animal lineage harbors 19,530 multicopy gene families, the largest number of any lineage sampled, followed by the plant lineage (Archaeplastida) with 6,495 multicopy gene families. Of particular importance for the present study, among the 24,571 multicopy gene families, we identified 1,823 that are present as multiple copies in at least one genome from all six supergroups and are thus potential candidates of gene duplications tracing to LECA. In order to distinguish between the possibility of 1) duplications within supergroups after diversification from LECA and 2) duplications giving rise to multiple copies in the genome of LECA, we used phylogenetic trees. \n Fig . 1 Distribution of multicopy genes across 150 eukaryotic genomes. All eukaryotic protein-coding genes were clustered, aligned, and used for phylogenetic inferences. The resulting gene families present as multiple copies in more than one genome are plotted (see Materials and Methods). The figure displays the 24,571 multicopy gene families (horizontal axis) and the colored scale indicates the number of gene copies in each eukaryotic genome (vertical axis). The genomes were sorted according to a reference species tree ( supplementary data 7) and taxonomic classifications were taken from NCBI. Animals and fungi together form the opisthokont supergroup. To infer the relative phylogenetic timing of eukaryotic gene duplication events, we focused our attention on the individual protein alignments and maximum-likelihood trees for all 24,571 gene families with paralogs in at least two eukaryotic genomes. We then assigned gene duplications in each tree to the most recent internal node possible, allowing for multiple gene duplication events and losses as needed (see Materials and Methods) and permitting any branching order of supergroups. This approach minimized the number of inferred duplication events and identified a total of 163,545 gene duplications, 160,676 of which generated paralogs within a single supergroup (inparalogs at the supergroup-level). An additional 2,869 gene duplication events trace to the common ancestor of at least two supergroups ( fig. 2 a and supplementary table 1 ). The most notable result however was the identification of 713 gene duplication events distributed in 475 gene trees that generated paralogs in the genome of LECA before eukaryotic supergroups diverged. For these 475 gene trees, the resulting LECA paralogs are retained in at least one genome from all six supergroups, as indicated in red in figure 2 a . The sample of 475 genes provides a conservative estimate of genes that duplicated in LECA. Among the 1,823 gene families having multiple copies in members of all six supergroups, note that only in 475 families (26%) do the duplications actually trace to LECA in the trees. These results indicate that most duplications in eukaryotes are lineage specific ( figs. 1 and 2 ), and furthermore raise caveats regarding earlier estimates of duplications in LECA ( Makarova et al. 2005 ; Vosseberg et al. 2021 ) based on more permissive criteria. \n Fig . 2 Distribution of paralogs descending from gene duplications across six eukaryotic supergroups. ( a ) The figure shows the distribution of paralogs resulting from gene duplications in eukaryotic-specific genes (E-O) and eukaryotic genes with prokaryotic homologs (E-P) (see Materials and Methods for details). Duplicated genes refer to the numbers of gene trees with at least one duplication event with descendant paralogs across the supergroups (filled circles in the center). Number of duplication events refers to the total number of gene duplications. The red row circles indicate gene duplications with descendant paralogs in species from all six supergroups and, thus, tracing to LECA regardless of the eukaryotic phylogeny. An early study assigned 4,137 duplicated gene families to LECA but attributed all copies present in any two major eukaryotic groups to LECA ( Makarova et al. 2005 ). In the present sample, we find 2,869 gene duplication events that trace to the common ancestor of at least two supergroups. Our stringent criterion requiring paralog presence in all six supergroups leaves 713 duplications in 475 gene families in LECA. ( b ) Rooted phylogeny of eukaryotic supergroups that maximizes compatibility with gene duplications. Gene duplications mapping to five edges are shown (b 1 , b 2 , … , b 5 ). The tree represents almost exactly all edges containing the most duplications, the exception is the branch joining Hacrobia and SAR because the alternative branch joining SAR and Opisthokonta is better supported. However, the resulting subtree ((Opisthokonta, SAR),(Archaeplastida, Hacrobia)) accounts for 249 duplications, fewer than the (Opisthokonta,(Archaeplastida,(SAR, Hacrobia))) subtree shown (262 duplications). The position of the root identifies additional gene duplications tracing to LECA ( table 1 and supplementary table 4 ). LECA’s Duplications Constrain the Position of the Eukaryotic Root The six supergroups plus LECA at the root represent a seven-taxon tree with the terminal edges bearing 97% of gene duplication events ( fig. 2 ). Gene duplications that map to internal branches of the rooted supergroup tree can result from duplications in LECA followed by vertical inheritance and differential loss in some supergroups, or they result from more recent duplications following the divergence from LECA. Branches that explain the most duplications are likely to reflect the natural supergroup phylogeny, because support for conflicting branches is generated by random nonphylogenetic patterns of independent gene losses ( Van De Peer et al. 2009 ). There is a strong phylogenetic signal contained within the eukaryotic gene duplication data ( fig. 2 ). Among all possible internal branches, those supported by the most frequent duplications are compatible with the tree in figure 2 b , which places the eukaryotic root on the branch separating Excavates from other supergroups, as implicated in previous studies of concatenated protein sequences ( Hampl et al. 2009 ; He et al. 2014 ). However, massive gene loss in specific supergroups (in excavates, e.g., see fig. 1 ) could impair identification of the eukaryotic root ( Zmasek and Godzik 2011 ; Ku et al. 2015 ; Albalat and Cañestro 2016 ). Indeed, the high frequency of duplications that trace to LECA readily explains why resolution of deep eukaryotic phylogeny or the position of the eukaryotic root with traditional phylogenomic approaches ( Ren et al. 2016 ) is so difficult (see also supplementary table 2 ): LECA was replete with duplications and paralogy. Paralogy imposes conflicting signals onto phylogenetic systematics, but gene duplications harbor novel phylogenetic information in their own right ( fig. 2 ), as shared gene duplications discriminate between alternative eukaryote supergroup relationships. Eukaryotic Duplications Are Not Transferred across Supergroups Like the nucleus, mitochondria, and other eukaryotic traits ( Speijer et al. 2015 ; Gould et al. 2016 ; Zachar and Szathmáry 2017 ; Barlow et al. 2018 ; Betts et al. 2018 ; Imachi et al. 2020 ), the lineage-specific accrual of gene and genome duplications distinguish eukaryotes from prokaryotes (Ohno 1917; Scannell 2006 ; Hittinger and Carroll 2007 ; Van De Peer et al. 2009 ; Treangen and Rocha 2011 ). Nonetheless, one might argue that the distribution of duplications observed here does not reflect lineage-dependent processes at all, but lateral gene transfers (LGTs) among eukaryotes instead ( Andersson et al. 2003 ; Keeling and Palmer 2008 ; Leger et al. 2018 ). That is, a duplication could, in theory, originate in one supergroup and one or more gene copies could subsequently be distributed among other supergroups via eukaryote-to-eukaryote LGT. However, were that theoretical possibility true then neither duplications, nor any trait, nor any gene could be traced to LECA because all traits and genes in eukaryotes could, in the extreme, simply reflect 1.6 Byr of lineage-specific invention within one supergroup followed by lateral gene traffic among eukaryotes rather than descent with modification ( Andersson et al. 2003 ; Keeling and Palmer 2008 ; Leger et al. 2018 ). However, the present data themselves exclude the deeply improbable eukaryote-to-eukaryote lateral duplication transfer theory in a subtle but strikingly clear manner. How so? Figures 1 and 2 a show that 30,439 gene lineages bearing duplications (93% of the total) are restricted in their distribution to “only one supergroup,” whereas only 2,245 (7% of the total) are shared among two to five supergroups. That is, only 7% of the duplications are shared across supergroups, hence they are the only possible candidates for LGT among supergroups. For the sake of argument, let us entertain the extreme assumption that all 2,245 patterns of shared but nonuniversal duplications involved intersupergroup LGT, recalling that there is no intersupergroup LGT in 93% of the genes ( fig. 2 and supplementary table 1 ). With that generous assumption, the intersupergroup LGT frequency would be maximally 7%. That is an extreme upper bound, though, because the observed 93% frequency for duplicates that are supergroup specific and thus have absolutely no observable intersupergroup LGT should apply equally to the 7% of duplications shared across supergroups. Thus, the more realistic maximum estimate is that 0.49% of duplications (7% of 7%) might have been generated by intersupergroup LGT. This estimate is based solely upon the distribution of the duplicates and the premise that eukaryote supergroups are monophyletic. As it concerns the 475 genes with duplications that trace to LECA ( fig. 2 and supplementary table 1 ), this means that 0.49% out of 475, or about 2.3 genes in our data might have been caused by intersupergroup LGT. That is a very low frequency and is consistent with independent genome-wide phylogenetic tests presented previously ( Ku et al. 2015 ) for the paucity of eukaryote-to-eukaryote LGT. If we count duplication events ( fig. 2 a , right panel) rather than gene lineages ( fig. 2 a , left panel), the picture is even more vertical, because 98% of the events are supergroup-specific, hence lacking any patterns that could reflect LGT, meaning that maximally 0.04% (2% of 2%) or 0.19 duplications among 475 (which rounds to zero genes) could be the result of lateral transfer. The supergroup-specific distributions of duplications themselves thus provide very strong evidence that the distribution of duplicated genes in eukaryotes is not the result of eukaryote-to-eukaryote LGT phenomena ( Andersson et al. 2003 ; Keeling and Palmer 2008 ; Leger et al. 2018 ) but the result of vertical evolution within supergroups accompanied by gene birth, death ( Nei et al. 1997 ), and differential gene loss ( Ku et al. 2015 ). LECA’s Duplications Support an Early Mitochondrion Arguably, the timing of mitochondrial origin is the central so far unresolved issue at the heart of eukaryote origin. Several alternative theories for eukaryogenesis have been proposed (reviewed in Martin et al. 2001 ; Embley and Martin 2006 ; Poole and Gribaldo 2014 ; López-García and Moreira 2015 ; Eme 2017 ). Symbiogenic theories posit a causal role for mitochondrial endosymbiosis at the origin of cellular eukaryotic complexity ( Lane and Martin 2010) with the host being a garden variety archaeon ( Martin and Müller 1998 ). Gradualist theories posit an autogenous origin of eukaryote cell complexity with little or no contribution of the mitochondrion to eukaryogenesis ( Cavalier-Smith 2002 ; Gray 2014 ). Intermediate theories posit the existence of endosymbioses prior to the origin of mitochondria. These include an endosymbiotic origin of the nucleus ( Lake and Rivera 1994 ), an endosymbiotic origin of peroxisomes ( de Duve 2007 ), an endosymbiotic origin of flagella ( Margulis et al. 2000 ), the lateral acquisition of the cytoskeleton ( Doolittle 1998 ) or, more liberally, additional symbioses preceding the mitochondrion in unconstrained numbers, as long as each symbiosis “explains the origin of any eukaryotic innovation as a response to an endosymbiotic interaction” ( Gabaldón 2018 ). Most current theories posit an origin of the host from archaea ( Martin et al. 2015 ; Spang et al. 2015 ; Zaremba-Niedzwiedzka et al. 2017 ; Imachi 2020 ), though theories for eukaryote origins from actinobacteria ( Cavalier-Smith 2002 ), and planctomycetes ( Cavalier-Smith and Chao 2020 ) are discussed. Notwithstanding such diversity of views, the main divide among theories for eukaryote origin remains the relative timing of mitochondrial origin, that is did the mitochondrion initiate or culminate eukaryote origin ( Martin et al. 2001 ; Embley and Martin 2006 ; Poole and Gribaldo 2014 ; López-García and Moreira 2015 ; Eme et al. 2017 )? Alternative theories for eukaryote origin generate distinct predictions about the nature of gene duplications in LECA. Gradualist theories entailing an archaeal host ( Cavalier-Smith 2002 ; Booth and Doolittle 2015 ; Pittis and Gabaldón 2016 ; Hampl et al. 2019 ) predict genes of archaeal origin and eukaryote-specific genes to have undergone numerous duplications during the origin of eukaryote complexity, prior to the acquisition of the mitochondrion. In that case, the mitochondrion arose late, hence bacterial-derived genes would have accumulated fewer duplications in LECA than archaeal-derived or eukaryote-specific genes ( fig. 3 a ). Models invoking gradual lateral gene transfers (LGT) from ingested (phagocytosed) food prokaryotes prior to the origin of mitochondria ( Doolittle 1998 ) also predict more duplications in archaeal-derived and eukaryote-specific genes to underpin the origin of phagocytotic feeding, but do not predict duplications specifically among acquired genes (whether from bacterial or archaeal food) because each ingestion contributes genes only once. \n Fig . 3 Alternative models for eukaryote origin generate different predictions with respect to duplications. In each panel, gene duplications during the FECA to LECA transition (boxed in upper portion) are enlarged in the lower portion of the panel. ( a ) Cellular complexity and genome expansion in an archaeal host predate the origin of mitochondria. ( b ) Mitochondria enter the eukaryotic lineage early, duplications in mitochondrial-derived, host-derived, and eukaryotic-specific genes occur, genome expansion affects all genes equally. ( c ) Gene transfers from a resident endosymbiont generate duplications in genes of bacterial origin in an archaeal host. ( d ) Observed frequencies from gene duplications that trace to LECA (see supplementary table 1 ). BE refers to eukaryotic genes with bacterial homologs only; AE refers to eukaryotic genes with archaeal homologs only; and Euk refers to eukaryotic genes without prokaryotic homologs. ( e ) Schematic representation of serial gene transfers from the mitochondrion (white arrows) to the host’s chromosomes. By contrast, transfers from the endosymbiotic ancestors of organelles continuously generated gene duplications in the host’s chromosomes ( Timmis et al. 2004 ; Allen 2015 ), a process that continues to the present day in eukaryotic genomes ( Timmis et al. 2004 ; Portugez et al. 2018 ). Symbiogenic theories posit that the host that acquired the mitochondrion was an archaeon of normal prokaryotic complexity ( Martin and Müller 1998 ; Lane and Martin 2010 ; Gould et al. 2016 ; Martin et al. 2017 ; Imachi et al. 2020 ) and hence lacked duplications underpinning eukaryote complexity. There are examples known in which bacteria grow in intimate association with archaea ( Imachi et al. 2020 ) and in which prokaryotes become endosymbionts within other prokaryotic cells ( Martin et al. 2017 ). However, there are two different ways in which mitochondria could promote the accumulation of duplications. If energetic constraints ( Lane and Martin 2010) were the sole factor permitting genome expansion, duplications would accrue in all genes regardless of their origin, such that gene duplications in the wake of mitochondrial origin should be equally common in genes of bacterial, archaeal, or eukaryote-specific origin, respectively ( fig. 3 b ). If, on the other hand, the role of mitochondria in gene duplications was mechanistic rather than purely energetic, genes of mitochondrial origin should preferentially undergo duplication. This is because the mechanism of gene transfers from resident organelles involve endosymbiont lysis and the “copying” ( Allen 2015 ) of organelle genomes to the host’s chromosomes followed by recombination and mutation ( Portugez et al. 2018 ). Gene transfers from resident endosymbionts specifically generate duplications of endosymbiont genes because new copies of the same genes are recurrently transferred ( Timmis et al. 2004 ; Allen 2015 ) ( fig. 3 c ). The duplications in LECA reveal a vast excess of duplications in LECA’s bacterial-derived genes relative to archaeal-derived and eukaryote-specific genes ( fig. 3 d ). Of all gene families tracing to LECA, 26% experienced at least one duplication event during the transition to LECA from FECA. Notably, the excess proportion of duplicates among genes of bacterial origin is significant as judged by the two-tailed binomial test ( P =1.3×10 −10 ; proportion of duplicates at 95% CI=[35–44%]; df=1). On the other hand, genes of archaeal origin show significantly fewer duplicates ( P =8.4×10 −7 ; proportion of duplicates 95% CI=[8–17%]; df=1) with the proportion of duplicates being similar to eukaryote-specific genes ( fig. 3 d ). Do Bacterial Genes in LECA Stem from the Mitochondrion? If bacterial genes in LECA stem from the mitochondrion, as opposed to 1) eukaryote-to-eukaryote gene transfers, which were already excluded for >99% of the families with duplications in this data on the basis of their distributions alone, or 2) multiple lineage-specific acquisitions from bacteria via LGT, then the bacterial genes should trace to the eukaryote common ancestor. That is, the eukaryotes should form a monophyletic clade in gene trees that connect prokaryotic and eukaryotic genes. To test this, we generated clusters, alignments, and trees for genes shared by prokaryotes and eukaryotes from 22,471,723 million genes from 5,655 genomes and including 150 eukaryotes (see Materials and Methods). The results from the 2,575 trees that contained at least five prokaryotic and at least two eukaryotic sequences are summarized in figure 4 . As with the duplications themselves, eukaryote gene evolution is again vertical. Out of the 2,575 trees only 475 did not recover eukaryotes as monophyletic. However, none of these 475 trees rejected eukaryote monophyly using the Shimodaira–Hasegawa (SH) test (see Materials and Methods) and only 25 trees (1% of the total) rejected eukaryote monophyly using the Kishino–Hasegawa (KH) test. Applying the approximately unbiased (AU) test, only three trees out of 475 rejected eukaryote monophyly. This traces gene origin of ≥1,649 out of the 2,575 genes shared by prokaryotes and eukaryotes to LECA, and the origin of ≤926 genes to the archaeplastidal ancestor because the latter trees contain only photosynthetic eukaryotic lineages ( fig. 4 a ). \n Fig . 4 Identification of prokaryotic sisters in 2,575 eukaryotic–prokaryotic gene trees. ( a ) The individual trees were rooted on the branch leading to the largest prokaryotic clade deriving the sister group to eukaryotes. The average number of sequences in the eukaryotic clade, sister group, and outgroup are indicated. ( b ) The list of bacterial (top) and archaeal (bottom) phyla occurring in the trees exclusive to plant lineages (right) and all other trees (left). Archaeal and bacterial phyla with less than five representative species in the data set were collapsed into “other archaea” and “other bacterial” groups. P mono refers the proportion of trees with a branch (split) separating the species of the phylum from the others; S non refers to the number of occurrence of the phylum only in the outgroup clade; S mix refers to the number of occurrences of the phylum as a mixed sister (more than one phylum in the clade); S pure refers to the number of occurrences of the phylum as pure sister (as the single phylum); S p.avg shows the average size of the sister group when the phylum occurs as a pure sister clade. N trees show the number of occurrences of the phyla across all trees. ID gen refers to the total number of species in each phylum. The 1,649 trees that trace prokaryotic gene origins to LECA fall into two classes with regard to the sister group of the eukaryotic gene: 966 in which the prokaryotic sister group to eukaryotes contained members of only one phylum (a “pure” sister, S pure in fig. 4 , 59% of the trees) and those in which the sister to the eukaryotes contained members of more than one phylum (a “mixed” sister, 41% of the trees). The only way to obtain a mixed sister topology of prokaryotic sequences for a eukaryotic gene is via LGT among prokaryotes ( Ku and Martin 2016 ). If we exclude the reality of LGT among prokaryotes, and interpret mixed sister topologies at face value, they would suggest that eukaryotes arose before the diversification of the diverse prokaryotic phyla present in our sample, which would be incompatible with accounts of eukaryote age ( Parfrey et al. 2011 ; Betts et al. 2018 ), and would furthermore have LECA arising at different times, depending on the membership in the sister group. LGT among the prokaryotic reference sequences in the mixed sister cases ( Ku and Martin 2016 ; Nagies et al. 2020 ) is clearly the simpler explanation. The pure sister was bacterial in 49% of the trees and archaeal in only 9.5% of the trees. Only in 115 trees (7.0%) was the bacterial pure sister clade alphaproteobacterial. These 115 trees are readily explained because they stem from the mitochondrion, even though the alphaproteobacterial-derived genes in eukaryotes do not all reside in the “same” alphaproteobacterial genome as previously observed ( Ku et al. 2015 ; Nagies et al. 2020 ), requiring LGT among alphaproteobacteria, at least, to account for the topology. Yet, the crucial and previously underinvestigated issue concerns the remaining 695 pure sister bacterial origin cases (86%) that trace to LECA but reside in a genome that does not carry an alphaproteobacterial taxon label ( fig. 4 ), as recently set forth in a study that examined the phylogeny of only the more conserved fraction of genes shared by prokaryotes and eukaryotes ( Nagies et al. 2020 ). There are two general ways to explain the 86% of nonalphaproteobacterial genes that trace to LECA. The first is to take one specific aspect of the trees—namely, the taxon label of the sister group—at face value and interpret the data as evidence for independent individual contributions to eukaryotes (via LGT or via multiple resident symbionts) by all of the bacterial phyla in the sample. At the level of the taxa listed in figure 4 , that would mean 26 different bacterial donors to LECA in addition to the alphaproteobacterial contribution, and donations from 13 different archaeal host taxa. With 39 donor phyla, LECA already looks like a grab bag of genes. At the level of genus, the taxon labels of the trees would mean 794 different bacterial donors to LECA under permissive models ( Gabaldón 2018 ), followed by a particularly ad hoc sudden stop of gene influx to eukaryotes after the FECA to LECA transition, because the eukaryotes are monophyletic in these trees. The suggestion of symbiont acquisition and gene transfers without constraints ( Gabaldón 2018 ) carries a hidden and seldom spelled out corollary ( Martin 1999 ). Namely, it entails the strict condition that all of the nonalphaproteobacterial bacterial genes in question not only resided in the genome of members of the 27 different phylum level bacterial taxa at the time of donation to LECA ( fig. 4 ) but furthermore, and crucially, that those genes evolved “vertically” within the chromosomal confines of those respective phyla during the 1.6 Byr since eukaryotes arose. Such unrestricted donor theories ( Gabaldón 2018 ) assume that the present-day phylum taxon label on the gene accurately identifies the donor phylum at the time of transfer. But that is true “if and only if” the gene has been vertically inherited within that phylum (no interphylum LGT) since its donation to LECA ( Martin 1999 ; Esser et al. 2007 ). Such theories of unrestricted LGT to eukaryotes with strictly vertical gene evolution among prokaryotes are unlikely and resoundingly rejected by the data. If we look beyond the mere taxon label of the sister group ( fig. 4 ), we see that the putative 27 bacterial donor lineages themselves do not evolve in a vertical manner. The average level of monophyly for bacterial phyla in the 1,649 trees that trace to LUCA is 47% ( P mono in fig. 4 ). Alphaprotebacteria were monophyletic in only 27% of the trees in which they occurred, as were generalists with large genomes such as betaproteobacteria (27%) and actinobacteria (33%). Specialists like chlorobi or chlamydia with more restricted pangenomes were more monophyletic (80% and 72%, respectively). Halophilic archaea, which are known to have acquired many genes from bacteria ( Nelson-Sathi et al. 2012 ), are the least monophyletic prokaryotes sampled (halobacteriales, 16%, fig. 4 ). For the 926 genes that, based on their distribution, trace to the archaeplastidal common ancestor ( fig. 4 , right panel), the bacterial phyla have a higher proportion of monophyly ( P =0.006, V = 67, using two-tailed Wilcoxon signed-rank test) than for those genes that trace to LECA. Plastids are younger than mitochondria, hence the genes from the ancestral plastid genome have had less time to migrate across prokaryotic genomes than genes from the ancestral mitochondrial genome. For the prokaryotic genes and phyla in question, evolution is not a vertical process. The bacterial reference system against which to infer the origin of eukaryotic genes that stem from the mitochondrion (or the plastid) is a system of mosaic ( Martin 1999 ) or fluid ( Esser et al. 2007 ) chromosomes. These findings are fully consistent with a recent larger scale investigation of gene verticality across genomes ( Nagies et al. 2020 ). If we accept the evidence that LGT in prokaryotes is real and if we accept the evidence that mitochondria were once endosymbiotic bacteria, then the expectation for the phylogeny of a gene that was acquired from the mitochondrion is that it traces to a single origin in LECA, which the genes in this study do, but “not” that it traces to alphaproteobacteria. This is because LGT among prokaryotes preceding and subsequent to the origin of mitochondria generates the illusion of many donors by shuffling the taxon labels attached to genes in mosaic bacterial chromosomes ( Martin 1999 ). Most current studies still equate mitochondrial origin with an alphaproteobacterial sister group relationship ( Vosseberg et al. 2021 ), but if we look at all the data, it is clear that such an interpretation is too strict. For example, Vosseberg et al. (2021) found that about 7% of the eukaryotic protein-domains that they examined branched with alphaproteobacterial homologs. But looking beyond the eukaryotic branch, Nagies et al. (2020) found that only about 35% of alphaproteobacterial genes recover alphaproteobacteria monophyly to begin with, and only 16% of the 220 trees in which alphaproteobacteria appeared as the sole sister of all eukaryotes recovered aphaproteobacteria as monophyletic among prokaryotes. To investigate mitochondrial origin from the standpoint of genes, it is not enough to identify the relationship of eukaryote genes to prokaryotic homologs. One has also to investigate the relationship of prokaryotic homologs to each other, because they are the reference system for comparison. It is because of LGT among prokaryotes that many different groups are implicated as donors of genes to LECA ( fig. 4 ; see also Nagies et al. 2020 ). There is no evidence independent of gene phylogenies to suggest or support theories for the participation of spirochaetes ( Margulis et al. 2006 ), actinobacteria ( Cavalier-Smith 2002 ), cyanobacteria ( Cavalier-Smith 1975 ), deltaproteobacteria ( López-García and Moreira 1999 ), planctomycetes ( Cavalier-Smith and Chao 2020 ), or multiple donor lineages ( Gabaldón 2018 ) at eukaryote origin ( Embley and Martin 2006 ). One could of course argue that those conflicting theories for contributions from many different prokaryotic lineages are all simultaneously true, but then theories for eukaryogenesis would no longer be constrained by observations in data, and any assertion about eukaryote origin would be permissible as a line of evidence, an untenable state of affairs. The same sets of considerations apply to the cyanobacterial origin of plastids ( fig. 4 ). If we let go of the belief that sister group relationships between eukaryotic genes and prokaryotic homologs ( fig. 4 ) identify the prokaryotic lineages that donated genes ( Martin 1999 ; Nagies et al. 2020 ), and take into account the functions encoded by nuclear genes of bacterial origin that were duplicated in LECA ( figs. 2 and 4; table 1 ), the simplest interpretation of the data in our view is that the bacterial duplicates in LECA were donated by the mitochondrion. Other more complicated interpretations are imaginable, but these interpretations do not simultaneously account for the phylogenetic behavior of the bacterial reference phylogeny set, which we have done here and elsewhere ( Nagies et al. 2020 ). Our data furthermore show that eukaryotic genes are of monophyletic origin. With large genomic samples spanning thousands of reference prokaryotic genomes, eukaryotic gene evolution is clearly vertical, both in terms of lineage-specific distribution of gene duplications ( fig. 1 ) and in terms of likelihood ratio tests ( Nagies et al. 2020 ). Table 1 Functional Categories of Genes Duplicated in LECAa Category b ( n ) Bacterial Archaeal Universal Eukaryotic Metabolism (141) 64 2 58 17 Protein modification, folding, degradation (89) 30 8 30 21 Ubiquitination 3 1 — 9 Proteases 9 1 7 1 Kinase/phosphatase/modification 12 6 19 9 Folding 6 — 4 2 Novel eukaryotic traits (61) 8 4 12 37 Cell cycle 1 1 2 5 Cytoskeleton 4 — 1 19 Endomembrane (ER; Golgi; vesicles) 2 2 8 10 mRNA splicing 1 1 1 3 Mitochondrion (47) 29 — 9 9 Carbon metabolism (37) 26 — 11 — Glycolysis 10 — 5 — Reserve polysaccharides, other 16 — 6 — Cytosolic translation (36) 15 7 10 4 Nucleic acids (55) 13 7 15 20 Histones — — 2 8 RNA 8 3 6 4 DNA 5 4 7 8 Membranes (excluding endomembrane) (46) 18 1 12 15 Transporters, plasma associated 8 1 9 14 Lipid synthesis 10 — 3 1 Redox (15) 11 — 4 — Hypothetical (229) 81 9 61 78 Total 295 38 222 201 \n Note .— n , number of duplicated genes in the corresponding category. a About 475 genes duplicated in LECA and present in all six supergroups plus 281 genes with duplications tracing to the common ancestors of excavates and other supergroups. The annotation, source (bacterial, archaeal, present in bacteria and archaea, eukaryote specific), and the numbers of duplications for each cluster are given in supplementary tables 3 and 4 . All categories listed had representatives on both the 475 and the 281 list except mRNA splicing, present in the 475 list only. b The categories do not strictly adhere to KEGG or gene ontology classifications, instead they were chosen to reflect the processes that took place during the FECA to LECA transition. The largest number of duplications in LECA for any individual gene was 12, a dynein chain known from previous studies to have undergone duplications in the common ancestor of plants animals and fungi ( Kollmar 2016 ). Can Positive Selection Explain Excess Bacterial Duplications? The vast excess of bacterial duplications ( fig. 3 ) and the phylogenies of 2,575 genes that would address the question of gene origin ( fig. 4 ) speak in favor of bacterial acquisition in LECA from a single-resident endosymbiont, the mitochondrion, prior to the origin of eukaryote complexity. Yet one could still imagine numerous individual gene acquisitions in LECA from different donors with a blanket ad hoc hypothesis of “positive selection” increasing the copy number of bacterial-related functions to account for the excess of bacterial-derived duplications ( table 1 ). However, the selection proposal would not explain the excess of bacterial over archaeal or eukaryote-specific genes with the same functional category, as is widely observed in table 1 . That is, selection would have to be invoked as a special plea on a bacterial-gene-for-bacterial-gene basis, requiring yet one additional corollary of positive selection for each duplication. Because we observe over 900,000 duplications in the present data, the selection theory to account for duplications carries a burden of too many corollary assumptions. On the other hand, it is possible that duplications are fundamentally mechanistic in origin, via chromosome mispairing, translocations, genome duplications, or via duplicative transfers from a resident endosymbiont as we argue in this paper. In a context of mosaic, fluid bacterial genomes ( Martin 1999 ; Esser et al. 2007 ) permitting LGT among prokaryotes ( fig. 4 ) ( Nagies et al. 2020 ), we would require no corollary assumptions of ad hoc selection. The mechanism of transfer from the endosymbiont generates the excess of bacterial duplications and does so across all functional categories ( table 1 ). The Functions of Bacterial Duplicates Polarize Events at LECA’s Origin Gene duplications speak to more than phylogeny. Gene duplications are a standard proxy for the evolution of complexity, as diversification of function and form is canonically underpinned by gene family expansion ( Ohno 1970 ). Accordingly, we observe that the morphologically most complex multicellular eukaryotes—plants, animals, and fungi—harbor the largest numbers of duplications ( fig. 1 ). As outlined above, the simplest interpretation of the present data is that complexity started with the mitochondrion. That is not only true for the present data on duplications, is also true from a purely physiological standpoint ( Martin et al. 2017 ) and a bioenergetic standpoint ( Lane and Martin 2010) . The functions of genes that were duplicated in LECA help to polarize events in LECA’s evolution. For example, LECA had a mitochondrion. LECA’s gene duplications in 47 genes with mitochondrial functions include pyruvate dehydrogenase complex, enzymes of the citric acid cycle, components involved in electron transport, a presequence cleavage protease, the ATP–ADP carrier, and seven members of the eukaryote-specific mitochondrial carrier family that facilitates metabolite exchange between the mitochondrion and the cytosol ( table 1 and supplementary tables 3 and 4 ). A recent study estimated that some genes for mitochondrial function were probably duplicated in LECA, but interpreted the data as evidence for mitochondria-intermediate hypothesis ( Vosseberg et al. 2021 ). The methodology used in Vosseberg et al. has major limitations because: 1) the timing of gene duplications was inferred using an approach that equates branch-lengths from phylogenetic trees to time, which is expected to be valid “only if” the evolutionary rate is constant across genes (substitutions and gene loss, for example); 2) prokaryotic sequences were arbitrarily removed from gene trees, inflating the estimates of duplications in genes of archaeal origin; 3) the use of trees for which the same gene sequence can be represented simultaneously in multiple trees, biasing the estimates of duplications and their origin; and 4) the use of too liberal thresholds for gene clustering which result in aberrantly large gene families (see supplementary fig. 5 , Supplementary Material online), a potential source of tree reconstruction errors. By contrast, we do not infer time from branch lengths, we did not remove sequences that did not fit our expectations, and gene membership in our gene families is always unique. Our findings clearly indicate that canonical energy metabolic functions of mitochondria were established in LECA, underscored by additional functions performed by mitochondria in diverse eukaryotic lineages: ten genes for enzymes of the lipid biosynthetic pathway (typically mitochondrial in eukaryotes; Gould et al. 2016 ), the entire glycolytic pathway (mitochondrial among marine algae; Río Bártulos et al. 2018 ), and 11 genes involved in redox balance are found among bacterial duplicates. The largest category of duplications with annotated functions concerns metabolism and biosynthesis ( table 1 ). Many products of bacterial-derived genes operate in the eukaryotic cytosol ( Martin et al. 1993 ; Esser et al. 2004 ). This is because at the outset of gene transfer from the endosymbiont, there was no mitochondrial protein import machinery ( Martin and Müller 1998 ; Dolezal et al. 2006 ), and no nucleus, such that the products of genes transferred from the endosymbiont were active in the compartment where the genes were cotranscriptionally translated ( French et al. 2007 ). Gene transfers in large, genome sized fragments from the endosymbiont, as they occur today ( Timmis et al. 2004 ; Portugez 2018 ), furthermore, permitted entire pathways to be transferred, because the unit of biochemical selection is the pathway and its product, not the individual enzyme ( Martin 2010 ). In the absence of upstream and downstream intermediates and activities in a pathway, the product of a lone transferred gene is generally useless for the cell, expression of the gene becomes a burden, and the transferred gene cannot be fixed ( Martin 2010 ). Bacterial-derived duplications are present in functions that underpinned the origin of cell compartmentation in LECA ( table 1 ). LECA possessed an endomembrane system consisting of bacterial lipids, as symbiogenic models predict ( Gould et al. 2016 ). Bacterial duplicates, not archaeal duplicates, dominate lipid synthesis and membrane biogenesis ( table 1 ). Functions of bacterial duplicates are also involved in mRNA splicing, a selective force at the origin of the nucleus ( Garg and Martin 2016 ; Eme et al. 2017 ). The origin of protein import into mitochondria was essential to mitochondrial origin ( Dolezal et al. 2006 ) and encompasses many bacteria-derived duplicates ( table 1 ). LECA’s duplicates of bacterial origin are also involved in the origin of eukaryotic-specific traits, including the cell cycle, the cytoskeleton, endomembrane system, and mRNA splicing ( table 1 ). Eukaryote complexity required intracellular molecular movement in the cytosol, which is realized by motor proteins. The protein with the most duplications found in LECA is a light chain dynein with 12 duplications ( supplementary table 3 ), in agreement with previous studies of dynein evolution that document massive dynein gene duplications early in eukaryote evolution ( Kollmar 2016 ). Notably, ten of the 20 genes encoding cytoskeletal functions that were duplicated in LECA ( supplementary tables 3 and 4 ) encode dynein or kinesin motor proteins (see also Tromer et al. 2019 ). The bacterial duplicate contribution vastly outnumbers the archaeal contribution to these categories, which are dominated by eukaryote-specific genes, indicating that eukaryotes not only acquired genes, but they also invented new ones as well ( Lane and Martin 2010) . Duplications in LECA depict bacterial carbon and energy metabolism in an archaeal host supported by genes that were recurrently donated by a resident symbiont, in line with the predictions of symbiotic theories for the nature of the first eukaryote ( Martin and Müller 1998 ; Martin et al. 2017 ; Imachi et al. 2020 ). The functions of duplications are consistent with the predictions of symbiogenic theories but contrast with gradualist theories positing eukaryote origin from an archaeal lineage that attained eukaryote-like complexity in the absence of the mitochondrial endosymbiont ( Cavalier-Smith 2002 ; Booth and Doolittle 2015 ; Pittis and Gabaldón 2016 ; Hampl et al. 2019 ). What Does This Say about the Biology of LECA? Gene transfers from the mitochondrion can generate duplications of bacterial-derived genes. What mechanisms promoted genome-wide gene duplication at the prokaryote–eukaryote transition? Population genetic parameters such as variation in population size ( Zachar and Szathmáry 2017) apply to prokaryotes and eukaryotes equally, hence they would not affect gene duplications specifically in eukaryotes, but recombination processes ( Garg and Martin 2016) in a nucleated cell could. Because LECA possessed meiotic recombination ( Speijer et al. 2015 ), it was able to fuse nuclei (karyogamy). Karyogamy in a multinucleate LECA would promote the accumulation of duplications in all gene classes and promote genome expansion to its energetically permissible limits ( Lane and Martin 2010) because unequal crossing between imprecisely paired homologous chromosomes following karyogamy generates duplications ( Ohno 1970 ; Scannell et al. 2006 ; Hittinger and Carroll 2007 ; Van De Peer 2009 ). At the origin of meiotic recombination, chromosome pairing and segregation cannot have been perfect from the start; the initial state was likely error-prone, generating nuclei with aberrant gene copies, aberrant chromosomes, and even aberrant chromosome numbers. In cells with a single nucleus, such variants would have been lethal; in multinucleate (syncytial or coenocytic) organisms, defective nuclei can complement each other through mRNA in the cytosol ( Garg and Martin 2016) . Multinucleate forms are present throughout eukaryotic lineages ( fig. 5 ), and ancestral reconstruction of nuclear organization clearly indicates that LECA itself was multinucleate ( fig. 5 and supplementary fig. 1 , Supplementary Material online). The multinucleate state enables the accumulation of duplications in the incipient eukaryotic lineage in a mechanistically nonadaptive manner, whereby duplications are implicated in the evolution of complexity ( Ohno 1970 ; Scannell et al. 2006 ; Hittinger and Carroll 2007 ; Van De Peer 2009 ), as observed in the animal lineage ( fig. 1 ). The syncytial state presents a viable intermediate state in the transition from prokaryote to eukaryote genetics. \n Fig . 5 Ancestral state reconstruction for nuclear organization in eukaryotes. Presence and absence of the multinucleate state in members of the respective group are indicated. Resolution of the branches (polytomy vs. dichotomy) does not alter the outcome of the ancestral state reconstruction, nor does position of the root on the branches leading to Amoebozoa, Excavata, or Opisthokonta. LECA was a multinucleate, syncytial cell, not uninucleate (see supplementary fig. 1 , Supplementary Material online). Together with mitochondrion and sex, the multinucleate state is ancestral to eukaryotes and fostered accumulation of duplications (see text). Conclusion Serial transfers of mitochondrial DNA to the chromosomes of the host are not only a mechanism of gene duplication, they are a form of endosymbiont genome duplication in which an original copy is retained in the organelle and remains functional. Gene duplications in LECA support an early origin of mitochondria and record the onset of the eukaryotic gene duplication process, a hallmark of genome evolution in mitosing cells ( Ohno 1970 ; Scannell et al. 2006 ; Hittinger and Carroll 2007 ; Van De Peer 2009 ; Treangen and Rocha 2011 )."
} | 13,022 |
33830652 | PMC8601198 | pmc | 1,011 | {
"abstract": "Summary Obligate methanotrophic bacteria can utilize methane, an inexpensive carbon feedstock, as a sole energy and carbon substrate, thus are considered as the only nature‐provided biocatalyst for sustainable biomanufacturing of fuels and chemicals from methane. To address the limitation of native C1 metabolism of obligate type I methanotrophs, we proposed a novel platform strain that can utilize methane and multi‐carbon substrates, such as glycerol, simultaneously to boost growth rates and chemical production in Methylotuvimicrobium alcaliphilum 20Z . To demonstrate the uses of this concept, we reconstructed a 2,3‐butanediol biosynthetic pathway and achieved a fourfold higher titer of 2,3‐butanediol production by co‐utilizing methane and glycerol compared with that of methanotrophic growth. In addition, we reported the creation of a methanotrophic biocatalyst for one‐step bioconversion of methane to methanol in which glycerol was used for cell growth, and methane was mainly used for methanol production. After the deletion of genes encoding methanol dehydrogenase (MDH), 11.6 mM methanol was obtained after 72 h using living cells in the absence of any chemical inhibitors of MDH and exogenous NADH source. A further improvement of this bioconversion was attained by using resting cells with a significantly increased titre of 76 mM methanol after 3.5 h with the supply of 40 mM formate. The work presented here provides a novel framework for a variety of approaches in methane‐based biomanufacturing.",
"introduction": "Introduction The increasing threat posed by climate change as a result of combustion of fossil fuels has drawn much attention to bio‐based production of value‐added chemicals and fuels. Methane is among the most potent greenhouse gases, with a short‐term global warming potential much greater than that of CO 2 (Clomburg et al., 2017 , Hwang et al., 2018 ). On the contrary, methane is the main component of natural gas and biogas, and its high availability and low cost make it a promising carbon substrate for industrial biotechnology (Haynes and Gonzalez, 2014 ; Strong et al., 2016 ). Methanotrophic bacteria can utilize methane as a sole carbon and energy source, giving them the potential to serve as promising hosts for methane sequestration and conversion to diverse biofuels and chemical intermediates (Nguyen et al., 2020a ). Despite many attempts to employ methanotrophic bacteria as biocatalysts for methane bioconversion, many technical challenges and limitations to our fundamental knowledge remain to be addressed to this important microbial group (Peyraud et al., 2011 , Kalyuzhnaya et al., 2015 ). Recent efforts to engineer methanotrophs have successfully produced some chemicals and fuels, but only at titres and productivity levels far below those required by industry (Nguyen et al., 2018 ). Novel engineering strategies are therefore required. \n In silico simulation of genome‐scale metabolic models in methanotrophs suggests that the reason for the lower yields of target products is not only due to the low efficiency of methane assimilation but also the unbalanced reducing equivalents used in methane oxidation by methane monooxygenase (Akberdin et al., 2018 , Bordel et al., 2019 ). In the other hand, although methane is the most reduced carbon, the supply of electrons from methane to the production of reduced products is limited due to the electron requirements for oxygen splitting during methane oxidation to methanol (Haynes and Gonzalez, 2014 ). Moreover, the redox‐neutral conversion of methane to formaldehyde in methanotrophs results in a 36% energy loss which causes slow growth and low cell yield of methanotrophs on methane (Haynes and Gonzalez, 2014 ). In addition, the oxidative phosphorylation under aerobic conditions competes reducing power pool in the form of NADH against reduced fuels and chemicals biosynthesis (Bennett et al., 2018 ). One possible solution to overcome this issue is to introduce methane utilizing pathway in fast‐growing bacteria like E. coli in which sugar substrate metabolism might supply reducing power for reduced products. Even though many efforts have been made to heterologously express methane monooxygenase, methanol dehydrogenase, formaldehyde dehydrogenase in a non‐native host, this approach faces engineering challenges for chemicals production due to the difficulties in operation of a methanotrophic lifestyle involving multiple‐enzyme expression levels, gene regulation and autocatalytic cycle constraints (West et al., 1992 , Woolston et al., 2018 , Kim et al., 2019 , Keller et al., 2020 ). To date, no heterologous host strain can replace methanotrophic bacteria to efficiently utilize methane as a carbon source for cell growth and production. Therefore, it is important to reconstruct co‐metabolism of another reduced substrate together with methane in methanotrophs to boost bacterial growth and provide extra reducing power pool to production of reduced fuels and chemicals. Enabling methanotrophs as a novel ‘platform strain’ for production of fuels and chemicals in a more reduced state by co‐utilizing methane and reduced multi‐carbon substrates would not only increase growth rates, but also provide reducing equivalents to enhance the yields of fuels and chemicals in a more reduced state. The reduced multi‐carbon substrate should be inexpensive and abundant in order to be economically feasible substrate. In the current context, glycerol, a by‐product of biodiesel production, is a suitable candidate. As biodiesel production increases, crude glycerol as 10% of the total production is generated (He et al., 2017 ). In fact, the global annual production of glycerol is estimated to reach about 4.2 million tons in 2020, and while the demand for glycerol is predicted to be less than 3.5 million tons (Li et al., 2018 , Nomanbhay et al., 2018 ), this makes crude glycerol simply industrial waste which not only reduces market price but also causes environmental issue. In addition to its high abundance and low cost, high reduced level of glycerol compared with other sugars can deliver more reducing equivalents for a metabolic system (Durnin et al., 2009 , Zhang et al., 2013 ). Moreover, crude glycerol contains a significant level of methanol (Yang et al., 2012 , He et al., 2017 ), which can also be utilized by methanotrophs. It would be advantageous if methanotrophs could utilize crude glycerol without extra purification cost. Thus, the development of methanotrophic platform strain capable of utilizing methane and glycerol can address issue on the limitation of reducing power for the production of target products in a more reduced state and provide solutions to valorize waste materials such as waste methane gas and crude glycerol into value‐added products. To demonstrate the potential of this conceptual design for methanotrophic platform strain, two distinct categories are illustrated in this study. For the first category, we demonstrated that efficient utilization of glycerol as the co‐substrate could provide extra reducing power for enhancing reduced product formation. To meet this objective, we reconstructed a mixed heterotrophic mode in Methylotuvimicrobium alcaliphilum 20Z for co‐metabolism of methane and glycerol to enhance the growth rate and productivity of 2,3‐butanediol as a model compound that requires additional NADH in the biosynthetic pathway. For the second category, co‐substrate, glycerol, plays a growth‐supporting role that results in the novel framework for the bioconversion of methane to methanol. In this approach, we took advantages of methane monooxygenase machinery for alkane oxidation reaction, not whole C1 assimilation metabolism. In this platform strain, glycerol was utilized to sustain cell growth, while methane was directly converted to methanol with complete removal of methanol dehydrogenase (MDH) to prevent further oxidation. In the conventional approach of methane‐to‐methanol bioconversion, various chemical inhibitors should be added to inhibit MDH, which also inhibit other key enzymes for bioconversion and increase process cost. In our engineered strain, we completely deleted genes encoding MDH that enabled the methane‐to‐methanol bioconversion without addition of any chemical inhibitor. This study proposes a promising engineering strategy for development of an efficient methanotrophic biocatalyst for sustainable methane‐based biorefinery.",
"discussion": "Results and discussion Establishing the effective glycerol‐utilizing strain of M. alcaliphilum 20Z Two alternative pathways have been proposed for glycerol utilization in E. coli , one involving the GlpK‐GlpD route under respiratory conditions (Booth, 2005 ) and the other comprising the GlpA‐DhaK route under fermentative conditions (Gonzalez et al., 2008 ). Glycerol crosses the cell membrane by passive diffusion or via a glycerol‐uptake facilitator such as GlpF (Voegele et al., 1993 , Maurel et al., 1994 ). For the respiratory pathway, glycerol is phosphorylated by glycerol kinase (GlpK) to form glycerol 3‐phosphate (G3P), which is then oxidized to dihydroxyacetone phosphate (DHAP), simultaneously generates FADH 2 by membrane‐bounded FAD + ‐dependent G3P dehydrogenase (GlpD) and transfers electron to quinone pool in electron transport chain (Booth, 2005 ). The GlpA‐DhaK route is the preferred respiratory pathway for glycerol utilization in terms of providing excess reducing equivalents. Glycerol is dehydrogenated to dihydroxyacetone, and NAD + is reduced to NADH by NAD + ‐dependent glycerol dehydrogenase (GlpA). DHA is then phosphorylated by ATP‐ or PEP‐dependent DHA kinases (DhaK) to generate DHAP (Gonzalez et al., 2008 ). \n M. alcaliphilum 20Z is an obligate methanotrophic bacterium that is naturally unable to utilize multi‐carbon substrates (Akberdin et al., 2018 ). The strain also does not possess a pathway for glycerol utilization or a glycerol repressor (GlpR). In fact, the wild‐type strain was cultured on 0.1% (v/v) glycerol as a sole carbon source; however, no growth was observed (Fig. S1 ). To enable a synthetic glycerol‐utilizing pathway in M. alcaliphilum 20Z, a glycerol transporter (GlpF), together with both GlpK‐GlpD and GlpA‐DhaK routes, has been separately constructed into a pAWP89 backbone vector under control of a P tac promoter (Fig. 1 ). In both pathways, catabolic flux of glycerol is expected to be incorporated into central metabolism at the DHAP node and then bifurcated towards the ribulose monophosphate and tricarboxylic acid cycles via the gluconeogenic and glycolytic fluxes as a result of fructose‐bisphosphate aldolase (FbaA) and triosephosphate isomerase (Tpi) activity, respectively (Fig. 1 ). The results showed that M. alcaliphilum 20Z strain harbouring a GlpA‐DhaK route did not grow on glycerol. This may be due to the fact that the GlpA‐DhaK route is functionally active in E. coli under anaerobic conditions; thus, the GlpA‐DhaK route may be inactive under aerobic conditions associated with cultivation of M. alcaliphilum 20Z. Although glycerol can be utilized under both aerobic and anaerobic conditions, the process of co‐utilizing glycerol with methane to enhance growth rates and productivity needs to be carried out under aerobic conditions because of methane oxidation, and the fermentative pathway is not suitable for methanotrophs in this context. Unlike the GlpA‐DhaK route, growth of the engineered strain (20Z_FKD) with GlpK‐GlpD routes was observed in 0.1% (v/v) glycerol, with a relatively low optical cell density (OD 600 ) of 0.3 after 40 h (Fig. S1 ). Previously, the poor growth on glycerol of some bacteria was also reported, and it was suggested that this glycerol utilization pathway might be toxic for cells due to unbalanced redox power (Kang et al., 2014 ), allosteric inhibition (Applebee et al., 2011 , Zhan et al., 2018 ) or intermediate toxicity (Rittmann et al., 2008 ). In E. coli , the glpFK operon is tightly controlled by the Glp repressor, which is not present in M. alcaliphilum 20Z, and expression of two‐gene cluster glpFK may not be the main reason for poor growth on glycerol in this engineered strain (Weissenborn et al., 1992 ). It has been inferred that conversion of G3P to DHAP is a rate‐limiting step because of unfunctional expression of a recombinant, membrane‐binding enzyme, FAD + ‐dependent GlpD, in an ultrastructure membrane system of methanotrophs due to saturation of the capacity of the Sec‐translocon in native strain (Zhang et al., 2015 ). It has been reported that accumulation of G3P inhibits cell growth under high activity of glycerol kinase, a cytoplasmic enzyme (Applebee et al., 2011 ). In addition, DHAP was excluded because this is a branch point in central metabolism of methanotrophs, and cells have mechanisms that convert further DHAP. Therefore, to improve carbon flux for biomass synthesis, generate reducing equivalents, especially NADH, and enhance oxidation of G3P to DHAP, it was necessary to express another alternatively soluble enzyme. We found a version of NAD + ‐dependent G3P dehydrogenase (GpsA) that is encoded by gpsA gene from M. alcaliphilum 20Z (MEALZ_0434). This gene was co‐overexpressed with three genes glpFKD , resulting in the 20Z_FKDA strain, which exhibited a higher growth rate on glycerol than in 20Z_FKD, reaching a final OD 600 of 1.4 and consuming ˜ 1 g l −1 glycerol (Fig. 2A ). In summary, overexpression of gpsA gene along with glpFKD enabled sustainable growth of M. alcaliphilum 20Z on glycerol. Although enzyme activity of GpsA has not been assayed in M. alcaliphilum 20Z, the existence of an orthologous gene in E. coli , as shown by experimental evidence at the protein level, indicated that G3P can be converted into DHAP and generate NADH simultaneously (Rho and Choi, 2018 ). Fig. 1 Metabolic engineering strategy of 2,3‐BDO production by co‐utilization of methane and glycerol using engineered M. alcaliphilum 20Z. Two pathways are proposed for glycerol utilization in methanotrophs, involving the GlpK‐GlpD route and the GlpA‐DhaK route. Glycerol crosses the cell membrane by passive diffusion or via a glycerol‐uptake facilitator (GlpF). In cells, glycerol is incorporated into central metabolism as dihydroxyacetone phosphate (DHAP), a metabolite that can participate in both gluconeogenic and glycolytic processes. Both pathways were constructed in a pAWP89 vector under control of the P tac promoter. For 2,3‐BDO production, three‐gene cluster budABC encoding acetolactate decarboxylase (BudA), acetolactate synthase (BudB) and acetoin reductase (BudC) were integrated into the chromosome of M. alcaliphilum 20Z under control of P tac promoter. Ru5P: ribulose 5‐phosphate, H6P: hexulose 6‐phosphate, F6P: fructose 6‐phosphate, KDPG: 2‐keto‐3‐deoxy 6‐phosphogluconate, F1,6BP: fructose 1,6‐bisphosphate, DHAP: dihydroxyacetone phosphate, G3P: glyceraldehyde 3‐phosphate, PEP: phosphoenolpyruvate, OAA: oxaloacetic acid, G3P (convert from glycerol): glycerol 3‐phosphate, DHA: dihydroxyacetone. ICM: intracytoplasmic membrane. Fig. 2 Glycerol utilization of M. alcaliphilum 20Z. A. Growth rate of glycerol‐utilizing strain 20Z_FKDA and glycerol consumption. B. Effect of glycerol concentrations on growth of the 20Z_FKDA strain. C and D. Effect of adaptive laboratory evolution on the growth rate of 20Z_FKDA at higher concentration of glycerol. More than 50 repetitions of sub‐culturing for adaptive laboratory evolution were carried out with gradually increased glycerol concentration until reaching 1% (v/v) ˜ 10 g l −1 . Data represent mean and standard derivation (SD) of two independent experimental replicates. Given that glycerol as an abundant, cheap substrate, it is significant to screen strains which can grow on the maximum glycerol concentration to enhance cell growth. To define allowing elimination for growth and effective metabolism on glycerol, the 20Z_FKDA strain was cultured on medium containing glycerol concentrations of 0.5% (v/v) and 1% (v/v), but the growth rate was even lower than that on the 0.1% (v/v) glycerol concentration and resulted in dead phase earlier than usual (Fig. 2B ). It is inferred that glycerol could significantly influence on the osmotic potential and inhibit growth (Szymanowska‐Powałowska, 2015 ). To overcome this issue and adapt to increased glycerol concentrations, we deployed adaptive laboratory evolution for the 20Z_FKDA strain. After 50 cycles of sub‐culturing for adaptive laboratory evolution, the 20Z_FKDA strain could grow on 1% (v/v) glycerol and achieved a significantly higher growth rate compared with cells grown on 0.1% (v/v) glycerol after 60 h (Fig. 2C and D ). Improvement of 2,3‐BDO production by enabling the growth of M. alcaliphilum 20Z on mixed substrate of glycerol and methane Because of the limited growth rates and low product yields of methanotrophs on methane compared with other host strains, commercial production of value‐added compounds is restricted in methanotrophs. In addition to a systematic approach to improving carbon flux towards target products, co‐utilization of alternative, inexpensive substrates represents a feasible strategy to achieve high productivity or yield of desired products. Given its availability, low prices and a high degree of reduction, glycerol is an ideal feedstock to enhance carbon flux in building blocks as well as reducing equivalents for enhancing titre or productivity of value‐added chemical production by methanotrophs. 2,3‐BDO is a bulk platform chemical with a variety of applications in the chemical, plastic, pharmaceutical, cosmetic and food industries (do Carmo Dias et al., 2018 ; Yang and Zhang, 2018 ). Our previous work demonstrated a systematic engineering approach for 2,3‐BDO production in a triple‐mutant strain of M. alcaliphilum 20Z by expressing acetolactate decarboxylase (BudA), acetolactate synthase (BudB), and acetoin reductase (BudC) from K. pneumoniae in a plasmid‐based expression system which produced 68.8 mg l −1 of 2,3‐BDO from methane in batch culture (Nguyen et al., 2018 ). We also found that NADH supply is a key limiting step of 2,3‐BDO production in engineered M. alcaliphilum 20Z (Nguyen et al., 2018 ). It is inferred that the unbalanced reducing equivalents used in methane oxidation by methane monooxygenase cause the low productivity of compounds that requires additional NADH in the biosynthetic pathway (Akberdin et al., 2018 , Bordel et al., 2019 ). Utilizing co‐substrates that provide extra reducing power would therefore be a feasible strategy to compensate the deficiency of reducing power in the biosynthetic pathway to achieve high productivity or yield of reduced compounds like 2,3‐BDO, lactate, ethanol and alanine (Durnin et al., 2009 , Li et al., 2018 ). Theoretically, 20Z_FKDA strain generates excess NADH by activity of NAD + glycerol‐3 phosphate dehydrogenase (GpsA), while the last step of 2,3‐BDO biosynthesis requires NADH as a cofactor for conversion of acetoin to 2,3‐BDO. The coupling of these reactions can self‐balance intracellular NADH/NAD + ratio which affects the yield of 2,3‐BDO. To determine whether glycerol does indeed enhance the reducing power pool in cells, 20Z_FKDA strain was separately cultured on 50% (v/v) methane, 0.1% (v/v) glycerol, and co‐substrates including 50% (v/v) methane and 0.1% (v/v) glycerol to quantify cellular concentrations of NADH, NAD + and NADH/NAD + ratio. To avoid the measurement difference of intracellular concentrations of NADH and NAD + due to the different cell densities, samples were collected at mid‐log phase with the same optical density. The results showed that cytoplasmic NADH concentration on co‐substrates utilization was 0.21 mM which was significantly higher than that of sole methane utilization of 0.17 mM (Fig. 3A ). Similarly, NADH/NAD + ratio significantly increased from 0.59 to 0.93 on methane and co‐substrates utilization, respectively (Fig. 3B ). Moreover, a higher concentration of NAD + was observed on sole methane utilization (Fig. 3A ). Obviously, these results demonstrated that utilization of glycerol could generate extra cytoplasmic NADH and therefore provide more reducing power for 2,3 BDO production. Fig. 3 NADH and NAD + redox status in cell growth of 20Z_FKDA strain on 50% (v/v) methane, 0.1% (v/v) glycerol, and co‐substrates including 50% (v/v) methane and 0.1% (v/v) glycerol at mid‐log phase. (A) NADH, NAD + concentration, (B) NADH/NAD + ratio. (* P < 0.05 methane versus co‐substrates groups by Student’s t ‐test). Data represent mean and standard derivation (SD) of two independent experimental replicates. Although plasmid‐based heterologous expression was promising, the metabolic burden and genetic instability of the plasmid precluded industrial applications (Rugbjerg et al., 2018 ; Rugbjerg and Sommer, 2019 ). Therefore, chromosomal integration of heterologous metabolic pathways is optimal for industrially relevant fermentation (Rugbjerg et al., 2018 ; Rugbjerg and Sommer, 2019 ). Moreover, because 20Z_FKDA already harboured an IncP‐based broad‐host‐range plasmid for expression of glpF, glpK, glpD, and gpsA genes, we integrated a 2,3‐BDO biosynthesis pathway including three‐gene cluster budABC into the chromosome of M. alcaliphilum 20Z to circumvent such problems (Fig. 1 ). The chromosomal integration was conducted to replace the glgA1 loci. The 2,3‐BDO gene cluster from K. pneumoniae under control of the P \n tac \n promoter were amplified from pBudK.p and subsequently ligated into a pCM351‐ glgA1 vector, generating pCM351‐ budABC . The pCM351‐ budABC vector was transformed into M. alcaliphilum 20Z via electroporation, and production of 2,3‐BDO was monitored in flask cultures. The resulting strain produced 17.6 mg l −1 2,3‐BDO in shake flasks after 144 h using methane as a sole carbon source (Fig. 4B ). Likewise, 46 mg l −1 2,3‐BDO was accumulated when glycerol was used as a sole carbon source. Interestingly, there was a dramatic increase in growth rate of 20Z_FKDA_ budABC with mixed substrates, compared to that with a single substrate (Fig. 4A ). The doubling time of 20Z_FKDA_ budABC strain with mixed substrates was 13 h (specific growth rate = 0.052 h ‐1 ), which is twofold faster than that of the 20Z_FKDA_ budABC strain cultured solely on methane, 28 h (specific growth rate = 0.024 h ‐1 ) and 1.2‐fold faster than that of the 20Z_FKDA_ budABC strain cultured solely on glycerol, 16 h (specific growth rate = 0.043 h ‐1 ). In agreement with this observation, 68 mg l −1 2,3‐BDO was produced when 20Z_FKDA_ budABC strain cultured on methane plus glycerol after 144 h with 90 mg methane and 2.5 g glycerol were concurrently consumed for co‐utilization (Fig. 4C ). Fig. 4 Effects of co‐substrate utilization on growth and 2,3 BDO production of the 20Z_FKDA_ budABC strain. A. Growth of engineered strain on sole substrate and co‐substrates, (B) 2,3‐BDO production on sole substrate and co‐substrates. C. The correlation of methane and glycerol consumption to growth and 2,3 BDO production when culturing on co‐substrates. The headspace of the shake flask was refreshed with 50% (v/v) methane every 24 h, and 0.35% (v/v) glycerol (3.5 g l −1 ) was initially supplied for utilization of sole glycerol and co‐substrates. Data represent mean and standard derivation (SD) of two independent experimental replicates. It should be noted that the 2,3‐BDO production when the 2,3‐BDO biosynthesis gene cluster was integrated into the genome was much lower compared to a plasmid‐based system (Nguyen et al., 2018 ), probably due to the lower copy number of the gene of interest on chromosome compared to the case of a multi‐copy broad‐host‐range plasmid (Mustakhimov et al., 2016 ). Likewise, the transcriptional levels of three genes budA , budB, and budC via plasmid‐based expression might be quantitatively higher than that of genome‐integrated expression. To assess this assumption, the transcriptional analysis of budA , budB and budC was conducted to determine the effect of heterologous protein expression manners. The results showed that the transcriptional level of three genes budA , budB, and budC via plasmid‐based expression was higher than that of genome‐integrated expression with statistical significance ( P < 0.05) (Fig. 5 ). Therefore, the development of other replicable plasmids in M. alcaliphilum 20Z is required to facilitate the metabolic engineering in this particular strain. Moreover, some systematic approaches including RBS optimization as well as gene knockout might be useful for further improvement of 2,3‐BDO titre (Nguyen et al., 2018 ). By using a similar approach, glycerol also was co‐utilized with carbon dioxide in an obligate photoautotrophic cyanobacterium to enhance 2,3‐BDO production, the engineered strain of Synechococcus elongatus PCC 7942 produced 761 mg 2,3‐BDO/L, a 290% increase over the control strain under continuous light conditions (Kanno and Atsumi, 2017 ). Although our titres and productivity levels of 2,3‐BDO are still far below those required by industry, the results demonstrated a proof of concept for utilizing glycerol as a co‐substrate to enhance the titre of not only 2,3‐BDO, but other industrial chemicals in a methanotroph‐based biorefinery. Fig. 5 Comparison of real‐time qPCR analysis of budA , budB, and budC gene expression via plasmid‐based expression and genomic‐integrated expression in M. alcaliphilum 20Z . This bar graph represents the fold changes of gene expression quantified by normalization to the RNA polymerase sigma factor rpoD (* P < 0.05, plasmid‐based expression versus genome‐integrated expression groups by Student’s t test). Data represent mean and standard derivation (SD) of three independent experimental replicates. Development of one‐step biocatalyst for direct conversion of methane to methanol Methane‐to‐methanol conversion has received research attention due to the demand for sustainable technologies of gas‐to‐liquid conversion. Recently, the ‘methanol economy’ has become a promising alternative solution to fossil fuels because it is suitable for use in the current transportation fuel infrastructure, has a greater energy density, and burns with fewer toxic by‐products. Due to the high bond energies of the C–H bond in methane among organic substrates, the development of efficient catalysts for direct conversion of methane to methanol remains one of the most challenging tasks in catalytic chemistry (Kamachi and Okura, 2018 ). Enzyme‐catalyzed oxidation of methane to methanol has several advantages over thermochemical oxidation reactions, including higher selectivity, improved process efficiency and safer, milder reaction conditions and energy savings, all leading to associated economic benefits (Hwang et al., 2018 ). Conversion of methane to methanol using methanotrophs has been considered as an efficient model for methane liquefaction compared with traditional methods (Hwang et al., 2018 ). To date, whole‐cell biocatalysts are a preferred option for bioconversion of methane to methanol due to the relatively low cost of this method. In brief, these bioconversions involve culturing cells on methane during the growth phase and supplying suitable bioconversion inhibitors for MDH. It is clear that accumulation of methanol cannot be achieved by wild‐type methanotrophs because methanol is an intermediate metabolite in methane metabolism. To prevent oxidation of methanol, MDH inhibitors are absolutely needed such as phosphate, ethylenediaminetetraacetic acid, cyclopropanol, and high concentrations of NaCl (Hwang et al., 2014 , Sirajuddin et al., 2014 , Miyaji et al., 2019 ). Despite being the currently preferred option, whole‐cell biocatalysts suffer from several disadvantages, including difficulties at high cell densities, biphasic growth processes for cell growth, and methane conversion by the addition of chemical MDH inhibitors that can also inhibit other key enzymes. On the other hand, whole‐cell bioconversions require supplemental exogenous NADH to enhance MMO activity (Haynes and Gonzalez, 2014 ; Hwang et al., 2014 ), reducing its cost‐effectiveness at commercial scales (Weckbecker et al., 2010 , Uppada et al., 2014 ). An endogenous regeneration system of reduced nicotinamide cofactors is therefore crucial. Theoretically, utilization of co‐substrates including glycerol and methane can circumvent this issue. To be more precise, glycerol metabolism in our engineered strain can generate reducing equivalents in terms of NADH and FADH 2 , which deposit their electrons to quinone pool via complex I of electron transport chain or membrane‐bound FAD + ‐dependent G3P dehydrogenase and subsequently used to activate particulate methane monooxygenase (pMMO) (Fig. 6 ). Fig. 6 One‐step biological conversion of methane to methanol. To accumulate methanol without using inhibitors, mxaF and xoxF genes of a glycerol‐utilizing strain of M. alcaliphilum 20Z were knocked out to prevent expression of methanol dehydrogenase. To support growth of the mutant strain, glycerol is used as a co‐substrate with methane and plays a main role in cell growth, and methane is directly converted to methanol by pMMO. In addition, glycerol metabolism is proposed to provide reducing equivalents in terms of NADH and FADH 2 , which deposit their electrons to quinone pool via other membrane components such as the complex I/NADH dehydrogenase (NDH‐I) or FAD + ‐dependent G3P dehydrogenase and are subsequently used to activate particulate methane monooxygenase (pMMO). UQ: Ubiquinone, UQH 2 : Ubiquinol, ICM: intracytoplasmic membrane. Gk: glycerol kinase, FAD‐Gpdh: FAD + dependent glycerol 3‐phosphate dehydrogenase, NAD‐Gpdh: NAD + dependent glycerol 3‐phosphate dehydrogenase, Mdh: Methanol dehydrogenase, DHAP: dihydroxyacetone phosphate, G3P: glycerol 3‐phosphate. Our engineered methanotroph strain possessed broader metabolic capacities with the expanded substrate spectrum from C1 (methane) to C3 (glycerol) compounds, which endowed a possibility of removing methanol dehydrogenase that is essential enzyme of C1 metabolism in methanotrophs (Fig. 6 ). In which, glycerol was utilized to sustain cell growth and provide reducing equivalents for pMMO activity, while methane or ethane was directly converted to methanol or ethanol without further oxidation. In M. alcaliphilum 20Z, methanol oxidation is catalyzed by the periplasmic pyrroloquinoline quinone‐linked methanol dehydrogenase including calcium‐dependent MDH and lanthanum‐dependent MDH (Vuilleumier et al., 2012 ). Genes encoding calcium‐dependent MDH, mxaFI , and the lanthanum‐dependent MDH, xoxF from M. alcaliphilum 20Z were completely knocked out (Fig. S3 ), resulting in the 20Z_M2 strain. As we expected, the 20Z_M2 strain no longer grew on C1 substrates (Fig. S2 ), indicating complete deletion of methanol dehydrogenase. We successfully carried out one‐step conversion of methane to methanol using the 20Z_M2 strain without the addition of any chemical inhibitor of MDH and exogenous NADH. The 20Z_M2 was cultured on medium containing glycerol for 24 h, and methane was introduced to the headspace at a concentration of 50 % (v/v). As a result, methanol was accumulated to a concentration of 11.6 mM after 3 days with a yield of 0.484 g methanol g −1 methane (Fig. 7A , Table S2 ). The methanol production obtained here was higher than methanol production achieved from whole‐cell M. alcaliphilum 20Z biocatalyst which showed the titre was only 0.022 mM (Patel et al., 2016 ). Fig. 7 (A) Conversion of methane to methanol using living cells of 20Z_M2 strain cultured in a shake flask. Cells were cultured on glycerol for 24 h; the headspace of the shake flask was refreshed with 50% (v/v) methane every 24 h for 72 h. B. Conversion of methane to methanol using resting cells of 20Z_M2 strain. The reactions were carried out using 2.29 g DCW l −1 (OD 600 : 6.8) resting cell with 30% (v/v) methane in the presence of 40 mM formate, buffer pH 7. Data represent mean and standard derivation (SD) of two independent experimental replicates. Even though we successfully demonstrated that bioconversion of methane to methanol using microbial cell factory system without the addition of any chemical inhibitor of MDH and reducing power‐generating sources such as formate, the yield of this bioconversion was rather lower than expected. In general, it is difficult to gain theoretical yield due to several reasons. Firstly, it should be considered that methane‐to‐methanol conversion also requires NADH as a second substrate while reducing power generated from glycerol metabolism is not only used for methane conversion but also for making ATP via electron transport chain to maintain cell metabolism. Even though methane is captured into cell cytoplasm, but it does not mean that 100% percentage of methane is trapped by methane monooxygenase and converted into methanol due to deficiency of reducing power as well as the distribution of MMO units on intracytoplasmic membrane system in case of cell growth on glycerol. Therefore, supply of reducing equivalents‐generating sources is useful to enhance activity of MMO (Hwang et al., 2015 , Oh et al., 2019 ). Secondly, it might be due to methanol excretion rate constraint as well as non‐specific oxidation of produced methanol by certain alcohol dehydrogenases which can convert methanol to downstream products, although the catalytic efficiency is low compared to methanol dehydrogenase. Therefore, reaction systems mimicking ideal conditions such as the cell‐free system in the absence of other enzymes that degrade the product need to be developed to enhance the yield of products. To overcome drawbacks from using living cells as a biocatalyst, we also designed a whole‐cell biocatalyst to convert methane to methanol using non‐growing resuspended cells or resting cells of 20Z_M2 strain. This approach could support facilitating product separation or preventing secondary reactions or undesired metabolites during cell growth, which could affect the bioconversion (Jackson et al., 2019 ). To do this, mutant cells were firstly cultured on glycerol to gain high optical density, and then, resting cells were used as a biocatalyst. Since glycerol is only used for cell growth phase not for biotransformation phase, this might cause the lack of NADH source for bioconversion of methane to methanol. To overcome this issue, 40 mM sodium formate was added to replenish NADH. Reaction process was carried out in 20 ml serum bottles with 30% (v/v) methane in the headspace. As a result, conversion yield was enhanced up to 79% with 76 mM methanol accumulated during 3.5 h (Fig. 7B ). Obviously, the use of resting cells may be a good alternative to gain higher bioconversion yield which is limited by living cells. Therefore, the optimization of some factors such as additional reducing power source, increased optical density or downstream processing might help to further enhance bioconversion yield of methane to methanol using living cells for fermentative process. In summary, we demonstrated a novel framework for direct conversion of methane to methanol in the absence of inhibitors and exogenous NADH. We expect this strategy to be applied for conversion of other alkanes that correspond to primary alcohols using methanotrophic biocatalyst."
} | 8,891 |
34841054 | PMC8614107 | pmc | 1,012 | {
"abstract": "The extraction and\nsubsequent separation of individual rare earth\nelements (REEs) from REE-bearing feedstocks represent a challenging\nyet essential task for the growth and sustainability of renewable\nenergy technologies. As an important step toward overcoming the technical\nand environmental limitations of current REE processing methods, we\ndemonstrate a biobased, all-aqueous REE extraction and separation\nscheme using the REE-selective lanmodulin protein. Lanmodulin was\nconjugated onto porous support materials using thiol-maleimide chemistry\nto enable tandem REE purification and separation under flow-through\nconditions. Immobilized lanmodulin maintains the attractive properties\nof the soluble protein, including remarkable REE selectivity, the\nability to bind REEs at low pH, and high stability over numerous low-pH\nadsorption/desorption cycles. We further demonstrate the ability of\nimmobilized lanmodulin to achieve high-purity separation of the clean-energy-critical\nREE pair Nd/Dy and to transform a low-grade leachate (0.043 mol %\nREEs) into separate heavy and light REE fractions (88 mol % purity\nof total REEs) in a single column run while using ∼90% of the\ncolumn capacity. This ability to achieve, for the first time, tandem\nextraction and grouped separation of REEs from very complex aqueous\nfeedstock solutions without requiring organic solvents establishes\nthis lanmodulin-based approach as an important advance for sustainable\nhydrometallurgy.",
"conclusion": "Conclusion We have demonstrated a reusable, all-aqueous REE extraction and\nseparation platform using an REE-selective protein chelator immobilized\non a biorenewable support. This concept represents a crucial step\ntoward sustainable REE production and minimizing global dependence\non primary REE resources. Immobilized LanM retained its remarkable\nREE selectivity and facilitated near-quantitative REE separation from\nnon-REE impurities. Followed by a sequential pH gradient or mild chelator\ntreatment, coextracted Nd/Dy, the most critical REE pair for E-waste\nrecovery, can be separated to high purity (>99.9% purity) within\n1\nor 2 adsorption/desorption cycles depending on the feed ratio. We\nfurther demonstrated the application of LanM for REE extraction and\nseparation into heavy and light REE groups in a single adsorption/desorption\ncycle, even starting from a highly complex feed solution with low\nREE content. Our continued efforts are directed at increasing the\nadsorption capacity to improve productivity, improving selectivity\nagainst process impurities, and fine-tuning the REE separation process.\nHowever, our process already has several unique advantages over prior\nart, including its compatibility with low-grade feedstock leachates,\nits lack of organic solvents, and its ability to achieve high-purity\nseparation of certain critical REEs while using up to ∼90%\ncolumn capacity. Consequently, this work establishes lan-modulin as\nthe basis for an ecofriendly process for not only REE recovery but\nalso tandem separation that is broadly applicable to both high-grade\n(e.g., scrap NdFeB magnets) and low-grade (e.g., fly ash) REE feedstocks.\nSustainable hydrometallurgical approaches such as the one proposed\nhere will be necessary to implement a coherent transition from the\nfossil fuel era to low-carbon energies.",
"introduction": "Introduction Rare\nearth elements (REEs), comprising the lanthanides, yttrium,\nand scandium, are essential for the transition from the fossil fuel\nera into the low-carbon era. 1 Five REEs\n(Tb, Dy, Eu, Nd, and Y) in particular have been highlighted by the\nU.S. Department of Energy for supply vulnerability and criticality\nfor clean energy technologies, such as electric vehicles, wind turbines,\nand LEDs. 2 Conversely, however, current\nREE extraction and separation processes require high energy consumption\nand pose severe environmental burdens that impede the development\nof a diversified REE supply chain and undercut the environmental benefits\nof clean energy technologies. 3 , 4 To meet the REE demands\nof the emerging clean energy technology market, it is thus imperative\nto develop new processing methodologies that enable environmentally\nfriendly REE extraction from current and future resources. Owing\nto their similar chemical properties and co-occurrence in\nREE-bearing deposits, 5 , 6 the separation among REEs is particularly\ndifficult, accounting for ∼30% of the total environmental impact\nduring REE production. 3 , 7 , 8 Currently,\nREE separation is dominated by organic solvent-intensive hydrometallurgy\nprocesses that involve a primary REE separation from impurities in\nacid leachate solutions and subsequent group or individual REE separation\nby liquid–liquid extraction. 6 In\norder to achieve high-purity single REEs, liquid–liquid extraction\nmay require hundreds of stages (i.e., mixer-settlers or pulsed columns),\nwhich requires a large process footprint and high investment cost\nand generates a large volume of secondary liquid waste. Furthermore,\nfor environmental and economic reasons, liquid–liquid extraction\nis typically not directly compatible with dilute REE leachate solutions\n(REEs < 1%) from unconventional sources, such as industrial wastes\n(e.g., coal fly ash, mine tailings) and end-of-life consumer electronics\n(e.g., electronic waste, or “E-waste”), given the challenge\nof REE preconcentration from these sources relative to high-grade\nores. In addition, such dilute leachate solutions require a higher\naqueous to organic phase ratio, resulting in large separation units\n(i.e., high capital expenditure), longer equilibration time, higher\nenergy cost, and losses of extractant dispersed into aqueous systems. 9 To reduce solvent losses, the liquid–liquid\nextraction process has been adapted for column chromatography by dissolving\nREE-selective ligands in organic solvent and loading within a solid\nsupport. 10 − 12 However, this physical impregnation strategy still\ninevitably results in leaching of the stationary liquid phase, causing\ncross-contamination and limited reusability. 13 Ion exchange chromatography (IX) is an alternative to liquid–liquid\nextraction for intra-REE separation. 14 In\nconventional IX, REEs are separated based on differences in the metal\nion migration rates in the presence of a chelating eluent [e.g., citrate\nor ethylenediaminetetraacetic acid (EDTA)] or by using a selective\nstationary phase (e.g., iminodiacetic acid-bound silica). IX approaches\ncan achieve ultrapure (99.9999%) individual REEs in a single run with\nhigh yield 15 but are typically restricted\nto applications where small quantities of ultra-high-purity REEs are\nrequired (e.g., electronics or analytical applications) given the\nhigh cost and low productivity. 15 However,\nrecent advances in process simulations and design methods have facilitated\nIX methods with higher productivity. 16 For\nexample, a two-zone ligand-assisted displacement approach separated\na ternary REE mixture (Pr, Nd, and Dy with a composition typical of\nmagnet waste) into individual components (∼99%) at a production\nscale that is competitive with solvent extraction. 17 Nevertheless, the application of IX for intra-REE separation\nfrom feedstocks containing significant non-REE impurities remains\na challenge considering the low REE selectivity of the resins. Solid–liquid extraction, whereby chemical ligands with high\nREE affinity are covalently anchored onto solid support resin, provides\nan effective means to achieve REE recovery from complex feedstock\nleachates. 18 , 19 Here, REEs are selectively adsorbed\nto the resin until the column is saturated, at which point a stripping\nsolution is applied to elute the REE concentrate. This process offers\nseveral advantages relative to liquid–liquid extraction, 18 , 20 − 22 including faster phase separation between the solid\nadsorbent and REE-bearing solution and high stability for reuse. 23 However, most of the solid-phase adsorbents\nemployed for REE separation are based on existing chemical ligands,\nwhich limits the intra-REE separation potential of the process. 24 The incorporation of biological ligands\ninto a solid–liquid\nextraction process offers the potential for novel chemistries and\nenvironmentally sustainable REE separation process development. 25 , 26 For example, lanthanide binding tags (LBTs), short peptides that\nhave been engineered for affinity and selectivity toward REEs, have\nbeen displayed on biomaterial surfaces (cells, curli fibers, etc.)\nand employed in solid–liquid extraction for selective recovery\nof middle and heavy REEs from various feedstock leachates. 27 − 29 However, the low selectivity against Cu 2+ and negligible\nREE binding below a pH of 5 limit the feedstock compatibility of LBT. 27 , 30 More recently, the discovery of lanmodulin (LanM), a small (12 kDa)\nprotein that is involved in lanthanide trafficking in methylotrophic\nbacteria, holds promise as a new ligand for process development. 31 Biochemical and biophysical characterization\nof LanM revealed remarkable selectivity for REEs against non-REE cations,\nthe ability to bind REEs down to pH ≈ 2.5, and uncommon robustness\n(relative to other proteins) to repeated acid treatment cycles. 32 Despite the protein’s high affinity,\nREE desorption can be induced by relatively mild treatments (lowering\npH or common chelators like water-soluble carboxylates), allowing\nmultiple cycles of binding and desorption. Furthermore, LanM also\ndisplayed an unusual preference toward middle–light REE over\nheavy REE; although modest [ K d(Dy) / K d(Nd) ∼ 5, ( K d = dissociation constant)], such a preference might allow for efficient\nseparation of light/heavy REE groups or even important REE pairs. 31 Here, we report a bio-material-based,\nall-aqueous REE extraction\nand separation scheme using the REE-selective LanM protein chelator.\nTo enable facile protein reuse, we immobilized LanM onto porous agarose\nmicrobeads, which are biorenewable and commercially available. 33 The resulting biomaterial allowed effective\ngrouped REE extraction from unconventional, low-grade REE feedstocks.\nFurthermore, by exploiting the preference of LanM for middle–light\nREE, we demonstrate proof-of-concept high-purity separation between\nREE pairs (Nd/Dy, and Y/Nd) and grouped separation between heavy REEs\n(HREE, Tb–Lu + Y) and light REEs (LREE, La–Gd). As such,\na key advantage of this approach over liquid–liquid extraction\nis the combination of primary REE extraction from non-REEs and secondary\nseparation between heavy and light REEs within a single, all-aqueous\nadsorption/desorption cycle. This advance greatly simplifies further\nprocessing and provides the basis for a full protein-based scheme\nfor efficient REE extraction, concentration, and separation from both\nhigh- and low-grade feeds.",
"discussion": "Results and Discussion Immobilization of LanM\nonto Agarose Microbeads To facilitate\nthe application of LanM for REE recovery in a flow-through format,\nwe immobilized a LanM variant containing a C-terminal cysteine residue\nwith a GSG spacer (hereafter LanM) on agarose microbeads using thiol-maleimide\nclick chemistry ( Figure 1 A). 34 , 35 Compared with other immobilization strategies,\nsuch as physical binding and entrapment, site-specific covalent attachment\nenables a stable and surface-accessible protein display, which is\ndesirable for repeated cycles of adsorption/desorption under harsh\nconditions (e.g., low pH and high ionic strength). Thiol-maleimide\nchemistry was pursued given the ease of encoding a terminal cysteine\nresidue and the linkage’s stability at low pH ( vide\ninfra ). This construct exhibited REE binding consistent with\nthe wild-type protein ( Figure S1 ). Purified\nLanM was handled at low pH (see the Methods section in the SI) to preclude oxidation while avoiding the need\nfor a reducing agent, which we found to adversely affect LanM immobilization\n(data not shown). Maleimide and LanM functionalization of the agarose\nbeads was confirmed by FT-IR ( Figure 1 B; Supporting Information ), and the immobilization kinetics were monitored by quantifying\nthe free LanM concentration in the conjugation solution over a 16\nh conjugation reaction. Approximately 97% of the added LanM was loaded\nwithin 3 h ( Figure 1 C), resulting in an immobilization density of 2.47 ± 0.54 μmol\nLanM/mL agarose. To visualize the distribution of LanM within the\nagarose bead, a fluorescently tagged variant of LanM (FITC-LanM, green)\nwas incorporated during immobilization. Confocal microscopy imaging\nconfirmed a homogeneous distribution of FITC-LanM within the agarose\nmicro-beads ( Figure 1 D). Figure 1 LanM immobilization and characterization. (A) Overall process for\nimmobilization of LanM protein onto agarose microbeads through thiol-maleimide\nclick chemistry. (B) Fourier transform infrared spectroscopy (FTIR)\nspectra of amino-agarose, N -succinimidyl 4-(maleimidomethyl)\ncyclohexane-1-carboxylate (SMCC), maleimide-agarose, and LanM-agarose\n(see the Supporting Information for further\ndetails of interpretation). (C) LanM immobilization kinetics, mean\n± standard deviation for 3 independent measurements. (D) Fluorescence\nmicroscopy image confirming LanM immobilization onto agarose microbeads\nusing FITC-LanM. Scale bar is 100 μm. Bottom images are split\nchannels of the top image. Immobilized LanM Retains the Ability to Bind REEs at Low pH\nand Is Stable for Reuse To test the efficacy of the immobilized\nLanM for REE extraction under flow-through conditions, fixed-bed columns\nwere packed with the LanM conjugates, and influent breakthrough behavior\nwas assessed with synthetic REE-containing solutions. Nd 3+ was selected as a representative model REE due to its abundance\nin REE deposits and high criticality for renewable energy technologies. 2 The Nd breakthrough point (at pH 5) occurred\nafter ∼25 bed volumes, in contrast to 1 bed volume with nonconjugated\nagarose beads, which is due to the passage of the void volume, indicating\nthat LanM retains high affinity for REEs upon immobilization ( Figure 2 A). The adsorption\ncapacity of the LanM column was 5.77 ± 0.67 μmol/mL, which\ncorresponds to a ∼2:1 stoichiometry of Nd per immobilized LanM.\nInterestingly, in solution, LanM binds 3 equiv of REEs ( Figure S1 ), suggesting that one metal site is\ndestabilized or inaccessible upon immobilization, or in the column\nformat. We suggest that this observation reflects the greater lability\nof metal ions bound to EF hand 1, especially at pH ≤ 5, 36 which may lead this site to not be stably occupied\nunder flow-through conditions. The lability of EF hand 1 may also\nexplain the requirement that LanM be added at a 1:2 (REE:protein)\nrather than 1:3 stoichiometry to achieve quantitative REE extraction\nin prior experiments with industrial feedstock leachates. 32 Figure 2 Immobilized LanM enables high-selectivity REE extraction\nat low\npH. (A) Effect of pH on Nd breakthrough. Experimental conditions:\n0.2 mM Nd in 10 mM glycine buffer (pH > 2.2 condition; for pH <\n2.1 conditions, Nd was diluted in HCl solution), 0.5 mL/min flow rate.\nThe control experiment was performed at pH 3 by using a column packed\nwith maleimide-functionalized agarose. (B) Effect of HCl concentration\non Nd desorption. Columns were preadsorbed with 40 bed volumes of\n0.2 mM Nd at pH 3. (C) Column reusability shown by 0.2 mM Nd breakthrough\ncurves for 10 consecutive adsorption/desorption cycles at a flow rate\nof 0.5 mL/min. Desorption condition: 10 bed volumes of pH 1.5 HCl.\n(D) Independent single-element breakthrough curves of Y, La, Nd, Dy,\nand Lu at pH 3. Feed: 0.2 mM. (E) Nd selectivity of the LanM column\nagainst non-REEs. Metal ion breakthrough curves using synthetic feed\nsolution containing 0.2 mM Nd and order-of-magnitude higher concentrations\nof each non-REE at pH 3 (the detailed feed composition and uncertainties\nare listed in Table S1 ). (F) Desorption\nprofile of metal ions following treatment with pH 1.5 HCl. For panels\nA–D, 1 bed volume = 0.94 mL. For panels E and F, 1 bed volume\n= 0.80 mL. Given that our prior work revealed\nthat solubilized LanM can bind\nREEs at a pH as low as 2.5–3, we tested the effect of influent\npH on Nd extraction performance over the pH range 1.7–5 ( Figure 2 A). Our results indicate\nthat immobilized LanM can effectively bind Nd down to pH 2.4, consistent\nwith the results for solubilized LanM. 32 Nd binding to immobilized LanM is reduced by 50% at pH 2.2 and becomes\ninsignificant at pH ≤ 1.7. Taking advantage of the pH\ndependence of REE binding, we tested\nNd desorption by pumping HCl solutions through Nd-saturated columns.\nA sharp desorption peak was observed between 1 and 6 bed volumes with\npH ≤ 1.7 with Nd concentrated by over an order of magnitude\nrelative to the feed solution ( Figure 2 B). In contrast, a pH 2.0 HCl solution yielded a tailed\ndesorption profile with 16.5 bed volumes required to desorb >95%\nof\nthe Nd. Importantly, the LanM-based sorbent was resilient to repeated\nlow pH exposures given that 10 consecutive absorption/desorption cycles\n(pH 3.0 and 1.5, respectively) yielded no reduction in adsorption\ncapacity ( Figure 2 C).\nTo confirm the generalizability of low-pH binding, breakthrough curves\nwith other representative REEs (i.e., Y, La, Dy, and Lu; Figure 2 D) were performed\nat pH 3.0 and yielded indistinguishable results from that of Nd. Overall,\nthe results demonstrate effective and reversible REE binding by immobilized\nLanM under low-pH conditions. LanM Enables High-Purity\nRecovery and Concentration of REEs To test the REE selectivity\nof immobilized LanM, Nd (0.2 mM) breakthrough\nexperiments were conducted with a synthetic feed solution at pH 3.0\nthat contained millimolar concentrations of Mg 2+ , Al 3+ , Ca 2+ , Co 2+ , Ni 2+ , Cu 2+ , and Zn 2+ , which are abundant in REE-containing\nfeedstocks. 15 The breakthrough of Nd occurred\nafter 24 bed volumes ( Figure 2 E), whereas all non-REEs emerged with the void volume. Importantly,\nthe Nd breakthrough curve and subsequent desorption curve were indistinguishable\nfrom the curves observed with a synthetic solution lacking non-REEs\n( Figure 2 E,F), indicating\nthat the behaviors of the REE and non-REE are completely decoupled\nin the process owing to the selectivity of LanM. Lastly, whereas the\nlimited solubility of Fe 3+ precluded its use in the multielement\nexperiment, a breakthrough experiment using a binary Nd/Fe solution\ncontaining citrate to maintain Fe solubility through coordination\nconfirmed the selectivity of LanM for Nd 3+ over Fe 3+ ( Figure S2 and Table S2 ). Collectively,\nthese results show that immobilized LanM retains the high REE selectivity\nof LanM in solution and can be employed to separate REEs from non-REEs. Ligand Competition with Nonimmobilized LanM Reveals Potential\nfor Intra-REE Separations LanM exhibits an inverse affinity\ntrend relative to most REE ligands (e.g., diglycolamides 37 ) with a preferential response to LREEs. 38 However, despite our studies of REE extraction\nin solution 32 and on column ( vide\nsupra ), we had not yet attempted to exploit these affinity\ndifferences for intra-REE separations ( vide infra ). We reasoned that competition with a mild chelator with a preference\nfor HREEs, such as citrate, 39 could allow\nfor enhanced selectivity in a desorptive process. To establish this\nprinciple, we first investigated ligand competition with nonimmobilized\nLanM. We took advantage of our observation that the fluorescence of\nthe sole tyrosine residue in LanM (Y96), which is adjacent to EF hands\n2 and 3, is quenched upon REE binding, which may reflect differences\nin solvent exposure and hydrogen bonding. 31 Titrations of LanM with REE displayed a maximal change in tyrosine\nfluorescence at 2 equiv of REE, which was counteracted upon binding\nof the third equivalent ( Figure 3 A). Our recent work has identified EF1 as the weakest\nof LanM’s three lanthanide binding sites; therefore, we suggest\nthat the first two equivalents of metal bind to EF2/3, and the third\nequivalent (to EF1) perturbs the environment of Y96. 36 This fluorescence change provides a convenient handle with\nwhich to assay conditions for REE desorption from LanM. These studies,\nusing titrations with LanM bound to individual REEs, revealed that\ncitrate selectively outcompeted LanM for HREEs before LREEs, as expected\ngiven LanM’s apparent K d ’s\n( Figure 3 B). For example,\n10 mM citrate is sufficient to fully desorb Dy from LanM, whereas\nNd is still mostly bound to the protein. The sharp desorption profiles\n(occurring over a 4-fold concentration range) are consistent with\ncooperativity in LanM’s metal binding. 31 These results demonstrate that, despite LanM’s high REE affinity,\nmild chelators can be used to desorb REEs. Furthermore, and importantly,\nan adjustment of conditions can be used to selectively desorb HREEs\nfrom LanM without destabilizing LREE-LanM complexes, setting the stage\nfor on-column separations. Figure 3 Spectrofluorometric titrations of LanM. (A)\nFluorometric titration\nof LanM (20 μM) with Nd 3+ , Dy 3+ , and Y 3+ , showing quenching of tyrosine fluorescence upon metal binding.\n(B) Desorption of LanM bound to 2 equiv of each metal using citrate,\nfollowed fluorometrically. Excitation at 278 nm, emission at 307 nm,\npH 5.0. All data points represent mean ± standard deviation for\n3 independent experiments. On-Column Separation between REE Pairs Next, we examined\nwhether judicious selection of desorption conditions could allow on-column\nseparation among REEs, first using individual REE-loaded LanM columns\nover a range of pH ( Figure 4 A) and citrate steps ( Figure S3A ). A representative set of REEs (Y, La, Pr, Nd, Dy, and Lu) was tested\nindependently to cover the entire ionic radius range of lanthanides\n(La and Lu) and based on their criticality for renewable energy technologies\n(Y, Pr, Nd, and Dy). Distinct elution profiles were observed for each\nREE using both desorption options ( Figure 4 A and Figure S3A ). For citrate, the order of REE elution was correlated with the\nionic radius and closely matched the solution data ( Figure 3 ). A similar trend was observed\nfor pH, with the notable exception that Nd required a lower pH for\ndesorption compared to La. This result is suggestive of higher relative\nstability for the LanM-Nd complex compared to LanM-La and is consistent\nwith a local maximum in stability for LanM in the Nd/Pr range. 31 , 32 We anticipated that the differences in stability among tested LanM\nREE complexes could be exploited to achieve separation between certain\nREE pairs. Figure 4 LanM enables high-purity separation of Nd/Y and Nd/Dy pairs. (A)\nAccumulative desorption profiles of single-element independently loaded\ncolumns using a stepwise pH scheme. REE ion desorption was normalized\nto the total REE desorbed. (B) Two-pH desorption scheme of a binary\nREE-loaded column. Experimental conditions: REE solutions (Nd:Y =\n78:22) at pH 3 were used to load the column to 75% saturation before\na pH 2.3 and pH 1.7 desorption step was carried out (feed loading\nand washing not shown). The values above each panel indicate the purity\nof REE over each elution zone. The duration of each pH step is depicted\nby the dark gray dashed line. Summary of the REE distribution in the\ninitial feedstock and three desorption regions relative to the total\nREE in feed solution using (C) a two-pH scheme and (D) citrate-pH\nscheme, respectively. (E, F) Demonstration of Dy/Nd separation using\ntwo coupled adsorption/desorption cycles. (E) A feedstock comprising\na 5:95 mixture of Dy:Nd was subjected to a two-step pH desorption\nscheme that was repeated 5 times ( Figure S4C ). The resulting Dy-rich desorption fractions were combined, adjusted\nto pH 3 by adding concentrated glycine buffer (1 M), and used as a\nfeed solution in a second adsorption/desorption cycle. (F) Two-step\npH desorption scheme with the combined 44% Dy/56% Nd solution. The\nzones of feed loading, washing, and each desorption step are divided\nby vertical dotted gray lines. The detailed feed composition and measurement\nuncertainties are listed in Tables S3 and S5 . For panels A and B, 1 bed volume = 0.94 mL. For panels E and F,\n1 bed volume = 1.0 mL. Guided by the above results,\nwe first tested the ability of LanM\nto separate Y from Nd ( Figure 4 B and Figure S3B ), as both REEs\nare abundant in primary REE deposits (e.g., those bearing monazite,\nxenotime, and/or allanite) and coal byproducts. 6 , 32 The\nproportion of Y and Nd in the feed solution was set to 22% and 78%,\nrespectively, to resemble the HREE and LREE compositions in typical\ncoal byproduct leachates. 40 , 41 The LanM column was\nloaded to ∼75% saturation (∼1 bed volume before breakthrough\npoint, C / C 0 = 0.05) followed\nby either a two-pH desorption scheme or by combining an initial citrate-mediated\nY desorption followed by pH-mediated elution of Nd. Two distinct peaks\nwere collected after the two-step pH desorption, with 95.6% Y purity\nand 99.8% Nd purity achieved, respectively ( Figure 4 C). A small overlap region (<25% of the\nadsorbed metals) of nearly identical composition as the influent feed\nsolution (20.9% Y + 79.1% Nd) was collected and can be recycled as\nfeed solution in a subsequent adsorption/desorption cycle. Remarkably,\ndesorption using 15 mM citrate eluted 95.8% of the adsorbed Y at 99.4%\npurity, while a subsequent pH desorption step eluted 99.7% of the\nNd at >99.9% purity ( Figure 4 D). Thus, with a single adsorption/desorption cycle with the\nLanM-based column, Y can be separated from Nd, supporting the feasibility\nof a HREE vs LREE separation step. We next tested the efficacy\nof the LanM column to separate the\nclean-energy-critical REE pair Dy/Nd by loading the column to ∼75%\nsaturation with a solution of Nd and Dy at a 50:50 molar ratio followed\nby a stepwise desorption process ( Figures S4A and S3C ). For the two-pH scheme, 76.2% of the Dy was eluted\nwith 99.9% purity with an initial pH 2.1 desorption, while 76.8% of\nthe Nd was eluted with 99.9% purity with a second pH 1.7 desorption.\nNotably, both products exceed the minimal purity threshold (99.5%\nrare earth oxide) for salability. 42 Less\nthan 24% of the loaded REE material was present in the peak overlap\nregion; considering the identical Nd/Dy composition as the feed solution\n(50.8% Dy; Figure S4A ), this fraction can\nbe combined with the initial feed solution and processed during a\nfuture purification cycle (hence, avoiding any loss of REE). The citrate/pH\ncombination eluted 94.2% of the adsorbed Dy at a purity of 99.1%,\nwhile a subsequent pH desorption step eluted the remaining Nd (99.2%)\nat a purity of 94.6% ( Figure S3C ). Collectively,\nthese results show that the LanM column can produce high-purity Dy/Nd\nproducts when starting with an equimolar mixture of both metal ions. To test with a feedstock composition that reflects NdFeB magnet-bearing\nE-waste, we loaded the column to ∼75% saturation with a feed\nsolution comprising 95% Nd and 5% Dy. 43 The use of Nd as a surrogate for the combined Nd/Pr content (typically\n3:1 Nd:Pr) is supported by the nearly identical desorption profiles\nof both metal ions as a function of pH or citrate concentration ( Figure 4 A and Figure S3A ). Successful separation of Dy from\nNd/Pr would enable the production of high-value dysprosium and didymium\n(NdPr) oxide products that could be fed back into the magnet manufacturing\nsupply chain. As shown in Figures S4B and S3D , both the two-pH and citrate desorption schemes yielded high-purity\nNd fractions (99.8% and 98.7%, respectively) while significantly upgrading\nthe Dy purity. For example, a pH 2.2 elution step yielded 88.6% of\nthe Dy content at 46.1% purity while the citrate desorption step yielded\n73.9% of the Dy at a purity of 50.3%. Given the earlier results with\n50:50 mixtures, we hypothesized that the roughly 50% purity Dy fractions\ncan be further upgraded to high purity by using a second adsorption/desorption\ncycle. To confirm this hypothesis, we repeated the Nd/Dy separation\nusing the 95% Nd/5% Dy feed 5 times to accumulate a sufficient volume\nof a 44% purity Dy desorption fraction to enable a second separation\nstep ( Figure 4 E and Figure S4C ). Despite the dilute nature of the\nNd/Dy feed solution, the LanM column exhibited a high loading yield\n(>99%) and achieved 88% of Dy at a purity of 99.2% and 82% of Nd\nat\na purity of 99.9%, respectively, following the pH desorption step.\nOur results suggest that the LanM column is able to effectively separate\nDy from Nd starting with low-purity Dy solutions typical of E-waste\nleachates. We anticipate that an even higher yield and product purity\nwill be achieved with higher operational volumes. In addition, further\nREE separations can likely be realized upon the fine-tuning of column\noperation conditions, such as pH, competing chelator identity and\nconcentration, flow rate, and column geometry. Such optimization will\nbe the subject of future studies. To facilitate a comparison\nof our LanM-based separation approach\nwith the prior art, we present the feed composition, operating conditions,\nand product purity achieved in several exemplary REE separation studies\nin Table S4 . When compared to liquid–liquid\nextraction and novel crystallization-based approaches, LanM enables\na comparable Nd/Dy separation efficacy in an equal or fewer number\nof steps without using organic solvent ( Table S4 ). For example, a liquid–liquid extraction process\nusing bis(2,4,4-trimethylpentyl) phosphinic acid (Cyanex 272) in kerosene\nfor Nd/Dy separation concentrated a feed solution containing 6.7%\nDy to ∼47% purity Dy using a single extraction stage with 92.3%\nDy extraction yield. 44 Similarly, a counter-current\nextraction configuration with 2-ethylhexyl phosphonic acid mono-2-ethylhexyl\nester (PC88A) extractant in kerosene required three extraction stages\nto upgrade a 16% Dy/84% Nd feed solution to a 74.5% purity Dy solution,\nwhile a fourth extraction stage only increased the Dy purity to 94.6%. 45 Additionally, novel tripodal ligands have been\nused to separate 50:50 and 25:75 Dy:Nd solutions into 95% purity products\nby leveraging differences in the solubility of the resulting Dy/Nd\ncomplexes in organic solvent (i.e., tetrahydrofuran, benzene, or toluene). 46 , 47 As such, our results suggest that immobilized LanM can be used as\nan effective and environmentally friendly alternative to organic solvent-based\nextraction processes for the production of high-purity (up to 99.9%)\nNd and Dy. Despite being column-based, these stepwise desorption\nschemes using\nimmobilized LanM are fundamentally different from traditional ion\nexchange chromatography. 48 , 49 In our experiments,\n∼75% LanM column capacity was utilized in a single adsorption/stepwise\ndesorption cycle, whereas only a small volume of feed solution (∼2–5%\nwith respect to column size) is processed in a typical chelation ion\nchromatography cycle, due to the requirement of repeatedly partitioning\n(adsorption/desorption) between the mobile and ligand-functionalized\nstationary phases, which yields higher-purity REEs but limits the\nthroughput of the separation process. 49 As such, our approach exceeds the capacity of traditional ion exchange\nchromatography while still achieving a high-purity product. In contrast,\nthe capacity of the LanM approach is lower, but purity is higher compared\nto the recently developed two-stage ligand-assisted displacement chromatography\nthat effectively separated a ternary REE mixture (Pr, Nd, and Dy)\ninto individual components at ∼99% purity. 17 However, it is unclear whether the ligand-assisted displacement\nprocess would be compatible with non-REE impurities present in the\ninitial feed. Based on the results in Figure 2 E, we hypothesized that LanM will select\nagainst such impurities in the adsorption step, thereby uniquely allowing\nthe REE extraction and selective desorption steps to be coupled. Grouped REE Extraction and Separation from a Low-Grade Feedstock\nLeachate Recovering REEs from abundant waste products, such\nas coal fly ash and red mud, provides a potential means to diversify\nthe REE supply while avoiding pollution inherent to mining. However,\nthe leachate solutions produced from such low-grade REE-bearing wastes,\nwhich contain high levels of metal ion impurities, such as Al 3+ , Ca 2+ , and Fe 3+ , are problematic for\ntraditional liquid–liquid extraction approaches. 6 , 15 For example, Al 3+ is commonly coextracted with REEs in\nliquid–liquid extraction, resulting in low REE purity and the\nformation of emulsions or a third phase via gelatinous hydroxides. 15 Similarly, other impurities, such as Ca 2+ , may cause fouling in liquid–liquid extraction processes\nthrough gypsum formation. 50 As such, leachate\nsolutions are commonly subjected to a pretreatment precipitation step\nto remove impurities before feeding into a liquid–liquid extraction\nunit. Although selective precipitation is effective for removing certain\nimpurities, such as Fe 3+ , the complete removal of Al 3+ presents a major challenge as REE hydroxides coprecipitate\nwith aluminum hydroxide. 15 , 51 , 52 Therefore, an REE extraction method with high REE selectivity over\nnon-REE impurities, particularly Al 3+ and Ca 2+ , is highly desirable. As an industrially relevant performance\ntest, we demonstrated LanM column-based REE extraction and separation\nconcepts using an exceptionally low-grade leachate (0.043 mol % REE,\nexcluding monovalent ions) prepared from Powder River Basin (PRB),\nUSA, fly ash. 53 , 54 The leachate contains ∼150\nμM total REEs compared with millimolar levels of Na, Mg, Al,\nCa, and Sr ( Table S6 ) and significant transition\nmetal content (e.g., Zn, Ni, Cu, and Mn). Using the LanM-based column,\nthe breakthrough of REEs occurred after 30 bed volumes, whereas non-REEs\nwere eluted in the void volume ( Figure 5 A). To assess the REE purity of the adsorbed metal\ncontent, the metal ion composition was determined following nonselective\ndesorption using a pH 1.5 solution. Over 96.5% of the REEs were desorbed\nwithin the most concentrated fractions (total 3.9 bed volumes, Figure 5 B and Figure S5 ). A total REE purity of 88.2% was observed\n( Figure 5 C), indicating\na striking 2040-fold increase in purity compared with the feed solution\n(0.043 mol % REEs). Importantly, the radionuclide uranium was not\nconcentrated from the PRB leachate. Figure 5 A LanM column facilitates REE recovery\nand grouped REE separation\nfrom a Powder River Basin (PRB) fly ash leachate solution. (A) Adsorption\nprofiles of metal ions. Adsorption condition: pH 5, 0.5 mL/min. One\nbed volume = 0.8 mL. (B) Metal compositions in PRB feed and recovered\nbiosorption solutions (desorption fractions 2–4 from Figure S5 ). Desorption was performed by using\nthe column used in panel A after 35 bed volumes of adsorption. Note\nthat, because the column adsorption was carried out slightly past\nthe breakthrough point (30 bed volumes), some displacement of the\nheaviest REEs (Ho–Lu) by LREEs had occurred (see Table S6 ). (C) Metal ion percentage in PRB feed\nand recovered biosorption solutions (excluding monovalent ions). (D)\nPurification factor for total REEs relative to selected non-REEs.\nThe values for liquid–liquid extraction were determined from\nthe data of Smith et al., 55 which used\ndi-2-ethyl-hexylphosphoric acid (DEHPA) to extract REEs from fly ash\ngenerated from Appalachian coal using a single-stage liquid–liquid\nextraction with kerosene. Values above the dotted line indicate REE\nselectivity over non-REE. (E) Selective desorption of HREE (Tb–Lu,\nY) and LREE (La–Gd) using a two-step pH scheme. The values\nabove panel D indicate the purity of LREE or HREE versus total REE\ncontents. Three elution zones are divided by vertical dotted lines.\nExperimental conditions: 29.1 bed volumes of PRB fly ash leachate\nwas pumped through a 0.94 mL column, followed by a washing step with\n10 bed volumes of pH 3.5 water. Desorption was performed with 16 bed\nvolumes of pH 2.3 HCl solution and 10 bed volumes of pH 1.7 HCl solution,\nsequentially. (F) HREE and LREE distribution in the PRB feed and each\nelution zone feed from panel E relative to the total REE in PRB. The\ndetailed ion composition and measurement uncertainty are listed in Tables S6–S8 . To benchmark the REE selectivity of the LanM-based solid-phase\nadsorption with a liquid–liquid extraction approach ( Figure 5 D), the purification\nfactor (see the Methods section in the SI for a definition) for total REEs relative to base metal ions was\ncompared with data recently generated by Smith et al. using the commercial\nextractant di-2-ethyl-hexylphosphoric acid (DEHPA), with coal fly\nash leachate of a comparable composition. 55 The LanM-based approach exhibited significantly higher selectivity\nagainst all non-REE impurities (purification factors of between 100\nand 500 000 for the LanM column, depending on the element,\nversus only 10–5000 for the liquid–liquid extraction\nprocedure) with the exception of Fe and Si. The high selectivity of\nLanM for REEs over Mg 2+ , Al 3+ , and Ca 2+ is particularly compelling considering their abundance in low-grade\nfeedstock leachates. 41 , 56 This improved selectivity represents\na key advance over our prior biosorption-based REE extraction method\nusing LBT-displayed Escherichia coli cells, which\nwas adversely affected by the high Ca/Mg content of the PRB leachate. 54 This selectivity advantage is expected to be\neven more pronounced at lower pH, where a higher solubility of metal\nion impurities (e.g., Al) is observed. We suspect that the lower relative\nselectivity against Si and Fe reflects the formation of unfilterable\ncolloid particles that accumulate on-column and are dissolved during\nthe low-pH desorption step. 57 As such,\nour current column-based approach necessitates precolumn methods to\nminimize Fe/Si content to maximize the effect of LanM. In sum, these\nresults highlight the ability of the immobilized LanM to selectively\nconcentrate REEs from low-grade leachates containing a wide array\nof metal ion impurities. Having demonstrated effective removal\nof the vast majority of non-REE\nimpurities during the adsorption stage, we next tested the ability\nof LanM to achieve grouped separation of the REEs adsorbed from the\nPRB leachate [72% LREE (La–Gd); 28% HREE (Tb–Lu, Y)]\nusing a two-pH desorption scheme. Grouped separation of REEs into\nHREE and LREE fractions is an important early step during liquid–liquid\nextraction and requires multiple extraction and stripping stages to\nenrich for HREEs, typically involving the use of a different extractant\nfrom that used in the non-REE removal stage. 6 In the case of the LanM-based column, two distinct peaks composed\npredominantly of either HREEs or LREEs were observed ( Figure 5 E); 82% of the HREEs were eluted\nwith 72.6% purity at pH 2.3, while 80% of the LREEs were eluted with\n98.8% purity at pH 1.7 desorption. Importantly, such group separation\nwas achieved with ∼90% REE loading (29.1 bed volumes). We anticipate\nthat even higher-purity LREE and HREE fractions could be achieved\nwith lower REE loading. In previously proposed methods, to achieve\na similar-purity LREE separation from an acid leachate generated from\nion adsorption ores, 11 stages of counter-current extraction through\na stepwise liquid–liquid extraction process involving the sequential\nuse of 2-ethyl-hexyl phosphonic acid mono-2-ethylhexyl ester (HEH(EHP),\nP507) and di(2-ethylhexyl)phosphoric acid (HDEHP, P204) were required. 58 Collectively, these results highlight an unprecedented\nadvantage of the LanM column: the ability to achieve both non-REE\nimpurity removal and grouped REE separation starting from low-grade\nleachates in a single adsorption/desorption step without using organic\nsolvent or hazardous chemicals. Based on our separation data with\nNd/Dy, and Nd/Y pairs, we anticipate that finer separation within\nthe HREE and LREE groups is possible by linking a small number of\nadsorption/desorption steps and/or by judicious incorporation of water-soluble\nchelators in the desorption step. The applicability of this\nLanM-based biosorption technology for\nindustrial-scale REE extraction and separation will require scale-up\nand increased column capacity. While commercial agarose resin was\nemployed in this proof-of-concept work, significant improvements in\nboth porosity and surface area for LanM attachment can likely be realized\nby using other substrates and/or modifications to LanM to increase\nmetal-binding stoichiometry. With respect to scaling, the high expression\nyield of LanM (80 mg/L with no optimization yet performed) offers\npromise for the development of a low-cost purification scheme, while\nthe high stability of the LanM column over multiple adsorption/desorption\ncycles ( Figure 2 C)\nsupports its reuse potential. A further step critical for scaling\nwill be the conversion from a semicontinuous to a continuous process\nby employing a rotating column operation that is commonplace in industrial\nadsorption/desorption operations. 59 In\nconclusion, while the current LanM-based process should not be considered\nas a direct competitor for liquid–liquid extraction—particularly\nwith regard to processing concentrated feed solutions from high-grade\nore sources—a scaled-up, continuous process would be uniquely\npositioned to unlock low-grade leachate solutions that are not currently\nprofitable with traditional hydrometallurgical methods."
} | 10,354 |
36864092 | PMC9981879 | pmc | 1,013 | {
"abstract": "Cyanobacterial biofilms are ubiquitous and play important roles in diverse environments, yet, understanding of the processes underlying the development of these aggregates is just emerging. Here we report cell specialization in formation of Synechococcus elongatus PCC 7942 biofilms—a hitherto unknown characteristic of cyanobacterial social behavior. We show that only a quarter of the cell population expresses at high levels the four-gene ebfG -operon that is required for biofilm formation. Almost all cells, however, are assembled in the biofilm. Detailed characterization of EbfG4 encoded by this operon revealed cell-surface localization as well as its presence in the biofilm matrix. Moreover, EbfG1-3 were shown to form amyloid structures such as fibrils and are thus likely to contribute to the matrix structure. These data suggest a beneficial ‘division of labor’ during biofilm formation where only some of the cells allocate resources to produce matrix proteins—‘public goods’ that support robust biofilm development by the majority of the cells. In addition, previous studies revealed the operation of a self-suppression mechanism that depends on an extracellular inhibitor, which supresses transcription of the ebfG -operon. Here we revealed inhibitor activity at an early growth stage and its gradual accumulation along the exponential growth phase in correlation with cell density. Data, however, do not support a threshold-like phenomenon known for quorum-sensing in heterotrophs. Together, data presented here demonstrate cell specialization and imply density-dependent regulation thereby providing deep insights into cyanobacterial communal behavior.",
"introduction": "Introduction Cyanobacteria are highly abundant photosynthetic prokaryotes that occupy diverse habitats. These microorganisms are responsible for ~25% of the global CO 2 converted to organic material and the accompanying O 2 released in the photosynthetic process 1 , 2 . Frequently, these photosynthetic prokaryotes are found in microbial assemblages known as biofilms or part of laminated biofilms, dubbed microbial mats 3 – 5 . Phototrophic biofilms are often associated with industrial problems 6 – 8 ; in contrast, such microbial consortia are beneficial, e.g., for effective biomass accumulation for the biofuel industry and for harvesting of secondary metabolites 9 – 12 . In-depth understanding of cyanobacterial biofilm development paves the way for inhibition of deleterious biofilms and promotion of beneficial ones. The mechanisms involved in cyanobacterial aggregation or biofilm formation started emerging only in recent years. For example, similarly to heterotrophic bacteria, cyanobacteria use the second messenger cyclic-di-GMP for regulating aggregated vs planktonic mode of growth 13 . Furthermore, the thermophilic cyanobacterium Thermosynechococcus vulcanus employs cyanobacteriochrome photoreceptors to mediate light-color input for controlling cell aggregation via c-di-GMP signaling 14 – 16 . Microbial cells within biofilms are encased in a self-produced matrix of hydrated extracellular polymeric substances (EPS) that allows multilayering of cells and structural stability and provides a protected environment. Numerous studies in diverse heterotrophic bacteria identified particular sugar polymers and protein filaments as matrix components 17 , 18 , however, little is known about the cyanobacterial biofilm matrix. Yet, extracellular polysaccharides were implicated in cyanobacterial biofilm formation, for example, studies of Synechocystis support involvement of extracellular polysaccharides in surface adhesion 19 and cell sedimentation 20 and cellulose accumulation is responsible for cell aggregation in T. vulcanus RKN 21 . The exo-protein HesF of Anabaena sp. PCC 7120 is required for aggregation and it was proposed that it interacts with polysaccharides 22 , however, its detailed role in aggregation is still unknown. Our previous studies revealed a biofilm self-suppression mechanism in S. elongatus that dictates planktonic growth of this strain (Fig. 1 ). Inactivation of gene Synpcc7942_2071, abrogates the biofilm inhibitory process and results in a biofilm-proficient strain in contrast to the planktonic nature of WT ( supplementary video ). This gene encodes a homologue of ATPases of type 2 secretion (T2S) systems or type 4 pilus (T4P) assembly complexes, thus, the mutant was initially designated T2EΩ but recently renamed PilB::Tn5 23 – 25 . The RNA chaperone Hfq and a conserved cyanobacterial protein (EbsA), which are part of the T4P complex, are also essential for the biofilm suppression mechanism 25 . In addition, we identified four small proteins, each with a double glycine secretion motif, that enable biofilm formation ( e nable b iofilm f ormation with a G G motif EbfG1-4). Fig. 1 Biofilm regulation in S. elongatus by an extracellular inhibitor that dictates transcription of the ebfG -operon. The type 4 pilus (T4P) assembly complex is involved in deposition to the extracellular milieu of biofilm inhibitor(s), which dictate transcription of the ebfG -operon. Low and high abundance of transcripts of this operon are indicated by thin and thick arrows, respectively. FM fresh medium, CM conditioned medium. The T4P complex is involved in deposition to the extracellular milieu of a biofilm inhibitor, a small MW (<3 kDa), heat stable and protease insensitive molecule, which has not yet been resolved to a specific compound 23 , 26 . This inhibitor serves for repression of the ebfG- operon, which is highly expressed in the PilB::Tn5 strain, where the biofilm inhibitory mechanism is abrogated. Expression is, however, low in WT cultures and in PilB::Tn5 cells inoculated into conditioned medium (CM) from a WT culture 23 , 26 . Mass-spectrometry (MS) analyses revealed the presence of EbfG1-4 in CM of PilB::Tn5 26 , 27 . Furthermore, we demonstrated that proteins PteB ( p eptidase tr ansporter e ssential for b iofilm formation), which belongs to the C39 peptidases family, and EbfE ( e nable b iofilm f ormation e nzyme), a homolog of microcin processing peptidases, are implicated in secretion of EbfG1-4 to the extracellular milieu 26 , 28 . The role of EbfG1-4 in biofilm formation, however, was unknown. Here, using a reporter construct we demonstrate that high expression of the ebfG -operon is limited to a small subpopulation of cells of PilB::Tn5. Further characterization indicates cell-surface and biofilm-matrix localization of EbfG4 and strongly supports amyloid nature of EbfG2. Together, the data indicate cell specialization and imply microbial cooperation for production of extracellular components beneficial for the whole population, known as “public-goods”. In addition, the response of the reporter strain to conditioned media harvested at different stages of logarithmic growth of the wild-type (WT) strain suggests a density dependent mechanism in regulation of S. elongatus biofilm development.",
"discussion": "Discussion Previous genetic and point mutation analyses demonstrated the requirement of the EbfG proteins for biofilm formation 23 , 26 . As manifested by analysis in individual cells, expression of the operon encoding these genes varied substantially more in the biofilm-forming mutant PilB::Tn5 than among WT cells (Fig. 2a, b and Supplementary Fig. 1 ). Although ~75% of PilB::Tn5 cells express the ebfG -operon at similar or lower level compared to WT (Fig. 2c ), ~90% of the cells of this strain are found in the biofilm (Supplementary Fig. 3 23 , 26 , 28 ). These findings are consistent with cell specialization during S. elongatus biofilm development. Given that EbfG1-3 are prone to form amyloids (Fig. 6 ) and TEM analysis revealed that EbfG2 forms fibrillar structures similar to amyloids, we propose that a relatively small subpopulation of PilB::Tn5 culture produces matrix components that support robust biofilm development by the majority of the cells. Taken together, we propose cell specialization in production of so-called public goods. Such a phenomenon may confer selective advantage because only a minority of the cells invest resources for the benefit of the population, which gains protection within the biofilm 29 . Cell specialization in microbes have been documented in heterotrophic bacteria, for example, during matrix formation in Bacillus subtilis 30 . Such differentiation, however, was not previously reported for cyanobacterial biofilm development, thus, this study, which uncovers cell differentiation suggests division of labor in communities of these ecologically important photosynthetic prokaryotes. Comparisons of expression of the ebfG -operon in fresh medium and under CM harvested at different time points along growth of WT culture indicate production and secretion of the inhibitor at early culture stages and suggest further accumulation with time and cell density. The amount of biofilm inhibitor present following 12 h growth is sufficient to affect reporter expression (Fig. 2b and Supplementary Fig. 1b ), yet is insufficient for biofilm inhibition (Fig. 2d ). Up to 24 h the inhibitor gradually accumulates (Supplementary Fig. 1b ) and at 48 h the inhibitor reaches levels that consistently inhibit biofilm development (Fig. 2d ). Together, data imply a density-dependent mechanism, however, a gradual impact of the inhibitor is observed rather than a “threshold-like” effect typical of quorum-sensing mechanisms known in heterotrophic bacteria, e.g., induction of the lux-operon 31 – 34 . Cyanobacterial quorum-sensing is largely unknown, although N-octanoyl homoserine lactone was suggested to be involved in quorum-sensing in Gloeothece PCC6909 35 . Biosynthetic LuxI-like proteins, which are responsible for production of acylated homoserine lactones in numerous heterotrophs, however, are not encoded in the majority of cyanobacterial genomes 36 , therefore, such molecules are not likely to represent a general mechanism for cyanobacterial intercellular communication. An additional study demonstrated a governing role of extracellular signals produced at high density on transcription of particular genes in low-density cultures of N. punctiforme PCC 73102. These data support the existence of a quorum-sensing-like mechanism(s), however, the nature of the signal(s) and the regulatory network have yet to be identified 36 . Microscopic analysis revealed that some of the cells of PilB::Tn5/EbfG4Ω/EbfG4::FLAG present the EbfG4 protein on their cell surface (Figs. 4 a and 5 ). Interestingly, EbfG4 is only observed on the surface of clustered cells and cells that lack EbfG4 labeling are dissociated from the clusters (Figs. 4 a and 5 ). Together, these observations are in accordance with an adhesion function of EbfG4. In addition, EbfG4 was observed in the intercellular space (Figs. 4 and 5 ), consistent with the hypothesis that it serves as a matrix component. It is possible that similarly to the adhesin protein SasG of Staphylococcus aureus 37 , EbfG4 is initially deposited to the cell surface and later is shed to the extracellular matrix. The role of EbfG4 in the matrix is unknown, however, although it does not form amyloids by itself (Fig. 6 ) it may be associated with amyloid structures formed by EbfG1-3. When establishing biofilms, microbes require a resilient scaffold on which the cells can settle. Bacteria from diverse ecosystems have solved this problem by producing and releasing functional amyloids into their environment 38 , 39 . Amyloid proteins are able to assemble into long and strong fibrils, which can withstand chemical and physical stresses 40 . However, the production of amyloids is a process that can easily get out of control, therefore, it requires a complex and dedicated machinery for appropriate manufacturing. Here, we have investigated the amyloid forming capabilities of the ebfG -operon proteins and found strong evidence supporting amyloid formation in EbfG1-3, and most manifestly in EbfG2. Modeling of the amyloid hotspot peptide within EbfG2 revealed arrangement in a steric zipper of antiparallel fashion (Supplementary Fig. 5 ), characteristic of amyloid proteins 41 . EbfG4, which has a prominent role in biofilm formation, however, did not spontaneously form amyloid fibrils. Consistent with homology in the amyloid hotspots, we hypothesize this could be related to a mechanism intended to control aggregation. By separating the amyloid nucleators, in this case EbfG1-3, from other components of the fibril, e.g. EbfG4, better control over the synthesis of amyloids could be achieved. This would be analogous to, for example, the functioning of CsgB and CsgA in the production of curli, the biofilm backbone in E. coli 42 . As observed in the TEM pictures, the EbfG2 fibrils were much shorter than the positive control and did not bundle together. Heterogeneous fibrils formed of several EbfG proteins could result in more stable fibrils, as seeding of amyloids composed of perfect repeats has been shown to cause fragmentation 43 . Given the amyloid nature of EbfG1-3 proteins one may speculate that these matrix components of S. elongatus biofilms assist in recruitment of additional cells of this cyanobacterium or of other microbes for establishment of multispecies biofilms. Recruitment of cells that have not yet initiated the synthesis of exopolysaccharides by a proteolysis product of the matrix protein RmbA was demonstrated in Vibrio cholerae 44 . Our findings pave the way for controlling formation of unwanted biomats, by using amyloid disrupting compounds, as already shown for other bacteria 45 , 46 . On the other hand, intentional use of protein seeds could facilitate stronger amyloids and hence elicit formation of beneficial biofilms."
} | 3,449 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.